DARLING: Ditch Network Automatic Reconstruction from Linear Incomplete Ditch Geolocation - Mapping the Water Harvesting Capacities at Regional Scale for Better Droughts Mitigation

International audience ; Ditches in agricultural and peri-urban landscapes form tree-like structures, directed from upstream to downstream, providing a combination of ecosystem services ranging from landscape amenities to water harvesting in case of check-dams in ditches. However, these functions often only make sense as a network at the landscape level and not at a single ditch level. For agricultural drainage networks, the network links ditches with different functions: ditches that are water collectors to ditches that are merely water carriers. Poorly considered elements of the landscape, ditch networks now have an increasing importance, particularly in terms of the functions they can perform (ecological habitats and corridors, water purification, interface between surface and groundwater) which are ancillary to those for which they were originally created (drainage, protection against soil erosion, etc). However, it has to be said that there are still very few exhaustive maps of these elements in national, let alone global, geodatabases. This is quite surprising since these ditch networks represent an upstream extension of the anthropogenic origin of hydrographic networks made of rivers, which are themselves exhaustively and globally mapped. Mapping the tree structure of these linear elements in the landscape therefore remains a challenge. In this work, we present the application of a large-scale, spatially explicit method of reconstructing probable ditch networks: the DARLING method. This method is based on the coherence of four sources of information: a land unit vector layer, a DTM at standard resolution (10-20 m), a small sample of identified but disconnected ditches, and a target network density to be reconstructed per landscape unit. This sample of disconnected ditches may come from non-exhaustive field surveys, recent or old aerial photos, or very high spatial resolution optical satellite images. Using a sequence of logical rules, the DARLING algorithm proposes a tree structure reconstruction scheme ...