19
 ACTA BIOLOGICA SLOVENICA 
 LJUBLJANA 2006 Vol. 49, [t. 1: 19–26
Sprejeto (accepted): 2006-12-28
Survey of the Lynx distribution in the French Alps: 2000–2004 population 
status analysis
Pregled razširjenosti risa v Francoskih Alpah: analiza statusa populacije za ob-
dobje 2000–2004
E. MARBOUTIN*, C. DUCHAMP, P. ROULAND, Y. LÉONARD, J. BOYER, D. MICHALLET, 
M. CATUSSE, P. MIGOT, J. M. VANDEL, & P. STAHL
*corresponding author: e-mail: e.marboutin@oncfs.gouv.fr
Abstract. Within the SCALP framework, the status of the pan-alpine population of Eurasian 
Lynx is assessed every 5 years, based on the compilation of national reports and standardized 
classifi cation of lynx presence signs according to data confi dence levels (C1, C2, C3). From 2000 
to 2004, the French national network of lynx experts collected N= 393 data, out of which 224 
(compared to only 69 in 1995–1999) were considered as robust enough to evidence the presence 
of lynx (C1 = 1%; C2 = 42%; C3 = 57%) and were used for further analysis. A majority of the 
signs concerned the northern part of the Alps, however, in mostly two regions (Chartreuse/Epine 
: 34% of the signs; Maurienne: 21%). Other data were more scattered over space, from the Cha-
blais region close to Switzerland down to the Haut-Verdon close to the Mercantour mountains. 
A negative trend was noticed from north to south in proportions of best quality signs (C1+C2), 
and a positive one in low quality ones – C3 – (χ² = 3.56, 1 df, p = 0.06), which could point at 
some methodological artefacts. Discarding C3 may however be too conservative a strategy to 
assess the species range and status. Using spatial recurrence and trend over time of all signs 
available (C1+C2+C3) could, therefore, provide the right balance between being too much 
versus not enough conservative. – When doing so, the area with lynx signs regularly detected 
sharply increased between 1996–1998 (100 km²), 1999–2001 (250 km²), and 2002–2004 (1195 
km²). The latter area is still quite small regarding what is required for a viable large carnivore 
population. A simple demographic model suggested that even a quite moderate proportion of 
immigrants (e.g. dispersal infl ow from neighbouring core areas – French Jura or Swiss Alps) could 
considerably decrease the theoretical demographic extinction risk of such a small population, but 
still depending upon adult survival rates, which also strongly infl uenced the extinction risk. The 
factors that may infl uence this sensitivity analysis (such as habitat connectivity and management 
of wooded corridors) should be evaluated within the Scalp framework.
Keywords: Lynx lynx, France, Alps, distribution, monitoring, population viability
Introduction
Standardized monitoring over countries that share large carnivore populations is obviously the fi rst 
step towards a common management of these species. Over Europe, such an international collaboration 
for population monitoring is now properly implemented only for the Eurasian lynx over the Alps within 
the SCALP framework (MOLINARI-JOBIN & al. 2003). The status reports about the national sub-units 
of this conceptual population build up a key-issue for assessing the overall status of the pan-alpine 
Acta Biologica Slovenica, 49 (1), 200620
“meta-population” (see Hystrix, vol. 12(2), 2001 for the 1995–1999 period), and regular meetings 
held under the auspices of SCALP yield valuable contributions (e.g. MOLINARI-JOBIN & al. 2005). 
The present paper provides the 2000–2004 French update, together with some simple demographic 
modelling to roughly enlighten the importance of dispersal and connectivity on the demographic via-
bility of the ‘French’ alpine sub-population. Dispersal, indeed, is a key-parameter when considering 
fragmented and/or small populations (see e.g. SCHADT 2002; ZIMMERMANN 2004). Factors affecting the 
habitat continuity – e.g. roads and traffi c volume, fencing – may, therefore, result in barrier effects to 
dispersal, and increased population extinction risk due to isolation (KLAR & al. 2006).  
