Nitrate reduction in geologically heterogeneous catchments - A framework for assessing the scale of predictive capability of hydrological models
Résumé
In order to fulfil the requirements of the EU Water Framework Directive nitrate load from agricultural areas to surface water in Denmark needs to be reduced by about 40%. The regulations imposed until now have been uniform, i.e. the same restrictions for all areas independent of the subsurface conditions. Studies have shown that on a national basis about 2/3 of the nitrate leaching from the root zone is reduced naturally, through denitrification, in the subsurface before reaching the streams. Therefore, it is more cost-effective to identify robust areas, where nitrate leaching through the root zone is reduced in the saturated zone before reaching the streams, and vulnerable areas, where no subsurface reduction takes place, and then only impose regulations/restrictions on the vulnerable areas. Distributed hydrological models can make predictions at grid scale, i.e. at much smaller scale than the entire catchment. However, as distributed models often do not include local scale hydrogeological heterogeneities, they are typically not able to make accurate predictions at scales smaller than they are calibrated. We present a framework for assessing nitrate reduction in the subsurface and for assessing at which spatial scales modelling tools have predictive capabilities. A new instrument has been developed for airborne geophysical measurements, Mini-SkyTEM, dedicated to identifying geological structures and heterogeneities with horizontal and lateral resolutions of 30-50 m and 2 in, respectively, in the upper 30 m. The geological heterogeneity and uncertainty are further analysed by use of the geostatistical software TProGS by generating stochastic geological realisations that are soft conditioned against the geophysical data. Finally, the flow paths within the catchment are simulated by use of the MIKE SHE hydrological modelling system for each of the geological models generated by TProGS and the prediction uncertainty is characterised by the variance between the predictions of the different models. (C) 2013 Elsevier B.V. All rights reserved.