Year 2009 PhD Thesis Development and validation of methods used to compute time values of indoor daylight illuminances Presented in front of: University of Ljubljana Faculty of Electrical Engineering L’Institut National des Sciences Appliquées de LYON Formation doctorale: Génie Civil Ecole doctorale: M.E.G.A. By: Matej Bernard KOBAV Presented on January 30, 2009 Reviewer Reviewer Thesis supervisor Thesis supervisor Member Member prof. dr. Peter Žunko prof. dr. Pierre Ineichen prof. dr. Grega Bizjak prof. dr. Dominique Dumortier prof. dr. Marc Fontoynont prof. dr. Stanislav Kovaèiè Research laboratories: Laboratoire des Sciences de l'Habitat, DGCB, Ecole Nationale des Travaux Publics de l'Etat, rue Maurice Audin, 69120 Vaulx-en-velin, France. Laboratory of the Lighting and Photometry, Faculty of Electrical Engineering, University of Ljubljana, Tržaška 25, SI-1000 Ljubljana, Slovenia Dedicated to Manja, Ema and Matija Acknowledgments Acknowledgments This PhD thesis is a result of my post-graduate study at Faculty of Electrical Engineering and very good collaboration with Laboratoire des Sciences de l’Habitat, Ecole Nationale des Travaux Publics de l’Etat, Vaulx-en-Velin. I wish to thank my two mentors for their support, patience and spent time. Their gentle but firm direction has been most appreciated. Dr. Grega Bizjak was particularly helping in guiding me toward the finish. Dr. Dominique Dumortier had always the right answers to all my questions and doubts. He was more that just a mentor when I was staying in Lyon and he lent me a working table in his office when I was at ENTPE. Since working conditions are always an important issue, I would like to thank also my mates in the office. Boštjan, Tomaž, Ambrož and Uroš, thank you. I would also like to thank Manja for being with me, giving me all the needed support and love and giving birth to two beautiful children, who fill me up with energy and love. Finally I would like to thank my parents and brother and sister for giving me support to continue with my postgraduate study. I Acknowledgments II Table of contents Table of contents ACKNOWLEDGMENTS ...................................................................................... I TABLE OF CONTENTS .................................................................................... III LIST OF ABBREVIATIONS AND SYMBOLS ........................................................ 1 ABSTRACT OF THE THESIS ............................................................................... 5 RESUME ........................................................................................................... 9 POVZETEK ...................................................................................................... 13 INTRODUCTION ............................................................................................. 17 1 DESCRIPTION OF THE SKY SCANNER ....................................................... 21 1.1 Sky scanner and measurements ............................................................ 21 1.2 Calibration of the sky scanner ............................................................... 23 1.2.1 Calibration of the sky scanner under artificial sky ....................................... 23 1.2.1.1 Zenith luminance ............................................................................... 24 1.2.1.2 Horizontal and vertical illuminances ..................................................... 26 1.2.1.3 Checking the Li-cor sensors ................................................................ 28 1.2.2 Calibration of the sky scanner under real sky ............................................. 30 1.2.2.1 Calculated vertical illuminances and measured data from IDMP station... 30 1.3 Setting up the sky scanner .................................................................... 33 1.3.1 Description of the IDMP station. . .............................................................. 33 1.3.2 Correction of orientation of vertical illuminance sensors .............................. 34 1.3.3 Correction of orientation of the sky scanner ............................................... 39 2 ANALYSIS OF THE SKY SCANNER DATA ................................................... 41 2.1 Description of the visualization ............................................................. 41 2.2 Defining CIE sky type based on CIE standard ........................................ 43 2.2.1 Determination of the gradation group ....................................................... 45 2.2.2 Determination of the indicatrix group ........................................................ 53 2.2.3 Calculation of the CIE sky type with gradation and indicatrix ....................... 60 2.3 Calculation of the CIE sky type based on Lz/Dv ratio ............................ 61 III Table of contents 2.3.1 Calculation of the diffuse horizontal illuminance (Dv) .................................. 65 2.3.2 Calculation of the vertical illuminances ...................................................... 68 2.4 Calculation of the CIE sky type with Tregenza statistical method ......... 69 2.5 Results of measurements ...................................................................... 70 3 USING DIGITAL CAMERA AS A SKY SCANNER .......................................... 75 3.1 Introduction ........................................................................................... 75 3.2 Transforming sky images to .05D files ................................................... 75 3.3 Linking sky scans and images from digital camera ................................ 77 3.3.1 Flipping and rotating an image ................................................................. 77 3.3.2 Flipping and rotating a sky scanner grid .................................................... 80 3.4 Comparing results derived with sky scanner and digital images ........... 81 3.5 Database of sky luminance measurements ........................................... 84 3.6 Conclusion on the use of digital images ................................................ 87 4 SKY LUMINANCE MODELS ........................................................................ 89 4.1 All weather model for sky luminance distribution - Perez ..................... 89 4.2 ASRC - CIE ............................................................................................. 92 4.3 Igawa All sky model ............................................................................... 92 4.4 Comparison of sky luminance models .................................................... 96 4.4.1 Results of comparison - all cases ............................................................ 101 4.4.2 Results of comparison dependent on CIE sky type .................................... 105 4.4.3 Results of comparison ............................................................................ 109 4.4.4 Results of comparison dependent on sky types and sun altitudes .............. 111 4.4.4.1 Sun altitudes 5 - 200 ........................................................................ 112 4.4.4.2 Sun altitudes 20 - 350 ....................................................................... 114 4.4.4.3 Sun altitudes 35 - 500 ....................................................................... 115 4.4.4.4 Sun altitudes over 500 ...................................................................... 117 4.4.4.5 Problems of Igawa model for intermediate skies ................................. 120 4.5 Conclusion on sky luminance models .................................................. 126 5 MEASUREMENTS IN SCALE MODEL ........................................................ 129 5.1 Introduction ......................................................................................... 129 5.2 Description of the model and method .................................................. 129 5.2.1 Illuminance and luminance measurements in scale model ......................... 132 IV Table of contents 5.2.1.1 Illuminance ..................................................................................... 132 5.2.1.2 Indoor luminances ........................................................................... 132 5.2.1.3 Outdoor luminances ......................................................................... 132 5.2.2 View from a scale model ........................................................................ 134 5.3 Calculation of directional daylight factors ........................................... 136 5.4 Calculation of illuminance values in scale model ................................. 144 5.4.1 Calculation of luminance values of sky elements ...................................... 144 5.4.2 Calculation of illuminance values in scale model with daylight factors ......... 146 5.4.2.1 Daylight factors with DIALux ............................................................. 146 5.4.2.2 Daylight factors with illuminance ....................................................... 147 5.4.3 Calculation of illuminance values in scale model from luminance values from sky scanner ........................................................................................... 148 5.5 Mean bias error between measured and calculated illuminance values in scale model .................................................................................................. 148 5.6 Database of scale model measurements ............................................. 149 5.7 Results of the scale model measurements .......................................... 152 5.8 Conclusions for scale model measurements ........................................ 156 6 CONCLUSION .......................................................................................... 157 CONTRIBUTIONS OF THE THESIS TO THE SCIENCE .................................... 159 STATEMENT OF THE AUTHORSHIP .............................................................. 161 REFERENCES ................................................................................................ 163 v Table of contents VI List of abbreviation and symbols List of abbreviations and symbols List of abbreviations IDMP CCD CIE DDF DF ENTPE FE LASH MA MBD MBE RMS RMSD SLR (camera) International Daylight Measurement Programme Charge-coupled device Commission Internationale de l'Eclairage (International Commission on Illumination) Directional daylight factor Daylight factor Ecole Nationale des Travaux Publics de l’Etat Faculty of Electrical engineering of University of Ljubljana Laboratoire des Sciences de l’Habitat Mean absolute Mean bias difference Mean bias error Root mean square Root mean square difference Single-lens reflex (camera) List of symbols used in chapter 1 fC Calibration factor of sky scanner (110.6), k Factor we defined based on measurements under artificial sky (1.65) SS L Luminance of a sky element in .05D file List of symbols additionally used in chapter 2 a,b,c,d,e Parameters defining gradation and indicatrix groups in Standard CIE S 011/E:2003 /ISO 15469.2004 Az A90 B,C,D,E Angular difference between azimuth of the sun and azimuth of the sky element, Difference in azimuth of ideal L90 sky element and solar meridian Standard parameters defining Lz/Dv curves 1 List of abbreviation and symbols D Diffuse horizontal illuminance V Eh Horizontal illuminance from sky scanner measurement EhCIEx Horizontal illuminance for CIE sky type x ÔE contribution to horizontal illuminance from a single sky patch contribution to vertical illuminance from a single sky patch f(x) Indicatrix function La Luminance of a sky element La ' Measured luminance of a sky element La_rel Relative sky element luminance Lp Luminance of a sky patch Lp Normalized luminance of a sky patch LpCIEx Luminance of a sky patch for CIE sky type x LpCIEx Normalized luminance of a sky patch for CIE sky type x LZ Zenith luminance L Z Measured zenith luminance Normalizing luminance, luminance of sky element with angular distance 900 towards sun Measured normalizing luminance L(90) l(%) Luminance of sky element with angular distance x towards sun l(%) Measured luminance of sky element with angular distance x towards sun nX>l5 Number of patches with scattering angle larger than 15 degrees rmsCIEx RMS error between measured values and CIE sky type x Z Zenith angle of a sky element ZS Solar zenith angle aS Solar azimuth a Azimuth of a sky element y Elevation angle of a sky element yS Solar altitude A or Angular difference in azimuth between sky patch and normal of the plane 2 List of abbreviation and symbols cp(Z) X da Altitude of the almucantar (imaginary circle on the celestial sphere, parallel to the horizon. The circle connects sky elements with same altitudes) Gradation function Scattering angle Solid angle of a sky patch List of symbols additionally used in chapter 4 ap , bp , cp , d p , ep a' , b', c' , d' , e' Ce Cle Ces Eed Ee Ee E eg K L Igawa pIgawa LpSS mbd Se m mbd rdIgawaCIEx rmsd Igawa Igawa rmsd IgawaCIEx rmsd Coefficients in Perez All weather model Coefficients in Igawa model Cloud ratio Cloudless index Standard cloud ratio Horizontal diffuse irradiance Global irradiance Normal incident direct irradiance Extraterrestrial direct normal irradiance Clear sky index Luminance of a patch in Igawa model, Luminance of a patch measured with sky scanner. Mean bias difference between Igawa model and sky scanner measurement, Standard global irradiance Optical air mass Mean bias difference between Igawa model and sky scanner measurement without elements on first almucantar and without element with ?<150 Relative difference for Igawa luminance model for CIE sky type X RMSD between Igawa model and sky scanner measurement RMSD for Igawa luminance model for CIE sky type X RMSD for best luminance model for CIE sky type X 3 List of abbreviation and symbols rmsd Igawa Si s A RMSD between Igawa model and sky scanner measurement and without elements on first almucantar and without element with ?<150 Sky index Sky brightness Sky clearness List of symbols additionally used in chapter 5 DDF Ap DDQ Ap DF E E Ap E AIntp E AExtp E diff E E sky13-A1 MBE sky13 RMSE sky13 Z Z p p ?Sp ?S Directional daylight factor related to a sky element for a point A in the model Directional daylight quotient related to a sky element p for a point A in the model Daylight factor for a given point Total illuminance at a given point Illuminance at a given point contributed by a sky element Indoor illuminance at a given point A, produced by sky element p Outdoor (unobstructed) illuminance at a given point A, produced by sky element p Diffuse illuminance on unobstructed horizontal plane Measured illuminance with Licor sensor in point A1 Calculated illuminance from 13 sky zone luminances in point A1 Mean bias error for sky modelled with 13 zones Root mean square error for sky modelled with 13 sky zones Incidence angle between surface in point and sky element Zenith angle of a sky element Solid angle of a sky element Solid angle of a sky element seen from indoor 4 Abstract of the thesis Abstract of the thesis The presented doctoral thesis concentrates on daylight availability calculations with an emphasis on models of the sky luminance distribution. The main goal of the thesis is to recommend the most appropriate sky luminance model used to calculate time values of indoor daylight illuminances and to define the needed accuracy of the sky luminance model, by the number of zones used in the model. Both conclusions are based on measurements. The first chapter introduces the equipment needed to perform sky luminance distribution measurements. The most important device is a sky scanner. The sky scanner that was used with our research was produced by company EKO and is capable to measure sky luminance distribution with accuracy of 145 sky zones in approximate 3 minutes. Before realization of the measurements the calibration of the sky scanner was carried out. Calibration of the sky scanner was carried out with different methods, firstly under artificial sky. Under artificial sky, zenith luminance and horizontal and vertical illuminances were checked. Second calibration was performed under real sky. In this case we compared calculated horizontal illuminance with illuminances measured on IDMP station. After the sky scanner was calibrated and set up on IDMP station it was possible to correct the orientation of the IDMP vertical sensors for illuminance, since they were not perfectly in line with correct azimuths. In the second chapter, we describe how to analyse luminance data gained with sky scanner. In this part also the CIE S 011/2003 standard is explained with an emphasis on gradation and indicatrix function. Determination of the gradation group is based on calculation of luminance ratio between two elements on every almucantar (imaginary circle on the celestial sphere, parallel to the horizon. The circle connects sky elements with same altitudes) with scattering angle as close as possible to solar zenith angle and zenith luminance. Determination of the indicatrix group is also based on luminance ratio. In this case in the numerator is luminance of an element on certain almucantar and in denominator we can find the normalizing luminance. Normalizing luminance is a luminance of a part of the sky with angular distance 90 degrees towards sun. There are 36 combinations of gradation and indicatrix groups and only 15 of them are listed in the CIE standard. Since some of combinations are really rare, we provide the table how to include 36 combinations in 15 standard CIE sky types. CIE sky type can be also defined with ratio between zenith luminance and diffuse horizontal illuminance. Later on in chapter two, we explain how to calculate diffuse illuminance from sky scan and how to derive CIE sky type from before mentioned ratio. Another method, which is also explained in this chapter, provides CIE sky type based on statistical method. The mentioned method is based on statistical approach. Luminances gained with sky scanner and the ones in the luminance model, described with a standard, are normalised with diffuse horizontal illuminance and compared with each other. If comparison is made between measurements and all 15 CIE sky types, the one with smallest discrepancy can be found. 5 Abstract of the thesis At the end of chapter two results are gathered and frequencies of CIE sky types derived with different methods are shown. Chapter three introduces a digital camera with fish-eye lens as a sky scanner. Our goal was to compare sky luminance values captured with sky scanner and the ones with digital camera and fish-eye. Since luminance values gained with sky scanner are gathered in a text file with sophisticated structure, also the luminances gained with digital camera were with help of Photolux software written in such a file. Both files (for the same time period) were processed in the same manner for 160 cases. As the luminance data were processed it was possible to compare significant sky type characteristics. When gained data were compared, we checked horizontal diffuse illuminance, zenith illuminance, Lz/Dv ratio, gradation group, indicatrix group, sky type based on Lz/Dv ratio, sky type based on statistical method and also all 145 luminance values were compared with each other. In this chapter a reader can find described also the whole procedure how to transform fish-eye image into luminance map and how to extract 145 luminances of the CIE sky partitioning, which is also used by the sky scanner. At the end of the chapter we conclude that a digital camera can be easily (with some limitations) used as a sky scanner. In chapter four, a description of different sky luminance models is given. The emphasis is on the following models: Perez all weather model, ASRC-CIE model and Igawa model. For the times, when we had sky scanner measurements (13.006 cases), sky luminance distributions were modelled with all three models. The modelled luminances are calculated with model's equations and irradiance data measured by IDMP station. In this chapter a problem of high luminance values of the solar corona is exposed. Neither sky luminance models nor sky scanner are capable to model or measure (truthfully) high luminance values of sky elements with low scattering angle towards sun. For this reason in our comparison, all sky elements with scattering angle smaller that 150 were eliminated. This decision is based on calculated mean bias error for all sky elements and only for those with scattering angle higher that 150. Later on, a comparison between three models and sky scanner measurements is introduced. Comparison was carried out with calculation of MB error and RMS error between each modelled and measured sky element. In similar way also CIE sky types were compared. A comparison was carried out for two methods used to derive sky scanner sky type; the one based on gradation and indicatrix group and the statistical one and for two sets of luminance values; for all sky elements without the ones with scattering angle less then 150 and the for all sky elements without the ones with scattering angle less then 150 and the ones on first almucantar. In continuation of chapter four, we present quality of luminance models depending on the sky type and on sun altitude. Conclusion on all comparisons in this chapter is that Igawa model is best for extremely cloudy sky types, Perez All weather model gives best results with intermediate sky types and ASRC-CIE model is in most cases in between upper two models, but always close to the best model. In the fifth chapter, we are talking about scale model measurements. Scale model is fully described and also all measuring problems are listed. In this chapter we introduce sky partitioning into different number of zones (212, 145, 97, 26 and 13) and we also introduce directional daylight factors (DDF) and traditionally used daylight factor. 