Methods
Lynx monitoring in France
The lynx monitoring in France is based on an extensive fi eld work by a national network of about 
850 lynx-experts who have been specially trained to collect possible presence signs (scats, tracks, 
visual observations, wild and domestic preys). All the data are validated by a single national expert 
(Offi ce National de la Chasse & de la Faune Sauvage) using a standardized grid of criteria that basi-
cally relies on the degree of convergence between technical characteristics within each presence sign 
(see VANDEL & STAHL 2005, for a detailed description). Such a centralized process ensures that any 
fi eld data is analysed in the same way, wherever it comes from and whoever collected it. The presence 
signs, once validated, are converted into C1, C2, C3 categories to fi t to the SCALP requirements: C1 
are hard facts such as captures, dead lynx, photos; C2 are data directly collected by lynx-experts and 
further confi rmed by the national expert; C3 are data indirectly collected by lynx-experts from the 
general public and confi rmed by the national expert. Biologists in charge of evaluating the lynx status 
usually devote most consideration to direct data fi rst (C1+C2).
Regarding range estimates, point data (i.e. defi ned by X, Y coordinates) were transformed fol-
lowing VANDEL & STAHL (2005)’s method: each data was attributed a spatial buffer of 81 km² grid 
area of theoretical lynx presence, made of nine 3 x 3 km elementary squares, centred on the given 
X,Y coordinates. The sum of the squares was the estimated overall range. When overlapping maps 
from different yearly periods, the elementary squares that were regularly “lynx-positive” made up the 
regularly occupied area, a conservative estimate of lynx distribution (since areas newly or irregularly 
detected were discarded). 
Demographic modelling
Because the French alpine population may be considered as small relative to other alpine ones 
(VON ARX & al. 2004).and may be demographically connected to those from the Jura Mountains and 
Swiss Alps, its long term viability may depend on immigration from these areas. Using Monte Carlo 
runs within the ULM package (LEGENDRE & CLOBERT 1995), a simple female-based model (with 3 age 
classes: kitten, sub-adult, adult; see the life cycle and structure of the model, Annex 1) with demo-
graphic stochasticity on vital rates was used to compute relative population viability analyses (PVA) 
according to the proportion of additional input from immigration. Mean survival and fecundity rates 
were from the literature (SCHADT 2002), and the infl uence of dispersal on the population extinction 
risk was modelled, step-by-step, by adding a given proportion of immigrant sub-adults to the initial 
population size. Because the colonizing process within the French Alps is still active over a very large 
un-colonized area, dispersal of local sub-adults out of the Alps was set to zero – i.e. immigration to 
but no emigration from the French Alps. Because the dispersal success may depend on habitat frag-
mentation, an additional barrier mortality was incorporated into the model, simulating either strong 
connectivity (i.e. weak additional mortality of 1/
3
) or weak connectivity (i.e. large additional mortality 
of ½). The extinction risk was estimated by the proportion of trajectories that went under a minimum 
of 1 individual within 1000 trajectories simulated over 100 years. 
21E. Marboutin & al.: Survey of the Lynx distribution in the French Alps: 2000–2004 population status ...
Annex 1: Life cycle and structure of the demographic lynx model 
 The model is run in the framework of demographic stochasticity on survival rates, to simulate the chance 
extinctions due to small numbers of individuals. Both the immigration and survival of sub adults while 
dispersing between sub populations are modulated. Transition probabilities between age classes are fe-
cundity and survival rates from the literature. Two level of habitat connectivity between populations are 
simulated: a weak connectivity associated to a large cost of dispersal (i.e. a strong additional mortality rate 
of 50%); a strong connectivity associated to a low cost of dispersal (i.e. a weak additional mortality rate. 
