Premium
From spatially variable streamflow to distributed hydrological models: Analysis of key modeling decisions
Author(s) -
Fenicia Fabrizio,
Kavetski Dmitri,
Savenije Hubert H. G.,
Pfister Laurent
Publication year - 2016
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1002/2015wr017398
Subject(s) - streamflow , hydrological modelling , hydrology (agriculture) , equifinality , variable (mathematics) , drainage basin , geology , climatology , computer science , mathematics , geography , cartography , geotechnical engineering , mathematical analysis , artificial intelligence
This paper explores the development and application of distributed hydrological models, focusing on the key decisions of how to discretize the landscape, which model structures to use in each landscape element, and how to link model parameters across multiple landscape elements. The case study considers the Attert catchment in Luxembourg—a 300 km 2 mesoscale catchment with 10 nested subcatchments that exhibit clearly different streamflow dynamics. The research questions are investigated using conceptual models applied at hydrologic response unit (HRU) scales (1–4 HRUs) on 6 hourly time steps. Multiple model structures are hypothesized and implemented using the SUPERFLEX framework. Following calibration, space/time model transferability is tested using a split‐sample approach, with evaluation criteria including streamflow prediction error metrics and hydrological signatures. Our results suggest that: (1) models using geology‐based HRUs are more robust and capture the spatial variability of streamflow time series and signatures better than models using topography‐based HRUs; this finding supports the hypothesis that, in the Attert, geology exerts a stronger control than topography on streamflow generation, (2) streamflow dynamics of different HRUs can be represented using distinct and remarkably simple model structures, which can be interpreted in terms of the perceived dominant hydrologic processes in each geology type, and (3) the same maximum root zone storage can be used across the three dominant geological units with no loss in model transferability; this finding suggests that the partitioning of water between streamflow and evaporation in the study area is largely independent of geology and can be used to improve model parsimony. The modeling methodology introduced in this study is general and can be used to advance our broader understanding and prediction of hydrological behavior, including the landscape characteristics that control hydrologic response, the dominant processes associated with different landscape types, and the spatial relations of catchment processes.