Print Email Facebook Twitter Accounting for the influence of vegetation and landscape improves model transferability in a tropical savannah region Title Accounting for the influence of vegetation and landscape improves model transferability in a tropical savannah region Author Gao, H. (TU Delft Water Resources) Hrachowitz, M. (TU Delft Water Resources) Sriwongsitanon, Nutchanart (Kasetsart University) Fenicia, F. (Swiss Federal Institute of Aquatic Science and Technology) Gharari, S. (TU Delft Water Resources; University of Saskatchewan; Luxembourg Institute of Science and Technology) Savenije, Hubert (TU Delft Water Resources) Date 2016 Abstract Understanding which catchment characteristics dominate hydrologic response and how to take them into account remains a challenge in hydrological modeling, particularly in ungauged basins. This is even more so in nontemperate and nonhumid catchments, where—due to the combination of seasonality and the occurrence of dry spells—threshold processes are more prominent in rainfall runoff behavior. An example is the tropical savannah, the second largest climatic zone, characterized by pronounced dry and wet seasons and high evaporative demand. In this study, we investigated the importance of landscape variability on the spatial variability of stream flow in tropical savannah basins. We applied a stepwise modeling approach to 23 subcatchments of the Upper Ping River in Thailand, where gradually more information on landscape was incorporated. The benchmark is represented by a classical lumped model (FLEXL), which does not account for spatial variability. We then tested the effect of accounting for vegetation information within the lumped model (FLEXLM), and subsequently two semidistributed models: one accounting for the spatial variability of topography-based landscape features alone (FLEXT), and another accounting for both topographic features and vegetation (FLEXTM). In cross validation, each model was calibrated on one catchment, and then transferred with its fitted parameters to the remaining catchments. We found that when transferring model parameters in space, the semidistributed models accounting for vegetation and topographic heterogeneity clearly outperformed the lumped model. This suggests that landscape controls a considerable part of the hydrological function and explicit consideration of its heterogeneity can be highly beneficial for prediction in ungauged basins in tropical savannah. To reference this document use: http://resolver.tudelft.nl/uuid:d3a901cc-36fa-4cb3-936b-afa345709a03 DOI https://doi.org/10.1002/2016WR019574 Embargo date 2017-05-01 ISSN 0043-1397 Source Water Resources Research, 52 (10), 7999-8022 Part of collection Institutional Repository Document type journal article Rights © 2016 H. Gao, M. Hrachowitz, Nutchanart Sriwongsitanon, F. Fenicia, S. Gharari, Hubert Savenije Files PDF gao_et_al_2016.pdf 1.99 MB Close viewer /islandora/object/uuid:d3a901cc-36fa-4cb3-936b-afa345709a03/datastream/OBJ/view