Ecology of leishmaniasis in the South of France. 22. Reliability and representativeness of 12 Phlebotomus ariasi, P. perniciosus and Sergentomyia minuta (Diptera: Psychodidae) sampling stations in Vallespir (eastern French Pyrenees region)

This study was conducted around Céret (Pyrénées-Orientales, mean elevation 200 m) to test the statistical reliability of 12 stations devoted to sampling the Leishmania infantum vectors Phlebotomus ariasi and P. perniciosus in the South of France. Each station included a retaining wall and the surrounding phytoecological environment (total area: 2,000 m2). The wall had rectangular drainage cavities (weep holes) in which flight interception traps (sticky paper) were inserted and stretched every 10 days from May to October. For both vector species, the statistical analysis of 10-day and annual frequencies led to the following conclusions: (1) P. ariasi densities were significantly higher than P. perniciosus densities, (2) densities per species were significantly different at the 12 stations : none of them could be considered as representative of local vector densities, which depend on the wall structure (exposure, shade, vertebrate hosts), (3) the 10-day variation trends were not significantly different between stations, indicating that these variations are not determined by the station structure but rather by a common external factor (likely meteorological) and (4) the phytoecological features at the stations were not correlated with the sandfly densities. Most of the observations obtained with P. ariasi and P. perniciosus are also relevant for the non-vectorial species S. minuta. In conclusion, future research on the dynamics of leishmaniasis outbreaks relative to climate change and agricultural-silvicultural modifications should be very cautiously carried out, while focusing especially on the vector sampling quality and the use of phytoecological maps as vector density indicators.


Introduction
For several decades, the epidemiology of leishmaniasis has benefited from ecology-based scientific concepts and methods. The aim of this approach was to investigate epidemiological cycles in terms of ''parasitic systems''. The ecoepidemiological method developed on this occasion [11,23] enabled identification of key factors that govern the structure and function of leishmaniasis outbreaks. The disciplines involved in this approach include: 1°mesology (biotopes, biogeography, bioclimatology, predators and pathogens), 2°taxonomy, genetics and evolution, 3°trophic and sexual behaviours of sandflies (hematophagy, trophogonic cycle, ''eurystenogamy'', fertility, fecundity and diapause). 4°morphophysiological modifications in parasites during the intravectorial cycle (multiplication, fusion, parietal attachment, metacyclogenesis), 5°Leishmania virulence in vertebrate reservoirs (receptivity, inoculation chancre, visceralization and immunity), 6°the last step of the approach, rational control, involves several techniques (physical, chemical and biological) targeting the cycle overall, i.e. the parasite, vectors and reservoirs.
This ecoepidemiological approach was applied in the Mediterranean region, which gave rise to the vector pre-eminence concept, i.e. the sandfly vector is the main factor responsible for the structure and dynamics of leishmaniasis outbreaks [23,28].
Global warming was recently taken into account with respect to the emergence and expansion of these outbreaks, thus giving new impetus to this type of research [5,14,16,26]. However, to clearly confirm the involvement of climatic factors, project leaders have stressed the need to very precisely determine the conventional zero point. In Languedoc-Roussillon, where mean annual temperatures increased by around 2°C between 1946 and 2004 (Figure 1), these recommendations prompted us to reassess the statistical quality of the trapping initiatives conducted in 1981 in the eastern Pyrenees region where both Phlebotomus ariasi and P. perniciosus, i.e. sandfly vectors of Leishmania infantum, are found. The aim of the present study was to detect potential sampling bias, which could call into question the conclusions of certain previous studies [22].
In the study area (mean elevation 200 m), 12 sampling stations located 2-3 km apart were selected using a purposive sampling method [12]. All stations except No. 1, 8 and 10 were located in wooded rural or periurban areas. Each had a retaining wall with drainage cavities (weep holes). Sandfly samples were obtained using flight interception traps (20 · 20 cm sticky paper that were stretched vertically in the weep holes) ( Figure 3). From 3 June to 12 November 1981, the 12 stations were sampled on a 10-day basis from March to November (total number of traps: 4,263; total area: 341.04 m 2 ; mean number of traps per station: 355.25; mean area per station: 28.42 m 2 ; mean number of traps per 10-day period: 284). Sampled sandflies were placed in a 90°alcohol solution and identified. For each species, imago densities were calculated on the basis of the number of sandflies (P# + $) counted per m 2 of sticky paper [2,23,27,32]. The statistical analysis was focused on P. ariasi and P. perniciosus, the only confirmed vectors sampled at the site, using Kolmogorof-Smirnov, Wilcoxon and Friedman, Cochran and McNemar tests (SYSTAT Ò 9 software) [33,35].

