Post-classification approaches to estimating change in forest area using remotely sensed auxiliary data

摘要: Multiple remote sensing-based approaches to estimating gross afforestation, deforestation, and net deforestation are possible. However, many of these have severe data requirements in the form long time series remotely sensed and/or large numbers observations land cover change train classifiers assess accuracy classifications. In particular, when rates small equal probability sampling is used, may be scarce. For situations, post-classification only viable alternative. The study focused on model-assisted model-based inference for estimation using Landsat imagery as auxiliary data. Emphasis was placed variances support construction statistical confidence intervals estimates. Both analytical bootstrap variance were used. a area Minnesota, USA, estimates not statistically significantly different from zero.