6 Abstract of the thesis Illuminances in the model were measured with illuminance sensors and later compared with calculated values. The calculated values of illuminances were calculated with directional daylight factors and averaged luminance of the sky zone and with daylight factors and diffuse illuminance measured by IDMP station. In continuation, we describe also the database of scale model measurements (available as an attachment on a CDrom) and the results. Results show us that the number of sky zones has practically no influence to MBE and after that we can conclude that 13 zones are enough to model sky luminance if we are considering only sky vault without direct sun. 7 Abstract of the thesis 8 Résumé Résumé Cette these de doctorat traite des calculs de disponibilité de la lumiere du jour a l’intérieur des bâtiments en s’intéressant particulierement aux modeles permettant de décrire la distribution des luminances du ciel. Son objectif principal est de recommander les modeles les plus appropriés au calcul des valeurs instantanées des éclairements lumineux intérieurs et de définir la finesse du maillage utilisé pour représenter les variations de luminance sur la voute céleste. Le premier chapitre présente le dispositif expérimental utilisé pour mesurer la distribution des luminances du ciel sur le site de l’ENTPE. L’appareil le plus important est le scanner de ciel. Celui que nous avons utilisé est de la marque EKO. Il est capable de mesurer les luminances de 145 points de la voute céleste en 3 minutes. Nous avons d’abord étalonné le scanner. Dans un premier temps, nous avons placé le scanner sous le ciel artificiel de l’ENTPE. Cette source de lumiere parfaitement contrôlée nous a permis de nous assurer de la cohérence des mesures effectuées par le scanner et des valeurs calculées a partir des mesures : luminances, luminance du zénith, éclairement horizontal, éclairements verticaux… Dans un deuxieme temps, le scanner de ciel a été installé sur le lieu de la station de mesure de la lumiere de l’ENTPE. Nous avons vérifié que les cellules de mesure des éclairements verticaux de la station étaient correctement orientées. Nous avons ensuite orienté le scanner puis nous l’avons étalonné de maniere a ce que toutes les informations calculables a partir de ses mesures de luminance soient cohérentes avec les mesures de la station. Le scanner de ciel a été programmé pour effectuer des mesures automatiques des luminances du ciel toutes les 10 minutes pendant un peu moins d’une année. Dans le deuxieme chapitre, nous décrivons comment les mesures de luminances provenant du scanner de ciel ont été analysées. Nous présentons la classification des ciels standardisée par la Commission Internationale de l’Eclairage (CIE) en 2003. Nous expliquons le principe des fonctions de gradation et de diffusion utilisées par la classification et nous montrons comment obtenir ces fonctions a partir des mesures. La détermination du groupe de gradation est basée sur la variation en fonction de l’angle au zénith, du rapport entre les luminances de deux points de mesure ayant le meme angle au zénith et un angle avec le soleil, le plus proche possible de l’angle au zénith du soleil. La détermination du groupe de diffusion est basée sur la variation en fonction de l’angle de diffusion (angle entre le soleil et un point du ciel) du rapport entre la luminance d’un point du ciel et la luminance moyenne des points du ciel situés a 90 degrés du soleil. La combinaison des 6 groupes de gradation et des 6 groupes de diffusion conduit a un total de 36 types de ciels dont 15 seulement font partie du standard CIE. Nous proposons donc une table d’équivalence. Certains auteurs ont proposé d’autres méthodes ne nécessitant pas les luminances de la voute céleste pour déterminer les types de ciel CIE. Nous les avons aussi testées. L’une est basée sur le rapport entre la luminance du zénith et l’éclairement horizontal diffus. L’autre est basée sur une approche statistique comparant la distribution des luminances mesurée avec la distribution des luminances calculée a partir de modeles n’utilisant que des 9 Résumé éclairements horizontaux. La fin de ce chapitre permet de comparer toutes ces méthodes sur la base des fréquences de ciels CIE obtenus pendant la période de mesures. Le chapitre 3 nous permet de présenter et de valider une nouvelle méthode de mesure des luminances du ciel basée sur l’utilisation d’un appareil photo numérique étalonné en luminances. Le systeme appelé PHOTOLUX et développé a l’ENTPE, avait jusqu’ici été utilisé pour l’évaluation du confort visuel dans les ambiances intérieures. Notre objectif était de l’utiliser a l’extérieur en parallele avec le scanner de ciel et de vérifier la qualité des informations qu’il était capable de produire. Nous présentons d’abord le systeme et sa mise en place sur la station. L’appareil est équipé d’un objectif fish-eye a 180° qui lui permet de voir l’intégralité de la voute céleste. Comme pour le scanner de ciel, sa mise en place s’est accompagnée d’une mise en cohérence avec les mesures de la station. Nous avons ensuite utilisé le systeme pendant des journées bien précises. Au total, nous avons effectué 160 mesures. La carte des luminances générée par PHOTOLUX (plus de 400000 pixels) a été utilisée pour calculer les éclairements lumineux horizontaux et verticaux mesurés sur la station. Pour pouvoir comparer ce systeme avec le scanner de ciel, nous avons calculé les luminances moyennes des 145 zones vues par le scanner. Nous avons aussi utilisé le systeme pour produire des informations liées a la classification des ciels de la CIE. Chacune des 160 mesures est présentée dans une fiche constituant une base de données utile pour comprendre la classification des ciels. La fin du chapitre nous permet de conclure que le systeme fournit des informations beaucoup plus riches qu’un scanner de ciel (photo et distribution tres fine des luminances) malgré quelques limitations liées a l’impact du soleil direct. Dans le chapitre 4, nous décrivons différents modeles de luminances de ciel. Nous nous intéressons aux modeles les plus récents ou les plus utilisés : « tout temps Perez », « ASRC-CIE Perez » et « Igawa ». Nous comparons leurs performances sur la base des 13000 mesures effectuées avec le scanner de ciel. Pour tenir compte du fait que ni les modeles, ni le scanner de ciel ne peuvent déterminer avec précision les luminances tres élevées de la zone circumsolaire, nous n’avons pas pris en compte les luminances correspondant aux zones dont l’angle par rapport au soleil était inférieur a 15°. Nous nous sommes intéressés a l’écart moyen et l’écart quadratique moyen entre les luminances des modeles et les luminances mesurées. Nous nous sommes aussi intéressés a la capacité de chacun des modeles a restituer les ciels types de la CIE. Nous présentons la performance de chacun des modeles en fonction du type de ciel et de la hauteur du soleil. Meme si le modele « Igawa » est le plus performant pour la modélisation des ciels couverts. Meme si le « tout temps Perez » est le plus performant pour la modélisation des ciels intermédiaires. Le modele « ASRC-CIE Perez » est souvent proche des deux autres dans les catégories citées précédemment et globalement le meilleur. Le cinquieme chapitre s’intéresse au calcul des éclairements intérieurs et notamment a l’influence de la finesse du maillage utilisé pour représenter les variations de luminance sur la voute céleste. Pour tester les calculs, nous avons mis en place un dispositif expérimental original s’appuyant sur une maquette placée a l’extérieur. La maquette représente un bureau avec une ouverture en façade. Les parois de la maquette ont été 10 Résumé peintes en noir pour supprimer au maximum les réflexions multiples. Nous nous sommes donc intéressés au calcul de la composante directe de la lumiere du jour résultant d’un ciel de distribution de luminance quelconque. Nous avons mesuré les éclairements lumineux sur les parois de la maquette (sol et murs). Nous avons utilisé deux systemes PHOTOLUX (donc deux appareils photos) : l’un pour produire la carte des luminances du ciel vus par l’ouverture, l’autre la carte des luminances des parois de la maquette. Nous avons calculé les éclairements lumineux directs correspondant aux points de mesure dans la maquette en utilisant la notion de facteur de lumiere du jour directionnel (DDF). Ce facteur permet de décrire la contribution d’une zone du ciel a l’éclairement d’un point de la piece. Nous l’avons calculé analytiquement pour chacun des points de mesure dans la maquette et pour chacune de zones du ciel d’un maillage variant de 13 a 212 zones. Nous les avons ensuite combinés avec les luminances moyennes de chaque zone du ciel pour déterminer l’éclairement a l’intérieur de la maquette aux différents points de mesure. Les mesures et les résultats des calculs sont documentées dans des fiches qui facilitent leur interprétation. La comparaison entre les mesures et les calculs montre que la méthode des facteurs de lumiere du jour directionnels fournit des meilleurs résultats que la méthode classique du facteur de lumiere du jour. Pour la configuration correspondant a notre maquette, un nombre minimum de 13 zones semble suffisant pour prendre en compte la diversité des distributions de luminances des ciels rencontrés dans la réalité 11 Résumé 12 Povzetek Povzetek Prièujoèa doktorska disertacija se osredotoèa na izraèun razpoložljive dnevne svetlobe s poudarkom na modelih svetlosti neba. Prvi cilj naloge je predlagati najbolj primeren model svetlosti neba za izraèun trenutnih vrednosti osvetljenosti z dnevno svetlobo v notranjih prostorih. Drugi cilj naloge pa je doloèitev potrebne natanènosti modela svetlosti neba. Natanènost modela svetlosti neba je doloèena s številom uporabljenih elementov v modelu. Oba zakljuèka sta osnovana na meritvah, ki so bile opravljene v Lyonu. V prvem poglavju je opisana oprema, ki je potrebna za opravljanje meritev porazdelitve svetlosti neba. Najbolj pomembna naprava je vsekakor merilnik porazdelitev svetlosti neba (sky scanner). Merilnik, ki smo ga uporabili pri meritvah, je znamke EKO in je sposoben meritve porazdelitve svetlosti neba v 145 toèkah v èasu približno treh minut. Pred opravljanjem meritev je bilo potrebno merilnik tudi kalibrirati. Kalibracija je izvedena z uporabo razliènih metod. Prva kalibracija je bila izvedena v simulatorju dnevne svetlobe. Pri tej kalibraciji smo preverjali meritve svetlosti zenita in izraèunane vrednosti vertikalnih in horizontalne osvetljenosti. Druga kalibracija je bila izvedena pod resniènim nebom. V tem primeru smo primerjali izraèunane vrednosti vertikalnih in horizontalne osvetljenosti z vrednostmi, ki so bile izmerjene z merilniki postaje IDMP (International Daylight Measuring Program), ki so namešèeni že od leta 1992. Ko je bil merilnik porazdelitve svetlosti neba umerjen in namešèen na IDMP postaji, smo lahko popravili usmerjenosti merilnikov vertikalnih osvetljenosti IDMP postaje. Ti merilniki namreè niso imeli popolnoma ustrezne usmerjenosti. V drugem poglavju je opisan postopek analize meritev porazdelitve svetlosti neba. V tem delu je opisan naèin doloèitve CIE tipa neba in tudi CIE standard S 011/2003 s poudarkom na funkcijah postopnosti (gradation) in razpršenosti (indicatrix). Doloèitev skupine postopnosti temelji na razmerju svetlosti dveh elementov na vsakem almukantarju (Krog na nebesni polobli, ki je vzporeden s horizontom. Krog povezuje toèke z enako kotno višino.) in svetlosti zenita. Elementa na vsakem almukantarju morate imeti kotno razdaljo do sonca èim bliže zenitnemu kotu sonca. Tako dobimo na vsaki strani sonènega poldnevnika en element, ki ustreza zahtevi. Prav tako tudi doloèitev skupine razpršenosti temelji na razmerju dveh svetlosti. V tem primeru je v števcu svetlost posameznega elementa na almukantarju, v imenovalcu pa se nahaja normalna svetlost almukantarja. Normalna svetlost almukantarja je svetlost elementa neba na almukantarju, ki ima kotno razdaljo 900 do sonca. Ker imamo šest skupin postopnosti in šest skupin razpršenosti, imamo skupno 36 kombinacij. Nekatere od kombinacij so izjemno redke in s pomoèjo tabele, ki je v nalogi podana, lahko vseh 36 kombinacij uvrstimo med 15 osnovnih tipov, kolikor jih je tudi v CIE standardu. CIE tip neba se lahko doloèi iz razmerja svetlosti zenita (Lz) in difuzne horizontalne osvetljenosti (Dv). V nadaljevanju drugega poglavja je opisan postopek izraèuna horizontalne osvetljenosti iz meritev porazdelitve svetlosti neba in postopek doloèitve CIE tipa neba iz omenjenega razmerja. 13 Povzetek V tem poglavju je opisan tudi postopek doloèitve CIE tipa neba s pomoèjo statistiène metode. Izmerjene vrednosti svetlosti elementov neba se normirajo z difuzno horizontalno osvetljenostjo. Na enak naèin se normirajo tudi vrednosti vseh 15 tipov neba, ki so opisani s standardom. Nato primerjamo izmerjene normirane vrednosti z normiranimi izraèunanimi vrednostmi za vseh 15 tipov. Tip neba, pri katerem je odstopanje najmanjše, je rezultat metode. Na koncu poglavja so podane tabele frekvenc pojavljanj posameznih tipov neba izraèunanih z razliènimi metodami. V tretjem poglavju predstavimo digitalni fotoaparat s širokokotnim objektivom kot merilnik porazdelitve svetlosti. Cilj, ki smo ga imeli, je primerjava porazdelitve svetlosti neba, ki jo dobimo s pomoèjo digitalnega fotoaparata s širokokotnim objektivom in porazdelitve, ki jo izmeri merilnik porazdelitve svetlosti neba. Meritve, izmerjene z merilnikom so zapisane v tekstovni datoteki s specifièno obliko zapisa, zato smo morali podatke dobljene z digitalnim fotoaparatom in obdelane s pomoèjo programske opreme Photolux, zapisati v datoteko z enako obliko zapisa. Obe datoteki (za isto èasovno obdobje) smo analizirali na enak naèin za vseh 160 primerov, kolikor smo imeli digitalnih fotografij. Z analizo vrednosti svetlosti elementov neba smo lahko primerjali tudi znaèilne lastnosti neba. Ko so bili vsi podatki zbrani, smo med seboj primerjali horizontalno difuzno osvetljenost, svetlost zenita, razmerje Lz/Dv, skupino postopnosti in razpršenosti, tip neba doloèen z razmerjem Lz/Dv in tip neba doloèen s statistièno metodo. Na koncu smo med seboj primerjali tudi pare vseh 145 svetlosti elementov. Na koncu poglavja zakljuèujemo, da lahko digitalni fotoaparat s širokokotnim objektivom uporabimo kot merilnik svetlosti, pri tem pa moramo upoštevati nekatere omejitve. V èetrtem poglavju so opisani modeli porazdelitve svetlosti neba. Poudarek je na modelih, ki se najveè uporabljajo: Perez-ov model za vse tipe neba, ASRC-CIE model in Igawa model. Za vse èase (13.006 meritev), ko je deloval merilnik svetlosti neba (april 2005 – december 2005) smo izdelali porazdelitve svetlosti neba za vse tri modele. Vrednosti svetlosti so bile izraèunane na podlagi modelov in meritev obsevanosti na IDMP postaji. V tem poglavju je opisan tudi problem visokih vrednosti svetlosti sonca in sonèeve korone. Niti modeli porazdelitve svetlosti neba niti merilnik porazdelitve svetlosti neba ne morejo modelirati oz. izmeriti visokih nivojev svetlosti elementov neba, ki imajo majhno kotno razdaljo do sonca. Prav zaradi tega razloga smo pri analizi podatkov izloèili vse elemente neba, ki imajo kotno razdaljo do sonca manjšo od 150. Ta odloèitev temelji na izraèunu povpreène razlike svetlosti elementov neba z upoštevanjem vseh elementov neba in le tistih s kotno razdaljo do sonca veèjo od 150. V nadaljevanju je predstavljena primerjava izraèunanih vrednosti svetlosti elementov neba z modeli in vrednosti izmerjenih z merilnikom porazdelitve. Primerjava je bila izvedena z izraèunom povpreène vrednosti razlike in efektivne vrednosti razlike med posameznim modelom in izmerjeno vrednostjo. Na podoben naèin smo primerjali tudi CIE tip neba. Primerjava je bila izvedena za dve metodi doloèitve CIE tipa neba, za doloèitev na podlagi postopnosti in razpršenosti in za doloèitev na podlagi statistiène metode. Primerjava smo izvedeli na dveh skupnah podatkov o vrednostih svetlosti; za vse vrednosti in le za tiste, katerih elementi imajo kotno razdaljo do sonca veèjo od 150 in višino veèjo od 60. Prvi elementi so bili izvzeti zaradi sonèeve korone, drugi pa zaradi 14 Povzetek ovir-stavb, ki so v okolici IDMP postaje. V nadaljevanju èetrtega poglavja je predstavljena kakovost modelov svetlosti neba v odvisnosti od tipa neba in višine sonca. Na koncu poglavja so podani zakljuèki, da je Igawa model najboljši zelo oblaène tipe neba, Perezev model je najboljši za srednje tipe neba, ASRC-CIE model neba pa je v povpreèju najboljši model. ta model je namreè v veèini primerov po kakovosti med prej omenjenima, vedno pa blizu najboljšega modela. Peto poglavje opisuje meritve na modelu prostora in analizo le-teh. Model prostora je natanèno opisan, prav tako pa so omenjeni tudi problemi, ki so nastali ob meritvah. Na tem mestu predstavimo tudi delitev neba na razlièno število elementov (212, 145, 97, 26 in 13), faktor dnevne svetlobe in direktni faktor dnevne svetlobe. Osvetljenosti v modelu so bile merjene z merilniki osvetljenosti, nato pa so bile te vrednosti primerjane z modeliranimi. Modelirane vrednosti so bile izraèunane na podlagi direktnega faktorja dnevne svetlobe in srednje vrednosti svetlosti elementa neba na podlagi faktorja dnevne svetlobe in izmerjenih vrednosti osvetljenosti na IDMP postaji. V nadaljevanju je opisana tudi baza podatkov, ki so bili izmerjeni z meritvami na modelu prostora. Rezultati meritev in primerjave kažejo, da natanènost modela (število elementov neba) praktièno ne vpliva na velikost napake pri izraèunu osvetljenosti. Tako zakljuèujemo, da je pri izraèunih, kjer ne upoštevamo direktne sonène svetlobe, dovolj delitev neba na 13 elementov. 15 Povzetek 16 Introduction Introduction Our life is closely connected to daylight. It has a bearing on our schedule, way of life and biological processes in the body. In fact, we cannot even imagine our life without daylight. Daylight constitutes a renewable source of energy. Consequently, we would like to exploit it to the greatest extent possible. If we strive for its rational exploitation, we need to provide for at least an approximate forecast. Thus, all over the world different methods for calculating the daylight contribution to the indoor illuminances are being developed. The sky is the main contributor to the indoor illuminances, whereas the smaller portion is contributed directly by the sun. The prerequisite for the calculation of indoor daylight illuminances is the knowledge of the distribution of sky luminances, which in turn enables the calculation of the indoor illuminances. First mathematical descriptions of the distribution of sky luminances can be traced as far back as beginning of 20th century, yet their focus was only on a cloudy sky [11, 12]. Complexity of the problem was the reason why the majority of researchers limited their work at the beginning to two basic sky types, i.e. clear and cloudy. It was only later that intermediate types were described as well. Researchers formulated mathematical equations for the description of sky luminances on the basis of local measurements. Most equations are consequently adapted to local conditions or sky types. 2003 saw the adoption of the standard by CIE - Commission Internationale de l'Eclairage (International Commission on Illumination). The standard [16] gives mathematical description of the distribution of luminances for 15 different sky types, five for a cloudy sky, five for a partly cloudy and five for a clear sky. These 15 sky types now constitute an instrument allowing for the worldwide description of the most frequent sky types. Various methods can be employed for the calculation of the daylight contribution to indoor illuminances on the basis of the recorded distribution of sky luminances. The most frequent method of calculation is the daylight factor. The calculation by way of computer simulations of the room conditions has been gaining ground recently. In 2003, our Laboratory of Lighting and Photometry developed a simulation procedure [17] whereby the calculation of indoor daylight illuminances employs photometry of the window through which the daylight comes into a room. Description of the problem The basis for the method developed at the Faculty of Electrical Engineering in Ljubljana is the sky division into 77 zones. The luminances of certain zones provide the basis for subsequent calculation of photometric data giving luminances for certain directions of the window through which the light comes into the premises. Similar methods have also been developed in other labs. ENTPE-LASH developed a method founded on sky division into 13 zones. It is presented in [1] and is also available at the website www.satel-light.com. The method is based on the calculation of the contribution which the individual sky zone makes to the illuminances of a certain point in a room. The method is also called Directional Daylight Factor – DDF. The methods used for the 17 Introduction calculation of the daylight contribution have unfortunately not yet been adequately verified, the primary reason being the lack of comparable measurements. The use of different software (e.g. Photolux enabling the measurement of distribution of sky luminances and the drawing of the luminance maps on the basis of the photographs taken by the digital camera and wide-angle lens) and measurements on a model hold promises for certain progress in the field. Rationale of the contribution to the science In 2003, CIE issued a standard which describes the distribution of sky luminances in different weather conditions. The standard provides for universal equations enabling the calculation of luminances of the sky element in an arbitrary direction and gives the coefficient table used for the determination of 15 different sky types, ranging from cloudy to clear skies. Not long ago after the standard had been issued, it became evident that the frequency of emergence of certain sky types at a certain location also needed to be determined so that the described model would be used correctly. Hence, the purpose of this doctoral thesis is to develop an adequate method which shall make this possible. Tables denoting the frequency of emergence of certain sky types at certain locations would be compiled in the easiest way by employing measurements of the distribution of sky luminances. Such measurements, however, are extremely rare since only 17 measurement stations in the world make use of the "sky scanner" for the measurements of sky luminances. There is not enough data on luminance distribution or, what is more, there is none for most of the locations. Therefore, the method in question shall be based on more comprehensive and accessible measurements. Satellite images obtained from geostationary satellites shall be used. The satellites provide permanent coverage of the surface of the Earth, with their resolution reaching the accuracy of up to 5 km. Each image of the Earth surface is made every 30 minutes, which allows for a great amount of data for the purpose of the analysis. Most of this data for Europe and Africa is to be accessed via the Internet. The accessibility of the data has led to the decision to develop a new method. This method shall allow us to determine the frequency of emergence of CIE sky types at a specific location on the basis of global and diffuse horizontal irradiances. We shall be able to carry out calculations to determine time values of indoor daylight illuminances by way of devised tables and known algorithms. Method description We shall carry out the method by comparing the measurements of sky luminance distribution, the measurements of spatial illuminance by way of the model, and the illuminance calculation by way of various simulation programmes. The measurements shall be carried out at ENTPE (Ecole Nationale des Travaux Publics de l’Etat) Vaulx-en-Velin, France, which is home to one of IDMP (International Daylight Measurement Programme) locations. Our efforts shall (simultaneously) focus on the following: 18 Introduction - measurement of sky luminance distribution by way of the sky scanner (a special measuring device); - measurement of zenith luminance (10 – spatial angle) by way of a luminance measuring device on a stand; - measurement of sky luminance distribution by way of a digital camera, wide-angle lens and Photolux program package, and the drawing of sky luminance maps; - measurement of luminance distribution within the scale model by way of a digital camera, wide-angle lens and Photolux program package, as well as the drawing of maps for wall luminance within the model; - measurements of illuminance in the model by way of Li-cor sensors. When processing the data, all additional data collected through calculations within the IDMP locations shall be taken into consideration. The corresponding sky type shall be determined for each scan of sky luminance distribution (obtained through sky scanner) according to CIE standards. As for classifying individual sky luminance scans, Kittler’s method shall be used. Furthermore, we shall also determine the corresponding sky type for both the sky scan and the measurements carried out at the IDMP location, which shall be measured at the same time as the scan of sky luminance distribution. In this case, the Perez method shall be used for determining the sky type. The method is instrumental in determining sky clearness and sky brightness and also the corresponding sky type on the basis of the measurements obtained from IDMP. The results of the measurements and both methods shall be duly compared. On the basis of the results, a new method for determining CIE sky types from satellites scans shall be devised. In the second part, we shall carry out computer simulation of illuminances in the model by way of various methods. As far as the calculation is concerned, the illuminances in the scale model shall be calculated with different sky luminance models and with daylight quotients. Input data for sky luminance models shall consist of the maps developed by a digital camera and PhotoLux programme. Computer simulations results shall be compared to the measurements on a model. Focus shall be placed on comparing the values of illuminances. This will allow for evaluating the applicability of the method as regards determining CIE sky type when calculating indoor daylight illuminances. 19 Introduction 20 Description of the sky scanner 1 Description of the sky scanner The sky scanner measures the spatial distribution of the luminance or radiance of the entire sky automatically. 1.1 Sky scanner and measurements For measuring purposes a sky scanner (Fig. 1.1) was kindly borrowed from Kyushu University (Japan) for one year (January 2005 - January 2006). Sky scanner was made by Japanese company EKO Instruments CO., LTD and it was already used for measurements in Japan. Its measurement head is mounted on a two axis turning table, the luminance sensor is a SI-photodiode with a V(?) filter. An amplifier and automatic temperature compensation is assembled in the sensor head. Sky scanner came together with acquisition system (I/O controller and portable computer) and power supply since the input voltage sky scanner acquisition system and computer is 110 V. Sky scanner measures luminance and radiance of 145 patches in sky hemisphere (Fig. 1.2), following the CIE recommendations [44]. The solid angle of each measured sky patch is 110. Measured value represents the average value of luminance and radiance in the solid angle of 110. Measurement takes place on eight almucantars (imaginary circle on the celestial sphere, parallel to the horizon. The circle connects sky elements with same altitudes) - altitudes (60, 180, 300, 420, 540, 660, 780, 900). Number of measurements at different altitudes is changing and depends on the size of virtual bend at that altitude (Table 1.1). Table 1.1. Number of measurements according to the altitude of the point Altitude (0) -Almucantar No. of measurements 6 30 18 30 30 24 42 24 54 18 66 12 78 6 90 1 SUM 145 Fig. 1.1. Sky scanner 21 Description of the sky scanner Fig. 1.2. 145 sky patches Measurement takes place every 10 minutes from sunrise to sunset. Time needed to perform one measurement of the sky luminance distribution is a little bit more than 3 minutes, which can cause problems under dynamic weather conditions, since a single moving cloud can cover different patches of sky in a single measurement. The sky scanner has two measuring heads, one for luminance and the other for radiance. After one measurement is finished, the computer creates two files with data, the first one is for luminance and the second one is for radiance. In our thesis, we used only data for luminance, but we stored also the one for the radiance. All files are in ASCII format and they include data from the scanner, date and time of the beginning and the end of the measurement, inserted constants and location of the measurement. (Fig. 1.3). Fig. 1.3. Example of the data file. 22 Description of the sky scanner Luminance data are written in a file .05D (in year 2005), where luminances of all 145 sky patches are listed one after another without any punctuation mark (Fig. 1.3 and Fig. 1.4). Luminances in a file have number format of X.XXX and they are not given in absolute values, but they are divided by constant k (Fig. 1.4). Factor k can be calculated with next equation: L = k-L L = l06/fC-kSS-LF ^k = i06/fC-k 1.1) C SS Where: L fC k SS L luminance of a sky patch, calibration factor of sky scanner (110.6), factor we defined based on calibration under artificial sky (1.65) Chapter 1.2) luminance in .05D file. (see Calculated value of the factor k is 14.918,63. File .05D 0.0000.0900.0820.0730.0730.07 60.0990. X Factor k 14 .918,6 3 Luminance of a sky zone [cd/m 2 ] r i i = i 1 r 13 421 12 23M10 89 108 9 11 33 12 38 ... Fig. 1.4. Calculation of luminances of sky patches from a .05D file 1.2 Calibration of the sky scanner The sky scanner was last used in Japan in 2000/2001 and it has not been calibrated since then. And so we had to calibrate it before performing any measurements. First calibration was done under artificial sky and second under real sky conditions. 