The potential research, policy and management applications of global satellite products place a high priority on rigorous, quantitative assessment of their accuracy. Such an assessment can be achieved by implementing validation methods employing design-based inference in which the independent reference data are selected via a probability sampling design. Well established methods have been developed for the validation of land cover maps at a variety of spatial scales, but these methods often consider only the spatial distribution of mapped classes, implicitly assuming that the map is static. This is not the case when mapping impermanent disturbances such as fires, forest cover loss and forest degradation. The presentation will focus on the rapidly changing state of the art in satellite product validation, with regards to tri-dimensional sampling designs where the temporal dimension is explicitly taken into consideration, and on the implications in terms of adequate sample size, and unbiased estimators of the accuracy metrics.
Data: Segunda-feira, dia 21 de novembro de 2022
Local: Lisboa
