Achieving land degradation neutrality (LDN) has been proposed as a way to stem the loss of land resources globally. To date, LDN operationalization at the country level has remained a challenge both from a policy and science perspective. Using an approach incorporating cloud-based geospatial computing with machine learning, national level datasets of land cover, land productivity dynamics, and soil organic carbon stocks were developed. Using the example of Botswana, LDN and proportion of degraded land were assessed. Between 2000 and 2015, grassland lost approximately 17% of its original extent, the highest level of loss for any land category; land productivity decline was highest in artificial surface areas (11%), whereas 36% of croplands show early signs of decline. With the use of national metrics (NM), degraded areas were found to be 32.6% compared to 51.4% of the total land area when global default datasets (DD) were used. Estimates of degraded land computed with NM and DD were validated in Palapye, an agro-pastoral region in eastern Botswana, where Composite Land Degradation Index (CLDI) field-based data exists. Comparing land degradation (LD) in the three datasets (NM, DD, and CLDI), NM estimates were closest to the field data. The extra efforts put into developing national level data for LD assessment in this study is, thus, well-justified. Beyond demonstrating remote sensing viability for LD assessment, the study developed procedures for generating and validating national level datasets. Using these procedures, LD monitoring will be enhanced in Botswana and elsewhere since these remote sensing datasets can be updated using freely available satellite datasets.
All Science Journal Classification (ASJC) codes
- Environmental Chemistry
- General Environmental Science
- Soil Science