1.2.1 Calibration of the sky scanner under artificial sky For calibration of sky scanner under artificial sky we used the artificial sky of ENTPE (Vaulx-en-Velin). This is a uniform sky with adjustable luminance. 23 Description of the sky scanner 1.2.1.1 Zenith luminance As a first calibration, we decided to compare the "zenith luminance" of the sky scanner with luminance of the artificial sky. Zenith luminance is the 145th patch, which is measured by the sky scanner. Luminance value of this patch was compared with value measured with Minolta handheld luminance meter LS-100 with 10 aperture and with the value derived with digital camera and software Photolux. The sky scanner was placed in the middle of a room with artificial sky and it was lifted to a height where luminance and radiance heads were facing mirror on side walls (Fig. 1.5, Fig. 1.6). The scanner was placed also in horizontal position with water level. When the scanner was placed in right position, the door was closed and the automatic measurement was started. Fig. 1.5. Diagram of the sky scanner under Fig. 1.6. Picture of the sky scanner under artificial sky artificial sky With the same luminance level, a picture with calibrated digital camera (Nikon Coolpix 5000) with fish eye lens was taken. Digital camera was placed on tripod in the middle of a room with artificial sky and it was lifted to a level, where top of the fish-eye lens was in the same height as luminance and radiance heads of the sky scanner. When the digital camera was placed in right position, the door was closed and the picture was taken. With the same luminance level also luminance of the "zenith" was measured with handheld luminance meter. The "zenith" zone was marked on ceiling with transparent adhesive tape, using plumb line and calculating the area which is covered at height 1.7 m with solid angle of 110 (Fig. 1.7) 24 Description of the sky scanner Transparent adhesive tape Fig. 1.7. Digital image of artificial sky with zoomed area of zenith luminance The same procedure was used for eight times with different luminance levels. Measured results are gathered in Table 1.2. The luminances of sky patch number 145, measured with the sky scanner were derived using calibration factor and data from files produced by the scanner. The luminances measured with digital camera and fish-eye lens were obtained with Photolux. Photolux is a software in a system including a digital camera calibrated in terms of luminance and a software which translated the image(s) produced by the camera into a luminance map. The measurements were performed on April 14th, 2005. Table 1.2. Results of "zenith" luminance measurements. Measurement No. (time) Zenith Luminance (EKO) [cd/m2] Zenith Luminance Minolta (cd/m2) Zenith Luminance Photolux (cd/m2) Coefficient Minolta/EKO Coefficient Photolux /EKO M1 (10:20) 651 1070 1080 1.644 1.659 M2 (10:30) 940 1540 1550 1.638 1.649 M3 (10:40) 1573 2580 2612 1.640 1.660 M4 (10:50) 2495 4140 4158 1.659 1.667 M5 (11:00) 2975 4920 4504 1.654 1.514 M6 (11:10) 3400 5620 5062 1.653 1.489 M7 (11:20) 3770 6400 5833 1.698 1.547 M8 (11:30) 3128 5150 4713 1.646 1.507 Average coefficient 1.654 1.576 From the measurements we calculated the coefficients between Minolta luminance meter and sky scanner and Photolux and sky scanner values. The average coefficient between Minolta luminance meter and the scanner is 1.654 and the average coefficient between Photolux and sky scanner is 1.576. 25 Description of the sky scanner As it is seen from the most right column in Table 1.2 deviation of values of coefficient between luminances calculated with Photolux and luminances from sky scanner are more fluctuating than values of coefficient between Minolta luminance meter and sky scanner. Due to this fact we decided to take coefficient 1.65 for further calculations. In this sense, we plotted diagram (Fig. 1.8), where almost perfect match between corrected values of sky scanner luminances and luminances measured with Minolta is seen. Fig. 1.8. Measured luminances under artificial sky 1.2.1.2 Horizontal and vertical illuminances The second calibration was done also under the artificial sky. At this time, we compared illuminances derived from the sky scanner and from the illuminance meter. We compared horizontal and 4 vertical illuminances. Illuminances from sky scanner are calculated by integrating of all luminances for horizontal illuminance (see chapter 2.3.1) and by integrating of only certain luminances for vertical illuminances. Measured illuminances were measured with Li-cor sensors and stored on logger. Li-cor sensors and a logger (Fig. 1.9) is a system capable of reading and storing different values depending on type of measurement we perform. In our case we were measuring illuminances, so illuminance sensors were attached to the logger. We used five illuminance sensors, one for horizontal illuminance and four for vertical illuminances. Sensors were fixed on a specially designed wooden cube (Fig. 1.10), which was painted black and fixed together with black wooden plate to a tripod (Fig. 1.11). 26 Description of the sky scanner Fig. 1.9. Li-cor illuminance sensors and data logger Fig. 1.10. Illuminance sensors fixed on a black wooden cube Fig. 1.11. Illuminance sensors with wooden cube fixed on a plate and on a tripod 27 Description of the sky scanner But before doing the comparison, Li-cor sensors had to be checked (see chapter 1.2.1.3). After the validation of Li-cor sensors we installed the sky scanner under the artificial sky in the same way as we did it when checking zenith luminance. When the scanner was in place we started measurement and when the measurement was finished, the sky scanner was removed from the artificial sky room and on the same position as there was the measuring head of the sky scanner we put the tripod with 5 sensors. One sensor was set up in horizontal position and the other four in four vertical positions (N, E, S, W). Then again the room with artificial sky was closed and from the data logger, which was placed outside, reading of all five illuminances from sensors was done. The same procedure was used for six different settings of illuminance levels under artificial sky. Illuminance values from sky scanner were obtained with integration of all or only certain luminances (for vertical illuminances) and then compared with measurement readings from Li-cor logger. The procedure of computation of horizontal and vertical illuminances from the sky scanner is explained in section 2.3.1 and 2.3.2. For all six different illuminance setups we calculated coefficients between illuminances from Li-cor sensors and illuminances calculated from sky scans. Coefficients were calculated for horizontal and for four vertical illuminances. For each measurement those five coefficients were averaged and values are collected in Table 1.3. Table 1.3. Average Li-cor/sky scanner coefficient for all measurements Measurement Average coefficient Li-cor/Sky scanner M1 1.626458 M2 1.653968 M3 1.697119 M4 1.62945 M5 1.679245 M6 1.627261 Average 1.65225 With averaging the average coefficient at all six measurements we got practically the same coefficient as we got it with luminance validation. So we can conclude that taking the coefficient 1.65 is the right decision 1.2.1.3 Checking the Li-cor sensors Li-cor sensors were last calibrated in July 2002 and they should be calibrated every two year. Instead of an absolute calibration we did a relative calibration and we compared sensors between each other. Comparison was done under artificial sky. All sensors were lied down on a horizontal plane and the artificial sky was switched on. At eight 28 Description of the sky scanner different luminance levels illuminance measurements were done and results are in table Table 1.4. After measurements the average illuminance and deviation of all measurements from average illuminance was calculated for all luminance levels. After that illuminance profiles on Fig. 1.12 were plotted. In the diagram, we can see the measured illuminances from all 5 sensors compared to average illuminance. Table 1.4. Calibration of illuminance sensors Probe El lx E2 lx E3 lx E4 lx E5 lx E6 lx E7 lx E8 lx Mean deviation % A-I1 1522 1977 2860 4059 5975 7120 8560 10150 -0.563 A-I2 1552 2006 2907 4170 6083 7280 8720 10360 1.401 A-I3 1508 1940 2815 4057 5895 7075 8430 9956 -1.786 A-I4 1556 1980 2853 4145 6050 7228 8630 10160 0.421 A-I5 1549 1988 2864 4146 6055 7305 8650 10070 0.528 Average 1537.4 1978.2 2859.8 4115.4 6011.6 7201.6 8598.0 10139.2 -¦—A-I1 *A-I2 ¦A—A-I3 A-I4 *—A-I5 0 2000 4000 6000 8000 10000 12000 Av e rage illuminance (lx ) Fig. 1.12. Measured illuminance under artificial sky. Deviation of illuminances when compared to average illuminance is less then 2 % and from that result, we can conclude, that sensors are coherent and they don't need calibration. 10000 - 8000 6000 4000 2000 29 Description of the sky scanner 1.2.2 Calibration of the sky scanner under real sky 1.2.2.1 Calculated vertical illuminances and measured data from IDMP station The same validation of illuminances as it was done under the artificial sky was also done under real sky conditions. After the installation of the sky scanner we compared four vertical illuminances calculated from sky luminance distribution and illuminances from IDMP station. With comparison it was possible to calculate the coefficient between IDMP measurements and sky scanner calculated values. The coefficient was calculated for each illuminance, but only for times when, there was no direct sun component since the sky scanner can't measure direct sun luminance. This fact is best viewed on next diagrams (Fig. 1.13 - Fig. 1.16), where sky scanner curve can not follow high values obtained from IDMP station for times with direct component of sun. Calculated coefficient was 1.62. Since almost the same coefficient was already calculated with measurements under artificial sky and it is 1.65, we decided to use 1.65 and plot diagrams of all four vertical illuminances for time from July 19 5:10:00 to July 22 14:50:00. Fig. 1.13. Diagram of North vertical illuminance measured on IDMP station and calculated from sky scanner data 30 Description of the sky scanner Fig. 1.14. Diagram of East vertical illuminance measured on IDMP station and calculated from sky scanner data 75 60 45 30 15 -¦ 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 Time Fig. 1.15. Diagram of South vertical illuminance measured on IDMP station and calculated from sky scanner data 0 31 Description of the sky scanner Fig. 1.16. Diagram of West vertical illuminance measured on IDMP station and calculated from sky scanner data 32 Description of the sky scanner 1.3 Setting up the sky scanner 1.3.1 Description of the IDMP station The IDMP (International Daylight Measuring Program) Vaulx-en-Velin station (Fig. 1.17) was set up with the IDMP project in 1992. On the station there are the following measurements (Fig. 1.18): Illuminances: - Global horizontal - Diffuse horizontal - North vertical - East vertical - South vertical - West vertical Irradiances: - Global horizontal - Diffuse horizontal Others: - Zenith luminance - Dry Bulb Temperature - Wind direction - Wind speed - Normal incidence direct solar radiation - UV-A, UV-B Fig. 1.17. Outlook of the IDMP station Vaulx-en-Velin West vertical Illuminance Global Horizontal Irradiance Zenith Luminance Sky scanner UV-A UV-B Diffuse horizontal Illuminance \ ^k ¦¦- - «3 i ' ^ Pyrhelio meter North Vertical Illuminance East Vertical Illuminance Global Horizontal Illuminance ^^H J r South Vertical Illuminance Diffuse Horizontal Irradiance Fig. 1.18. IDMP station measurements (3600 view) 33 Description of the sky scanner Beside these measurements, in April 2005, an EKO sky scanner was added. It was placed on the south east part of the station. 1.3.2 Correction of orientation of vertical illuminance sensors Although the station has been working more then ten years, the actual orientation of vertical illuminance meters was not checked in last two years. The orientation was checked before we did the comparison between vertical illuminances derived from the sky scanner and from the IDMP station. Precise positioning of measuring devices with normal magnetic compass is on the IDMP station almost impossible since the whole station is made of iron tubes. To define correct orientation of vertical illuminance meters, we compared data from IDMP station and data derived from a program with Perez – All weather sky luminance model [14]. The input data for the Perez model were global and diffuse illuminances and position of the Sun (date, time). In the sky luminance model, the whole hemisphere was divided into patches with a grid of 10. For all patches, luminance was calculated and from luminance values four vertical illuminances were calculated. To find the best matching orientation we calculated profiles for rotations from 100 counter-clockwise to 100 clockwise and in some cases up to 120 clockwise for every degree. Suggested corrections were calculated with root mean square (RMS) differences between measurements and values from Perez All weather sky luminance model. Comparison was done under clear sky conditions. For all four vertical illuminances RMS difference was calculated for different orientation setups for 4 cloudless or semi cloudless days (May 26, 27, 31 and July 3, 2005). In next tables (Table 1.5 - Table 1.8) we present diagrams with original setup, diagrams with calculated values if sensors would be rotated for 5 degrees in clockwise and 5 degrees in counter-clockwise and the last diagram is plotted with suggested correction of the direction of vertical sensors. All diagrams in tables (Table 1.5 - Table 1.8) present illuminance values for July 3, 2005. From the results for all four days we concluded that vertical illuminance meters should be rotated for: - North vertical illuminance: 40 counter-clockwise - East vertical illuminance: 50 counter-clockwise - South vertical illuminance: 2.50 counter-clockwise - West vertical illuminance: 120 clockwise 34 Description of the sky scanner Table 1.5. North Global Vertical Illuminance Original setup Rotation: 50 clockwise 30 20 ¦7 5:00 7:00 9:00 11:00 13:00 15:00 17:00 19:00 Tim IDMP Measurement Perez All weather model Result: Worse than original setup. Rotation: 50 counter-clockwise Suggested correction: 40 counter-clockwise 3 r v Li 5:00 7:00 9:00 11:00 13:00 15:00 17:00 19:00 Tim Perez All weather model Perez All weather model Result: Better than original setup Result: Best fit with least RMS difference 25 15 10 5 0 25 20 5 0 35 Description of the sky scanner Table 1.6. East Global Vertical Illuminance Original setup Rotation: 50 clockwise Result: Worse than original setup Rotation: 50 counter-clockwise Suggested correction: 50 counter -clockwise 90 i 60 50 40 30 20 - 10 0 5:00 \ j L \ / X / \ \ " \ ¦•' * : V •• [ f 1 ^^^»^ 7:00 9:00 11:00 Per ez Al l weat her model 13:00 15:00 17:00 19:00 Time Perez All weather model Result: Better than original setup 90 80 70 60 50 40 30 20 10 0 l\ \ i f[ \ \ \ : \ j \ / l / : i~,„...... i ^^^^, 9:00 11:00 13:00 15:00 17:00 19:00 Tim Per ez A l l weat her m odel Perez All weather model Result: Best fit with least RMS difference 80 70 5:00 7:00 36 Description of the sky scanner Table 1.7. South Global Vertical Illuminance Original setup Rotation: 50 clockwise Result: Worse than original setup. Rotation: 50 counter -clockwise Suggested correction: 2.50 counter clockwise 0 5:00 / /' "*\ // \ 1 i • V ^rfll ^^*« ^*-, 7:00 9:00 11:00 13:00 15:00 17:00 19:00 Tim Perez All weather model Perez All weather model Result: Better than original setup Result: Best fit with least RMS difference 50 40 30 20 10 37 Description of the sky scanner Table 1.8. West Global Vertical Illuminance Original setup Rotation: 50 clockwise Result: Better than original setup Rotation: 50 counter-clockwise Suggested correction: 120 clockwise 40 10 /y TE.-- /¦' V- /¦ v /.' \. i v. I- li \ rU' i. .,^-mr---- * 5:00 7:00 9:00 11:00 13:00 15:00 17:00 19:00 Time Perez All weather model Perez All weather model Result: Worse than original setup Result: Best fit with least RMS difference 20 0 38 Description of the sky scanner 1.3.3 Correction of orientation of the sky scanner The sky scanner was positioned on a new aluminium plate, which was fixed on the metallic frame of the IDMP station. The scanner was placed in horizontal position with spirit level. Orientation of the scanner was approximately defined and checked with compass. Setting the sky scanner in the right orientation with only compass is impossible since the sky scanner and also the frame of the IDMP station is made of metal. The influence of surrounding objects to the compass was noticeable up to a distance of 1 meter above metal frame of the station. Approximate orientation was set up with a handheld compass from a distance 1 meter from sky scanner. After the approximate set up, first measurement were performed under real sky. From measurements on a cloudless sky it was possible to check the orientation of the sky scanner. First orientation was not correct (Fig. 1.19). Calculated position of sun (yellow dot) which was calculated from date and time and location, was not on the brightest part of sky, which was measured with sky scanner. If the orientation of the sky scanner would be correct, the calculated position of the sun would be on the brightest part of the sky. After first positioning, sky scanner was rotated for 10 degrees counterclockwise. From image with new setup (Fig. 1.20), we can conclude that repositioning of the sky scanner was successful. Fig. 1.19. Measurements done with original Fig. 1.20. Measurements done with original setup setup with rotation 10 degrees counter- clockwise 39 Description of the sky scanner 40 Analysis of the sky scanner data 2 Analysis of the sky scanner data 2.1 Description of the visualization Data got from sky scanner are in raw format. With a Java program, we did visualisation of the data as shown on Fig. 2.1. First and most important matter was to present the sky luminance distribution in a graphical way. The whole hemisphere was flattened onto 2D circle. Every measured part of the sky is presented as trapezoid with curved top and bottom borders, except the last one, the zenith luminance is presented as a circle. All elements are coloured in shades of grey, where the brightest sky element is coloured in white and the darkest in black. On all elements there are values of luminance of that sky element. If the brightest sky element has a value less than 1100 cd/m2, the values are in cd/m2 otherwise values are in kcd/m2. On the luminance distribution map there is also a position of sun drawn. The position of the sun is calculated from date and time and the location of measurement. Azimuth of ideal L9 Azimuth of ideal L9 Azimuth of ideal L9 sky element on 300 sky element on 180 sky element on 60 almucantar almucantar almucantar Measurement data (location, date, start time, end time) Solar data (azimuth, altitude) Lz/Dv ratio Visualization of sky luminance distribution. Every patch is presented with trapezoid. Luminance of each patch is written on the trapezoid (in cd/m2 or kcd/m2). Intersection of virtual cone with hemisphere. Patch on 300 almucantar with angular distance to sun approximately Zs Position of the sun Solar meridian Fig. 2.1. Visualization of the sky scanner data 41 Analysis of the sky scanner data 2.2 Defining CIE sky type based on CIE standard Calculation of gradation and indicatrix for all sky scans was performed on the basis of the CIE and ISO standard CIE S 011/E:2003 /ISO 15469.2004 "Spatial distribution of daylight – CIE standard general sky" [16]. The standard defines 15 standard skies; 5 overcast, 5 intermediate and 5 clear skies. Skies are defined with next equations and parameters a, b, c, d and e that can be found in Table 2.7. a La-rel ------ L Z f(z)-(p(Z) f(ZS)-(p(0) f(x) = l + c. U<*) 2 + e-cos x (2.1) (2.2) Where: La-rel L lz fix) {wl2) = \ (2.7) And in zenith: ) (2.10) Equation 2.10 describes relative gradation, which is the fact of decision when searching gradation group. This equation can be applied on measured values on all almucantars and gradation can be calculated by equation 2.11. 4. \ J—» i •. 1 06 5 4 !» > Jn •-v-- 2 > :<\ 3 2 1 s *X( %. 1 ^ **^ . -j-- L — - .-.si 0 20 40 60 80 100 120 140 160 180 Scattering angle % (°) Fig. 2.12. Relative indicatrixes and standard indicatrix function groups Relative indicatrixes can now be compared to standard indicatrix groups. RMS difference between appropriate sky elements and standard indicatrix function is calculated and based on this result indicatrix group is defined. In our example indicatrix group with least RMS difference is number 5 (Table 2.6). Table 2.6. RMS differences for all indicatrix groups Standard gradation group 1 2 3 4 5 6 RMS difference 1.786 1.286 0.910 0.447 0.441 1.535 On Fig. 2.13 a sample of relative indicatrix diagram with standard indicatrix functions and measured data are shown. 56 Analysis of the sky scanner data Fig. 2.13. Determination of relative indicatrix group for 14th July, 2005 On the diagram of relative indicatrix, calculated values are plotted with colored small circles. The circles in blue color represent sky elements on 180 almucantar with its relative indicatrix and with their angular distance towards sun and the ones in green color they represent relative indicatrix and their angular distances towards for elements on 300 almucantar (same example is also shown on Fig. 2.12). 57 37 Analysis of the sky scanner data From the points on relative indicatrix diagram it is possible to read out the indicatrix group. In the shown sample indicatrix group is 5, and it means that luminance is gradually increasing towards the sun position. 2.2.3 Calculation of the CIE sky type with gradation and indicatrix CIE Standard [16] defines 15 sky types (Table 2.7). Although with six gradation and six indicatrix groups we would get 36 sky types. Some combinations of gradation and indicatrix are extremely rare and those combinations are not included in 15 CIE sky types. Table 2.7. 15 standard skies with gradation and indicatrix groups and parameters which define those groups a b c d e Description of luminance distribution 1 I 1 4 -0,7 0 -1 0 CIE Standard Overcast Sky, alternative form Steep luminance gradation towards zenith, azimuthal uniformity 2 I 2 4 -0,7 2 -1,5 0,15 Overcast, with steep luminance gradation and slight brightening towards the sun 3 II 1 1,1 -0,8 0 -1 0 Overcast, moderately graded with azimuthal uniformity 4 II 2 1,1 -0,8 2 -1,5 0,15 Overcast, moderately graded and slight brightening towards the sun 5 III 1 0 -1 0 -1 0 Sky of uniform luminance 6 III 2 0 -1 2 -1,5 0,15 Partly cloudy sky, no gradation towards zenith, slight brightening towards the sun 7 III 3 0 -1 5 -2,5 0,3 Partly cloudy sky, no gradation towards zenith, brighter circum solar region 8 III 4 0 -1 10 -3 0,45 Partly cloudy sky, no gradation towards zenith, distinct solar corona 9 IV 2 -1 -0,55 2 -1,5 0,15 Partly cloudy, with the obscured sun 10 IV 3 -1 -0,55 5 -2,5 0,3 Partly cloudy, with brighter circumsolar region 11 IV 4 -1 -0,55 10 -3 0,45 White-blue sky with distinct solar corona 12 V 4 -1 -0,32 10 -3 0,45 CIE Standard Clear Sky, low illuminance turbidity 13 V 5 -1 -0,32 16 -3 0,3 CIE Standard Clear Sky, polluted atmosphere 14 VI 5 -1 -0,15 16 -3 0,3 Cloudless turbid sky with broad solar corona 15 VI 6 -1 -0,15 24 -2,8 0,15 White-blue turbid sky with broad solar corona 58 Analysis of the sky scanner data Table 2.8. Matrix with gradation and indicatrix combinations for 15 CIE sky types. Gradation I II III IV V VI 1 1 3 5 9 9 9 2 2 4 6 9 9 9 3 2 7 7 10 10 12 4 8 8 8 11 12 14 5 8 8 8 11 13 14 6 8 8 8 13 13 15 Reduction from 36 sky types to 15 CIE sky types was done with Tregenza statistical [37] method for different sun altitudes which is shortly described in chapter 2.4. 