Monitoring forest cover from a mix of data sources: A contribution to landscape analysis

The importance of forest statistics, and especially the estimation of woodland cover over small administrative domains, is evident in spatial planning, environmental impact assessment, and official statistics. Despite significant improvements, small-scale indicators of forest cover show a variable accuracy depending on multiple affecting factors, and the effectiveness of automatic (or manual) recognition of land cover types is only one of the possible issues at stake. The present study adopted 7904 Italian municipalities as the elementary unit with the aim of testing the spatial variability of 16 (synchronic) forest cover indicators derived from national and international data sources with different spatial coverage and resolution, with the final objective of evaluating their impact on the overall estimation of forest cover at both regional and national scale. Through a methodological framework based on analysis of correlation and similarity between the municipal samples, the spatial variability in woodland cover, as well as the overall precision of forest statistics, was estimated, identifying the information sources most suited to meet data requirements. The precision of forest cover indicators was related to the final resolution of original data sources, although additional, nonnegligible sources of spatial heterogeneity were identified, suggesting the importance of a reasonable use of multiple data sources, especially remote-sensed, when assessing forest cover rates at small-area statistical domains.