Corine biotope habitats
A 2,000 m 2 area was delineated around each sampling wall to integrate the different natural or manmade habitats that could have an impact on sandfly abundance. These habitats (including walls and access roads) were classified according to Corine Biotope codes [1,36]. Most of them corresponded to synsystematic units (phytosociological classification of J. Braun-Blanquet). For each station, the cover of each habitat was expressed as a percentage occupation within the delineated areas. The weep-hole retaining walls and access roads accounted for 18% of the overall area.
Habitat/station/sandfly relationships were described by normalized principal component analysis (PCA) [  order to determine potential links between vegetation complexes and vector productivity, total sandfly densities were projected on the first PCA axis which pooled all of the main floristic characteristics in the records (ADE Ò 4 software).

Results and Discussion
1. At the end of the survey, 8,280 sandflies (P) were caught in a total of 4,263 traps (Table 2). When calculated in terms of m 2 of sticky paper (both sides), the annual cumulative frequency of P. ariasi, P. perniciosus and Sergentomyia minuta was 24.27 P/m 2 . S. minuta was more abundant (5,600 specimens: 16.42 P/m 2 ), followed by P. ariasi (2,297 specimens: 6.73 P/m 2 ), and P. perniciosus (383 specimens: 1.12 P/m 2 ). Stations with the highest abundance of P. ariasi (>8 P/m 2 , No. 3, 4, 11, 12) were south-facing. For S. minuta, stations 4 and 11 also had the highest abundances (44.83 P/m 2 and 257.37 P/m 2 , respectively). 2. The cumulative monthly total for the 12 stations was used to plot annual activity curves for both vectors. At Vallespir, P. ariasi thus had a bimodal distribution pattern, with the second peak in early autumn, whereas this species had a unimodal distribution in the Cévennes region ( Figure 4). This pattern was likely due to bioclimatic differences, as indicated by the presence of cork oak, since the Pyrénées-Orientales stations had much milder Fruit orchards 20 Robinia pseudacacia 10 Quercus suber Vegetated siliceous cliffs Native pine plantations

41.714
Quercus pubescens forest climatic conditions than the southern Cévennes. These differences could be explained by the fact that P. ariasi thrives in humid-subhumid climatic conditions [23]. A similar phenomenon was observed with P. perniciosus, which is the most thermophilic sandfly species. At Vallespir, as in the Cévennes, this species had a unimodal distribution, while it has a bimodal pattern in the vicinity of Tunis, where semiarid climatic conditions prevail ( Figure 5) [3]. However, there could have been a temporal shift in the maximum activity peaks between stations. At station No. 12, the peak for P. ariasi was thus around 20 June, whereas it occurred around 20 July at station No. 4 ( Figure 6). Excluding the fact that the corresponding walls had identical exposure, this shift could have been due to other factors, such as the extent of solar radiation or the presence of vertebrate hosts. 3. P. ariasi was clearly the most abundant of the two vectors present at the site: for 4,263 sticky paper traps (surface area 341.04 m 2 ), a total of 2,297 specimens (6.73 P/m 2 ) were sampled, as compared to only 383 for P. perniciosus      respectively, whereas the frequencies at the remaining stations ranged from 0.08 P/m 2 (No. 8) and 1.36 P/m 2 (No. 2) ( Table 2). 4. The statistical analyses confirmed these findings. The comparison of vector densities revealed significant differences between the 12 stations, i.e. each had a specific sandfly abundance pattern with respect to both P. ariasi ( Figure 7) and P. perniciosus. 5. Moreover, although the sandfly densities were specific to each station, this was not the case for their variations. Between samplings (10-day sampling periods), the variation trends were not significantly different: the curves were parallel and in the same direction for the 12 stations (Figure 8). 6. Projection of the vector abundances on the first PCA axis did not reveal any relationship between the sampled sandfly abundance of the trapping walls and their phytological environment, expressed in terms of the cover of the different Corine Biotope habitats (Figures 9, 10).