2.3 Calculation of the CIE sky type based on Lz/Dv ratio In several papers, Kittler [15, 22, 36, 38, 42, 43] proposed that CIE sky types from measurements should be considered based on Lz/Dv values. Lz/Dv curves are defined by next equation: LZ/DV = B ¦ (sin xS )C /(cos fS )D + E • sin yS 133.8- sinyS (2.18) Where: LZ DV ?S B,C,D,E Zenith luminance, Diffuse horizontal illuminance, Solar altitude, Standard parameters defined with Table 2.9 59 Analysis of the sky scanner data Table 2.9. Parameters B, C, D, E for Lz/Dv curves CIE Sky Type Gradation group Indicatrix group Description of luminance distribution Parameters B C D E 1 I 1 CIE Standard Overcast Sky, Steep luminance gradation towards zenith, azimuthal uniformity 54.63 1.00 0 0 2 I 2 Overcast, with steep luminance gradation and slight brightening towards the sun 12.35 3.68 0.59 50.47 3 II 1 Overcast, moderately graded with azimuthal uniformity 48.3 1.00 0 0 4 II 2 Overcast, moderately graded and slight brightening towards the sun 12.23 3.57 0.57 44.27 5 III 1 Sky of uniform luminance 42.59 1.00 0 0 6 III 2 Partly cloudy sky, no gradation towards zenith, slight brightening towards the sun 11.84 3.53 0.55 38.78 7 III 3 Partly cloudy sky, no gradation towards zenith, brighter circum solar region 21.72 4.52 0.64 34.56 8 III 4 Partly cloudy sky, no gradation towards zenith, distinct solar corona 29.35 4.94 0.70 30.41 9 IV 2 Partly cloudy, with the obscured sun 10.34 3.45 0.50 27.47 10 IV 3 Partly cloudy, with brighter circumsolar region 18.41 4.27 0.63 24.04 11 IV 4 White-blue sky with distinct solar corona 24.41 4.60 0.72 20.76 12 V 4 CIE Standard Clear Sky, low illuminance turbidity 23.00 4.43 0.74 18.52 13 V 5 CIE Standard Clear Sky, polluted atmosphere 27.45 4.61 0.76 16.59 14 VI 5 Cloudless turbid sky with broad solar corona 25.54 4.40 0.79 14.56 15 VI 6 White-blue turbid sky with broad solar corona 28.08 4.13 0.79 13.00 For all 15 sky types Lz/Dv curves are plotted and grouped for overcast (Fig. 2.14), intermediate (Fig. 2.15) and cloudless (Fig. 2.16) skies. 60 Analysis of the sky scanner data 0.50 0.45 - 0.40 0.35 0.30 0.25 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Solar altitude in deg I/1 I/2 II/1 II/2 III/1 III/2 Fig. 2.14. Lz/Dv curves for overcast skies 0.50 - 0.45 0.40 0.35 0.30 - 0.25 0.20 0.15 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Solar altitude in deg III/3 III/4 IV/2 IV/3 Fig. 2.15. Lz/Dv curves for intermediate skies 61 Analysis of the sky scanner data Fig. 2.16. Lz/Dv curves for cloudless skies Around every Lz/Dv curve a surrounding strip ±2.5% should be drawn (Fig. 2.17). This is the proposed band of trust. If measurement is inside the band, then the measurement can be trusted (Point 3 on Fig. 2.17). The problem arises, when a measurement is inside two or more bands of trust (Point 2 on Fig. 2.17). In this case we can find the nearest Lz/Dv curve, but we can not trust the measurement. It is the same if a measurement is not in any of those ±2.5% strips (Point 1 on Fig. 2.17). Also in this case Lz/Dv can be calculated and regarding this value and solar altitude, nearest Lz/Dv curve can be defined with relative RMS difference. Fig. 2.17. Lz/Dv curves for overcast skies with ±2.5% surrounding strips 62 Analysis of the sky scanner data 2.3.1 Calculation of the diffuse horizontal illuminance (Dv) Diffuse horizontal illuminance is calculated from sky scans with integration of illuminance contributions from every patch. For calculations of diffuse illuminance the following equation is used: SE = Lp-cos(Z)-d(o (2.19) Where: SE contribution to horizontal illuminance from a single sky patch. Lp luminance of a sky patch. Z zenith angle of sky patch. dco solid angle of a sky patch. Fig. 2.18. Contribution of a single sky patch to diffuse horizontal illuminance. Since we have to calculate diffuse horizontal illuminance, direct part of sunlight has to be removed. Direct component of sunlight can be removed with eliminating contribution of sky element with sun to illuminance. The corresponding sky element is defined with calculated sun position. Calculated illuminance values were compared with measured illuminance from IDMP station. The first calculations we performed overestimated diffuse horizontal illuminance, although sky element with sun was removed. Calibration of the sky scanner was performed under cloudy sky, where global and diffuse illuminances have practically the same value. In cases of cloudy conditions, horizontal illuminance was calculated correctly. Zenith luminance from sky scans was 63 Analysis of the sky scanner data calculated correctly in all types of sky conditions also in cases of intermediate and clear skies (Fig. 2.19), but on the other hand, diffuse illuminance on horizontal plane was overestimated. (Fig. 2.20) Fig. 2.19. Zenith luminance for a week in July 2005 Fig. 2.20. Diffuse horizontal illuminance for a week in July 2005 As it is seen from Fig. 2.20 the calculated diffuse horizontal illuminance is far too estimated when compared with measured values from IDMP station. The only reason for that could be solar corona, which is also scanned with sky scanner. From some sample sky scans it can be seen that two or even more sky elements near sun can have almost the same (very high) luminance. (Fig. 2.21). Removing only one sky element might not be enough. In next step we tried to near curves for diffuse horizontal illuminance from IDMP measurements and from sky scans. In calculations with sky scans sky patches near sun were excluded. Patches were selected on the basis of the scattering angle (?), which is defined with equation 2.4. For a cloudless day (June 1st 2005) we did comparison between measured diffuse horizontal illuminance and calculated one. When calculating diffuse horizontal illuminance from sky scans different number of sky elements near the sun was excluded. Number of excluded elements depends on angle?, which was changed from 8 degree to 15 degree. Results of different calculations can be seen on diagram on next page (Fig. 2.22). 64 Analysis of the sky scanner data Fig. 2.21. Number of sky elements with very high luminance around sun, can be more than one or two. 30 25 - 20 15 10 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 Dv IDMP-----------Dv X<8deg Dv X<10deg Dv X<12deg Dv X<15deg Time Fig. 2.22. Diffuse horizontal illuminance with different number of excluded sky elements near sun 5 0 65 Analysis of the sky scanner data For further calculations all sky elements with scattering angle below 15 degrees from sun were excluded in diffuse horizontal illuminance calculation. When we applied this method to data for week in July 2005 we got new diffuse horizontal illuminance diagram (Fig. 2.23). Fig. 2.23. Diffuse horizontal illuminance for a week in July with new calculation method When new method was applied to the data set from a week in July the illuminance values from IDMP station and calculated values from the sky scans are practically the same (Fig. 2.23). 2.3.2 Calculation of the vertical illuminances Vertical illuminances were calculated from sky scans with integration of illuminance contributions of all sky patches that are seen from the vertical plane. Visibility depends on the azimuth of the sky patch. North vertical illuminance is illuminance on a plane that is facing north. From this plane sky elements with azimuths higher than 2700 or smaller than 900 are visible. For visibility we have next conditions: (2.20) North \a > 270° {tf<90° East {015 rmsCIEx = . 1 nX>15 n 7 , \Lp LpCIEx ) (2.24) X>15 p=1 RMS error between measured values and CIE sky type x, number of patches with scattering angle larger than 15 degrees. Tregenza sky is defined by the smallest RMS error between measured values and CIE sky type. 2.5 Results of measurements When measurements with sky scanner in December 2005 were finished, calculations of CIE sky type were carried out for all measured cases. CIE sky type was defined with three previously mentioned methods. Together with these calculations CIE sky type was also calculated for data measured at IDMP station. From the IDMP data it was only possible to define CIE sky type based on Lz/Dv ratio. Although zenith luminance and diffuse illuminance are measured every minute, CIE sky type was calculated only for every tenth measurement. We decided to calculate CIE sky type only for times when sky scanner started to measure. The zenith luminances and diffuse illuminances we used for further calculation ware carried out every full ten minutes (00, 10, 20, 30, 40, 50). When all calculations were carried out we had to synchronize measurements database. In some cases sky scanner data were missing. This was due to problems with sky scanner. Sky scanner was checked every day in the morning and sometimes it was not working due to unknown reason. The sky scanner stopped also at all voltage drops and dips. Voltage drops were mostly problematic during weekends when it was not possible 68 Analysis of the sky scanner data to get to the IDMP station because of the security alarm which was switched on at corridors leading to the station. On the other hand in some cases zenith luminance and diffuse illuminance measurements were missing. Practically they were not missing, but they did not pass the quality control. For example if diffuse illuminance is higher than global illuminance, the measured data does not meet the quality control conditions. All cases when one of the measurements was missing were excluded. After the synchronization of the sky scanner and zenith luminance/diffuse illuminance database, the final database was created. In this database we have all together 14172 measurements. If we apply sun altitude limitations and limitation for Lz/Dv ratio to the dataset, we get fewer cases that meet the limitations (Table 2.10). The best quality data are the ones that meet both limitations (Table 2.11). Table 2.10. Number of cases with different limitations Sun altitude < 50 1317 > 50 12855 all cases 14172 Sky scanner data IDMP data Lz/Dv in ±2.5% band Lz/Dv in ±2.5% band YES 5905 5859 NO 8267 8313 all cases 14172 14172 Table 2.11. Number of cases in cross-section of sun altitude and Lz/Dv limitations Sun altitude Lz/Dv in ±2.5% band Sky scanner data IDMP data > 50 YES 5338 5452 NO 7516 7402 all cases 12854 12854 On next charts (Fig. 2.25 - Fig. 2.28) frequencies of CIE sky types are presented for data with sun altitude above 50 and Lz/Dv (calculated from sky scanner data) in ±2.5 % band. 69 Analysis of the sky scanner data C IE sky ty pe 15 12 9 CIE sky type Fig. 2.25. Frequencies of CIE sky types for sky scanner data (sun altitude ? 50, Lz/Dv in ±2.5 % band.) – based on Lz/Dv ratio. Fig. 2.26. Frequencies of CIE sky types for IDMP data (sun altitude ? 50, Lz/Dv in ±2.5% band.) – based on Lz/Dv ratio. 35 30 25 20 15 10 5 0 CIE sky type 35 30 25 20 15 10 CIE sky ty pe Fig. 2.27. Frequencies of CIE sky types for sky scanner data (sun altitude ? 50, Lz/Dv in ±2.5 % band.) – based on gradation and indicatrix groups Fig. 2.28. Frequencies of CIE sky types for scanner data (sun altitude ? 50, Lz/Dv in ±2.5% band.) – based on Tregenza method. 6 3 0 5 0 70 Analysis of the sky scanner data On next charts (Fig. 2.29 - Fig. 2.32) frequencies of CIE sky types are presented for all data with sun altitude above 50 regardless of Lz/Dv ratio. 15 -12 9 - 6 3 -0 CIE sky type 15 12 9 6 3 -0 C IE sky ty pe Fig. 2.29. Frequencies of CIE sky types for sky scanner data (sun altitude ? 50) – based on Lz/Dv ratio. Fig. 2.30. Frequencies of CIE sky types for IDMP data (sun altitude ? 50) – based on Lz/Dv ratio. 35 30 25 -20 15 10 5 CIE sky ty pe Fig. 2.31. Frequencies of CIE sky types for sky scanner data (sun altitude ? 50) – based on gradation and indicatrix groups Fig. 2.32. Frequencies of CIE sky types for scanner data (sun altitude ? 50) – based on Tregenza method. 0 71 Analysis of the sky scanner data On next charts (Fig. 2.33 - Fig. 2.34), the frequencies of CIE sky types are presented for data with sun altitude above 50 regarding and regardless of Lz/Dv ratio. On these charts, the difference between both selections can be seen. 15 i 12 9 6 3 0 CIE sky type A ll data In Lz/Dv band 15 12 9 C IE sky ty pe A ll data In Lz/Dv band Fig. 2.33. Frequencies of CIE sky types for sky scanner data (sun altitude ? 50) – based on Lz/Dv ratio for all data and data with Lz/Dv in ±2.5 % band. Fig. 2.34. Frequencies of CIE sky types for IDMP data (sun altitude ? 50) – based on Lz/Dv ratio for all data and data with Lz/Dv in ±2.5 % band. Based on graphs on Fig. 2.33 and Fig. 2.34 we can conclude that the method based on Lz/Dv gives similar frequencies of CIE sky types for data from IDMP station and from sky scanner. The gradation and indicatrix method and the Tregenza method, which could be only applied to sky scanner data, do not provide similar CIE sky frequencies as the Lz/Dv method. 0 72 Using digital camera as a sky scanner 3 Using digital camera as a sky scanner 3.1 Introduction Sky scanner that was installed at ENTPE IDMP station was borrowed from Kyushu University for a period of one year. Since one year is not especially long period for sky observations, we wanted to continue with measurements also without sky scanner. Instead of using sky scanner we would use digital camera with fish-eye lens and software to derive sky luminance distribution from digital images. On certain days when sky scanner was still installed, we were also taking pictures of sky vault with digital camera and fish-eye lens (Fig. 3.1). The main purpose of taking pictures at the same time as the sky scanner is working, was to evaluate sky images taken with digital camera. Fig. 3.1. Sample images for different sky types 3.2 Transforming sky images to .05D files To do the comparison between sky scans taken with sky scanner and fish-eye pictures it is necessary to transform sky images to the same data format as the sky scan data are. 145 averaged luminances extracted from the digital image were saved in the same format as the one used by the sky scanner. 73 Using digital camera as a sky scanner Sky images were taken with two different digital cameras. The first one was Nikon Coolpix 5000 owned by ENTPE and the second one was also Nikon Coolpix 5000 but owned by FE (Faculty of Electrical engineering of University of Ljubljana). After the fish-eye sky images were taken (Fig. 3.2a), the first step was to import them to Photolux and to calculate luminances (Fig. 3.2b). After the luminances were calculated we had to save them in an organized text format so we would be able to use them for next calculations. Fig. 3.2. Original image (a) and belonging to luminance map (b) for October 6th, 2005, 10:42 (GMT+1) In Photolux calculated luminance values can be saved in different format. The most useful format is Genelux format. The whole sky luminance map is covered with a grid. The grid divides the luminance map into zones with practically the same solid angle. Density of a grid can be set between 1 degree and 30 degrees in zenith angle. Together with the density of the grid also the number of zones and size of a zone is changed (Table 3.1). Table 3.1. Number of zones and solid angles with different Genelux grid. Genelux grid [0] 1 2 3 5 6 10 15 30 Number of zones 20673 5181 2310 835 583 212 97 26 Solid angle [sr] 0.000304 0.001213 0.002720 0.007525 0.010777 0.029638 0.064775 0.241661 74 Using digital camera as a sky scanner In the Genelux format the luminance values are listed in lines, with one luminance per line. In the same line there are also bordering azimuths and zenith angles (Fig. 3.3). Ce//# ZenI ZenS Azil A ziS SAngle Luminance Dg Dg DgN->E DgN->E Sr Cd/m2 7 0.00 1.00 0.00 90.00 0.0002392 3813 2 0.00 1.00 90.00 180.00 0.0002392 3936 3 0.00 1.00 180.00 270.00 0.0002392 3926 4 0.00 1.00 270.00 360.00 0.0002392 3766 5 1.00 2.00 0.00 36.00 0.0002871 3863 6 1.00 2.00 36.00 72.00 0.0002871 3850 7 1.00 2.00 72.00 108.00 0.0002871 3892 8 1.00 2.00 108.00 144.00 0.0002871 3883 9 1.00 2.00 144.00 180.00 0.0002871 3857 10 1.00 2.00 180.00 216.00 0.0002871 3908 11 1.00 2.00 216.00 252.00 0.0002871 3900 12 1.00 2.00 252.00 288.00 0.0002871 3927 13 1.00 2.00 288.00 324.00 0.0002871 3751 14 1.00 2.00 324.00 360.00 0.0002871 3888 Fig. 3.3. Sample data from Genelux format file (with 1 degree grid) Although we had the opportunity to save Photolux luminances also with CIE grid (145 sky zones), we did not use it because the camera was not perfectly in line with points of sky when taking pictures of sky. Since we were trying to get data as accurate as it was possible we decided to use the Genelux format with most dense grid: 1 degree. With this setting we exported Photolux luminances in Genelux files with 20673 luminance values, which were listed in lines together with bordering azimuths and zenith values (Fig. 3.3). 3.3 Linking sky scans and images from digital camera When the image is taken with a digital camera and a fish-eye lens, the picture is turned upside down. The picture is the same as we would lie on the floor and we would be facing towards zenith. If our body is in line with north-south line and our head is facing north, then west is on our right side and east is on our left side. With the sky scanner is the opposite. Image taken with sky scanner has the correct orientation and points of a compass are in line with our understanding (North-top, South-bottom, East-right and West-left). 3.3.1 Flipping and rotating an image As it is seen from the paragraph above, all images taken with digital camera and fish-eye lens have to be horizontally flipped. When the pictures were taken, the top of the camera was facing west, because this was the best way to get the camera in line with points of a compass, although it was not possible to get 100% in line. As we mentioned before, the top of the camera (and the picture) should be facing north. Besides flipping the image, the image should also be rotated. To define the right angle of rotation, images of a certain day were compared with the position of sun on sunny days and with surroundings obstacles. Before the first images were taken, azimuths of some higher and observable obstacles were measured (Fig. 3.4). 75 Using digital camera as a sky scanner 180 225 270 315 0 45 Fig. 3.4. Azimuths of some higher and observable obstacles Original image Horizontally flipped image Horizontally flipped image with azimuth mask Flipped and rotated image for 990 counter-clockwise Fig. 3.5. Image transformation 76 Using digital camera as a sky scanner Once the image is flipped and oriented a mask of 145 CIE patches has to be applied (Fig. 3.6). After the mask is applied, fields (calculation area) in Genelux grid have to be defined for all 145 sky scanner zones (Fig. 3.7). 15 16 17 46 45 44 , 47 43 48 72 73 74 49 71 97 ge on , 70 98 9tJ 94 76 99 117118119 ; u8101 r- 114 67102 113 1: :103 ,4„136 ' KM112 137138127 124 ™* „104 111 l* 125 — 81 105 110109126 87 35 64 106„„, „ 86 82 56 63 10710a 05 „ °*3, 57 62 61 84 - « 58 59 60 31 32 28 - 3 „ 2 1 30 Fig. 3.6. Flipped and rotated digital image with applied sky scanner mask. Fig. 3.7. Finding fields in Genelux grid for sky scanner zone number 1 When all zones from Genelux file that belong to one element in sky scanner grid are defined, an average luminance of all this Genelux zones is calculated. The calculated value represents luminance of a CIE sky patch derived with digital camera. The same procedure is used to define all 145 luminances of CIE grid. When luminance values are calculated they are divided by factor k and calibration factor for camera and written in a .05D file, together with official beginning "0.000" and ending of the file "0.0070.010110.60.00005/10/0610:00:0510: 00:28LYON". In last part of the file there are also a date and time when picture was taken. Those data are saved in Genelux file and could be read out and saved in .05D file. 77 Using digital camera as a sky scanner 3.3.2 Flipping and rotating a sky scanner grid Flipping an image and/or rotating an image in programs like Photoshop has irreparable consequences. When flipped and rotated file is saved, all information about the camera and exposure value are lost, what makes it impossible for use in Photolux. This is the reason why images are first handled in Photolux and then flipped and rotated. For further calculation also luminances in Genelux file should be "flipped and rotated". As at is almost impossible to flip and exactly rotate more then 20.000 luminances in Genelux file, it is easier to flip and rotate sky scanner grid. Flipped and rotated (Fig. 3.8) sky scanner grid can be applied when defining fields - calculation area for every sky scan zone. Fig. 3.8. Transformation of sky scanner grid This procedure was used to get luminance of all 145 sky patches. The image was imported in Photolux, and luminance values were saved in Genelux format. At the end flipped and rotated sky scanner grid was applied and average luminances were calculated. Calculated luminance values were divided and written in .05D file as mentioned in previous chapter. 78 Using digital camera as a sky scanner 3.4 Comparing results derived with sky scanner and digital images For evaluation purposes, we analyzed two sets of pictures. The first set consists of 112 digital images taken with ENTPE Nikon CoolPix 5000 camera on September 7th, 8th, 13th, 15th, 16th and October 6th, 2005. The second one consists of 48 digital images taken with FE Nikon CoolPix 5000 camera on November 30, December 1, 2 and 7. With the software we developed and described in chapter 2, we did analysis of all 160 .05D files derived from digital images and later on compared results with sky scanner results (Fig. 3.9). Results derived from digital image Results derived from sky scanner a b Fig. 3.9. Visualization of luminance distribution from digital image (a) and from sky scan (b) For visualization we also processed digital image in a similar way that was presented in previous paragraph, only in false colors (Fig. 3.10). 79 Using digital camera as a sky scanner Fig. 3.10. Visualization of luminance distribution from digital image in false colors For all 160 images we did comparison on all data derived from analysis (horizontal illuminance, four vertical illuminances, zenith luminance, Lz/Dv and 145 luminances of sky zones). Major differences in luminance values between digital images and sky scans were found on first almucantar (0-12 degrees). Differences on this almucantar are almost 3 times higher that on other almucantars (Fig. 3.11). The most probable reason for higher differences is the impact of the surroundings. For further analysis luminances of sky zones with zenith angles higher than 78 degrees were eliminated. Most of those sky zones are at least partly covered with surrounding obstacles and we also eliminate them when we do analysis sky scanner data. When creating .05D files we applied new factor k for each camera. The new value that was applied for ENTPE Nikon CoolPix 5000 camera was 15945.9 and it's 6,9% higher that the original value. The calculated average horizontal diffuse illuminance from digital image with the original value of factor k was 6.9 % higher than calculated horizontal diffuse illuminance from sky scans. Due to this we decided to change factor k. Due to the same reason new factor was applied also for FE Nikon CoolPix 5000 (kFE=18018.9). 80 Using digital camera as a sky scanner 40 35 30 25 20 15 10 5 0 6 18 30 42 54 66 78 90 Almucantar (deg) Fig. 3.11. Relative difference in luminances of sky zones on different almucantars. Table 3.2. A sample comparison table for measurement on September 8th, 2005, 9:50 (GMT+1) Digital image Sky scanner Difference Horizontal illuminance lx 22586 22810 - 0.98% Zenith Luminance cd/m2 8712 9100 + 4.45% Lz/Dv cd/m2 lx 0.386 0.399 - 3.28% Gradation 2 1 +1 Indicatrix 2 2 0 Lz/DV sky 2 2 0 Tregenza Sky 4 2 +2 115 sky zones luminance Average difference 0.77 % RMS bias difference 8.63 % For both sets of data we created tables with most important data (Table 3.2), and then summarized results from all tables like that in next tables (Table 3.3 and Table 3.4): 81 Using digital camera as a sky scanner Table 3.3. Average and RMS bias differences for first set of measurements Average difference RMS bias difference Horizontal illuminance lx + 0.48 % 5.06 % Zenith Luminance cd/m2 + 0.80 % 14.66 % Lz/Dv cd/m2 lx + 0.537 % 13.47 % 115 sky zones luminance + 3.25 % 13.42 % Table 3.4. Average and RMS bias differences for second set of measurements Average difference RMS bias difference Horizontal illuminance lx - 1.34 % 5.50 % Zenith Luminance cd/m2 - 4.60 % 9.51 % Lz/Dv cd/m2 lx - 3.31 % 7.14 % 115 sky zones luminance + 2.66 % 8.82 % 3.5 Database of sky luminance measurements In the database we include all relevant data of the whole sky measurements and analysis of the measured values. The database consists of two pages for each measurement. On the first page, one can find a fish-eye image of the sky vault and false colour luminance map for 145 sky patches created with Photolux and procedure described in previous chapter. On the second page complete analysis of the sky scan, illuminance and luminance measurements and CIE sky type classification are presented The measurements were carried out on September 7th, 8th, 13th, 15th, 16th, October 6th, November 30th, December 1st, 2nd and 7th. All together we performed 160 measurements. All measurements are collected in the database. 82 Using digital camera as a sky scanner Fish-eye picture of a sky vault. Picture is taken at the start of sky scanner measurement. False colour luminance map created from fish-eye image with a procedure described in previous chapter 83 Using digital camera as a sky scanner Complete analysis of the sky scanner measurement. Global horizontal illuminance, diffuse horizontal illuminance, vertical Illuminances (North, east, South, West), zenith luminance measured by the IDMP station. CIE sky type classification based on three different methods (gradation/indicatrix method, Tregenza statistical method, LZ/Dv method) from sky scanner measurements (pink) and from Fish-eye image (blue). Calculated CIE sky type with different sky models. Luminance values in models are based on global and diffuse horizontal irradiances from whole sky measurements (IDMP). 