Landsat-based approaches for mapping of land degradation prevalence and soil functional properties in Ethiopia

• Spatially explicit sampling for landscape-scale studies of soil and ecosystem health
• Novel approaches combining soil spectroscopy, remote sensing and ensemble learning
• Maps of soil function and land degradation risk to target land management practices
• Case study of four diverse sites in Ethiopia

Agriculture is the basis of the Ethiopian economy, accounting for the majority of its employment and export earnings. Land degradation is, however, widespread and improved targeting of land management interventions is needed, taking into account the variability of soil properties that affect agricultural productivity and land degradation risk across landscapes. In the current study we demonstrate the utility of Landsat ETM + imagery for landscape-level assessments of land degradation risk and soil condition through a combination of systematic field methodologies, infrared (IR) spectroscopy and ensemble modeling techniques. The approaches presented allow for the development of maps at spatial scales that are appropriate for making spatially explicit management recommendations. Field data and soil samples collected from 38 sites, each 100 km2, were used to develop predictive models that were applied as part of a case study to an independent dataset from four sites in Ethiopia. The predictions based on Landsat reflectance were robust, with R-squared values of 0.86 for pH and 0.79 for soil organic carbon (SOC), and were used to create predicted surfaces (maps) for these soil properties. Further, models were developed for the mapping of the occurrence of soil erosion and root depth restrictions within 50 cm of the soil surface (RDR50), with an accuracy of about 80% for both variables. The maps generated from these models were used to assess the spatial distribution of soil pH and SOC, which are important indicators of soil condition, and land degradation risk factors in order to target relevant management options.