Tree canopy cover estimation by means of remotely sensed data for large geographical areas: overview, available data, and proposal

Abstract

Climate change and global warming requires a strong boost to sustainable growth strategies. In particular, urban green management and planning is becoming a crucial and at the same time critical aspect. Therefore, urban green requires being accurately mapped, quantified and monitored over time. In this study we propose a cost-effective but reliable approach for the automatic classification and quantification of the tree canopy cover over extended geographical areas. The classification can also be used for estimating the number of trees, based on land use land cover (LULC) and the corresponding planting layout. The case study application is the Metropolitan City of Milan. Data used for classifying the tree canopy are based on high-resolution satellite imagery provided by the PlanetScope constellation. Based on the latter information, the work relies on the use of radiometric Vegetation Indices (VIs) to quantify the tree canopy. However, because the use of VIs can cause mixing of different types of vegetation, such as tree and grass, we used a stack of multi-temporal data from PlanetScope to retrieve per-pixel statistics for Red band and Normalized Difference Vegetation Index (NDVI). The hypothesis here is that during spring-summer season tree canopy provides less variability than grass and/or agricultural fields. The approach provides an improved vegetation index capable of separating potential canopy-tree from other vegetation types. The result of the accuracy assessment shows an overall accuracy of 78.33% and 71.5% for the whole Metropolitan City of Milan and the City of Milan respectively.