Integrating Spatially Exhaustive Mapping and Sampling for National and Regional Forest Change Assessments

Timely information on forest extent and change is required to support national sustainable development goals and to monitor national land use policy implementation and enforcement. Freely available moderate spatial resolution satellite data, such as Landsat imagery, enable global forest monitoring. However, available global products are not always consistent with the national forest definitions and may not be suitable for national reporting. Moreover, national forest monitoring agencies often lack the capacity for independent satellite data processing. Our research team developed a flexible methodology for operational wall-to-wall mapping and sample-based area estimation of forest extent and change that can be adapted to match country forest definition and used for national forest resources and GHG emissions reporting. To facilitate Landsat data archive analysis we developed a unique automated system for processing imagery into a set of spatially and temporally consistent Analysis Ready Data (ARD) time-series. The national forest mapping and monitoring algorithm using ARD data and purposely designed royalty-free software dramatically enhance the capacity of national monitoring agencies to operationally map forest extent and change. Integration of optical ARD with active remote sensing products (vegetation structure metrics derived from spaceborne and airborne lidar) using machine-learning algorithms was successfully implemented to produce spatially exhaustive maps of forest stricture attributes, such as tree canopy cover and height. The unbiased estimates of forest extent and change with known uncertainty are obtained through sample analysis. Samples selected using probability-based stratified design, where national forest extent and change maps are used as stratifiers increasing precision of the estimates and reducing sample interpretation cost. The ARD time-series data serve as one of the main inputs for sample interpretation. Both national maps and sample-based area estimates are used for national reporting. A number of countries adopted or are considering implementation of the ARD-based forest monitoring workflow, including Peru, Vietnam, Bangladesh, Nepal, Madagascar, Colombia, Cambodia, Guatemala, Cameroon, the DRC, and the Republic of the Congo.