Results
Lynx distribution
During the 2000–2004 period, N = 393 data have been collected, out of which 55% have been 
fi nally validated and used for further analysis. Despite this large number of data discarded, a sharp 
increase in the number of validated data is observed for the last pentad (Table 1). Although C3 are still 
in a majority, robust data about the lynx presence (i.e. C1+C2), are obviously increasing too. Most 
of the presence signs were, however, still concentrated over some very limited areas in the northern 
French Alps (Fig. 1), such as the Chartreuse / Epine massif, the Maurienne valley, and the Bauges 
massif (respectively n = 72, n = 45, and n = 20, i.e. 34%,  21%, and 9% of all signs of presence). North 
to Annecy and south to Grenoble, the data were more or less scattered over space, from the Chablais 
region down to the Haut-Verdon. Location of data (north to Grenoble vs. south to Grenoble) and data 
type (C1+C2 vs. C3) were not independent (Table 2, χ² = 3.56, 1 df, p = 0.06): there was a negative 
trend from north to south in proportions of C1+C2, and, conversely, a positive one in C3. 
Acta Biologica Slovenica, 49 (1), 200622
Fig. 1: Distribution of validated lynx signs (
= C1, 
= C2, = C3) collected from 2000 to 2004 in the French 
Alps; shaded areas represent altitudinal patterns (the darker, the higher).
Table 1: Numbers of lynx presence data, according to SCALP categories, validated over the French Alps.
Categories 1990–94 1995–99 2000–04 Total
C1  2  0  3  7
C2  5  7  92 103
C3 24 62 128 214
Total 31 69 224 324
Regarding range estimates, the area regularly occupied (using C1+C2+C3) increased from 100 
km² in 1996–1998, to 250 km² in 1999–2001, and up to 1195 km² in 2002–2004. When adding areas 
newly detected, for which no one knows whether they will fi nally contribute to the regular area of the 
species, the total estimate amounted to 4444 km². Because the latter value is based on large numbers 
of C3 detected for the very fi rst time in new areas, one would better consider the lower range estimate 
(1195 km²), computed only from those C1+C2+C3 that were recurrent over time.
Table 2:  Unbalanced numbers of lynx presence data, according to SCALP categories (C1+C2 versus C3), and 
to geographical location.
Categories North to Grenoble South to Grenoble
C1+C2  80 (46%) 16 (31%)
C3  93 (54%) 35 (69%)
Total 173 51
23E. Marboutin & al.: Survey of the Lynx distribution in the French Alps: 2000–2004 population status ...
Population dynamics modelling
Demographic parameters were derived from Schadt (2002). Survival rates were set at 0.50 (kit-
tens), 0.65 (sub-adults), 0.75–0.80 (adults); fecundity was 1 for the fi rst attempt to breed, and 2 for 
older females. When using such values within a simple matrix-based deterministic model, the yearly 
population growth rate was λ= 1.02 – 1.07 (i.e. 2 to 7% increase/year); λ was more sensitive to changes 
in adult survival rates (elasticity: 0.66) than to changes in any other vital rate (e.g. overall fecundity : 
0.17): a 10% increase in adult survival would yield a 10 x 0.66 = 6.6 % increase in λ, whereas a similar 
10% increase in fecundity would yield only a 10 x 0.17 = 1.7% increase in λ. Within the stochastic 
framework (Monte Carlo runs), the extinction risks were, therefore, modelled according to changing 
adult survival rates (0.75 or 0.80); the dispersal success between source and target populations was 
modulated too, using additional mortality rates of 1/
3
 or ½ as a simulation of differences in habitat 
connectivity due to e.g. fragmentation of wooded corridors [i.e. survival while dispersing within a 
patch: 0.65; survival while dispersing between patches: 0.65x(1–0.33) = 0.50 or 0.65x(1–0.50) = 0.325 
according to high vs. low habitat connectivity]. 