84 Using digital camera as a sky scanner 3.6 Conclusion on the use of digital images From those 160 comparisons, we can conclude that calibrated digital camera can be used as a sky scanner. Use of digital cameras as sky scanners has some limitations. The most important limitation is weather. Digital camera can not be used in rain or snow since cameras are not water resistant. This problem could be solved with a special spherical glass cover which should cover the whole camera and fish-eye lens, but it should not have any impact of image received to the camera CCD. The other and even more destructible problem is also connected with weather. The camera CCD sensor can be easily damaged if its sensor is exposed to direct sun light for a long time. The camera should be kept under protection screen in time between two pictures are taken or another type of camera should be used. Single-lens reflex (SLR) cameras have a different construction; their CCD is exposed to sunlight only when capturing the image. The SLR camera uses an automatic moving mirror system which permits the photographer to see exactly what will be captured by the film or digital imaging system, as opposed to non-SLR cameras where the view through the viewfinder could be significantly different from what was captured on film. 85 Using digital camera as a sky scanner 86 Sky luminance models 4 Sky luminance models Sky models generate continuous sky luminance patterns. Researchers all over the world have tried to describe and to model sky luminance distribution for more that 100 years. The first measurements were performed by Schramm [48] in 1901, by Khaeler [11] in 1908 and by Kimball and Hand [12] in 1921. The measurements were performed under overcast skies and the gradual decrease of luminance from zenith to horizon in ratio between 1:0.5 to 1:0.3 was noted. Moon and Spencer [41] surveyed and arranged the previous research work and in year 1942 proposed luminance distribution of the overcast sky as a standard. The CIE has adopted a simplified version of the proposed standard as a CIE overcast sky only in year 1955. Parallel to the measurements on overcast skies, in year 1929 Pokrowsky [48] found gradual increase in luminance from zenith to horizon in case of clear skies in ratio of 1:3.65 and more. On bases of work performed by Boldyrev [49] in year 1935 and Krat [50] in year 1943, Kittler [51] recommended standard clear sky in year 1967. Kittler's proposal was adopted in year 1973 as a CIE standard clear sky [52]. Since then several different approaches and concepts were used to describe sky luminance distribution under various intermediate skies. In year 2003 CIE adopted a standard "Spatial distribution of daylight" [16], which is describing 15 different sky types (five overcast, five intermediate and five clear). In this chapter we describe three sky models that are mostly used to model sky luminance distribution from simple parameters such as irradiances. Irradiances (global and diffuse) are measured within IDMP stations or could be derived from the images of geostationary satellites [29] These satellites provide a continuous coverage of the earth, at least every half hour, at a spatial resolution of 5 km or less. For western and central Europe, this data is available on a Web server (www.satel-light.com). 4.1 All weather model for sky luminance distribution - Perez Model [14] is based on 5 different parameters, which are related to darkening or brightening of the horizon ( ap ), luminance gradient near the horizon (bp ), relative intensity of the circumsolar region or solar aureole ( cp ), width of the circumsolar region ( d p ) and relative backscattered light ( ep ). This model is based on more than 16.000 all-sky scans recorded in Berkeley, California in years 1985 and 1986 which are covering a wide range of insolation conditions from overcast to clear through intermediate skies. The model is given by next equation: L g(r,z) L (tc 5_ 2'2 7S (4.1) g(r,x) l + a -exp ( b \) p sinf (l + cp-exp(d p-^)+ep-cos2^) (4.2) JJ _ 87 Sky luminance models Where a bp , cp , d p and ep are the distribution parameters and they describe atmospheric conditions. Parameters depend on sky brightness and sky clearness. Sky brightness and sky clearness are calculated from horizontal diffuse irradiance and normal incident direct irradiance with next equations: E E es +1+1.041Z3 s = ed 1+1.041Z3 A m-E ed E ed E Ees0-os(ZS) (4.3) (4.4) where: ? ? Eed Ees Z sky brightness sky clearness horizontal diffuse irradiance normal incident direct irradiance solar zenith angle Sky clearness and sky brightness describes two distinct characteristics of the atmosphere. The first one describes turbidity of a sky and the second one describes thickness of clouds. Regarding sky clearness, eight categories of skies are described (Table 4.1). Table 4.1. Discrete sky clearness categories ? category Lower bound Upper bound 1 Overcast 1.000 1.065 2 1.065 1.230 3 1.230 1.500 4 1.500 1.950 5 1.950 2.800 6 2.800 4.500 7 4.500 6.200 8 Clear 6.200 -- 88 Sky luminance models Distribution parameters are all calculated with equation of a same form. Next equation is an analytical form, using coefficient ap as an example: a =a1(s) + a2(s)-ZS+A[a3(s) + a4(s)-ZS] p V / V / L V / V / I (4 5) The terms ai{s) are discrete function of the parameter s represented by eight-term vectors corresponding to each s interval. Parameters ai{s) are given with a Table 4.2 Among all cases, there are only two exceptions. Those two exceptions are in the first s interval for coefficients cp anddp. The functions for those two coefficients are: cp = exp[(A(c! +c2-ZS ))c3 ]- c4 (4.6) dp = exp[A(d + d2 ¦ ZS )] + d3 - M4 (4.7) Table 4.2. Model coefficients for Perez All weather model. Sky clearness interval S from 1.000 1.065 1.230 1.500 1.950 2.800 4.500 6.200 to 1.065 1.230 1.500 1.950 2.800 4.500 6.200 ----------- Coefficients a1 1.3525 -1.2219 -1.1000 -0.5484 -0.6000 -1.0156 -1.0000 -1.0500 a 2 -0.2576 -0.7730 -0.2515 -0.6654 -0.3566 -0.3670 0.0211 0.0289 a3 -0.2690 1.4148 0.8952 -0.2672 -2.5000 1.0078 0.5025 0.4260 a4 -1.4366 1.1016 0.0156 0.7117 2.3250 1.4051 -0.5119 0.3590 b1 -0.7670 -0.2054 0.2782 0.7234 0.2937 0.2875 -0.3000 -0.3250 b 0.0007 0.0367 -0.1812 -0.6219 0.0496 -0.5328 0.1922 0.1156 b, 1.2734 -3.9128 -4.5000 -5.6812 -5.6812 -3.8500 0.7023 0.7781 b -0.1233 0.9156 1.1766 2.6297 1.8415 3.3750 -1.6317 0.0025 ci 2.8000 6.9750 24.7219 33.3389 21.0000 14.0000 19.0000 31.0625 c2 0.6004 0.1774 -13.0812 -18.3000 -4.7656 -0.9999 -5.0000 -14.5000 c3 1.2375 6.4477 -37.7000 -62.2500 -21.5906 -7.1406 1.2438 -46.1148 c4 1.0000 -0.1239 34.8438 52.0781 7.2492 7.5469 -1.9094 55.3750 d, 1.8734 -1.5798 -5.0000 -3.5000 -3.5000 -3.4000 -4.0000 -7.2312 d2 0.6297 -0.5081 1.5218 0.0016 -0.1554 -0.1078 0.0250 0.4050 d3 0.9738 -1.7812 3.9229 1.1477 1.4062 -1.0750 0.3844 13.3500 d4 0.2809 0.1080 -2.6204 0.1062 0.3988 1.5702 0.2656 0.6234 e1 0.0356 0.2624 -0.0156 0.4659 0.0032 -0.0672 1.0468 1.5000 e2 -0.1246 0.0672 0.1597 -0.3296 0.0766 0.4016 -0.3788 -0.6426 e3 -0.5718 -0.2190 0.4199 -0.0876 -0.0656 0.3017 -2.4517 1.8564 e4 0.9938 -0.4285 -0.5562 -0.0329 -0.1294 -0.4844 1.4656 0.5636 89 Sky luminance models 4.2 ASRC - CIE Perez et al [40] modified Matsuzawa's model to take into account the high turbid intermediate skies. This model is a linear combination of four skies - the CIE or Kittler clear sky, the Gusev turbid clear sky, the intermediate sky and the CIE overcast sky [23]. The coefficients of linear combination are computed using the sky clearness and the sky brightness factors [40, 29]. for ? < 1.2: L (I-a) L} L + a V z J CIE-Overcast L V z J Intermediate (4.8) where: a = min^ 1, max ?-1 ?-0.05 0.2 0.4 (4.9) for 1. 2< ? < 3.0: L 0-b) L V z J Intermediate + b L vLzyClear - Gusev where: b ?-1.2 1.8 (4.10) (4.11) for 3.0 < ? : L (l-c) + c V z J Clear-Gusev L L V z J Clear-Kittler (4.12) where: c = min 1. ?-3 (4.13) 4.3 Igawa All sky model This model is the most recent one and it was introduced in 2004 on basis of previous models for sky luminance and radiance from Matsuzawa, Igawa and Nakamura. Model was tested on IDMP data for Tokyo and Fukuoka [39]. Also in this model, the input data are irradiances from IDMP or any other weather station and zenith angle of the sun. Similar to other models, also in this case model we 90 Sky luminance models have two parameters which describe atmosphere conditions. First parameter is clear sky index, which is defined with next equation: (4.14) Kc Eeg = ------- seeg Where: Kc clear sky index, Eeg global irradiance, sees standard global irradiance. Standard global irradiance is a global irradiance of clear sky with TL (turbidity) of 2.5, and therefore defined with next equation: S = °'84"Ee0 • exp(-0.0675 • m) (4.15) m Where: Ees0 extraterrestrial direct normal irradiance, m relative optical mass. The relative optical mass can be calculated with next equation: 1 1 m =----------------------------------------------------T7^Z~------------ ( 4.16) cos(Z5) + 0.50572- (96.07995 -Zs-l cos(Z5) Second parameter is cloudless index. Since the lower bound values of the cloud ratio (4.17) appear in the clear sky and depend on the solar altitude, an index without solar altitude dependency is needed. C, =^- (4.17) Eeg Where: Ce cloud ratio, Eed horizontal diffuse irradiance. The cloudless indices of all the solar altitude are almost in the range from 1 to 0 though they vary a little when the solar altitude is low. The cloudless index proposed here can be used as the index without solar altitude dependency to classify sky conditions from clear sky to overcast sky, and can be calculated with next equation: 91 Sky luminance models I-Ce C le =e (4.18) Where: Cle cloudless index, C standard cloud ratio. Standard cloud ratio is a cloud ratio with TL of 2.5 and can be obtained with next equation: Ces = 0.01299 + 0.07698 • m - 0.003857 • m2 + 0.0001054 • m3 - 0.000001031 • m4 ( 4.19) With this two parameters sky conditions are given and sky luminance distribution can be calculated with equation that is the same as for CIE sky, only gradation and indicatrix functions are different and they are depended on the sky index. a-relLZ = f(ZS).15 Number of patches with scattering angle larger than 15 degrees In the first step for root mean square differences next equation was used (a sample for Igawa model): rmsdlgawa = 1 ^(r t V ----> [L T -L „„ I (4.27) 145 j=1y pg p ) Where: rmsdIgawa root mean square difference between Igawa model and sky scanner measurement, In the second step for root mean square differences next equation was used (a sample for Igawa model): rmsdlgaJ=, nx<15 nx<15 p=1 'YMjplgawa -Lpss) (4.28) Where: rmsdj ' root mean square difference between Igawa model and sky scanner measurement and without elements on first almucantar and without element with z<150 In all cases also relative differences were calculated. Relative differences were obtained with normalization of absolute differences with luminance of sky patches measured with sky scanner. 1 98 Sky luminance models 4.4.1 Results of comparison - all cases Comparisons were done on 13006 cases dating from May 18, 2005 to December 9, 2005. Results are gathered in next tables: Table 4.4. MBD for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Sky luminance Model Mean bias difference [cd/m2] All patches Without sky patches with ?<150 Without sky patches on 60 almucantar and ?<150 Igawa -1441.18 -422.91 -742.73 ASRC-CIE -1472.40 -429.89 -713.60 Perez-All weather -1356.70 -387.80 -778.87 All X>15 No X<15&H -200 -400 -600 -800 -1000 -1200 --1400 -1600 - Pe re z - A ll wea t he r a ASRC-CIE ¦ Igawa Fig. 4.7. MBD for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 In our case mean bias difference does not show the real situation and the results (Fig. 4.7) could mislead us if we would trust them without any reflection. The results are misleading because they show that calculations with sky patches on the 60 almucantar give better results than calculations without them. Since there are obstacles around IDMP station the sky scanner is not measuring sky patches on the 60 almucantar, but it is measuring luminances of façades of surrounding buildings. In most cases, especially on sunny days, luminances measured with the sky scanner are lower then the true sky luminances. This fact can be observed on Fig. 4.3 and Fig. 4.4, where sky scanner luminance for first 30 sky patches (6 degree almucantar) are practically always under modelled luminances. Modelled values can be treated as true values, measurements 99 Sky luminance models performed with sky scanner are underestimating luminance values because the surrounding obstacles. For almost all 30 sky patches on first almucantar the bias difference is positive (Fig. 4.6) while on the other hand for the most of the rest sky elements the difference is negative (sky models underestimate sky luminance). If we calculate mean bias difference with all sky patches the values is lower (because of the above mentioned fact) compared with the one when taking into account sky elements without horizon (6 degree almucantar). Since we are aware of problems regarding surrounding buildings, we can conclude that MBD cannot be trusted and we should move to RMSD. Table 4.5. RMS differences for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degrees almucantar and without sky patches with ?<150 Sky luminance Model Root mean square difference [cd/m2] All patches Without sky patches with ?<150 Without sky patches on 60 almucantar and ?<150 Igawa 9185.39 2335.20 2215.03 ASRC-CIE 9231.04 2240.76 2194.44 Perez-All weather 9016.78 2240.55 2227.75 10000 9000 - 8000 7000 -6000 5000 -4000 3000 -2000 1000 -0 All X>15 No X<15&H Perez - All weather ASRC-CIE Igawa Fig. 4.8. RMS differences for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degrees almucantar and without sky patches with ?<150 RMSD is more trustful, since problems with averaging the bias difference are eliminated. From Fig. 4.8 it can be read that the best results are obtained with elimination of first 30 sky patches and patches with scattering angle less than 150. 100 Sky luminance models Table 4.6. Relative MBD for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Sky luminance Model Relative MB difference [%] All patches Without sky patches with ?<150 Without sky patches on 60 almucantar and ?<150 Igawa -20.36 -9.22 -14.11 ASRC-CIE -20.38 -8.92 -13.57 Perez-All weather -16.99 -6.48 -14.45 0 -5 -10 -15 - -20 - -25 J All X>15 No X<15&H Perez - All weather ASRC-CIE Igawa Fig. 4.9. Relative MBD for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 101 Sky luminance models Table 4.7. Relative RMSD for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Sky luminance Model Relative RMS difference [%] All patches Without sky patches with ?<150 Without sky patches on 60 almucantar and ?<150 Igawa 128.02 42.75 40.03 ASRC-CIE 128.78 41.02 40.03 Perez-All weather 127.38 41.02 40.88 140 120 - 100 80 -60 40 -20 0 - All X>15 No X<15&H Perez - All weather ASRC-CIE Igawa Fig. 4.10. Relative RMSD for the whole sky hemisphere, for the sky hemisphere without sky patches with ?<150 and for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Comparison of relative values of MB and RMS difference give practically the same results as comparison of absolute values. If we look at relative MB (Fig. 4.9), we get the best results with elimination of sky patches with scattering angle less than 150. Since we have the same situation as for absolute values, we know that those results cannot be trusted and we should not ground any further decision on them. On the other hand the relative RMS difference results are more trustful and they suggest eliminating patches on first almucantar and those with scattering angle less than 150. If we consider all measurements with all sky patches except the ones with scattering angle below 150, than only Igawa model deviate from ASRC-CIE and Perez All weather model. From all measurements we can conclude that both ASRC-CIE and Perez All weather model lead to same results and to practically same RMS difference (Table 4.8) between modelled and measured values. 102 Sky luminance models Table 4.8. RMS difference for three sky luminance models Sky luminance Model RMS difference cd/m2 Relative RMS difference [%] Igawa 2335.20 42.75 ASRC-CIE 2240.76 41.02 Perez-All weather 2240.55 41.02 Mean luminance: 5462.74 cd/m2 4.4.2 Results of comparison dependent on CIE sky type When sky scanner data were processed, also a CIE sky type was determined. CIE sky type was defined on three methods, the first one was based on gradation on indicatrix group, the second one was based on Tregenza statistical method and the third was based on Lz/Dv ratio. In this chapter we are analysing relative RMS difference between sky luminance models for each CIE sky type. Since it is possible to analyse all sky scanner measurements only with the first two methods, we decided to gather sky scanner measurements in groups of CIE sky types only for method based on gradation and indicatrix group and on Tregenza statistical method. After the groups were formed for both methods, RMS difference was calculated for all groups of CIE sky types. Results based on gradation/indicatrix method for hemisphere without 6 degree almucantar and without sky patches with scattering angle below 150 are gathered in Table 4.9 and presented on Fig. 4.11, and in Table 4.10 and on Fig. 4.12, for the whole sky hemisphere without sky patches with scattering angle below 150. Results based on Tregenza statistical method for hemisphere without 6 degree almucantar and without sky patches with scattering angle below 150 are gathered in Table 4.11 and presented on Fig. 4.13. In Table 4.12 and on Fig. 4.14 are gathered results for statistical method for the whole sky hemisphere without sky patches with scattering angle below 150. Table 4.9. Relative RMS differences for different CIE sky types (based on gradation and indicatrix) for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Sky luminance Model Relative RMS difference for hemisphere without 6 deg. almucantar and without sky patches with ?<150 [%] CIE Sky type (based on gradation/indicatrix method) Number of cases Igawa ASRC-CIE Perez-All weather 1452 26.6 27.1 30.1 392 38.0 37.6 39.3 748 27.0 26.8 28.3 372 33.2 31.7 32.7 439 36.9 35.6 36.0 485 36.2 34.3 35.2 820 43.8 41.3 40.6 663 47.5 45.5 43.1 636 46.7 45.3 44.2 395 44.5 42.3 40.8 782 40.7 40.6 39.5 1024 42.9 43.3 43.2 500 39.7 41.4 43.4 4085 43.6 44.9 46.6 213 59.0 59.9 61.9 103 Sky luminance models CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 CIE14 CIE15 No. of cases Igawa ASRC-CIE Perez - All weather Fig. 4.11. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Table 4.10. Relative RMS differences for different CIE sky types (based on gradation and indicatrix) for the sky hemisphere without sky patches with ?<150 Sky luminance Model Relative RMS difference for hemisphere without sky patches with ?<150 [%] CIE Sky type (based on gradation/indicatrix method) Number of cases 1452 392 748 372 439 485 820 663 636 395 782 1024 500 4085 213 Igawa 28.9 42.0 29.4 373 39.0 40.3 50.7 54.9 48.4 50.9 45.0 44.7 41.1 44.0 57.9 ASRC-CIE 30.0 41.3 29.7 35.9 38.2 37.9 45.3 49.0 46.5 45.6 42.0 43.7 41.6 44.0 57.5 Perez-All weather 36.2 44.8 34.6 38.4 41.2 41.1 46.1 47.1 46.4 45.8 41.0 42.6 42.5 44.1 57.7 _ C IE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE 11 CIE12 CIE13 C IE14 CIE15 No. of cases Igawa ASRC-CIE Perez - All weather Fig. 4.12. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for the sky hemisphere without sky patches with ?<150 104 Sky luminance models Table 4.11. Relative RMS differences for different CIE sky types (based on Tregenza method) for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Sky luminance Model Relative RMS difference for hemisphere without 6 deg. almucantar and without sky patches with ?<150 [%] CIE Sky type (based on Tregenza method) Number of cases 1912 592 559 475 183 221 350 3890 49 45 505 13 3653 24 535 Igawa 24.5 34.7 36.3 36.3 54.8 48.7 48.1 39.8 67.8 57.9 56.0 48.0 44.5 52.8 46.3 ASRC-CIE 25.1 34.4 34.6 34.7 52.2 46.0 46.5 39.0 64.8 55.3 54.4 49.8 46.1 48.7 47.3 Perez-All weather 28.3 35.6 34.1 35.6 49.1 44.5 46.0 39.0 60.0 50.9 50.4 48.5 48.3 47.1 46.9 60 50 40 30 20 Y 10 CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 CIE14 CIE15 No. of cases Igawa ASRC-CIE Perez - All weather Fig. 4.13. Number of cases and relative RMS differences for different CIE sky types (based on Tregenza method) for the sky hemisphere without 6 degree almucantar and without sky patches with ?<150 Table 4.12. Relative RMS differences for different CIE sky types (based on Tregenza method) for the sky hemisphere without sky patches with ?<150 Sky luminance Model Relative RMS difference for hemisphere without sky patches with ?<150 [%] CIE Sky type (based on Tregenza method) Number of cases 1912 592 559 475 183 221 350 3890 49 45 505 13 3653 24 535 Igawa 26.7 38.5 37.5 38.8 57.6 50.5 50.7 43.6 73.5 62.6 66.7 53.9 44.4 64.1 49.5 ASRC-CIE 28.1 38.9 35.8 37.0 54.9 47.5 48.5 40.7 70.1 60.2 60.7 56.9 44.5 51.5 49.6 Perez-All weather 35.3 42.9 37.3 41.6 51.2 48.9 55.0 39.9 64.1 56.1 56.6 61.9 44.9 54.6 48.0 105 Sky luminance models Fig. 4.14. Number of cases and relative RMS differences for different CIE sky types (based on Tregenza method) for the sky hemisphere without sky patches with ?<150 Table 4.13. Relative RMS differences for different CIE sky types for both methods of CIE sky type determination and for different number of eliminated sky patches. 4200 - 3600 - 3000 2400 1800 1200 600 - ----------r 70 IE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 C E14 CIE15 No. of cases Igawa-Without X<15 Igawa-Without 6 deg almucantar and X<15 1—n n n In 1 m n-, 1 IT, 1 1 n_- -^nyjiv- rh n - Hr ^ - - - - - : : : : : ~ lllM__]n__nr L__].__..__ ..LI_____L' CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 CIE14 CIE15 ¦ Igawa-Without X<15 Igawa-Without 6 deg almucantar and X<15 4200 3600 3000 2400 1800 1200 600 CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 CIE14 CIE15 No. o f c as e s Perez All Weather-Without X<15 Perez All Weather-Without 6 deg almucantar and X<15 4200 3600 3000 2400 1800 1200 600 r r 1-1 -, r rfl p, 1 -i n_ r v f 1 I il Ir T T f 1| II Ir { 1 Ir i n C E01 CI E02 CIE03 CIE04 CI E05 CI E06 CI E07 CI 08 CI 09 CI 10 CI 11 CI 12 CIE13 IE14 CIE 15 CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CE07 CIE08 CIE09 CE10 CE11 CE12 CIE13 CIE14 CE15 No. of cases ASRC-CIE-Without X<15 ASRC-CIE-Without 6 deg almucantar a nd X<15 No. of cases ASRC-CIE-Without X<15 ASRC-CIE-Without 6 deg almucantar and X<15 Number of eliminated sky patches has a meaningful influence to RMS difference. As it is seen from diagrams in Table 4.13 the RMS is lowered if beside sky patches with scattering angle lower than 150 also sky patches on first almucantar are excluded from calculations. 4200 80 70 3600 60 2400 1800 - 30 1200 20 600 10 No. of cases 10 106 Sky luminance models If CIE sky type is defined with gradation/indicatrix method, then RMS is lowered if first almucantar is eliminated with sky types from 1 to 12. Surprisingly the RMS is higher in cases of clear skies. If the CIE sky type is defined with Tregenza statistical method, than RMS is lowered with elimination of first almucantar practically in all CIE sky types and all three sky luminance models, except for both Perez models (ASRC-CIE and All-Weather) for CIE sky type 13. 4.4.3 Results of comparison The quality of a model can be expressed as a relative difference between best model and certain model for each sky type. Relative difference is computed with next equation (a sample for Igawa model): rdIgawaCIEx = rmsd IgawaCIEx rmsd rmsd (4.29) Where: rd IgawaCIEx rmsd IgawaCIEx rmsd BestCIEx relative difference for Igawa luminance model for CIE sky type X RMSD for Igawa luminance model for CIE sky type X RMSD for best luminance model for CIE sky type X Quality of luminance models expressed in relative difference for different sky types is presented in Table 4.14 and plotted on Fig. 4.15, for selection based on gradation and indicatrix and in Table 4.15 and plotted on Fig. 4.16 for selection based on Tregenza method. In Table 4.14 and Table 4.15 the best model for certain CIE sky type has a value 0.0. Table 4.14. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) Sky luminance Model Relative difference [%] CIE Sky t ype Number of cases 1452 392 748 372 439 485 820 663 636 395 782 1024 500 4085 213 Igawa 0.0 1.8 0.0 3.9 2.3 6.2 11.8 16.7 4.2 11.7 9.9 5.0 0.0 0.0 0.6 ASRC-CIE 3.9 0.0 1.0 0.0 0.0 0.0 0.0 4.2 0.1 0.0 2.5 2.7 1.3 0.0 0.0 Perez-All weather 25.5 8.5 17.6 7.1 8.0 8.4 1.8 0.0 0.0 0.4 0.0 0.0 3.4 0.3 0.3 107 Sky luminance models 35,0 30,0 25,0 20,0 -15,0 10,0 5,0 0,0 1 2 3 4 5 6 7 8 9 10111213 1415 CIE sky type Igawa ASRC-CIE Perez - All weather Fig. 4.15. