Effect of historical land‐use change on soil erosion in a Mediterranean catchment by integrating 137 Cs measurements and WaTEM/SEDEM model

• Published in: Hydrological processes 2022-05, Vol.36 (5)
• Main Authors: Lizaga Villuendas, Ivan, Latorre, Borja, Gaspar, Leticia, Navas, Ana
• Format: Article
• Language: English
• ISSN: 0885-6087; 1099-1085
• DOI: 10.1002/hyp.14577

Soil erosion is especially severe in areas affected by intermittent heavy rainfalls after dry periods, and human practices such as deforestation. Mediterranean mountain environments underwent conversion of rangelands into croplands during the previous centuries to increase agricultural production but this process was reversed after land abandonment in the middle of the twentieth century allowing the natural revegetation. To understand the effect of the past practices and the current agricultural management, we have combined the strength of empirical data and spatially distributed modelling in a medium‐sized catchment representative of agroforestry landscapes of NE Spain. We developed an ensemble technique composed of 137 Cs‐derived soil redistribution rates as specific point values and as grid‐based setup calibration for the WaTEM/SEDEM. Thus, we overcame the point specific limitation of the 137 Cs measurement and the need for calibration of spatially distributed models that allow a spatial interpretation of soil redistribution in the catchment. We implemented an automated routine model tool to increase the calculation speed and calibrate the model thus, allowing estimates of the main model parameters for simulating a variety of temporal scenarios. With this method a good fit between modelled and measured soil redistribution rates ( R 2  = 0.82) was obtained (Nash–Sutcliffe efficiency: 0.79). Estimates were also consistent with the previous apportions of sediment provenance obtained by the fingerprinting technique. An increase of 300% in sediment export was predicted for the past scenario when most of the catchment was cultivated, while a reduction of 40% occurred in the future scenario of land abandonment. A high influence of input resolution was found, wherein model efficiency was reduced when pixel‐size was increased. Our study demonstrates that spatially distributed models combined with 137 Cs‐derived rates provide a powerful tool to understand the driving factors of erosion and delineate the hotspot areas that could suffer high erosion if submitted to certain management practices.