A rough and conservative estimate of lynx numbers in the French Alps may be obtained using an 
average winter density of 1 adult/100 km² together with 0.5 young/100 km² (HALLER & BREITENMOSER 
1986; BREITENMOSER-WÜRSTEN & al. 2001) over the estimated range (1195 km²). Assuming a balanced 
sex-ratio, half of the resulting value was used as an initial population size (i.e. 9 females) in Monte 
Carlo runs to simulate extinction of population trajectories. 
The extinction risk decreased sharply with increasing immigration rates, and reducing the level of 
theoretical mortality while dispersing from higher to lower values improved population persistence 
too (Figure 2A). This pattern was most pronounced when adult survival rate was lower: once this rate 
amounted 0.80, the extinction risk was moderate even with no input from immigration (Figure 2B). 
The infl uence of immigration on extinction risk logically depended on survival rates (of sub-adults and 
adults), but some kind of similar ‘threshold effect’ was observed with a 5–10% immigration rate.
Fig. 2: Extinction risk (y-axis) as a function of increasing (0 to 20%) immigration rates (x-axis); survival of 
dispersing sub-adults is modulated (A- barrier effect) together with that of philopatric adults [B- : 
S-ad = 0.80; : S-ad = 0.75].
Acta Biologica Slovenica, 49 (1), 200624
Discussion
During the 2000–2004 period, the strong increase in numbers of lynx signs collected is likely to 
refl ect both a higher sampling rate (quite a large number of new lynx-fi eld experts have been additionally 
trained to collect possible lynx signs), and an actual north-to-south colonizing process of the lynx. In 
spite of this active colonizing process, the detected distribution area of the species is still composite: 
north to Grenoble, the range is more or less continuous and documented by quite robust data (C1+C2), 
whereas, southward to this latitude, only islets of presence that are mostly C3-based have been detected 
so far. Because the lynx expert network is implemented now in the whole possible distribution area 
of the species, the latter trend (more and more C3 south to the core area) could illustrate sampling 
artefacts (C3 being more likely in those newly- or even non-colonized areas). This might therefore 
suggest that the lynx status over the French Alps be fi rst assessed in a conservative way, i.e. using 
preferably C1+C2 data only. However, within C3s collected at time t, those that were actual artefacts 
are unlikely to be spatially recurrent later on, whereas those that were not artefacts are likely to be next 
confi rmed either as C3s, C2s, or even C1s. The spatial recurrence of all data available (C1+C2+C3) 
could, therefore, be used as a complementary approach to assess the lynx status. 
The spatial patchiness in the distribution of lynx signs may refl ect a low effi ciency of the expert 
network to record these data under the alpine environmental conditions. The relationship between the 
locations of lynx signs of presence and the surrounding eco-variables (altitude, steepness, percentage 
of wood, distance to roads or cities) have been modelled using the ENFA method (HIRZEL & al. 2002; 
BASILLE 2004). The resulting map displayed a very patchy distribution of areas where lynx signs would 
likely be detected (Figure 3), and a methodological bias due to habitat accessibility was suspected (e.g. 
a negative relation was noted between signs of occurrence and increasing distance to roads). Contrary 
to the academic and biological fi ndings in ZIMMERMANN (2004), our map refl ects only the sub-sample 
of the potential distribution area for which the expert network could detect lynx signs of presence. 
The next issue is to improve the detection rate of such signs, based on e.g. an extensive use of remote 
camera traps or hair snares (see ZIMMERMANN & al. 2006, MARBOUTIN & al. 2005). Despite the possible 
under estimation of the range occupied, lynx presence signs are however found over larger and larger 
areas; the species is now well established and regularly detected in several mountainous geographic 
entities (see Table 2 in VANDEL & STAHL 2005 for a detailed review). Compared to the previous SCALP-
update (1995–1999, STAHL & VANDEL 2001), numbers of detected signs and corresponding areas are, 
from north to south: i) stable north to Annecy (Chablais, Chamonix, Glières-Aravis , Vuache-Salève); 
ii) stable (Belledone-Oisan-Taillefer) or increasing (Bauges, Maurienne, Chartreuse-Epine) between 
latitudes of Annecy and Grenoble; iii) stable but scarce and scattered (Dévoluy- Beauchêne, Valbonnais-
Valgaudemard, Briançon-Queyras) between latitudes of Grenoble and Gap; iv) still to be confi rmed 
(Monges, Embrunais-Ubaye, Haut-Var, Haut-Verdon-Canjuers) south to Gap (Fig. 1). Such a patchy 
distribution of lynx signs results in a small proportion of the total area being regularly occupied: in 
2002–2004, the overall range detected was about 4500 km² out of which only 1200 km² with regular 
presence. The corresponding population size (roughly estimated to less than 20 animals) can obviously 
not be considered a long term viable unit, from the demographic or genetic point of view.
