Abstract
AbstractGully erosion has been recognized as an important soil degradation process in rangelands of southwest Spain. In this paper, a relatively new data mining technique called Multivariate Adaptive Regression Splines (MARS) was applied to construct a model capable of predicting the location of gullies. A large database was used to support the model composed of a target variable (presence or absence of gullies) and 36 independent variables related to topography, lithology, soils, rainfall, land use and vegetation cover. The performance of the model was evaluated using the Receiver Operating Characteristic (ROC) curve for five external datasets. The model had high predictive power, with values for the area under the ROC curve of the external validation datasets varying from 0·75 to 0·98 (1·0 being perfect prediction). The most important variables explaining the spatial distribution of gullies were lithology and soil type. Finally the model was compiled and implemented into a geographical information system to obtain maps of susceptible areas for gully erosion. These maps show that approximately 7% of the study area presents favourable conditions for the development of gullies. The results demonstrate that MARS constitutes a valuable model in geomorphic research and could also be a useful tool for assessing the impacts of changing climate and land use on gully erosion. Copyright © 2009 John Wiley & Sons, Ltd.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have