Visual Validation of the e-RUSLE Model Application at the Pan-European Scale

Bosco, C., de Rigo, D., Dewitte, O., 2013. <strong>Visual Validation of the e-RUSLE Model Application at the Pan-European Scale</strong>. <em>Scientific Topics Focus 1</em>, MRI-11a13. Notes Transdiscipl. Model. Env., Maieutike Research Initiative. doi: 10.6084/m9.figshare.844627 <br>Version: DRAFT 0.2.1<br>This is a preliminary version. The views expressed are those of the authors and may not be regarded as stating an official position of mentioned organisations. <strong>Visual Validation of the e-RUSLE Model Application at the Pan-European Scale</strong> Claudio Bosco ¹ ² ⁴, Daniele de Rigo ² ³ ⁴ and Olivier Dewitte ⁵ 1 Loughborough University, Department of Civil and Building Engineering,<br>Loughborough, United Kingdom 2 Joint Research Centre of the European Commission,<br>Institute for Environment and Sustainability, Ispra, Italy 3 Politecnico di Milano, Dipartimento di Elettronica, Informazione e Bioingegneria, Milano, Italy 4 Maieutike Research Initiative, Milano, Italy 5 Royal Museum for Central Africa, Department of Earth Sciences, Tervuren, Belgium Detailed analysis and methods for modelling soil erosion dynamics are growingly available at local and catchment scales. The integrated assessment and mitigation of natural hazards often demands regional to continental scales to be considered. At these scales, geospatial methods and computational modelling need to deal with multiple sources of uncertainty in heterogeneous environmental systems which also involve multiple scientific disciplines. This complex modelling activity is required at the science-policy interface in order for appropriate management options to be assessed (Integrated Natural Resources Modelling and Management, INRMM). However, modelling strategies typical of the local scale may not be suitable for landscape analysis or at wider scales. The predictive power of existing models at regional and continental scale is still limited. Among those models, empirical approaches based on regressions may offer a viable option as their requirement in input data is less demanding than physically based models and their performances are often better. However, quantitatively validating at wide-scale the performance of a model– exploiting the procedures which are typical at local scales – is still extremely challenging. The common local-scale validation procedures are not technically and financially applicable for large spatial extents. Here we present a qualitative validation of a recent proposal aiming at improving the current computational modelling approaches to support wide-scale soil loss estimation. The approach proposed in Bosco et al.(under review) extends a well-established empirical model for estimating soil erosion by water. The extended model is based on the Revised Universal Soil Loss Equation (RUSLE). The proposed extension is the e-RUSLE, an array-based, semantically enhanced modification of the RUSLE in which the multiplicity intrinsic in large scale complex and uncertain problems is exploited. The e-RUSLE modelling architecture takes advantage from the semantic array programming (SemAP) paradigm. A first dimension of multiplicity focused on the multiplicative structure of the RUSLE model for suggesting a new factor specifically designed to consider soil stoniness. Spatial, geographic and climatic multiplicity was better recognised by merging different empirical rainfall-erosivity equations with a robust ensemble model based on climatic similarity. The rates of soil erosion by water were estimated in Europe exclusively relying on publicly available datasets. This validation work offers a detailed analysis of the resulting map of soil erosion by water with plausibility checks. They have been implemented by means of visual validation expert judgement. The second section of the report was generated by means of SemAP data-transformations (exploiting GNU Bash, GNU Octave, Latex languages and the Mastrave modelling library under GNU/Linux) from a concise text-based tabular file in which the whole validation database has been recorded. The results corroborate the map and the applied modelling methodology.