From a theoretical basic modelling, the infl uence of demographic stochasticity on extinction 
risk could be buffered fi rst with increasing adult survival rates, and with moderate immigration rates 
(5–10 %). Immigration also means that the local dynamics within the source population are very 
important too. Population simulations are projections rather than exact predictions, because they 
rely on the quality of both model structure and demographic data. They should mostly be used, as a 
result, to evaluate relative outputs of different scenarios. In the present case, reducing for example the 
theoretical mortality induced by the barrier effect from ½ to 1/
3
 , when dispersal rate is 0.15, would 
induce a 50% relative decrease in extinction risk (from 0.2 to 0.1). Such results should only be regarded 
relative values, as they are partly conditional on the structure of the model and parameters’ values. 
Increasingly powerful but complicated models are available (e.g. Schadt & al. 2002, Wiegand et al. 
25
Fig. 3: ENFA-based modelling of the potential distribution of detected lynx presence signs. The grey areas are those 
with higher detection likelihood, i.e. those where the lynx-experts network would likely collect presence 
signs given the presence of the species AND the environmental conditions (slope, altitude, wooded area, 
distance to roads).
E. Marboutin & al.: Survey of the Lynx distribution in the French Alps: 2000–2004 population status ...
2004), so the trade-off is now between richness of model structure and availability of fi eld estimates 
for their parameters. Above all, the present results should be analysed as an illustration that factors 
affecting dispersal patterns may be key-ones, but conditional on patterns in adult survival rates. When 
these vital rates are to fl uctuate over time/space (e.g. due to diseases, or man-induced mortality) 
the buffering infl uence of immigration on extinction risks should not be neglected. Some emphasis 
should also be put on the study of dispersal patterns since recent results have shown this phenomenon 
is area-specifi c (ZIMMERMANN & al. 2005). Factors that may improve the dispersal success, such as 
habitat connectivity and management, based on the conservation of e.g. wooded corridors, should 
therefore be evaluated as a possible key-issue for lynx conservation (ZIMMERMANN 2004, KLAR & al. 
2006). The SCALP approach perfectly fi ts into that framework since it makes use of trans-boundary 
monitoring of populations and management of key-factors as a basis for defi ning what could be a 
robust conservation biology strategy. 
Acknowledgements
The authors are indebted to the lynx experts from the national network for providing the fi eld data; 
many thanks to A. Molinari-Jobin, SCALP coordinator, for having created the pan-alpine scientifi c 
network and for fruitful discussions about related concepts. The authors are indebted to A. Molina-
ri-Jobin, P. Molinari & M. Wölfl  for comments on a fi rst draft. Many thanks to F. Zimmermann for 
advices regarding ENFA application.
Acta Biologica Slovenica, 49 (1), 200626
References
BASILLE, M. 2004: Le Lynx, l’ENFA, et le SIG. Msc. report, Univ. Lyon C. Bernard.
BREITENMOSER-WÜRSTEN, Ch., F. ZIMMERMANN, A. RYSER, S. CAPT, J. LAASS, A. SIEGENTHALER & U. 
BREITENMOSER 2001: Untersuchungen zur Luchspopulation in den Nordwestalpen der Schweiz 
1997–2000. KORA-Bericht 9: 1–88.
