Deriving forest cover rates from map sources: A contribution to official statistics and environmental reporting

Abstract

While forest statistics are currently released at NUTS-1 (macro-regions) or NUTS-2 (administrative regions) levels, advancements in remote sensing technology may improve their accuracy at smaller spatial units. To explore the potential contribution of remote sensing in downscaling forest cover rates to finer administrative levels, we run a quantitative analysis of the statistical relationship between selected indicators of forest cover derived from 16 independent (wall-to-wall) map-sources and 4 probabilistic sampling surveys (land cover/forest inventories) with the aim at verifying the consistency of their statistical distribution at the regional scale in Italy. The empirical results indicate that, given current technological capabilities and the standard land cover classifications used in each survey, only a limited number of map-based (local-scale) sources align with official sampling sources provided at large (regional) scale. Specifically, indicators from EFI, FAO, FROM, JRCLU and the JRC20 dataset exhibit compatibility with EFI and JRCLU, standing out as particularly reliable for spatial downscaling of sample-based forest indicators, being not affected by the Modifiable Areal Unit Problem. Forest cover rates derived from these sources can serve as valuable ancillary variables in spatial downscaling procedures of official (sample-based) forest estimates provided at a large scale, thus representing a reliable source of information for the routine production of official statistics at the level of small-area administrative units.