Characterization of watershed model behavior across a hydroclimatic gradient

Abstract

[1] A fundamental tradeoff exists in watershed modeling between a model's flexibility for representing watersheds with different characteristics versus its potential for overparameterization. This study uses global sensitivity analysis to investigate how a commonly used intermediate-complexity model, the Sacramento Soil Moisture Accounting Model (SAC-SMA), represents a wide range of watersheds with diverse physical and hydroclimatic characteristics. The analysis aims to establish a detailed understanding of model behavior across watersheds and time periods with the ultimate objective to guide model calibration and evaluation studies. Sobol's sensitivity analysis is used to evaluate the SAC-SMA in 12 Model Parameter Estimation Experiment (MOPEX) watersheds in the US. The watersheds span a wide hydroclimatic gradient from arid to humid systems. Four evaluation metrics reflecting base flows, midrange flows, peak flows, and long-term water balance were used to comprehensively characterize trends in sensitivity and model behavior. Results show significant variation in parameter sensitivities that are correlated with the hydroclimatic characteristics of the watersheds and time periods analyzed. The sensitivity patterns are consistent with the expected dominant processes and demonstrate the need for moderate model complexity to represent different hydroclimatic regimes. The analysis reveals that the primary model controls for some aspects of the simulated hydrograph are different from those typically assumed for the SAC-SMA. Results also show that between 6 and 10 parameters are regularly identifiable from daily hydrologic data, which is about twice the range that is often assumed (i.e., 3 to 5). Synthesized results provide comprehensive SAC-SMA calibration guidance, demonstrate the flexibility of the model for representing multiple hydroclimatic regimes, and highlight the great difficulty in generalizing model behavior across watersheds.

Multiobjective sensitivity analysis to understand the information content in streamflow observations for distributed watershed modeling

“…There is an interesting difference between the sensitivity to RMSE for event scales versus longer time scales. The latter was analyzed for the lumped SAC-SMA model by van Werkhoven et al [2008b], who found that the controlling parameters were located in both the upper and the lower zone. This is not the case for the shorter time scale.…”

“…The ''total'' Sobol' sensitivity index used in this study reflects the combined effect of the parameter alone (i.e., individual sensitivity) plus its interactions with all other parameters in the analysis. For more details on Sobol''s method see Sobol' [1993Sobol' [ , 2001, Saltelli et al [1999], or van Werkhoven et al [2008b.…”

“…

[1] In a previous paper, van Werkhoven et al (2008b) demonstrated that the information content of streamflow observations at a watershed outlet is a dynamic entity and is dependent on the spatiotemporal dynamics of the causal precipitation event. This result has important consequences for distributed hydrological model calibration strategies and for the design of observation networks.

…”

Global sensitivity analysis in hydrological modeling: Review of concepts, methods, theoretical framework, and applications

“…However, we also argue that SA is a useful perspective for conceptualizing and understanding hydrological models for several reasons. As indicated by Rakovec et al (2014), SA can be used to: (a) detect when increasing model complexity can no longer be supported by observations and whether it is likely to affect model predictions (e.g., Saltelli et al, 1999;van Werkhoven et al, 2008a;Doherty and Welter, 2010;Rosolem et al, 2012;Gupta et al, 2012;Foglia et al, 2013); (b) reduce the time required for model calibration by focusing estimation efforts on parameters that are important for calibration metrics and predictions (e.g., Anderman et al, 1996;Hamm et al, 2006;Zambrano-Bigiarini and Rojas, 2013); (c) determine the priorities for theoretical and site-specific model development (e.g., Hill and Tiedeman, 2007;Saltelli et al, 2008;Kavetski and Clark, 2010); and (d) identify the advantageous placement and timing of new measurements (e.g., Tiedeman et al, 2003Tiedeman et al, , 2004.…”

Using an informational entropy-based metric as a diagnostic of flow duration to drive model parameter identification

“…from 30 to 70 percentile flows) and either construct objective measures (i.e. volumetric bias, error in the slope) for each segment (van Werkhoven et al, 2008) or select evaluation points along the FDC and thereby evaluate the deviation between simulated and observed discharge. Therefore, current FDC calibration methods have a limited ability to extract the distributional information contained in the probability density function (pdf) of the flow signal.…”