Implementation of a multi-layered land degradation monitoring and forecasting framework toward integrated countermeasure planning in drylands

Various earth observation (EO) data and related spatial technologies are employed to monitor land degradation (LD) and its impacts on the drylands. However, a comprehensive framework that integrates multiple data sources and metrics to generate a consistent output still needs to be explored. Furthermore, the lack of consistency and the limited reliabilities of existing approaches make it challenging to identify proper countermeasures (avoid, reduce, reverse LD) and support decisions to combat LD in line with the Land Degradation Neutrality (LDN) Framework. The "Multi-Layered Land Degradation Tracer (ML-LDT)" is a framework that addresses the above research gaps. It is based on a comprehensive framework that integrates data extraction, LD monitoring, and forecasting capacities and consists of three functional layers, that is, a base data processor, monitoring/modeling components, and a forecasting layer employing machine learning (ML) algorithms. ML-LDT has a state-of-art capacity to model wind erosion together with sand and dust storms (SDS), making it particularly suited to the arid lands most prone to these processes. It is implemented over cloud processors with open-source components based on industry standards. Therefore, the routines are easily adaptable to user requirements in LDN planning. We evaluated the ML-LDT framework in the Indian Thar Desert and Inner Mongolian drylands. The results showed the overall robustness of the procedures and that ML-based models provide good insights to forecast future LD/SDS patterns that can be used to simulate LDN scenarios and targets. We validated the forecasting results against LD maps and remote sensing observations.