top of page
Vigie-Chiro
Method
How to read the maps
Results
Predictive distribution maps
Method
Predictive distribution maps are generated thanks to random forest models, that aim at predicting bat activity counts (the response variable) using different descriptive variables such as habitat, bioclimatic conditions, ground elevation, proximity to roads, wind turbines or artificial light at night.
Step 1: Species activity counts are predicted according to
a forest of regression trees based on descriptive variables:
Species activity counts
and more...
Road network
Bioclimatic conditions
Habitat
(Corine Land Cover)
Wind farms
1/23
Step 2: Thanks to the forest of decision trees built in step 1, predictions are madeon a systematic grid of 300,000 points covering France.
Step 3 : For each species groups that are acoustically close, mistnetting data (CACCHI program) are used according to the same process in steps 1 and 2 to create predictive maps of the proportion of the species among its acoustic group (Plecotus, Myotis myotis/blythii, etc). Steps 1 and 2 are done separately for each species.
The results depend on the sampling effort and the performance of the identification software Tadarida. We are constantly trying to reduce the identification errors, using the method published by Barré et al. (2019).
How to read the maps
The method used to create the maps can be summarised in this way: for example, knowing that Daubenton's bat was contacted much more in wetlands than in other habitats, the probability to contact this species in wetlands is relatively high, and the model will thus predict a higher density of Daubenton's bat in wetlands than in other habitats. However, Daubenton's bat being rarely contacted in areas with artificial light at night, the model will predict a lower density of the species in wetlands lighted at night.
The colour scale is adjusted for each species and represents the number of bat passes per night. The dark blue zones are supposed to be unfavourable for the species while yellow zones are supposed to be favourable.
Results are presented after step 2 if the species is not in an ambiguous acoustic group, or by multiplying the results from step 2 with the results from step 3 in the other case.
The maps below are predictions for the month of August.
Acoustic data (Vigie-Chiro) only
Acoustic data (Vigie-Chiro) + Mistnetting data (CACCHI)
Maps for spatial planning
Spatial planning helps prioritise biodiversity stakes to designates areas of conservation priority and areas of vulnerability to human activities. To know more about this, visit the dedicated page here:
Spatia planning
Recommanded citation :
Bas Y, Kerbiriou C, Roemer C & Julien JF (2022, March) Maps of predicted bat distribution. Muséum national d'Histoire naturelle. Retrieved from https://croemer3.wixsite.com/teamchiro/maps-predicted-activity
bottom of page