Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4574429 | Geoderma | 2009 | 8 Pages |
Human-induced soil erosion has severe economic and environmental impacts throughout the world. It is more severe in the tropics than elsewhere and results in diminished food production and security. Kenya has limited arable land and 30% of the country experiences severe to very severe human-induced soil degradation. The purpose of this research was to test visible near infrared diffuse reflectance spectroscopy (VNIR) as a tool for rapid assessment and benchmarking of soil condition and erosion severity class. The study was conducted in the Saiwa River watershed in the northern Rift Valley Province of western Kenya, a tropical highland area. Soil 137Cs concentration was measured to validate spectrally derived erosion classes and establish the background levels for different land use types. Results indicate VNIR could be used to accurately evaluate a large and diverse soil data set and predict soil erosion characteristics. Soil condition was spectrally assessed and modeled. Analysis of mean raw spectra indicated significant reflectance differences between soil erosion classes. The largest differences occurred between 1350 and 1950 nm with the largest separation occurring at 1920 nm. Classification and Regression Tree (CART) analysis indicated that the spectral model had practical predictive success (72%) with Receiver Operating Characteristic (ROC) of 0.74. The change in 137Cs concentrations supported the premise that VNIR is an effective tool for rapid screening of soil erosion condition.