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) Table 4.15. Quality of luminance models expressed in relative difference for different sky types (based on Tregenza method) Relative difference [%] Sky luminance Model CIE Sky ty pe Number of cases 1912 592 559 475 183 221 350 3890 49 45 505 13 3653 24 535 Igawa 0.0 0.0 4.7 5.1 12.3 6.3 4.6 9.3 14.6 11.6 17.9 0.0 0.0 24.5 3.1 ASRC-CIE 5.4 1.1 0.0 0.0 7.1 0.0 0.0 1.9 9.4 7.2 7.3 5.4 0.2 0.0 3.2 Perez-All weather 32.4 11.4 4.2 12.6 0.0 3.0 13.4 0.0 0.0 0.0 0.0 14.8 1.2 5.9 0.0 35,0 30,0 - 25,0 - 20,0 - 15,0 -10,0 -5,0 ' 0,0 ¦ Ä \ A - /\ l A A A /^ \Ä / \ \/.x "v_y a \ / x \ / vi * „ „/i 4 5 6 7 8 9 1011 12 \ \// \ ^ 1 2 3 13 14 15 CIE sky type ¦ Igawa ASRC-CIE Perez - All weather Fig. 4.16. Quality of luminance models expressed in relative difference for different sky types (based on Tregenza method) 108 Sky luminance models Table 4.16. Average and maximum value of relative difference for sky luminance models Sky luminance Model Relative difference [%] Average value Maximum value Gradation/indicatrix method Tregenza statistical method Gradation/indicatrix method Tregenza statistical method Igawa 4,9 7,6 16,7 24,5 ASRC-CIE 1,0 3,2 4,2 9,4 Perez-All weather 5,4 6,6 25,5 32,4 In this model quality evaluation all sky scans (13006) were analysed. In the comparison all sky patches, except the ones with scattering angle lower then 150 were used. In Table 4.16 the results are summarized. As it can be seen from the mentioned table, the smallest average relative difference is for ASRC-CIE model. The difference is the smallest for both methods of CIE sky type decision. ASRC-CIE model is the best model for eight CIE sky types (CIE 2, CIE 4, CIE 5, CIE 6, CIE7, CIE 10, CIE 14 and CIE 15) in case of gradation/indicatrix method and for five CIE sky types (CIE 3, CIE 4, CIE 6, CIE 7, CIE 14) in case of Tregenza statistical method. Although in case of Tregenza statistical method Perez-All weather model seem to be better (best model for six CIE sky types), it can not be treated as the best model because of higher relative differences in cases where it is not the best sky model. ASRC-CIE sky luminance model also has smallest maximum value of relative difference. Igawa model gives best results with extreme cloudy skies (CIE 1, CIE 2 and CIE 3) and worse results with intermediate skies (CIE 7, CIE 8, CIE 10 and CIE 11). On the other hand Perez-All weather model provides best results for intermediate and intermediate-sunny skies (CIE 7, CIE 8, CIE 9, CIE 10 and CIE 12) and worse results for extreme cloudy skies (CIE 1, CIE 2 and CIE3). ASRC-CIE model is in most cases in between Igawa and Perez-All weather model, but always close to the best model. When taking into account also CIE sky types, we can conclude from Table 4.14 and Table 4.15 and from Fig. 4.15 and Fig. 4.16 that the best results are provided with ASRC-CIE sky luminance model. From both diagrams, especially from Fig. 4.15, it's obvious that the ASRC-CIE model has no peaks of high levels of relative difference and gives best results in eight sky types when CIE sky type is selected based on gradation and indicatrix. And also in cases of sky types where is not providing the best results, the relative difference is below 5% and it never provides worse results. No matter what method was used to define CIE sky type, the ASRC-CIE luminance model provides results with the best average relative difference and the smallest values of maximum relative difference (Table 4.16). 4.4.4 Results of comparison dependent on sky types and sun altitudes Similar comparison as we did it in previous chapter was done also in next comparison. In this comparison measured cases were arranged by sun altitude into four bins (Table 4.17). And for each sun altitude bin relative difference for each CIE sky type was calculated. 109 Sky luminance models Table 4.17. Bins of sun altitudes Bin Sun altitudes [0] Bin 1 5 - 20 Bin 2 20 - 35 Bin 3 35 - 50 Bin 4 > 50 For all four bins, CIE sky types were derived from sky scanner measurement with gradation and indicatrix method. In the comparison all sky patches, except the ones with scattering angle lower then 150 were used. 4.4.4.1 Sun altitudes 5 - 200 Table 4.18. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes between 5 and 20 degrees. Sky luminance Model Relative RMS difference for sun altitudes between 5 and 20 degrees [%] CIE Sky type Number of cases 488 96 270 93 175 126 179 153 114 101 175 211 112 1140 49 Igawa 26.4 43.2 29.2 38.7 39.7 39.4 58.7 67.7 56.7 57.0 59.9 51.5 56.9 51.6 80.9 ASRC-CIE 27.0 42.4 28.8 37.2 38.4 38.3 52.3 62.3 54.9 51.1 56.9 52.1 58.8 52.2 80.6 Perez-All weather 36.4 53.5 37.4 40.6 45.2 47.9 58.6 63.6 60.5 58.9 56.5 51.2 58.8 51.9 81.7 CI E01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE 11 CIE12 CIE13 C IE14 CIE15 No. of cases Igawa ASRC-CIE Perez - All weather Fig. 4.17. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes between 5 and 20 degrees. 110 Sky luminance models Table 4.19. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) for sun altitudes between 5 and 20 degrees. Sky luminance Model Relative difference [%] CIE Sky type Number of cases 488 96 270 93 175 126 179 153 114 101 175 211 112 1140 49 Igawa 0.0 2.0 1.3 3.9 3.3 3.0 12.1 8.8 3.3 11.6 6.0 0.6 0.0 0.0 0.5 ASRC-CIE 2.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.6 1.7 3.4 1.2 0.0 Perez-All weather 38.1 26.2 29.8 9.2 17.7 25.2 12.0 2.1 10.2 15.2 0.0 0.0 3.3 0.7 1.4 40,0 35,0 30,0 25,0 20,0 15,0 - 10,0 5,0 0,0 —x—*^-B a-^ 11 13 CIE sk ^* 15 pe Igawa ASRC-CIE Perez - A ll we a t he r Fig. 4.18. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) for sun altitudes between 5 and 20 degrees. As it can be read from Fig. 4.18 ASRC-CIE is the best model for practically all CIE sky types. Perez-All weather model is good for clear sky and extremely bad for cloudy skies. Relative difference for Perez-All weather model for cloudy sky types is almost in all cases higher than 25 %. Igawa model is acceptable for cloudy and sunny sky types, but not so good for intermediate sky types. Relative difference of Igawa sky luminance model for intermediate sky types is around 10 %. It's interesting that Igawa sky luminance model has quite low relative difference for CIE sky type 9. 111 Sky luminance models 4.4.4.2 Sun altitudes 20 - 350 Table 4.20 Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes between 20 and 35 degrees. Sky luminance Model Relative RMS difference for sun altitudes between 20 and 35 degrees CIE Sky type Number of cases Igawa ASRC-CIE Perez-All weather 459 27.8 29.6 36.1 100 45.0 43.9 47.5 260 29.3 30.0 34.0 108 38.4 37.1 42.6 125 39.4 38.8 40.6 121 43.7 39.8 43.0 235 56.6 48.3 49.7 188 62.4 53.0 48.7 132 53.3 50.7 52.2 100 63.1 53.6 50.4 149 59.7 52.6 49.7 255 54.7 51.8 48.6 134 1320 47.7 47.2 47.0 48.5 47.7 47.5 56 68.5 67.1 64.7 CI E01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE 07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 C IE14 CIE15 No. of cases Igawa ASRC-CIE Perez - All weather Fig. 4.19. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes between 20 and 35 degrees. Table 4.21. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) for sun altitudes between 20 and 35 degrees. Sky luminance Model Relative difference [%] CIE Sky t ype Number of cases 459 100 260 108 125 121 235 188 132 100 149 255 134 1320 56 Igawa 0.0 2.3 0.0 3.5 1.5 9.7 17.2 28.1 5.1 25.2 20.2 12.5 1.5 2.3 5.9 ASRC-CIE 6.7 0.0 2.2 0.0 0.0 0.0 0.0 8.8 0.0 6.4 5.9 6.6 0.4 0.5 3.7 Perez-All weather 29.9 8.1 15.9 14.8 4.6 8.0 2.9 0.0 3.0 0.0 0.0 0.0 0.0 0.0 0.0 112 Sky luminance models 35,0 30,0 -25,0 20,0-------- 15,0 10,0 5,0 0,0 11 13 CIE sk 15 pe Igawa ASRC-CIE Pe rez - A ll we a t he r Fig. 4.20. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) for sun altitudes between 20 and 35 degrees. Also for sun altitudes between 20 and 35 degrees the ASRC-CIE is the best model for practically all CIE sky types. Igawa model is good for cloudy and sunny sky types, but not so good for intermediate sky types. Perez-All weather model is the worst for cloudy sky types, but the best model for sunny sky types. Also here we can find an obvious drop in relative difference for Igawa model for intermediate sky types in CIE sky type 9. 4.4.4.3 Sun altitudes 35 - 500 Table 4.22. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes between 35 and 50 degrees. Sky luminance Model Relative RMS difference for sun altitudes between 35 and 50 degrees CIE Sky type Number of cases Igawa ASRC-CIE Perez-All weather 265 29.3 31.4 35.8 82 40.3 40.8 41.4 129 30.0 30.9 32.8 104 35.3 34.2 35.2 86 39.5 38.8 38.3 131 42.5 40.3 41.1 186 46.6 42.5 40.0 122 55.1 48.0 43.1 130 49.9 47.5 45.0 79 48.8 44.0 41.4 130 45.5 42.2 38.7 270 39.2 38.4 38.1 113 36.1 36.5 37.5 937 37.0 37.4 37.9 32 43.0 44.3 46.3 1 3 5 7 9 113 Sky luminance models 1400 1120 840 560 280 100 80 60 40 20 00 CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 CIE14 CIE15 No. of cases Igawa ASRC-CIE Perez - All weather Fig. 4.21. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes between 35 and 50 degrees. Table 4.23. Quality of luminance models expressed in relative difference for different sky types based on gradation and indicatrix) for sun altitudes between 35 and 50 degrees. Sky luminance Model Relative difference [%] CIE Sky t ype Number of cases 265 82 129 104 86 131 186 122 130 79 130 270 113 937 32 Igawa 0.0 0.0 0.0 3.2 3.3 5.2 16.6 27.7 10.9 18.0 17.6 2.9 0.0 0.0 0.0 ASRC-CIE 7.0 1.4 3.1 0.0 1.4 0.0 6.1 11.4 5.5 6.2 9.2 0.6 1.1 0.9 3.0 Perez-All weather 21.9 2.7 9.4 2.8 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 2.3 7.6 30,0 25,0 20,0 15,0 10,0 5,0 0,0 1 11 13 CIE sk 15 pe Igawa ASRC-CIE Perez - All weather Fig. 4.22. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) for sun altitudes between 35 and 50 degrees. 3 9 5 7 114 Sky luminance models Also for sun altitudes between 35 and 50 degrees the ASRC-CIE is the best model for practically all CIE sky types. Igawa model is best for cloudy and sunny sky types, but not so good for intermediate sky types. Perez-All weather model is the worst for cloudy sky types, but the best model for intermediate sky types. Also here we can find an obvious drop in relative difference for Igawa model for intermediate sky types in CIE sky type 9. 4.4.4.4 Sun altitudes over 500 Table 4.24. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes over 50 degrees. Sky luminance Model Relative RMS difference for sun altitudes over 50 degrees CIE Sky type Number of cases Igawa ASRC-CIE Perez-All weather 233 36.0 35.7 36.9 114 39.7 38.3 37.6 85 29.6 29.9 30.6 66 36.7 34.8 33.7 53 35.4 34.8 34.3 106 34.5 32.2 30.7 218 41.2 38.8 37.3 197 37.7 35.5 35.0 254 41.2 39.8 37.8 114 36.2 34.8 33.1 323 30.0 29.0 29.6 283 35.9 35.2 34.9 139 26.1 26.7 29.1 666 31.5 31.7 32.6 76 41.4 41.1 41.8 1400 1120 840 560 280 100 No. of ca s e s Igawa D ASRC-CIE Perez - All weather 80 60 40 20 0 I; ;;; ;;; ;;; ;;; ;| 0 CIE01 CIE02 CIE03 CIE04 CIE05 CIE06 CIE07 CIE08 CIE09 CIE10 CIE11 CIE12 CIE13 CIE14 CIE15 Fig. 4.23. Number of cases and relative RMS differences for different CIE sky types (based on gradation and indicatrix) for sun altitudes over 50 degrees. 115 Sky luminance models Table 4.25. Quality of luminance models expressed in relative difference for different sky types based on gradation and indicatrix) for sun altitudes over 50 degrees. Relative difference [%] Sky luminance Model CIE Sky type Number of cases 233 114 85 66 53 106 218 197 254 114 323 283 139 666 76 Igawa 0.6 5.6 0.0 8.7 3.4 12.4 10.6 7.8 9.0 9.3 3.6 2.7 0.0 0.0 0.6 ASRC-CIE 0.0 1.9 0.7 3.1 1.5 4.8 4.0 1.4 5.3 5.1 0.0 0.9 2.1 0.7 0.0 Perez-All weather 3.3 0.0 3.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0 11.4 3.6 1.6 14,0 - 12,0 - 10,0 - 8,0 - 6,0 -4,0 -2,0 -0,0 a A / \/ V / ' / \ V / m \ ^^^. / ^\ f\/y\............A?\ J 3 5 7 9 11 13 15 CIE sky type ¦ Igawa ASRC-CIE Perez - All weather Fig. 4.24. Quality of luminance models expressed in relative difference for different sky types (based on gradation and indicatrix) for sun altitudes over 50 degrees. For sun altitudes above 500 it is seen from Fig. 4.23 that relative difference is practically the same for all sky types. Relative difference does not differ much from sky type to sky type (from 26% to 41%). The difference is much lower as it is at lower sun altitudes (up to 80% with CIE sky type 15 at sun attitudes between 5 and 200). With higher sun altitudes relative difference between models fades away. The difference in worst case (Igawa model for CIE sky type 6) is only 12,4%. Although the differences between models are small, we can conclude that Igawa model is good for cloudy and best for sunny sky types. Perez-All weather model is best for intermediate sky types and ASRC-CIE gives best average results. Similar to previous case, where all measurements were grouped only by CIE sky type, also in this case we group measurements by CIE sky type and also by sun altitude. From Fig. 4.18, Fig. 4.20, Fig. 4.22 and Fig. 4.24 it is seen that also in this comparison ASRC-CIE sky luminance model provides best results. Also when we consider average relative difference (Table 4.26) Perez All weather and ASRC-CIE sky luminance share the same number of best results. First one is better for sun altitudes above 500 and the second one is better for sun altitudes below 500. But a note should be added, that Perez All weather model has high average relative 116 Sky luminance models difference for sun altitudes below 500 and on the other hand ASRC-CIE's average relative difference for sun altitudes above 500 is practically in line with Perez All weather model (Fig. 4.25). When maximum relative difference (Table 4.27) is considered, at all sun altitudes ASRC-CIE luminance model is the best choice. This models highest maximum relative difference is in bin with sun altitudes between 350 and 500 and has a value of 11.4%. Highest maximum relative differences for Igawa and Perez All weather models are higher and have values of 28.1 and 38.1%. Table 4.26. Average relative difference for three luminance models dependent on sun altitudes Sky luminance Model Average relative difference [%] sun altitude 50 – 200 sun altitude 200 – 350 sun altitude 350 – 500 sun altitude above 500 Igawa 3.8 9.0 7.0 5.0 ASRC-CIE 0.6 2.8 3.8 2.1 Perez-All weather 12.7 5.8 3.5 1.7 1 1 2 4 6 8 0 2 5 - 20 deg. 20 - 35 deg. 35 - 50 deg. Above 50 deg. Sun altitude Igawa ASRC-CIE Perez - All weather Fig. 4.25. Average relative difference for three luminance models dependent on sun altitudes Table 4.27. Maximum relative difference for three luminance models dependent on sun altitudes Sky luminance Model Maximum relative difference [%] sun altitude 50 – 200 sun altitude 200 – 350 sun altitude 350 – 500 sun altitude above 500 Igawa 12.1 28.1 27.7 12.4 ASRC-CIE 3.4 8.8 11.4 5.3 Perez-All weather 38.1 29.9 21.9 11.4 117 Sky luminance models 4.4.4.5 Problems of Igawa model for intermediate skies Since Igawa model provides worse results for CIE sky type 8, we decided to search further what is happening with this model in combination with CIE sky type 8. In Table 4.10 relative RMS differences for different CIE sky types (based on gradation and indicatrix) for the sky hemisphere without sky patches with ?<150 are gathered. If we extract from it only data for CIE sky type 8, we get Table 4.28. Table 4.28. Relative RMS differences for CIE sky type 8 (based on gradation and indicatrix) for the sky hemisphere without sky patches with ?<150 Sky luminance Model Relative RMS difference for hemisphere without sky patches with ?<150 Number of cases 663 Igawa 54.9 ASRC-CIE 49.0 Perez-All weather 47.1 700 - 600 500 - 70 60 50 400 300 200 -100 0 40 30 20 10 0 No. of cases Igawa Perez - ASRC Perez - All weather Fig. 4.26. Number of cases and relative RMS differences for CIE sky type 8 (based on gradation and indicatrix) for the sky hemisphere without sky patches with ?<150 From the matrix with gradation and indicatrix groups and CIE sky types (Table 2.8), it can be read that CIE sky type 8 can be achieved with six different combinations of gradation and indicatrix groups. These combinations are (gradation/indicatrix): I/6, II/5, II/6, III/4, III/5, III/6. All measurements were grouped in bins of combinations of gradation and indicatrix groups and next table was filled. 118 Sky luminance models Table 4.29. Number of cases and relative RMS differences for different combinations of gradation and indicatrix groups for CIE sky type 8 Sky luminance Model Relative RMS difference [%] Gradation/Indicatrix combination I/6 II/5 II/6 III/4 III/5 III/6 Number of cases 23 55 13 353 189 30 Igawa 52.1 58.2 53.6 54.5 54.3 60.0 ASRC-CIE 49.0 52.4 51.9 48.6 48.0 53.5 Perez-All weather 51.7 51.8 50.9 46.8 44.7 50.9 _ 60 50 40 30 20 I/6 II/5 II/6 III/4 III/5 III/6 Gradation/Indicatr ix No. of cases Igawa Perez - ASRC Perez - All weathe r Fig. 4.27. Number of cases and relative RMS differences for different combinations of gradation and indicatrix groups for CIE sky type 8 Most cases of CIE sky type 8 (53.2%) have gradation III and indicatrix 4. From our whole database of measurements we have chosen out four cases of CIE sky type 8 (Table 4.30). The presented four cases were chosen because the RMS difference between Igawa model and measurement was much higher than for other two models. Table 4.30. Relative RMS differences for three luminance models and four cases of measurement Sky luminance Model Relative RMS difference for hemisphere without sky patches with ?<150 [%] Date and time June 15, 2005 18:10 July 30, 2005 11:20 October 2, 2005 10:40 November 1, 2005 9:50 Igawa 143.48 117.08 129.37 131.16 ASRC-CIE 83.708 40.823 61.585 75.865 Perez-All weather 56.54 26.388 44.402 47.426 119 Sky luminance models We were intended to plot luminance profiles for all four cases and look for specialities, but in three cases out of previously mentioned four, there was no profile for Igawa luminance model. Luminances of all sky patches on the hemisphere were zero. In Igawa luminance model, the parameters, that describe luminance distribution of the sky vault, are dependent on sky index. Sky index was especially calculated for the mentioned four cases (Table 4.31). Table 4.31. Sky index used in Igawa model for four cases of measurement Date and time of case measurement June 15, 2005 18:10 July 30, 2005 11:20 October 2, 2005 10:40 November 1, 2005 9:50 Sky index 1.9386 2.1006 2.1417 2.1440 If we take a closer look into equation for parameter c' of Igawa luminance model (Fig. 4.28), we can see that the equation has a "Not a number" value for all cases where sky index is higher than 2.1. c'= 1.77?(1.22?Si)356 ?exp(0.2?Si)?(2.1-Si)°-8 4.30) From the equation it is obvious that in all cases in our database, where sky index is higher than 2.1, values of luminances in Igawa model are zero. Fig. 4.28. Igawa coefficients against sky index Sky index was calculated for all 13006 measurements in our data base and grouped in bins of an interval 0.1. Numbers of cases in certain bins are collected in Table 4.32 and presented on Fig. 4.29. 120 Sky luminance models Table 4.32. Sky index bins and number of cases in bins Sky index interval 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 No. of cases 445 889 873 711 526 424 398 391 306 276 231 196 178 241 No. of cases [%] 3.42 6.84 6.71 5.47 4.04 3.26 3.06 3.01 2.35 2.12 1.78 1.51 1.37 1.85 Sky index interval 1.4 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 No. of cases 294 379 635 1239 2428 1720 188 30 4 1 0 2 1 No. of cases [%] 2.26 2.91 4.88 9.53 18.67 13.22 1.45 0.23 0.03 0.01 0.00 0.02 0.01 20.00 18.00 16.00 14.00 12.00 10.00 8.00 6.00 4.00 2.00 0.00 Sky index int erv al Fig. 4.29. Sky index bins and number of cases in bins There are not many cases (126 or 1.7%) with sky index higher than 2.0. but for all this cases value of c' is questionable. In equation 4.31 a new formulation for coefficient c' is given. c'* = 6 1 + 0.5?exp(40?(Si-2)) + 6 4.31) If we do not apply a new formulation for c', at least we should limit value of sky index for calculations of coefficient c' to value 2.1. Original formulation and new formulation of coefficients are plotted on Fig. 4.30 121 Sky luminance models 18 16 - 14 - 12 - 10 - 6 - 4 - 2 0 -2 -•—•- 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 2,2 2,4 2,6 Sky index origina l c' new formulation c' Fig. 4.30. Original and new formulation of coefficients against sky index If the sky index is limited to value 2.1 in cases where the computed value is higher than 2.1, the luminance distribution can be calculated and luminance profiles can be plotted (Fig. 4.31 -Fig. 4.34). 100,0 90,0 80,0 70,0 60,0 50,0 40,0 30,0 20,0 10,0 0,0 10 20 30 40 50 60 70 80 90 Perez-All weather Igawa ASRC-CIE 100 110 120 130 140 Sky patch Sky scanner Fig. 4.31. Luminance profile for June 15, 2005 at 18:10 (Si=1.94) 8 0 122 Sky luminance models 50, 0 45, 0 40, 0 35, 0 30, 0 25, 0 20, 0 15, 0 10, 0 5, 0 0, 0 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 Perez-All weather Igawa ASRC-CIE Sky scanner Sky patch Fig. 4.32. Luminance profile for July 30, 2005 at 11:20 (original Si=2.101) 60,0 50,0 40,0 30,0 20,0 10,0 0,0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 Sky patch Perez-All weather Igawa ASRC-CIE Sky scanner Fig. 4.33. Luminance profile for October 2, 2005 at 10:40 (original Si=2.14) 0 123 Sky luminance models Fig. 4.34. Luminance profile November 1, 2005 at 9:50 (original Si=2.144) All luminance profiles on Fig. 4.31 are in line. Luminances modelled with Igawa model are overestimated, especially around position of sun, but the shape of the profile is same as the shape of the measurements. On the other hand, the Igawa profile on Fig. 4.32, Fig. 4.33 and Fig. 4.34 is totally different when compared to profiles of Perez All weather, ASRC-CIE or measurements. Sky index is computed from global and diffuse irradiance and in first case (Fig. 4.30, June 15, 2005 at 18:10) has a value of 1.9386. Sky indices for the other three cases were limited to 2.1 when computing luminances, since their value calculated from global and diffuse irradiances were higher than 2.1. We can conclude, that Igawa model does not provide comparable luminances for sky index values around 2.1 and higher. 4.5 Conclusion on sky luminance models Sky luminance models have been tested in three different ways: - All measurements together, - Measurements sorted by CIE sky type: o with gradation/indicatrix method o with Tregenza statistical method - Measurements sorted by CIE sky type and sun altitude. After comparing sky luminance models in all ways, we can conclude that on average ASRC-CIE sky luminance model provides best results for all CIE sky types. Igawa luminance model gives usually good results for cloudy and sunny skies but extremely bad results for intermediate skies. On the other hand Perez-All weather model provides 124 Sky luminance models usually good results for intermediate skies and extremely bad results for cloudy skies. Perez-All weather model cannot model the standard overcast sky and with that is unable to predict sky type 1, which is CIE standard overcast sky. In any way, ASRC-CIE sky luminance model provides best results for all sky types and we suggest the use of this sky luminance model for calculations of frequencies of occurrence of CIE sky types from satellite images. 125 Sky luminance models 126 Scale model measurements 5 Measurements in scale model 5.1 Introduction So far we have defined the way to produce luminance maps from simple parameters (global and diffuse irradiances). Since this data are available at least for Western and Eastern Europe on the SATEL-LIGHT Web server (www.satel-light.com) it is very simple to produce luminance maps with use of ASRC-CIE sky luminance model, which was recognised as the best model to obtain the most frequent skies for any location. When sky luminance distribution is known, a new question about resolution of the luminance map arises. Resolution of the CIE sky grid is 145 sky elements, but in our work we would like to conclude if this number is the right one or maybe the CIE grid should be more condensed (212 elements) or if the grid could be simplified to a smaller number of elements (97, 26 or even 13). We decided to base our conclusions on measurements made inside a scale model. Measured illuminances will be compared with calculated values with different density of the luminance map. 5.2 Description of the model and method Measurements on a scale model were performed in July 2005. The scale model, that we used, was a wooden box with inner sizes (WxDxH): 80x80x60 cm (Fig. 5.1). All inside walls were painted with mat black colour, with measured reflectance of 6%. For wall with window we used two different plates. The first one had 40 cm high and 60 cm wide window opening (no glazing) and the dimensions of the second window opening (no glazing) were 20 cm by 40 cm (Fig. 5.2). Window opening was in the centre of the wall. Orientation of the window was west for morning measurements and east for afternoon measurements. In our work we do not take into account direct sunlight, since it can not be modelled with sky luminance models and with organizing these two orientations, direct sunlight is prevented to come directly in the scale model. Fig. 5.1. Scale model side walls 127 Scale model measurements In the model, measurements of illuminances were carried out at ten measuring positions with Li-Cor sensors (Fig. 5.2). Sensors were connected to two data loggers (five sensors per logger). For measurements of luminances on indoor walls, a digital camera with a fish-eye lens was fixed just under the window opening (Fig. 5.2 and Fig. 5.3). With the camera and Photolux we were able to produce luminance maps for practically all indoor walls except the one with window. a linance Camera b Fig. 5.2. Scale model with illuminance sensors and fish-eye lens Fig. 5.3. Position of the digital camera with fish-eye lens (black surface is the opening (60 cm x 40 cm) into the scale model) At the same time also luminance of a view from a window was taken. For this, we used another digital camera also with a fish-eye lens. A Nikon CoolPix 990 digital camera was fixed on the box on the right at the height of the middle of the window (Fig. 5.4). 