HALLER, H. & U. BREITENMOSER 1986: Zur Raumorganisation der in den Schweizer Alpen wiederan-
gesiedelten Population des Luchses (Lynx lynx). Z. Säugetierkunde 51: 289–311.
HIRZEL, A., J. HAUSSER, D. CHESSEL & N. PERRIN 2002: Ecological-niche factor analysis: how to compute 
habitat suitability maps without absence data? Ecology 83: 2027–2036.
KLAR, N., M. HERMANN & S. KRAMER-SCHADT 2006: Effects of roads on a founder population of lynx 
in the biosphere reserve “Pfälzerwald-Vosges du Nord”. Natureschutz und Landschaftsplanung 
38: 330–337.
LEGENDRE, S. & J. CLOBERT 1995: ULM, a software for conservation and evolutionnary biologists. J. 
Appl. Stat. 22: 817–834.
MOLINARI-JOBIN, A., P. MOLINARI, C. BREITENMOSER-WÜRSTEN, M. WOELFL, C. STANISA, M. FASEL, P. 
STAHL, J.M. VANDEL, L. ROTELLI, P. KACZENSKY, T. HUBER, M. ADAMIC, I. KOREN & U. BREITENMOSER 
2003: Pan-Alpine Conservation Strategy for the Lynx. Nature and environment 130, Council of 
Europe Publishing, 20 pp.
MARBOUTIN, E., A. LAURENT, C. REGAZZI, F. LÉGER, P. MOISSON, M. LAMBRECH, L. BALESTRA, J.P. HENRI, 
M. BASILLE, L. TOUZAIN & D. MICHALLET 2005: Tests de nouvelles méthodes pour le suivi des po-
pulations de Lynx en France: le piégeage photographique en coulées et les pièges à poils. Oncfs 
Rapport Scientifi que 2004: 18–21. 
MOLINARI-JOBIN, A. 2005: Proceedings of the 2nd conference on the Status and Conservation of the 
Alpine Lynx Population, 7th to 9th May 2003, Amden Switzerland. Environmental encounters 
58: 1–105. 
SCHADT, S. 2002: Scenarios assessing the viability of a lynx population in Germany. Ph-D thesis, 
Univ. Munich, 116p. 
SCHADT, S., REVILLA E., WIEGAND T., KNAUER F., KACZENSKI P. & BREITENMOSER U. 2002: Assessing 
the suitability of central European landscapes for the reintroduction of Eurasian lynx. Journal of 
Applied Ecology 39: 189–203.
VANDEL J.M. & P. STAHL 2005: Distribution trend of the Eurasian lynx (Lynx lynx) populations in 
France. Mammalia 69 (2): 145–158.
VON ARX M., C. BREITENMOSER-WÜRSTEN, F. ZIMMERMANN & U. BREITENMOSER 2004: Status and con-
servation of the Eurasian Lynx (Lynx lynx) in Europe in 2001.  KORA report Nr. 19.
WIEGAND T., REVILLA E. & KNAUER F. 2004: Dealing with uncertainty in spatially explicit population 
models. Biodiversity and Conservation 13: 53–78.
ZIMMERMANN, F. 2004: Conservation of the Eurasian lynx (Lynx lynx) in a fragmented landscape – habitat 
models, dispersal, and potential distribution. Ph-D thesis, Univ. Lausanne, 179p. 
ZIMMERMANN, F., C. BREITENMOSER- WÜRSTEN & U. BREITENMOSER 2005: Natal dispersal of Eurasian 
Lynx (Lynx lynx) in Switzerland. Journal of Zoology, London 267: 381–395.
ZIMMERMANN, F., A. MOLINARI-JOBIN, P. MOLINARI, A. SIEGENTHALER  & U. BREITENMOSER 2006: Bericht 
zum Luchs-Monitoring mittels Fotofallen in den Nordwestalpen und der Zentralschweiz West im 
Winter 2005/06. KORA –Berichte.