The Effect of Forest Mask Quality in the Wall-to-Wall Estimation of Growing Stock Volume

Information about forest cover and its characteristics are essential in national international inventories, monitoring programs, reporting activities. Two of the most common variables needed to support sustainable management practices area growing stock volume (GSV m3 ha?1). Nowadays, inventories (NFI) complemented by wall-to-wall maps which rely on models auxiliary data. The spatially explicit prediction GSV is useful for small-scale estimation aggregating individual pixel predictions a model-assisted framework. Spatial knowledge land an prerequisite. This information contained mask (FM). number FMs increasing exponentially thanks wide availability free data, creating doubts best-suited specific purposes such as estimation. We compared five available entire Italy examine their effects clarify product this purpose. considered were mosaic local produced Italian regional authorities; FM from Copernicus Land Monitoring System; JAXA global FM; hybrid Schepaschencko et al., estimated Corine Cover 2006. used out non-forest pixels map constructed using inventory remotely sensed accuracies first evaluated against independent dataset 1,202,818 NFI plots four accuracy metrics. For each masked maps, pixel-level calculate regional-level estimates. with respect coefficient correlation (?) between estimates they (both terms mean total within boundaries) official At levels, based closest ? = 0.986 0.972, GSV, respectively. found negative differences estimate, demonstrating that choice plays important role when estimator.