Impact of data sources to DEM construction and application to runoff and sediment yield modelling using LISEM model

Abstract

The use of interpolated digital elevation models (DEM) obtained from free low-resolution satellite images is a typical solution for soil erosion modelling. However, this method may result in loss of surface details, especially in small catchments. Thus, planialtimetric surveys at the study site may improve quality of terrain description, although it requires high costs and time. Therefore, this study aims to evaluate the impact of two DEM sources on the results of a process-based erosion model. The study was developed in the Lajeado Ferreira creek catchment (1.23 km2) in southern Brazil. Two interpolated 5-m resolution DEM maps obtained from: (i) remote sensing source Shuttle Radar Topography Mission (SRTM), and (ii) field topographic survey with Global Navigation Satellite System using Real Time Kinematic (GNSS/RTK); were used as input in the Limburg Soil Erosion Model (LISEM) to simulate runoff and sediment transport. Two other maps required for modelling were developed from each DEM source: the flow direction and drainage network map. Six monitored rainfall events were calibrated for the variables, timing of peak flow (Qtime), peak flow (Qpeak), surface runoff coefficient (C), total surface runoff volume (Qtotal) and sediment yield (SY). Most variables calibrated were within the acceptable statistical ranges. The Qtime and Qpeak simulated for all events were close to the measured values with minor modification of the input parameters while using GNSS/RTK. However, even in maps with better accuracy of relief description (GNSS/RTK), the erosion parameters adjusted for calibration were not within an acceptable physical limit. For example, soil cohesion values had to be multiplied for at least eight times their original value. Hydrograph and sedimentgraph shapes calibration did not reach a satisfactory statistical level, even with the GNSS/RTK database and the incorporation of relevant features of the landscape based on local observation. Although GNSS/RTK database resulted in gains for some variables calibrated with the LISEM model, the advantages were not significant for the conditions of this study.