A multi-scale approach for erosion impact assessment in the Andes

Erosion processes and their impact are driven by a multitude of biophysical factors that are poorly understood in the Andean highlands. The first attempt to assess soil erosion rates in Peru was made by Felipe-Morales et al. (1977) using runoff plots. Since then, very few erosion investigations have followed and reference to Andean soil erosion is often criticized because of the lack of quantitative data. Lack of understanding the causes and effects of erosion also hamper the development of appropriate conservation strategies. Obviously, there is a need for a better quantitative understanding of erosion processes at the hillslope scale for on-site impact assessment and at the watershed scale for off-site impact assessment. A powerful modern tool to achieve such understanding is the physically based WEPP-model that has replaced older empirical approaches such as the USLE. However, the high data demand of this model combined with the lack of local data hampers the application of WEPP. A new cost-effective three-step methodology for multiscale data collection is proposed and being tested in La Encañada watershed (Cajamarca Peru). The first step consists of a broad investigation of all WEPP input parameters at the watershed scale followed by an uncertainty analysis. The first results show that only 12 % of the rainfall events are larger than 10 mm. However, this category represents 46% of the precipitation. Showers greater than 10 mm can be very erosive, and soil loss seems more related with physical soil properties than with slope angle or slope length. Whether WEPP predictions are well in absolute quantitative terms is not so relevant in this stage. Our specific area of uncertainty analysis will determine which parameters require further field study and with what accuracy (and cost) these parameters need to be collected. In the second and third steps of the procedure, measured runoff and soil loss will then be used for validating the model in absolute terms. 1 Universidad Nacional Agraria-La Molina, Lima, Peru. E-mail: [email protected] 2 Wageningen University and Research Center. E-mail: [email protected] INTRODUCTION In many countries, soil erosion is the most serious form of soil degradation, with water being the major driving force. The loss of soil from a field, the breakdown of soil structure and the decline in organic matter and nutrients, result in a reduction of soil fertility (Morgan, 1995). In Peru, water erosion is a problem in agricultural production and has reduced the amount of cultivable land (Felipe-Morales et al., 1977). However, remarkably little quantitative data on erosion are available, and there is little understanding of the processes and causes underlying the erosion in the Andean highlands (Stroosnijder, 1997). Type and quality of data differ strongly per country. In Peru, data from Felipe-Morales et al. (1977) are almost the only available data, reporting 5 Mg ha yr from runoff plots, as a maximum loss in the mountain area. However, sources from other Andean countries report a variation in the erosion rate from 1 to 800 Mg ha yr under the highly Proceedings The Third International Symposium on Systems Approaches for Agricultural Development 2 diversified ecosystems of the Andes. Erosion rate seems to depend on land use more than on soil type. Fallow, pastures and abandoned fields sometimes enhance runoff and erosion risk relative to tilled fields due to low vegetation cover, crust formation and compaction of soils (Stroosnijder, 1997). It is known that soil loss is related to rainfall partly through the detaching power of raindrops striking the soil surface and partly through the contribution of rain to runoff. Intensity is generally considered the most important rainfall characteristic that influence particle detachment and splash (Hillel, 1998; Morgan, 1995). In the Andes area, several authors point out the torrential character of the rainfall. However, rainfall erosivity is unknown since analysis of data is not carried out on a routine basis. Obviously, there is a need for a quantitative understanding of erosion processes for impact assessment. A powerful modern tool to achieve such understanding is the physically based Water Erosion Prediction Project (WEPP)-model. The US Department of Agriculture developed WEPP for the quantitative prediction of erosion from hillslopes and small to medium-sized basins. The WEPP model is starting to replace the empirical models previously used by the USDA, such as the USLE (Nearing et al. 1994). However, it is still in the testing and evaluation phase and it should be used with caution until the types of environment for which it gives reliable results have been clearly identified (Soto, 1997). The WEPP model describes the processes of soil particle detachment, transport and deposition due to hydrologic and mechanical forces acting on a hillslope or in a basin. However, the high data demand of this model combined with inaccessibility or lack of local data hampers the application of WEPP. A special three-stage methodology was developed for data collection and model validation. This is being tested in La Encañada watershed in Cajamarca, Peru. Step one consists in the broad investigation of all WEPP input parameters at the watershed scale followed by an uncertainty analysis. This analysis provides the decision support for the second step that consists in the validation of the hillslope version of WEPP. In the third step, the watershed version is validated using the knowledge and experience gained in the previous steps. The WEPP is then used to recommend appropriate conservation strategies for the region. In this paper, we describe the first step of our multi-scale approach. We performed a reconnaissance inventory of all WEPP input parameters in La Encañada watershed. The uncertainty of the model was tested for the observed domain of the input variables. An uncertainty analysis differs from a sensitivity analysis in that only the local variation of the input variables is used. Therefore, our uncertainty analysis is location specific whereas a sensitivity analysis would have been independent of the site application. An uncertainty analysis allows a realistic assessment of the effect of an input parameter on the predicted soil loss estimates provided by the model.