Rapid prediction of tropical soil degradation using diffuse reflectance spectroscopy: Method verification in the Saiwa River basin, western Kenya

Kenya is a poor country with an exploding population growth outpacing food production. Millions of Kenyans regularly experience starvation and/or malnutrition. In the highlands of western Kenya, land conversion from indigenous forests to agriculture results in reduced soil quality through declining soil fertility and increased soil erosion, two principal components of soil degradation. To understand and effectively address the impacts of anthropogenic soil degradation, reliable spatial and temporal soil quality assessment is needed.;This research used Near Infrared (NIR) Diffuse Reflectance Spectroscopy (DRS), Classification and Regression Tree (CART) analysis, stable and radioactive isotopes, satellite imagery interpretation, and Geographic Information System (GIS) to accurately categorize the associations between land cover change, soil fertility and soil erosion.;NIR DRS can quickly index soil fertility and soil erosion. General soil types and conditions were distinguished by mean spectrum grouping and quantification of albedo differences. Calibration and cross validation statistics show that NIR DRS accurately analyzes and predicts soil nutrient and physical parameters. Using spectral properties, soils were successfully differentiated into a three class soil fertility index (SFI) and a two class erosion index (SEI). CART analysis shows these two indices have a high predictive performance with low misclassification results and can be replicated. Significant differences in soil parameter values illustrate that the methods are effective for watershed scale soil degradation assessment and monitoring programs. Analyses of carbon isotope ratios, SOC, and satellite imagery demonstrate the association of deforestation and soil quality change. Historic carbon sources have more influence on current SOC concentrations than recent carbon sources. Grasslands converted to agriculture have the lowest SOC concentration, unconverted forests contain the most, and mixed systems have more SOC than unmixed agricultural systems. Soil erosion and topographic position significantly affect SOC content.;Together, these analytical methods form a recommended methodology for soil degradation evaluation in tropical, remote areas of the world. Accurate landscape scale, soil degradation models are valuable tools to quickly identify target areas in which to focus limited intervention funds and efforts. The results can bolster agricultural land management knowledge and help improve land management policy development intended to increase food production and security.