Antananarivo - 2022 Land cover map

We publish two land cover maps of the city of Antananarivo produced with data acquired in 2022 using a methodology combining machine learning and object-based image analysis (OBIA). This work follows on from work that resulted in a map of the same area in 2017.

The maps are produced by processing satellite images using the Moringa processing chain developed in our lab. We use a Pleiades very high spatial resolution (VHSR) image, a time series of Sentinel-2 images, a digital terrain model (DTM) and a reference database.
The Pleiades image is used to generate a layer of objects using a segmentation algorithm. Each object is then classified using the variables from the THRS image, the time series and the DTM information. The hierarchical nomenclature used consists of four levels with a number of classes ranging from 4 to 19.

We only publish here the most detailed map (level 4) which contains, however, in the attribute table, the information of the other levels.
The overall accuracy of the maps ranges from 93% to 84%. Such land cover products are very rare in Madagascar, so we decided to make them open access so that they can be used by land managers and researchers.
Warning, since December 2, 2022 we publish a new version to limit the effects related to flooding (the wetland class was overestimated on the previous version).

• The map entitled "version 1" is produced with the SRTM digital surface model (DSM) with a spatial resolution of 30m.
• The map entitled "version 2" is produced with a DTM and a LiDAR DSM with a spatial resolution of 1m (over a reduced area)

A technical report describing the method implemented and the statistics obtained at the validation step is available here : https://agritrop.cirad.fr/602680 (2022-10-20)