Soil erosion risk modeling within upland landscapes in Vietnam using remotely sensed data and the RUSLE model

Huong Tra is one of the mountainous districts of Thua Thien Hue province, Vietnam, located between two large rivers which ultimately flow into the Tam Giang Lagoon, the largest Lagoon in Southeast Asia with an area of over 220 km2. Water erosion creates negative impacts on agricultural production, aquaculture, infrastructure, and water quality throughout the region. The Revised Universal Soil Loss Equation (RUSLE) is a well tested model for erosion prediction at the field scale. The integration of the RUSLE and remotely sensed data provide a useful tool for regional erosion risk assessment. The aim of this research is to assess soil erosion risks for upland landscapes using the RUSLE model and remotely sensed data within a GIS and to propose several land use scenarios which are able to reduce soil losses. Each erosion factor of the RUSLE model was computed. The Tropical Rainfall Measurement Mission (TRMM) data, provided by point scale, was examined to calculate rainfall and runoff erosivity (R) factors. The monthly TRMM rainfall data and measured rainfall were significantly correlated with a regression correlation coefficient of approximately 0.9. Annual R values range from 960 to 1033 MJ mm ha -1 hr-1, whereas the R values of the wet season are double that of the dry season. Additionally, the TRMM data describe the spatial variation of rainfall in the region better than measured rainfall at meteorological stations. The Digital Elevation Model (DEM) of Shuttle Radar Topographic Mission (STRM), NASA was validated with measured elevation values in official topographic maps. Using this DEM, the topographic LS factors were computed from slope and flow accumulation algorithms. Flow accumulation illustrates the impact of upslope contributing areas to sediment detachment and transportation, so it better reflects the effects of concentrated flow on increased erosion on sloping areas. Landsat ETM+ images were used to calculate C factors by using the Linear Spectral Mixture Analysis (LSMA) model which determines the proportion of each land use type within each pixel. The comparison between bare soil and erosion resistant covers (vegetation cover and non-photosynthetic materials) resulted in C factors for pixels. Soil erodibility factors were computed from readily available soil maps by using a soil erodibilty nomograph. Lastly, soil loss rates were computed for pixels, and erosion risk classes were identified by reclassification. Erosion risks on rice cultivation land were moderate, whereas they were quite high on dry crop, protection forest, and unused land types. The most severe erosion rates occurred on production forest land which had poor vegetation cover. The RUSLE model was also used for developing forest land planning in which the balance between low erosion land area (protection forest) and severe erosion land area (production forest) was taken into account. Erosion risk and slope maps are useful to identify and delineate the spatial allocation of protection forest. In summary, the integration of the RUSLE model and remotely sensed data provides an effective tool for assessing soil erosion risks and selecting appropriate land use scenarios which can reduce soil losses in a large scale.