Large-area mapping of active cropland and short-term fallows in smallholder landscapes using PlanetScope data

• DOI: 10.1016/j.jag.2022.102937
• Publication Date: 2022-07-30T11:20:36Z
• Authors: Philippe Rufin; Adia Bey; Michelle Picoli; Patrick Meyfroidt
• Keywords: Land Cover

Cropland mapping in smallholder landscapes is challenged by complex and fragmented landscapes, labor-intensive unmechanized land management causing high within-field variability, rapid dynamics shifting cultivation systems, substantial proportions of short-term fallows. To overcome these challenges, we here present a large-area framework to identify active cropland fallows for the 2020/2021 growing season at 4.77 m spatial resolution. Our study focuses on Northern Mozambique, an area comprising 381,698 km2. The approach based Google Earth Engine time series PlanetScope mosaics made openly available through Norwaýs International Climate Forest Initiative (NICFI) data program. We conducted multi-temporal coregistration using seasonal Sentinel-2 base images derived consistent gap-free metrics classify An iterative learning Random class probabilities was used training rare classes uncertain regions. map accurate (area-adjusted overall accuracy 88.6% ± 1.5%), with main error type being commission cropland. Error-adjusted estimates extent (61,799.5 km2 4,252.5 km2) revealed that existing global regional cover products tend under-, or over-estimate extent, respectively. Short-term occupied 28.9% our reference sample (13% mapped cropland), consolidated agricultural regions showing highest shares relies cloud-based processing Engine, which minimizes financial constraints maximizes replicability methods. All code maps were further use.