128 49 Scale model measurements Fig. 5.4. Digital camera with fish-eye lens for outdoor luminance measurements The wooden box with the Li-Cor sensors, the two data loggers and the two cameras with fish-eye lens was placed on the roof of ENTPE building. Because of the surrounding obstacles we wanted to place the scale model as high as possible. The best choice was to place it on a wooden base, which was fixed in place using spirit level, on the concrete ventilation shaft (Fig. 5.5). Fig. 5.5. Scale model was positioned on ventilation shaft 129 Scale model measurements 5.2.1 Illuminance and luminance measurements in scale model 5.2.1.1 Illuminance Indoor illuminance was measured at ten places with Li-Cor sensors. Measurements were performed on six different points on floor of the scale model, two on back wall and one on each side wall. Sensors were connected to data logger and were stored every minute. Every stored measurement is an average value of 1 s measurements. 5.2.1.2 Indoor luminances For indoor luminance purposes a digital camera Nikon CoolPix 5000 with fish-eye was installed under the window opening. For every luminance measurement at least four or five pictures were taken (Fig. 5.6a). Span between the highest and the lowest luminance in model was so high that pictures with different exposure values were needed. All fish-eye images were processed with Photolux (Fig. 5.6b). Luminance measurements were performed when sky scanner started to run (every ten minutes) and when it returned to home position (three minutes after the start). For every sky scan two luminance maps are available for all indoor walls. For comparisons we used the average of both measurements. Fig. 5.6. Fish-eye image of indoor (a) and luminance map of indoor walls (b) 5.2.1.3 Outdoor luminances For outdoor luminance purposes a digital camera Nikon CoolPix 990 with fish-eye was installed on the side of the window opening. All fish-eye images (Fig. 5.7a) were processed with Photolux (Fig. 5.7b). Luminance measurements were performed in the same manner as for indoor luminance measurements, one at the time when sky 130 Scale model measurements scanner started to run and one when it returned to home position For every sky scan two luminance maps are available for the view from the window. For comparisons we used the average of both measurements. Fig. 5.7. Fish-eye image of a view from the window (a) and luminance map of the same view (b) 131 Scale model measurements 5.2.2 View from a scale model View from a scale model is dependant on position of observation. A view from each point is differently obstructed with wall surrounding the window opening. For all ten points in which we measured illuminances, we calculated vertical and horizontal angles of view through the window opening. Calculated angles are listed in Table 5.1. Angles listed in the table are defined with Fig. 5.8 and Fig. 5.9. Table 5.1. Angles of view through the window for a window opening 60x40 cm Point of measurement Coordinates in a room Angles of view for a window opening 60x40 cm Angles of view for a window opening 40x20 cm X (m) Y (m) Z (m) Vl (o) (o) V x (o) yV2 (o) Vl (o) CV 2 (o) V x (o) yV2 (o) A1 0 -0.1 0 -68.53 68.53 45.00 76.72 -59.46 59.46 63.43 73.56 A2 0 -0.3 0 -43.33 43.33 18.43 57.54 -32.17 32.17 33.69 51.52 A3 0 -0.5 0 -30.08 30.08 11.31 43.99 -21.11 21.11 21.80 37.68 A4 0 -0.7 0 -22.68 22.68 8.13 34.85 -15.57 15.57 15.95 29.12 A5 -0.3 -0.4 0 -55.14 0.00 14.04 50.10 -50.10 -14.04 26.57 43.74 B1 0.3 -0.4 0 0.00 55.14 14.04 50.10 14.04 50.10 26.57 43.74 B2 0 -0.8 0.1 -20.14 20.14 0.00 26.06 -13.74 13.74 7.13 20.14 B3 0 -0.8 0.3 -20.14 20.14 -13.74 13.74 -13.74 13.74 -6.97 6.97 B4 0.4 -0.4 0.3 14.04 59.16 -25.57 25.57 26.57 55.14 -13.45 13.45 B5 -0.4 -0.4 0.3 -59.16 -14.04 -25.57 25.57 -55.14 -26.57 -13.45 13.45 a b Fig. 5.8. Horizontal angles defining view from a window with sample angles for point A4 (a-opening 60x40 cm, b-opening 40x20 cm) 132 Scale model measurements a b Fig. 5.9. Vertical angles defining view from a window with sample angles for point A4 (a-opening 60x40 cm, b-opening 40x20 cm) Points of measurements can be divided into three groups. In the first group are the points where we measured the horizontal illuminance. This group consists of points A1, A2, A3, A4, A5 and B1. In the second group are two points (B2 and B3) on the wall opposite the wall with the window opening. And in the last group there are the remaining two points (B4 and B5), which are placed on side walls. At these last four points, we measured the vertical illuminance. For all measuring points we also created images of view from a point through the window opening. Images are presented in Fig. 5.10. 133 Scale model measurements View from the window View from the window with angles View from point A1 through the window View from point A2 through the window View from point A3 through the window View from point A4 through the window View from point A5 through the window View from point B1 through the window View from point B2 through the window View from point B3 through the window View from point B4 through the window View from point B5 through the window Fig. 5.10. Images of view from the window opening for all points of measurement for a window opening 60x40 cm 5.3 Calculation of directional daylight factors In our work, we calculated directional daylight factors (DDF) for different celestial hemisphere partitioning. The hemisphere can be divided into different number of discrete sky elements. To investigate the influence of the number of sky elements on the error in illuminance calculations, we calculated DDF's and latter illuminance values for the hemisphere divided into 212, 145, 97, 26 and 13 sky elements (Fig. 5.11). All the grids except the 13 elements grid, are of the Genelux type. Genelux grid is characterized by a constant interval in zenith angle and in quasi constant solid angles. The numbers of sky elements are defined with the density of the grid. In the first case, the interval in zenith angle is 10 degrees, in second 12 degrees then 15 and 30 degrees respectively. Grid with 13 elements is not of the Genelux type. In this case the hemisphere is divided into three different altitude zones: a horizontal band, a zenith 134 Scale model measurements zone and intermediate band. The horizontal band is divided into 8 zones and the intermediate band into 4 zones. Each zone has the same zenith angle. Celestial hemisphere divided into 212 elements Celestial hemisphere divided into 145 elements Celestial hemisphere divided into 97 elements Celestial hemisphere divided into 26 elements Celestial hemisphere divided into 13 elements Fig. 5.11. Different sky partitions with different grid density (view to the east) The quantity of daylight that penetrates into indoor premises can be represented in different ways. The most common way is to represent it with daylight factor. Daylight factor is ratio of the illuminance at a point on a given plane due to the light received directly or indirectly from a sky of assumed or known luminance distribution, to the illuminance on a horizontal plane due to an unobstructed hemisphere of this sky. The contribution of direct sunlight to both illuminances is excluded. Glazing, dirt effects, etc. are included. When calculating the lighting of interiors, the contribution of direct sunlight must be considered separately. Unless stated otherwise the sky luminance distribution is assumed to be that of the CIE standard overcast sky. 135 Scale model measurements In our work we would like to evaluate the contribution of individual sky elements to the illuminance at a point on a given plane. Since daylight factor describes the contribution of the whole sky, we should implement more sophisticated daylight representation -directional daylight factor. Directional daylight factor (DDF) is the daylight factor at a point P due to the light received from only a patch of the sky. In other words: DDF is the ratio between the illuminance at a given point due to a sky element over the illuminance produced outdoor by the same sky element (eq. 5.1 ). DDF. =_^* 5.1) ^AExtp Where: DDFAp directional daylight factor related to a sky element p at a given point A Ejantp indoor illuminance at a given point A, produced by sky element p EAExtp outdoor (unobstructed) illuminance at a given point A, produced by sky element p As seen, the DDF related to a sky element at a given point, can be expressed with indoor and outdoor illuminances. Indoor and outdoor illuminances are defined with equations 5.2 and 5.3. The equation 5.2 describes only the direct component of daylight. Indirect component in this equation is neglected because the walls of the scale model were painted in black and there are no multiple reflections inside the scale model. Besides neglecting the indirect component, we should not forget the glazing. In the scale model we do not have any glazing and the luminance of a sky element that is seen from inside, is the same as its true luminance. In real buildings the luminance of a sky element should be corrected with the influence of the transmission factor and the incidence angle between the sky element and glass surface. EAintp=Lp-cos(Zp')-AS' 5.2) EAEXtp=Lp-cos(Zp)-AS 5.3) Where: Lp mean luminance of a sky element Zp incidence angle between surface in point and sky element ASp solid angle of a sky element seen from indoor Zp zenith angle of a sky element AS solid angle of a sky element 136 Scale model measurements If we combine equations 5.1, 5.2 and 5.3, DDF related to a sky element at a given point can be expressed: EAintP Lp ¦ cos(Zp ' ) • AS' cos(Zp ' ) • AS" --------------- zz ------------------------------------------- zz --------------------------------- eaexp V cos(z. ) 'M cos(Zp)-AS DDFAp = ^ = ?------* =-----* 5.4) In our scale model all six points, that were taken latter into account for evaluation, are on horizontal plane – floor. The floor plane is parallel to ground and because of that the incidence angle between surface in a point and sky element is the same as the zenith angle of a sky element: Zv=Zv 55) pp For the six points on the floor the DDF notation can be simplified to: AS' DDFAv =----- 5.6) AS A reader can see from equation 5.6 that DDF for six points in the scale model is just a ratio between a solid angle of a sky element seen from indoor and solid angle of a sky element. If a whole sky element is seen from the point in the scale model, that DDF equals 1, if a sky element is not seen from a point, the DDF equals 0. If only a part of a sky element is seen from the point, the value of the DDF is between 0 and 1. The sky is divided into zones of altitude and azimuth, and the directional daylight factor can be computed for each zone. The total illuminance at a point is the sum of the products of the mean luminance of each sky element, the solid angle of the each sky element, the cosine of the mean zenith angle of each sky element and the corresponding directional daylight factor (eq. 5.7): EA=±DDFAp.ASp-cos(zp).Lp 5.7) p=\ Where: EA total illuminance at a given point (in this example for point A) DDFAp directional daylight factor related to a sky element p for a point A in the model ASp solid angle of a sky element Zp zenith angle of a sky element Lp mean luminance of a sky element Zenith angle of a sky element and solid angle of a sky element are unique for each sky element. Regarding this fact a product of DDF, solid angle and zenith angle can be replaced with so called directional daylight quotient (DDQ): DDQA = DDFAp ¦ ASp ¦ cos(Zp ) 5.8) 137 Scale model measurements DDQ is unique for each sky element. Illuminance values in different points in our scale model will be calculated directly from luminance values of sky elements. For these calculations it's possible to rewrite equation 5.2. In equation 5.9 a product of DDF, solid angle and zenith angle is replaced with directional daylight quotient (DDQ): With new notation the total illuminance at a given point can be computed with: EA= njDDQAp-Lp 5.9) p=\ Where: EA total illuminance at a given point (in this example for point A) DDQAp directional daylight quotient related to a sky element p for a point A in the model Lp mean luminance of a sky element DDF is without units and is expressed in per cents, on the other hand the DDQ has a unit lx'm Cd . Directional daylight quotients describe contribution of each sky element to the illuminance at a given point in the model. When directional daylight quotients were calculated we were taking into account the window opening and position of a point in a model. Since most of the sky elements are obstructed with walls of the model (window only on one wall) there are only few sky elements that penetrate light into the model. The numbers of sky elements that are visible from measuring points in the model depend on the sky partitioning and position of a point in the model. Some elements that are seen from the points in the model are seen totally and on the other hand some elements are seen only partly. In the calculation of a solid angle ?SAp we have taken this into account. 138 Scale model measurements Fig. 5.12. Fish eye images of view from points of measurement on floor with a mask of sky model with 26 sky elements (view to the east) for a window opening 60x40 cm For each sky element a percentage of the element that is seen from a measuring point (Fig. 5.12) in the model was calculated. Then those percentages were multiplied with solid angles of the sky elements, the results are “visible solid angles” of sky elements from a measuring point in the model. For all sky partitioning, tables similar to Table 5.2 were designed. For the evaluation purposes of the solid angle calculation we also calculated the sum of solid angles of all sky elements viewed from measuring points on floor of the model. As we supposed the sum should be practically the same for all sky partitioning. The results are listed in Table 5.3 and presented in diagrams Fig. 5.13a and Fig. 5.13b. As it can be read from the diagrams the solid angles do not differ much regarding the sky partitioning. The highest difference was seen at sky partitioning into 26 elements, smaller window opening and measuring point A1. The most probable reason for the difference lies in small zenith angle and the fact that window opening is covering parts of only two sky elements. 139 Scale model measurements Table 5.2. Percentages and solid angles of a sky element visible from measuring points in the model for a window opening 60x40 cm (sky partitioned into 26 elements) Sky element Solid angle Percentage of sky element visible from point in a model Solid angle of sky element visible from point in a model Sr A1 A2 A3 A4 A5 B1 A1 A2 A3 A4 A5 B1 % Sr 1 0.210447 42.44 0 0 0 0 0 0.0893 0 0 0 0 0 2 0.210447 42.44 0 0 0 0 0 0.0893 0 0 0 0 0 3 0.210447 0 0 0 0 0 0 0 0 0 0 0 0 4 0.210447 0 0 0 0 0 0 0 0 0 0 0 0 5 0.255534 23.16 0 0 0 8.61 0 0.0592 0 0 0 0.0220 0 6 0.255534 50 76.49 23.41 5.12 67.00 0 0.1278 0.1955 0.0598 0.0131 0.1712 0 7 0.255534 50 91.80 46.63 13.21 16.75 50.25 0.1278 0.2346 0.1192 0.0338 0.0428 0.1284 8 0.255534 48.16 30.59 0.09 0 0 42.11 0.1231 0.0782 0.0002 0 0 0.1076 9 0.255534 0 0 0 0 0 0 0 0 0 0 0 0 10 0.255534 0 0 0 0 0 0 0 0 0 0 0 0 11 0.255534 0 0 0 0 0 0 0 0 0 0 0 0 12 0.255534 0 0 0 0 0 0 0 0 0 0 0 0 13 0.255534 0 0 0 0 0 0 0 0 0 0 0 0 14 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 15 0.241661 0 12.13 0 0 39.42 0 0 0.0293 0 0 0.0953 0 16 0.241661 0 38.57 52.09 41.48 53.20 0 0 0.0932 0.1259 0.1002 0.1286 0 17 0.241661 0 38.57 62.30 72.90 13.30 39.90 0 0.0932 0.1506 0.1762 0.0321 0.0964 18 0.241661 0 31.42 20.95 5.03 0 53.20 0 0.0759 0.0506 0.0122 0 0.1286 19 0.241661 0 0 0 0 0 12.83 0 0 0 0 0 0.0310 20 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 21 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 22 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 23 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 24 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 25 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 26 0.241661 0 0 0 0 0 0 0 0 0 0 0 0 Table 5.3. Table of solid angles of view from measuring points for different sky partitioning Sky partitioning Solid angles of view with window opening 60x40 cm Solid angles of view with window opening 40x20 cm A1 A2 A3 A4 A5 B1 A1 A2 A3 A4 A5 B1 13 0.636 0.789 0.520 0.343 0.495 0.495 0.129 0.237 0.190 0.116 0.154 0.154 26 0.616 0.800 0.506 0.335 0.492 0.492 0.094 0.244 0.170 0.119 0.141 0.141 97 0.628 0.796 0.523 0.338 0.499 0.499 0.140 0.253 0.175 0.115 0.154 0.154 145 0.631 0.796 0.520 0.340 0.503 0.503 0.138 0.254 0.176 0.114 0.153 0.153 212 0.635 0.795 0.521 0.339 0.506 0.506 0.134 0.257 0.176 0.115 0.153 0.153 20763 0,637 0,798 0,523 0,340 0,505 0,505 0,134 0,256 0,177 0,115 0,154 0,154 CIE145 0,629 0,796 0,520 0,340 0,503 0,503 0,138 0,254 0,176 0,114 0,153 0,153 140 Scale model measurements 0,90 0,80 0,70 0,60 0,50 0,40 0,30 0,20 0,10 0,00 Window opening 60x40 cm A1 A2 A3 A4 A5 B1 Sky 13 Sky 26 Sky 97 Sky 145 Point in Sky 212 CIE145 Sky 20763 a model Window opening 40x20 cm Sky 13 Sky 212 Sky 26 Sky 97 Sky 145 CIE145 Sky 20763 a b Fig. 5.13. Solid angle of view from measuring points for different sky partitioning (a-window opening 60x40cm, b- window opening 40x20cm) With the calculated solid angles it was possible to calculate DDQ's with use of equation 5.8 for all sky partitioning, all measuring points, all sky elements and both window opening. As a sample we show a table for DDQ's for sky partitioning into 26 elements with window opening 60x40 cm. 0,30 0,25 0,20 0,15 0,00 A1 A2 A3 A4 A5 B1 Point in a model 141 Scale model measurements Table 5.4. Directional daylight quotients for sky partitioning into 26 elements with window opening 60x40 cm. Sky element Directional daylight quotient for points in a model A1 A2 A3 A4 A5 B1 /x m2/cd 1 0.08627 0 0 0 0 0 2 0.08627 0 0 0 0 0 3 0 0 0 0 0 0 4 0 0 0 0 0 0 5 0.04185 0 0 0 0.01556 0 6 0.09034 0.13821 0.04230 0.00926 0.12106 0 7 0.09034 0.16587 0.08426 0.02387 0.03027 0.09080 8 0.08702 0.05528 0.00017 0 0 0.07609 9 0 0 0 0 0 0 10 0 0 0 0 0 0 11 0 0 0 0 0 0 12 0 0 0 0 0 0 13 0 0 0 0 0 0 14 0 0 0 0 0 0 15 0 0.00759 0 0 0.02466 0 16 0 0.02412 0.03258 0.02594 0.03327 0 17 0 0.02412 0.03897 0.04560 0.00832 0.02496 18 0 0.01965 0.01310 0.00315 0 0.03327 19 0 0 0 0 0 0.00803 20 0 0 0 0 0 0 21 0 0 0 0 0 0 22 0 0 0 0 0 0 23 0 0 0 0 0 0 24 0 0 0 0 0 0 25 0 0 0 0 0 0 26 0 0 0 0 0 0 5.4 Calculation of illuminance values in scale model Illuminance values were calculated for all points in the scale model for all cases of scale model measurements. Values are presented in a database of scale model measurements. Illuminance calculations were done with equation 5.9. Since illuminance values depend on luminance of sky elements it was necessary to first calculate luminance values of sky elements for all sky partitioning. 5.4.1 Calculation of luminance values of sky elements Luminance values of sky elements were defined with digital image of the view from the window. All images were edited in Photolux software and luminances saved in Genelux format. Using the Genelux format we kept 20673 patches of the view. Each patch is defined with left and right bordering azimuth, upper and lower bordering zenith angle, solid angle of the patch and its luminance. Unfortunately azimuths and zenith angles 142 Scale model measurements would have the right meaning if the picture would be taken horizontally, but the pictures were taken vertically. To get the azimuths and zenith angles with the right meaning we had to transform vertical values into horizontal values (Fig. 5.14). Fig. 5.14. Angles defining sky element at horizontal and vertical camera position Cartesian coordinates of a sky element are defined with next equation: Xh = Yh = Zh = cos a -sin© sin a -sin© cos© 5.10) From the Fig. 5.14 next relation between angles can be read: X =-Y Zv = Yv = Xh -Z 5.11) With a combination of equations 5.10 and 5.11 it is possible to express vertical azimuth and zenith angles with horizontal ones: © arctan(sinah-tan0h) arccos(cosah-sin0h) 5.12) When the equation 5.12 was applied to all 20673 sky elements, bordering angles of 20673 sky elements and bordering angles of different sky partitioning were in the same system of coordinates and luminance values could be calculated for all sky partitioning. Luminance values were then multiplied with directional daylight quotients (equation 5.9). 143 Scale model measurements 5.4.2 Calculation of illuminance values in scale model with daylight factors For comparison reasons we calculated the illuminance values in the scale model also with daylight factors. Daylight factors were calculated in two different ways. The first way was using the lighting software DIALux and the second one was with DDF for 20673 sky elements. Illuminance values in measuring points are calculated with next equation: EA = DFA ? Ediff 5.13) total illuminance at a given point daylight factor for a given point diffuse illuminance on an unobstructed horizontal plane 5.4.2.1 Daylight factors with DIALux In the DIALux lighting software we defined our scale model with both window openings (Fig. 5.15). The walls of the scale model were in black colour with approximate reflectance 6%. Measuring points in the scale model were defined in DIALux as daylight calculation points. To get the whole image of daylight factors on floor, also a Daylight Quotient Surface was added to the floor. a b Fig. 5.15. 3D view of a scale model with both window openings in DIALux (a - window opening 40x20 cm, b - window opening 60x40 cm ) 144 where: E DFA Eäw Scale model measurements Daylight calculations in DIALux are performed in accordance with DIN 5034, this means that the sky is overcast. Calculation results (daylight factors) for measuring points are listed in Table 5.5, results for calculation surfaces are shown on Fig. 5.16. Table 5.5. Daylight factors calculated with DIALux Point in scale model DIALux daylight factors [%] Window opening 60x 40 cm Window opening 40x 20 cm A1 18 5.85 A2 11 4.43 A3 4.92 1.87 A4 2.51 0.83 A5 5.07 1.78 B1 5.15 1.73 a b Fig. 5.16. Results of daylight calculations for calculation surfaces (a - window opening 40x20 cm, b - window opening 60x40 cm) 5.4.2.2 Daylight factors with illuminance Uniform hemisphere was divided into 20673 elements with constant interval in zenith angle (10) and quasi constant solid angles. The partitioning was the same as it was used when calculating directional daylight factors. With the same method as it was used in section 5.2, percentage of each sky element that is visible from point in a model was calculated. Percentage was multiplied with solid angle of a sky element and the result (modified solid angle) was used in equation 5.9. After the summation of contributions of all 20763 sky elements to the illuminance at a given point, the result was divided by theoretical illuminance on horizontal plane under uniform sky. 145 Scale model measurements So called illuminance daylight factors are listed in Table 5.6. Table 5.6. Illuminance daylight factors Point in scale model Illuminance daylight factors [%] Window opening 60x 40 cm Window opening 40x 20 cm A1 17.03 3.96 A2 14.73 5.45 A3 7.42 2.76 A4 3.86 1.40 A5 8.11 2.78 B1 8.11 2.78 5.4.3 Calculation of illuminance values in scale model from luminance values from sky scanner During scale model measurements, the sky scanner was also performing measurements of luminance values of the sky. The EKO sky scanner we used is measuring luminance of 145 sky elements. For this reason we had to calculate DDF also for CIE partitioning of the sky. With the same method as used in sections 5.2 and 5.3 the illuminance values were calculated for all measuring points in the scale model. 5.5 Mean bias error between measured and calculated illuminance values in scale model In the scale model illuminance values were measured with Licor sensors. Calculated values (from sky scanner luminances, fish-eye luminances averaged over different number of sky zones and from daylight factors) were compared with measured ones and mean bias error was calculated with next equation (shown for fish-eye luminances averaged over 13 sky zones): Esky13-A1 -ELicor-A1 Esky13-A2 -ELicor-A2 Esky13-A5 -ELicor-A5 Esky13-B1 -ELicor-B1 + +....