Geomorphological variables to predict the spatial distribution of plant species in agricultural ditches
Résumé
Background : In Mediterranean cultivated lands, exogenous variables controlling spatial distribution
of non-cropped plant species at the landscape scale are habitat variables, seed-sources proximity,
and lastly human practices. In order to manage ecosystem services provided by head catchments
drainage networks (ditch networks), we need to predict spatial distribution of plant species living in
them. Geomorphological variables have shown to be important habitat variables in other
ecosystems (riverine, wetlands,...), but whether such variables can be used to predict plant
distribution in ditch drainage networks is unknown. An arising question is then : which is the
importance of geomorphological landforms in driving plant species distribution at the ditch network
scale ?
Methodology : We collected presence and absence data for 10 plant species (herbaceous) on a 35-
km drainage ditches network in a Mediterranean agricultural landscape. We simulated their spatial
distribution using geomorphological and land-use derivative variables (representative of seed-
sources proximity) with Maxent and GLM model. Each plant species was considered separately and
each model was validated using k-fold cross-validation (k=4). We compared mean AUC (Area
Under the Curve) values obtained for each model and relative importance of explanatory variables.
We analysed compiled false positive and false negative residual maps.
Results : Mean AUC values for Maxent model ranged from 0.92 for Mentha aquatica to 0.67 for
Rubus fruticosus. Mean AUC values for GLM ranged from ranged from 0.86 for Lythrum salicaria
to 0.61 for Rubus fruticosus and Sorghum halepense. Both geomorphological and land-use
derivative variables were important to explain Asparagus Acutifolius, Equisetum arvense, Mentha
aquatica, Mentha suaveolens, Rumex crispus, Scirpus holoschoenus distribution and only
geomorphological variables well explained Lythrum salicaria distribution. The models failed to
explain accurately Sorghum halepense, Rubus fruticosus and Elytrigia juncea distribution.
Compiled residual maps showed a false positive predictions pattern at the South of the catchment.
Conclusion : We highlighted the importance of considering geomorphological variables to explain
spatial distribution of plant species in Mediterranean head catchments drainage ditches. Results
suggested that part of the remaining non-explained variability for some species was induced by
human practices. Agro-ecological measures aiming at managing ecosystem services provided by
ditch networks have to consider plant specific predictors at the local and landscape scale in order to
be successful. They further have to investigate human practices effects on ditch plant distribution.