An optimized Enhanced Vegetation Index for Sparse Tree Cover Mapping across a Mountainous Region

In mountainous areas with sparse woody vegetation, tree cover is among the most crucial indicators for landscape conservation and ecosystem stability. It is thus essential to be continuously monitored, which is yet seriously hurdled by harsh topography and inaccessibility. Field-based methods are time consuming and costly and are thus particularly prohibitive across large areas. Multispectral remote sensing data sources have been historically used via deriving various vegetation indices to enhance the visual interpretability as well as to deliver practical products from vegetation cover. Whereas indices like NDVI or even more empirically-parametrized indices provide convenient and easy-to-use proxies for large area cover mapping, their robustness to issues raised by shadow, atmospheric and illumination conditions and thus their interpretability for regional cover mapping in sparse vegetation cover are highly questioned. Here, we introduce an efficient method to optimize the enhanced vegetation index (EVI) by particle swarm optimization (PSO). We compared the efficiency of the optimized EVI against the NDVI and the traditional EVI across a portion of remote, sparsely vegetated and mountainous Zagros region in western Iran. The results based on Sentinel 2 sensor data show clear advantages of locallyoptimized EVI over the conventional indices in both visual interpretability and explained variance.