Conclusion and prospects
The present results led to the following conclusions: -At the study site, the per-station P. ariasi abundances were significantly higher than those of P. perniciosus. However, this species was not as scarce as it was in the southern    and Sergentomyia minuta (c) densities at sampling stations on the normalized PCA F1 axis. Among the readings grouped at the origin on the F1 axis, there are two habitat groups that are opposed in Figure 9 (circle of correlations). There is therefore no gradual relationship between the vector abundance and the environmental (habitat) trends. Corine Biotope habitats thus do not seem to be suitable for drawing up a vector sampling plan.
Cévennes region. This difference could be explained by the bioclimatic conditions, i.e. humid Mediterranean climate in the Cévennes and subhumid at Vallespir. These climatic differences could also explain the P. ariasi variation patterns, i.e. monophasic in the Cévennes and diphasic at Vallespir. -The significant between-station differences in sampled sandfly abundance were due to the specific environmental conditions at each station (geopedology, orientation and shadows cast on the trapping wall, animal occupation of the weep holes, nearby livestock) [18,28]. Note that the weep holes served as roosting sites, not as breeding sites -During many surveys, our attempts to isolate larvae or nymphs in weep-hole soil were always fruitless. However, the constant high abundance sampled in some especially attractive walls could be explained by the high movement capacity of imagos [30]. -The parallelism in the 10-day variations could have been due to a single factor, i.e. likely weather related (storm, heat wave, etc.), taking place accidentally and sporadically at all stations. -The multivariate analysis did not reveal any association between sandfly frequencies and the phytoecological environment of the sampling stations (Figures 9 and 10). As we noted, these associations occurred with certain physical or biotic properties specific to each wall. More generally, the cartographic typology of the area according to the Corine Biotope system or its derivatives (Corine Land Cover, EUNIS [15,17]) cannot be recommended as a vector abundance indicator, especially in the definition of zero points. By their phytosociological nature, the corresponding maps should be used for studies on landscape modifications resulting from climate change, fire, flooding, cropping or non-native plant invasions. However, these maps have a real value in applied research, e.g. in forest management (see French Forestry Office). Finally, they can still be used as density indicators to characterize certain zoological groups on the condition that they are closely tailored to specific botanical taxa or syntaxa. This category includes strict pests, specific pollinators, specialized herbivores and animals with a narrow ecological niche. In other cases, caution is necessary [21], especially since Corine Biotope maps are not the only way to express vegetation-indicator relationships. This was the case in studies carried out in Morocco to identify the preferred bioclimatic conditions of sandflies. In this country, the only phytoclimatic map, drawn up on the basis of vegetation layers (from humid to arid), made it possible to attribute a bioclimatic value to each inventoried sandfly species, and especially to specify the current or future geographical distribution (global warming) of Leishmania spp. vectors [23]. Otherwise, in metropolitan France, a very successful phytoecological map was drawn up of biotopes for pre-adults of Aedes spp. (Diptera-Culicidae). The detection of eggs in the litter of halophyte plants revealed a close relationship between some plant species (Salicornia spp., Scirpus maritimus) and certain Aedinae species (Aedes caspius, A. detritus). This relationship was dependent on the egg-laying behaviour of gravid females, which were found to often only lay eggs on one or two species of plants growing along narrow strips in lagoon environments. The focus has thus readily shifted from ''egg laying sites'' to the phytoecological mapping of ''breeding sites'', which has turned out to be a remarkable operational tool that has been very successful in the control of pest mosquitoes in Mediterranean coastal regions [10,24,25]. -Finally, the current results should be considered as a starting point for further research. At Vallespir, after more than 30 years, the same protocol should be applied at the same sampling stations. This type of operation was recently undertaken in the lower Cévennes region [19,31], which generated promising results. At Vallespir, if the sandfly density modifications and geographical distributions noted here are confirmed, it would be of interest to supplement this work by more in-depth taxonomic and genetic analyses, both in terms of specific taxa (P. sergenti is already present in Pyrénées-Orientales region [29]), and within the same species (modification in population-based polymorphism, selection of thermophilic variants, drift [30], etc.). Future epidemiological research focused on the impact of climate change or of agronomic-silvicultural modifications should be very cautiously carried out, especially with respect to vector sampling and the use of phytoclimatic maps as vector density indicators.