On the performance of conceptual and physically based modelling approach to simulate a headwater catchment in Brazil

Abstract

Conceptual large-scale distributed hydrological models (e.g. MHD-INPE) were developed to not only be used with limited available data but also to match the scale of atmospheric models. However, it is unknown if it can be representative for small watersheds in rural mountainous regions, which are crucial not only for agriculture, but also for water supply for several uses. Therefore, the objective of this study is to assess MHD-INPE performance in a headwater catchment with complex terrain alongside to a distributed physically based model (DHSVM). The Lavrinha Watershed (LW) has a drainage area of 6.7 km2, being 63% of the catchment occupied by native vegetation (Atlantic Rainforest), while the remaining area is mainly occupied by pasture for livestock farming. MHD-INPE and DHSVM were applied using a soil moisture zone map derived from the height above the nearest drainage (HAND) algorithm. The runoff simulated by the MHD-INPE better fits the observed data, with a validation Nash-Sutcliffe efficiency (NSE) of 0.70 for the daily scale, compared to a 0.55 in DHSVM. In the evapotranspiration simulation, both models showed similar trends, being 49% of the precipitation in the MHD-INPE and 46% in the DHSVM, while the observed value was 49%. For the baseflow, the MHD-INPE fitted better to the observed streamflow, whereas the DHSVM underestimated it during the dry season. Thus MHD-INPE was able to accurately simulate the streamflow in a mountainous headwater catchment in southeast Brazil, despite its large spatial scale.