Abstract

Precipitation is the most important input to hydrological models, and its spatial variability can strongly influence modeled runoff. The highly dense station network WegenerNet (0.5 stations per km2) in southeastern Austria offers the opportunity to study the sensitivity of modeled runoff to precipitation input. We performed a large set of runoff simulations (WaSiM model) using 16 subnetworks with varying station densities and two interpolation schemes (inverse distance weighting, Thiessen polygons). Six representative heavy precipitation events were analyzed, placing a focus on small subcatchments (10–30 km2) and different event durations. We found that the modeling performance generally improved when the station density was increased up to a certain resolution: a mean nearest neighbor distance of around 6 km for long-duration events and about 2.5 km for short-duration events. However, this is not always true for small subcatchments. The sufficient station density is clearly dependent on the catchment area, event type, and station distribution. When the network is very dense (mean distance < 1.7 km), any reasonable interpolation choice is suitable. Overall, the station density is much more important than the interpolation scheme. Our findings highlight the need to study extreme precipitation characteristics in combination with runoff modeling to decompose precipitation uncertainties more comprehensively.

Highlights

  • Heavy precipitation events can have significant impacts on society and ecosystems by causing severe floods and landslides

  • The deviations observed in the calibration/validation period exhibited a maximum deviation of 0.04/0.02 in Nash–Sutcliffe model efficiency coefficient (NSE), 0.07/0.13 in logarithmic Nash–Sutcliffe efficiency (logNSE), 0.08/0.04 in Kling–Gupta efficiency (KGE), and 11/14% for percent bias (PBIAS), respectively

  • The best model performance was obtained with the parameter set of krec = 0.8, dr = 40, kd = 1.5, and ki = 2. This setup resulted in a model performance for the river runoff in the calibration period of summer 2009 with an NSE of 0.80, logNSE of 0.76, KGE of 0.77, and PBIAS of 8%

Read more

Summary

Introduction

Heavy precipitation events can have significant impacts on society and ecosystems by causing severe floods and landslides. Despite the availability of remote-sensing data, ground-based precipitation measurement tools are still widely used in hydrological modeling [6,22] Many studies, such as those by Lopez et al [22], Goovaerts [23], and Zeng et al [6], pointed out the advantages of dense and regularly distributed precipitation station networks. Huang et al [12] used a lumped and a distributed hydrological model to study the sensitivity of model performance to spatial rainfall resolution They identified temporal resolution as the most important aspect, observing better model performance at higher temporal resolutions. We go one step further, applying our initial study approach and focus to study the impact of such precipitation uncertainty on hydrologic simulation results, and especially on runoff peaks, using a combination of station densities and interpolation methods

Study Area
Model Setup and Calibration
Experimental Design
Selection
Selection of Precipitation Events
17 September 2010
Precipitation time series of the “short-1 event”
Spatial Interpolation Schemes
Runoff Analysis Approach
Results for Individual Example Events
Peak Flow Deviation
Threshold
Influence of Station Location
Effect of Timing of Peak Flow
Comparison of Interpolation Schemes IDW and TP
Conclusions
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call