+ + MBEskyii= E E E E ( 5.14) Where: MBE sky13 E sky13-A1 E mean bias error for sky modelled with 13 zones calculated illuminance from 13 sky zone luminances in point A1 measured illuminance with Licor sensor in point A1 6 146 Scale model measurements 5.6 Database of scale model measurements In the database we include all relevant data of whole sky measurements and all data measured and calculated for inside of the scale model. The database consists of three pages for each measurement. On the first page, one can find data about outdoor situation (fish-eye pictures, sky luminance maps and illuminance data), on the second page there are data about indoor situation (luminance maps of walls and measured and calculated illuminance values in all 6 points) and on the last page, there is an analysis of the calculated data. The measurements were carried out on July 25th, 26th, 27th and 28th, 2005. All together we performed 153 measurements. All measurements are collected in the database, although some of the data are missing at certain measurements. Fish-eye picture of a view seen from the window opening. Picture is taken at the start of sky scanner measurement. Luminance map of the fish-eye view from the window opening at the start of sky scanner measurement. Fish-eye picture of a view seen from the window opening. Picture is taken at the end of sky scanner measurement. Luminance map of the fish-eye view from the window opening at the start of sky scanner measurement. Outdoor illuminance values on the scale model wall with the window opening vertical. Values are calculated from luminance map. In the left column illuminance is calculated with all elements of luminance map (also ground) and in the right column we take into account only elements above horizon. Illuminance values calculated from sky scanner luminances Global horizontal, diffuse horizontal and vertical Illuminance measured by IDMP station 147 Scale model measurements Fish-eye luminance map of scale model surfaces. Fish-eye images are taken at the start and at the end of sky scanner measurement. Illuminances inside the scale model measured with Licor sensors Illuminances inside the scale model calculated with DDFCIE145 from sky scanner luminances Illuminance values inside the scale model Illuminances inside the scale model calculated with DDF's from fish-eye luminance maps Illuminances inside the scale model calculated with daylight factors and IDMP diffuse horizontal illuminance 148 Scale model measurements Graphical presentation of measured and modelled illuminances inside the scale model. Left diagram presented the conditions at the start of the sky scanner measurement and the right diagram presents the conditions at the end of the sky scanner measurement. Table with mean bias errors of luminance based models and models based on daylight factors Graphical presentation of mean bias errors of luminance based models and models based on daylight factors. Left diagram presents the MBE for the conditions at the start of the sky scanner measurement and the right one presents the MBE for the conditions at the end of the sky scanner measurement. 149 Scale model measurements 5.7 Results of the scale model measurements Our expectations were simple. We expected that the error would be inversely proportioned to the number of elements used at sky partitioning. So we expected that the error would be the smallest with the sky partitioning into 212 sky zones. From all measurements where calculation of the MBE was possible, we calculated average MBE. Average MBE are collected in next table and presented in Fig. 5.17. Table 5.7. Table of mean bias error Fish-eye luminances averaged over # sky zones Daylight factors 212 145 97 26 13 Dialux Illumin. MBE for T0 [%] 11,29 11,29 11,54 0,19 2,15 45,36 88,84 MBE for T3 [%] 9,45 9,46 9,66 -1,28 0,85 44,43 87,34 Fig. 5.17. Mean bias error To present results in a more evident way, diagrams of cumulative MBE frequency distribution were also plotted (Fig. 5.18 - Fig. 5.24). In this graphs mean bias errors are sorted by the height of the error. 140 Sky 212 T0 140 Sky 212 T3 a b Fig. 5.18. Cumulative MBE frequency distribution for sky modeled with 212 zones (a - time T 0 min, b – time T 3 min) 120 120 100 100 80 80 60 60 40 40 i 11,29 % 20 20 9,45 % -20 -20 -40 -40 150 Scale model measurements 140 Sky 145 T0 140 Sky 145 T3 a b Fig. 5.19. Cumulative MBE frequency distribution for sky modeled with 145 zones (a - time T 0 min, b – time T 3 min) Sky 97 T0 140 Sky 97 T3 a b Fig. 5.20. Cumulative MBE frequency distribution for sky modeled with 97 zones (a - time T 0 min, b – time T 3 min) Sky 26 T3 a b Fig. 5.21. Cumulative MBE frequency distribution for sky modeled with 26 zones (a - time T 0 min, b – time T 3 min) 120 120 100 100 80 80 60 60 40 40 11,29 °/o 20 20 9,46 % -20 -20 -40 -40 140 120 120 100 100 80 80 60 40 40 20 - 11,54 °/o 20 9,66 °/o -20 -20 -40 -40 160 140 120 120 100 100 80 80 60 60 40 - 20 20 -20 -40 151 Scale model measurements 140 Sky 13 T0 2,15 % Sky 13 T3 a b Fig. 5.22. Cumulative MBE frequency distribution for sky modeled with 13 zones (a - time T 0 min, b – time T 3 min) Dialux DQ TO Dia lux DQ T3 a b Fig. 5.23. Cumulative MBE frequency distribution for sky modeled with Dialux daylight factors (a - time T 0 min, b – time T 3 min) Illum ina nce DQ T3 a b Fig. 5.24. Cumulative MBE for sky modeled with illuminance daylight factors (a - time T 0 min, b – time T 3 min) 160 140 120 120 100 100 80 80 60 60 40 40 20 20 0,85 °/o -20 -20 -40 -40 140 140 120 120 100 100 80 60 60 44,43 % 40 40 - 20 20 -20 -20 -40 -40 160 140 140 120 100 80 80 60 60 40 40 - 20 20 -20 -40 -40 152 Scale model measurements One can see that the graphs of cumulative frequency distribution of mean bias error for both daylight factors are quite full and also the mean bias error (Table 5.7) is not in favour of the daylight factors. Daylight factor in all cases overestimate illuminances in the model. The reason is that daylight factors are calculated with overcast sky in case of Dialux factors and with uniform sky in case of illuminance factors. Average value of the MBE can be misleading, since the average MBE can be close to zero, but the individual values could be far from zero. To eliminate this possible misleading, also RMSE was calculated and results are presented in Table 5.8 and on Fig. 5.25. Relative RMSE was calculated with next equation (shown for fish-eye luminances averaged over 13 sky zones): RMSEsty13 = F -F Sky13-A2 -sh)13-A5 -sh)13-B1 ( 5.15) Where: RMSE sky13 E sky13-A1 E Licor-A1 root mean square error for sky modelled with 13 sky zones Calculated illuminance from 13 sky zone luminances in point A1 Measured illuminance with Licor sensor in point A1 Table 5.8. Table of RMS error Fish-eye luminances averaged over # sky zones Daylight factor 212 145 97 26 13 Dialux Illumin. RMSE for T0 [%] 25,21 25,37 25,42 24,28 23,69 75,31 97,98 RMSE for T3 [%] 22,56 22,69 22,65 22,42 21,84 74,23 95,91 Fig. 5.25. Root mean square error As can be seen from Table 5.7 and Table 5.8, the results are not in line with our expectations. As mentioned in the beginning of this paragraph the error should be E E E LLCor-A2 LICOF-A5 Ücor-B1 + + + ....+ E E E E Licor-A1 ÜCOF-A2 Licor-A5 Licor-B1 6 153 Scale model measurements inversely proportional to the number of sky zones used in the model. Mean bias error has practically the same value for sky models with 212, 145 and 97 zones and is around 11,5 % for measurements performed at the start of the sky scanner measurements and around 9,5 for measurements performed at the end of the sky scanner measurements (three minutes after the first measurements). Surprisingly the values of MBE with sky models with 26 and 13 zones are lower and they are less than 2,5% for both measurements (T0 and T3). Illuminances modelled with Dialux and illuminance daylight factors are over estimated in more than 85% cases for Dialux and in 95% for illuminance daylight factors (Fig. 5.23 and Fig. 5.24). The results for daylight factors did not surprised us, since the daylight factors are supposed to be used with uniform and overcast skies, what was not a case with our measurements. If we base our conclusion on root mean square error the figure is similar but not the same. Practically all scale model illuminances based on sky partitioning give the same RMS error, which is around 25 % for T0 measurements and 23 % for measurements performed at the end of the sky scanner measurements. Also with RMS error we get much higher values for both methods with daylight factors. RMS error is around 75 % for Dialux daylight factors and around 95% for illuminance quotient. 5.8 Conclusions for scale model measurements As already mentioned we get most of the light in indoor premises from the sky and not from the sun. If there is direct component of the sun penetrating into premises it is usually screened with blinds or other similar devices. In our research we wanted to investigate the influence of the number of sky elements in the sky model on the accuracy of indoor illuminance calculations. All the measurements that were performed with scale model were without direct sun and for that the difference in luminance of the darkest and brightest part of the sky was not great. If also direct sun would be taken into account, then the difference could be larger. As it can be carried out from the scale model measurements, in cases of homogeneous sky or sky without extreme differences in luminance, the number of sky elements is not playing an important role. Surprisingly the mean bias error is not inversely proportional with the number of the elements used in the sky model and from results (Table 5.7 and Table 5.8) we can conclude that the most convenient number of sky zones used for sky models is 13. Results are pretty the same for 26 and 13 sky zones, but a smaller number of zones, means a shorter computing time. In this place we should also think about the simplifications we made. One of the most important simplifications is that in our scale model we did not use any glazing. Since the transmission factor of the glazing depend on the direction of the light penetrating into the rooms, it is difficult to say for certain that 13 elements are enough. 154 Conclusion 6 Conclusion The main topics of the thesis are daylight measurements and calculations connected with daylight availability. The most important contribution of the thesis is selection of the sky luminance model that can be used to derive frequencies of appearance of CIE sky types from satellite images (global and diffuse horizontal irradiances) and to calculate time values of indoor daylight illuminances. Selection of the best performing sky luminance model is based on sky luminance distribution measurements performed in Lyon between April 2005 and December 2005. Sky luminance distribution measurements were carried out with an EKO sky scanner, borrowed from Kyushu University (Japan). This scanner measures sky luminance distribution with accuracy over 145 sky zones in approximately 3 minutes. The sky scanner was calibrated under artificial sky and under real sky conditions before performing measurements. Together with the installation of the sky scanner we carried out also some modifications to vertical illuminance sensors on ENTPE IDMP station. Measurement data were analysed in different ways. The most important analysis was deriving the CIE sky type. CIE sky type was defined firstly on basis of indicatrix and gradation groups, which are defined with CIE standard CIE S 011/2003. Secondly the CIE sky type was defined with ratio between zenith luminance and diffuse horizontal illuminance. The last method we used was a statistical method. With all three methods, for every sky scan, one of the 15 CIE sky types was determined. At the end of chapter two, we present frequencies of CIE sky types derived with different methods are shown. All the data were analysed and the complete database is available online at http://lrf.fe.uni-lj.si/ekoscan.htm. There are only a few working sky scanners in the world. The main reason is that this equipment is expensive and it needs regular attendance and servicing. We tried to find a new, more effective method, for a reasonable price by using a digital camera with a fish-eye lens. We compared more than 160 pairs of digital images and sky scans. The results show that a digital camera with a fish-eye lens can be used as a sky scanner with some limitation. For instance there are problems with rain, temperature variations, and it’s easy to burn a CCD if we do not use a single-lens reflex (SLR) camera. When selecting the sky luminance model that can be used to derive frequencies of occurrence of CIE sky types from satellite images, we have reviewed the following sky luminance models: Perez all weather model, ASRC-CIE model and Igawa model. For the times, when we had sky scanner measurements (13.006 cases), sky luminance distributions were produced with all three models. The modelled luminances are calculated with model’s equations and irradiance data measured by IDMP station. Based on all comparisons, we can conclude that Igawa model is best for extremely cloudy sky types, Perez All weather model gives best results with intermediate sky types and ASRC-CIE model is in most cases in between upper two models, but on average the best model. 155 Conclusion In last part of the thesis, we draw conclusions on needed density of the grid of sky luminance model. To come to those conclusions we performed measurements on scale model. Measurements were carried out on scale model with black indoor walls to eliminate multiple indoor reflections. In the scale model, illuminances were measured and the same illuminances were modelled with different methods. One method was by means of directional daylight factor (DDF). DDFs were calculated for different densities of sky luminance model grid (212, 145, 97, 26 and 13 zones). The other method was with traditional daylight factor. Results show that the number of sky zones has almost no influence to MBE and after that we can conclude that 13 zones are enough to model sky luminance if we are considering only sky vault without direct sun. Conclusions about the use of digital camera as a sky scanner and needed accuracy of the sky luminance model and with that needed number of sky zones that should be used in a model are based on analysis of a large amount of measured data. The measured data and the complete analysis are gathered in two databases. The first one is a database of outdoor measurements (comparison of sky scanner measurements and images taken with digital camera) and the second one is a database of scale model measurements (measurements of illuminances in the scale model accompanied with pictures of the sky vault and luminance maps in the scale model). Both databases are available to public on CD upon request. During our research, we have realized that there is still much work that could be done in this field. When we were talking with the daylight designer, we got the impression that one of the first things that could be done is a table with frequencies of occurrences of CIE sky types for at least capital cities of EU countries. In next iteration we could broaden this table with all bigger cities in EU and in countries which are covered with Meteosat geostationary satellite (http://www.satellight.com). Performing measurements on scale models is easy and cheap if we compare it with measurements in real buildings. But in our future work we should change model with real building. Those measurements should be long-term and during the measurements we should consider also different sorts of glazing and different sorts of blinds. When working with real buildings or bigger scale models (size of a window should be compared to real window situation) also the luminance intensity distribution could be measured with an automated goniophotometer on the inner side of a window. In our work we conclude that a digital camera could be used as a sky scanner with some limitations, we could also perform long-term measurements of sky luminance distribution with digital camera. A camera should be put in a tempered box, covered with a spherical glass cover, such as used for collectors with light-pipes. A problem with burning CCD could be eliminated with use of single-lens reflex (SLR) camera. 156 Contributions of the thesis to the science Contributions of the thesis to the science Main contribution of the thesis to the science is a selection of a sky luminance model that allows the termination of frequency of emergence of CIE sky types at a specific location on the basis of global and diffuse horizontal irradiances. The model is defined and it can be used on data available from satellite images. Another contribution to the science is definition of partitioning of the sky vault. In the thesis we suggest the number of zones that should be used when calculating daylight availability without direct sun. Beside the selection of the sky luminance model and suggested number of sky zones the contribution to the science are also next large databases: - database of all sky scanner measurement that were performed at ENTPE in year 2005. In the database a reader can find measured sky luminance values together with the analysis of the sky luminance distribution (CIE sky type, indicatrix and gradation group…) - database of outdoor measurements (comparison of sky scanner measurements and images taken with digital camera) - database of scale model measurements (measurements of illuminances in the scale model accompanied with pictures of the sky vault and luminance maps in the scale model. 157 Contributions of the thesis to the science 158 Statement of the authorship Statement of the authorship Undersigned Matej B. Kobav, born on March 21st 1974 in Ljubljana, declare that I have written the PhD thesis "Development and validation of methods used to compute time values of indoor daylight illuminances" on my own under the supervision of dr. Grega Bizjak and dr. Dominique Dumortier. Help of other collaborator is indicated in acknowledgements. Matej B. Kobav 159 Statement of the authorship 160 References References [1] Dumortier D., Mesure, analyse et modélisation du gisement lumineux. Application a l'évaluation des performances de l'éclairage naturel des bâtiments. (1995) PhD Thesis, Université de Savoie, France, 350 pages. [2] Darula S., Kittler R., Kmeto P., New CIE General Sky Defining Luminance Distributions, SUSTAINABLE BUILDING & SOLAR ENERGY 2001, Brno, 2001 [3] Nakamura H., Oki M., Hayashi Y., Luminance distribution of intermediate sky, J. Light and Vis. Env., 9, 1, pp. 6-13. [4] Igawa N., Nakamura H., Matsuzawa T., Koga Y., Goto K., Kojo S., Sky luminance distribution between two CIE standard skies 1,2, Pros. Lux Pacifica, pp. 359-373. [5] Igawa N., Nakamura H., Matsuura K., Sky luminance distribution model for simulation of daylit environment, BS99 (Building Simulations 1999), 1999, Kyoto. [6] CIE, Spatial Distribution of Daylight - Luminance Distributions of Various Reference Skies, Pub. CIE No.110 1994 ISBN 3 900 734 52 6, 1994. [7] CIE, Daylight International recommendations for calculation of natural daylight, Pub. CIE no. 16 (E-3.2) 1970. [8] ISO 15469/CIE S003, Spatial Distribution of Daylight – CIE Standard overcast and clear sky, 1996. [9] http://genelux.entpe.fr/sommaire.html [10] http://www.satellight.com/guide/advday.htm. [11] Kaehler K., Flächenhelligkeit des Himmels und Beleuchtungsstärke in Raumen. Meteorol. Zeitschr., 25, 2, pp 52-57, 1908. [12] Kimball H., Hand I. F. Sky brightness and daylight illumination measurements. Monthly Weather Rev. 49, 9, pp. 481-488, 1921 [13] Moon P., The scientific basis of illuminating engineering, Dover publications, New York, 1961. [14] Perez R., Seals R., Michalsky J., All Weather Model for sky luminance distribution – preliminary configuration and validation, Solar Energy, Volume 50, Issue 3, PP 234-245, 1993. [15] Kittler R., Darula S., Perez R., A set of standard skies, characterizing daylight conditions for computer and energy conscious design, American-Slovak grant project, US-SK 92 052, 1998. [16] CIE DS 011.1/E:2003 Spatial distribution of daylight – CIE standard general sky, Standard, CIE Central Bureau, Vienna, 2003. [17] KOBAV Matej Bernard, Razvoj nadomestnega svetlobnega vira za izraèun osvetljenosti prostora z dnevno svetlobo, magistrsko delo, Fakulteta za elektrotehniko Univerza v Ljubljani, Ljubljana, 2003. 161 References [18] Michalsky J., Perez R., Seals R., Evaluation of Algorithms for Sky Luminance Distribution Prospects for Performance Improvements, ISES World Congress, Denver Colorado, 1991. [19] Esteban C., Torres A., Fasulo A., Tocho J, Adjusted Model of Linke's Turbidity Factor, EuroSun 98, Portorož, Slovenia, 1998. [20] Roy G. G., Ruck N., Reid G., Winkelmann C. F., Julian W., The Development of Modelling Strategies for Whole Sky Spectrum under Real Conditions for International Use, University of Sydney, 1995. [21] Daylight Simulation: Methods, Algorithms and Resources, A report of IEA SHC 21/ECBCS ANNEX 29, December 1999, Berkeley, California. [22] Kittler R., Perez R., Darula S., A New Generation of Sky Standards, Proceedings Conference Lux Europa 1997, pp 359-373, Amsterdam, 1997. [23] Perez R., Ineichen P., Seals R., Michalsky J., Stewart R., Modeling daylight availability and irradiance components from direct and global irradiance, Solar Energy, Vol. 44, Issue 5, PP 271-289, 1990. [24] Perez R., Ineichen P., Moore K., Kmiecik M., Chain C., George R., Vignola, F., A new operational model for satellite-derived irradiances: description and validation Solar Energy, Vol. 73, Issue 5, PP 307-317, 2002. [25] Chain C, Dumortier D., Fontoynont M., A Comprehensive Model of Luminance, Correlated Colour Temperature and Spectral Distribution of Skylight: Comparision with Experimental Data, Solar Energy, Volume 65, Issue 5, PP 285-295, 1999. [26] Ineichen P., Molineaux B., Perez R., Sky luminance data validation: Comparison of seven models with four data banks, Solar Energy, Volume 52, Issue 4, PP 337-346, 1994 [27] Rutten AJF, Sky luminance measurements for design and control of indoor daylight illumination, Lighting Research and Technology, Volume 22, PP 189–192, 1990 [28] Littlefair PJ, A comparison of sky luminance models with measured data from Garston, United Kingdom. Solar Energy, Volume 53, PP 315–322, 1994 [29] Dumortier, D., Van Roy F., Daylighting information throughout Europe from the SATEL-LIGHT and SODA Internet servers. CIE 25th Conference San Diego, 2003, 1, D3-4. [30] Karayel M, Navvab M, Ne’eman E, Selkowitz S, Zenith luminance and sky luminance distributions for daylighting calculations, Energy and Buildings, Volume 6, PP 283–291, 1984. [31] Moon P, Spencer DE, Illumination from a non-uniform sky, Illuminating Engineering, Volume 37, PP 707–726, 1942 [32] Kittler R., Ruck N., Definition of typical and average exterior daylight conditions in different climatic zones, Energy and Buildings, Volume 6, Issue 3, PP 253-259, 1984. [33] Tregenza P.R., Subdivision of the sky hemisphere for luminance measurements, Lighting Research and Technology, Volume 19, Issue1, PP. 13–14, 1987. 162 References [34] Kittler R., Darula S., Perez R., Advantages of new sky standards: more realistic modelling of daylight conditions in energy and environmental studies. International Journal of Energy, Environment and Economics, Volume 8, Issue 1, PP 65–71, 1999. [35] Bartzokas A., Sky luminance distribution in Central Europe and the Mediterranean area during the winter period, Journal of Atmospheric and Solar-Terrestrial Physics, Volume 65, Issue 1, PP 113-119, 2003. [36] Darula S., Kittler R., Parametric definition of the daylight climate, Renewable Energy, Volume 26, Issue 2, PP 177-187, 2002. [37] Tregenza, P.R., Analysing sky luminance scans to obtain frequency distribution of CIE Standard General Sky. Lighting Research and Techology, 2004. 36(4): pp 271-281 [38] Kittler, R. Darula, S., New trends in daylighting theory based on the new ISO/CIE sky standard. 1. Zenith luminance on overcast skies. Building research journal, Vol. 52, No. 3, 2004 [39] Igawa, N. Koga, Y. Matsuzawa, T. Nakamura, H. Models of sky radiance distribution and sky luminance distribution. Solar Energy Vol. 77, No. 4, 2004: pp. 137-157. [40] Perez, R. Michalsky, J. Seals, R., Modelling sky luminance angular distribution for real sky conditions. Expiremental evaluation of existing algorithms, J. Illumin. Eng. Soc. 21(2), pp 84-92. 1992. [41] Moon, P. Spencer, D. E., Illumination from a nonuniform sky, Illum. Eng. (N.Y), 37, pp. 707-726,1942. [42] Kittler, R. Darula, S., New trends in daylighting theory based on the new ISO/CIE sky standard. 2. Typology of cloudy skies and their zenith luminance. Building research journal, Vol. 52, No. 4, 2004 [43] Kittler, R. Darula, S., New trends in daylighting theory based on the new ISO/CIE sky standard. 2. Zenith luminance formula verified by measurement data under cloudless skies. Building research journal, Vol. 53, No. 1, 2005 [44] Tregenza, P.R., Perez, R., Seals, R., Michalsky, J., Guide to recommended practice of daylight measurement, CIE TC 3.07, 1993 [45] Kittler, R., Relative scattering indicatrix: derivation from regular radiance/luminance distribution sky scans, Lighting Res. Technol. Vol. 25, No. 3, pp 125-127, 1993 [46] Kittler, R., Hayman, S., Ruck, N., Julian, W., Daylight measurement data: Methods of evaluation and representation, Lighting Res. Technol. Vol. 24, No. 4, pp 173-187, 1992 [47] Schramm, W., Über die Verteilung des Lichtes in der Atmosphäre. Schriften d. Naturw.Veriens f. Schl.-Holst., Vol. 12, No. 1, 1901, pp 81-129 [48] Pokrowski, G. I., Über die Helligkeitsverteilung am Himmel. Phys. Zeitsch, Vol. 30, No. 20, 1929, pp 697-700 [49] Boldyrev, N. G., O raspredeleniyi jarkosti po nebu (About the distribution of luminance on sky), Svetotekhnka, Vol. 6., 1935, pp 16-18 [50] Krat, V.A., Indikatrisa rasseyaniya sveta v zemnoj atmosfere. (Indictrix of light diffusion in earth atmosphere), Astronom. J., Vol. 20, 1943, pp 5-6. 163 References [51] Kittler, R., Standardisation of the outdoor conditions for the calculation of the Daylight factor with clear skies, Sunlight in buildings, Bouwcentrum Rotterdam, 1967, pp 273-286. [52] CIE, Standardization of luminance distribution on clear skies, Pub. CIE No. 22, 1973. 164