Testing the ephemeral gully erosion model (EGEM) in mediterranean environments

*J. Nachtergaele, J. Poesen, L. Vandekerckhove, and D. Oostwoud Wijdenes; Laboratory for Experimental Geomorphology, K.U. Leuven, Redingenstraat 16, 3000 Leuven, Belgium; M. Roxo, Departemento de Geografia e Planeamento Regional, Universidade Nova de Lisboa, Avenida de Berna, 26C, 1050 Lisboa, Portugal. *Corresponding Author: jeroen.nachtergaele@geo.kuleuven.ac.be ABSTRACT Few models can predict ephemeral gully erosion rates (e.g. CREAMS, WEPP, EGEM). The Ephemeral Gully Erosion Model (EGEM) was specifically developed to predicted soil loss by ephemeral gully erosion. Although EGEM pretends to have a great potential in predicting soil losses by ephemeral gully erosion, it has never been thoroughly tested. An EGEM-input data set for 86 ephemeral gullies was collected: 46 ephemeral gullies were measured in intensively cultivated land in Southeast Spain and another 40 ephemeral gullies were measured in both intensively cultivated land and in abandoned land in Southeast Portugal. Together with the EGEM-input parameters, the eroded volume for each gully was determined, so that the EGEM performance in predicting ephemeral gully erosion could be tested. A very good relationship between predicted and measured ephemeral gully volumes was found (R2 = 0.88). But as ephemeral gully length is an EGEM input parameter, both predicted and measured ephemeral gully volumes have to be divided by this ephemeral gully length. The resulting relationship between the predicted and measured ephemeral gully cross-section is not significant (R2 = 0.27). It can therefore be concluded that EGEM is not capable of predicting ephemeral gully erosion for the given Mediterranean areas. A second conclusion is that ephemeral gully length is a key parameter in determining the ephemeral gully volume. Regression analysis shows that a very significant relation between ephemeral gully length and ephemeral gully volume exists (R2 = 0.91). Accurate prediction of ephemeral gully length is therefore crucial.