Key Words:
Sediment yield estimation, spectral unmixing, leaf area index

Abstract:
Vegetation and soil properties and their associated changes through time and space affect the various stages of erosional processes. This paper discusses the application of remote sensing techniques in the retrieval of vegetation and soil parameters necessary for the distributed soil loss modeling in small agricultural catchments. To account for the compositional nature of the ground surface as depicted on remotely-sensed data, a linear spectral mixture modeling (LSMM) approach is used to parameterize vegetation and soil optical properties. Results of these parameter estimates were coupled to a DEM-based distributed hydrologic rainfall-runoff model to simulate overland flow and sediment yield for a given rainfall event. Field observations were undertaken to gather spectral and physical measurements of vegetation and to obtain data for soil hydraulic properties. Results of the sediment yield model indicate strong relationship between vegetation abundance and erosion by soil detachment. A general agreement between the simulated and measured sediment discharges is also observed. The method provides a physical basis for incorporating the spatial and temporal variability of various vegetation and soil condition to dynamic processes such as soil erosion and may be applicable to monitor other non-point source pollutants from agricultural watersheds.

1. Introduction
The risk of soil erosion by water, varies as a function of many factors, but the degree of protection provided by vegetation is one of the most important. Usually erosion rates are computed by means of empirical methods or if treated in terms of physically-based models, are highly idealized, such that effects of vegetation presence become trivialized, mainly due to its high spatial and temporal variation. Leaf area index (LAI) and percentage vegetation cover (PVC), two parameters routinely derived from remote sensing, can provide the necessary details to incorporate vegetation effects to soil losses.

Spectral mixture models are among the popular methods to resolve the optical components of surfaces with diverse land cover types especially useful when the analysis is constrained by the spatial resolution of the available dataset (van Leeuwen et. al, 1997). A desirable feature of mixture models is that they are able to estimate the fractional abundance of vegetation and soils simultaneously, appropriate for purposes which require both information at the same instant such as in the case of erosion analysis. LSMM is most appropriate for crop plants since leaves are relatively dominant, soil reflectance is uniform, topography is minimal and most crowns can be modeled with simple geometric shapes, the favorable conditions to achieve linear spectral mixing.

This study was conducted to investigate the effect of applying vegetation indexes, particularly LAI and PVC, obtained through spectral mixture analysis of remotely-sensed data, to quantify soil losses, particularly sediment discharge from a predominantly agricultural area during rainfall events with consideration to development stage of vegetation.