C-Factor mapping using Remote Sensing and GIS : a case study of Lom Sak / Lom Kao, Thailand

Abstract

Attempts to study land degradation processes and the necessity of degradation prediction have resulted in the creation of erosion models. The cover management factor is one of the most important parameters of the Universal Soil Loss Equation (USLE) since it measures the combined effect of all interrelated cover and management variables and it is the factor which is most easily changed by men. One of the major problems with modelling is how to obtain the necessary information. Data requirements are large and include information on vegetation cover and soil properties, which can only be measured directly in the field or can be derived from other kinds of data such as supplied by remote sensing. However, field surveys are labour intensive and expensive and yield normally only information of one geographical location. The presented study tries to investigate the use of digital satellite maps for obtaining quantitative information parameters for the use of erosion modelling. The study area is located in the Lom Sak and Lom Kau districts in the north of Thailand. Different methods to C-factor mapping using remote sensing have been attempted of which three were applied in this study. The first approach considers C-factor mapping as a special type of land cover mapping. A land cover classification map is produced first and corresponding C-factors to the identified land cover are taken from literature. The final land cover map for the study area was obtained by a combination of a visual interpretation and a supervised classification using the maximum likelihood algorithm. The second and third applied methods are following a completely different approach, where spectral indices are used in order to obtain direct information on C-factors from digital images. For the Lom Sak case the NDVI and the transformation index were applied. Both methods are based on two reference samples of pure vegetation (forest) and bare soil. Based on the assumption that there is a linear relation between the indices and the C-factor linear models were created and used to assess the C-factors for each pixel of the image. The contrasting point between these two methods is that in order to distinguish between the reflectance parameters of pure vegetation and bare soil different TM band combinations are used. The correlation between the two indices based C-factor maps was large (0.97). Hence, the different methods of combining spectral information to optimise the contrast between green vegetation from bare soil yielded in very similar results. The comparison between the C-factor map produced by land cover mapping and the indices based C-factor maps showed lower correlation. However, analysing the spatial variation within the image showed that all three methods classified bare land into the same class (0.9-1.0). Conversely vegetated areas as forests, shrubs and bush land were assigned different C-factors. These results should not lead to the conclusion that one or the other approach is not useful for the extraction of vegetation parameters for erosion modelling. In order to assess each of the here applied methods detailed ground information is needed, which was not available for this study.