Global sensitivity analysis results based on IVARS50 for various metrics and modelsNash-Sutcliffe (NS)

to reflect HIGH flowsNS of Log of flows (NS_Log)

to reflect LOW flowsPercent Bias (PBIAS)

To reflect flow VOLUME

MESH

SWAT

HGS

Sensitivity Analysis and Insights into Hydrological Processesand Uncertainty at Different Scales

MODEL

PARAMETER

SETSMETRICS

Sensitivity analysis (SA) is an essential tool for providing insight into model behavior, calibration and uncertainty assessment. It is often overlooked that the metric choice and study scale can significantly change the assessment of model sensitivity. In order to identify important hydrological processes across various case studies, we conducted a multi-model multi-criteria sensitivity analysis using a novel and efficient technique, Variogram Analysis of Response Surfaces (VARS). The analysis was conducted using three physically-based hydrological models, applied at various scales ranging from small (hillslope) to large (watershed) scale. In each case, the sensitivity of simulated streamflow to model processes (represented through parameters) were measured using different metrics selected based on various hydrograph characteristics such as high flows, low flows, and volume.

INTRODUCTION CONCLUSIONS

1. Metric choice has a significant influence on SA results and must be aligned with study objectives.

• For NS and NS_Log, runoff and routing parameters are more important; but for PBIAS, soil and vegetation properties become more important.

2. Performance metric has a dominant influence and scale has a secondary effect on SA results across the case studies here.

• At the smaller scale, soil infiltration parameters are more important for all metrics; but at the larger scale, routing parameters become more important except for PBIAS.

REFERENCES

Razavi, S., & Gupta, H. V. (2015). What do we mean by sensitivity analysis? The need for comprehensive characterization of ‘‘global’’ sensitivity in Earth and Environmental systems models, Water Resources Research, 51(5), 3070–3092.

Razavi, S., & Gupta, H. V. (submitted a). A new framework for comprehensive, robust, and efficient Global Sensitivity Analysis: Part I - Theory, Water Resources Research.

Razavi, S., & Gupta, H. V. (submitted b). A new framework for comprehensive, robust, and efficient Global Sensitivity Analysis: Part II - Application, Water Resources Research.

CONTACT

Amin Haghnegahdar, PhDPost-doctoral Fellow

Global Institute for Water Security (GIWS)University of Saskatchewan, Saskatoon, SK Canadaamin.haghnegahdar@usask.cahttp://researchers.usask.ca/amin-haghnegahdar

1. HydroGeoSphere (HGS): 3-D fully-integrated surface and subsurface flow simulator, ~1200 m2, 10 parameters

2. Soil and Water Assessment Tool (SWAT): semi-distributed hydrological model, ~860 km2, 14 parameters

3. Modélisation Environmentale–Surface et Hydrologie (MESH): semi-distributed land surface-hydrological model, ~2700 km2, 40 parameters

Variogram Analysis of Response Surfaces

STAR-based sampling

STAR-VARS products

1. Illustrate the effect of metric choice on sensitivity assessment

2. Identify important hydrological processes across various models and scales

Paper Number: H23F-1638

Amin Haghnegahdar1, Saman Razavi1, Howard Wheater1, Hoshin Gupta2

1 Global Institute for Water Security, University of Saskatchewan, Saskatoon, Saskatchewan, Canada 2 Department of Hydrology and Water Resources, University of Arizona, Tucson, Arizona, USA

ACKNOWLEDGEMENT

Special thanks to Aquanty Inc. for support withHydroGeoSphere for this study.

CASE STUDIES

OBJECTIVES