How would peak rainfall intensity affect runoff predictions using conceptual water balance models?

B Yu

doi:10.5194/piahs-371-109-2015

Abstract

Abstract. Most hydrological models use continuous daily precipitation and potential evapotranspiration for streamflow estimation. With the projected increase in mean surface temperature, hydrological processes are set to intensify irrespective of the underlying changes to the mean precipitation. The effect of an increase in rainfall intensity on the long-term water balance is, however, not adequately accounted for in the commonly used hydrological models. This study follows from a previous comparative analysis of a non-stationary daily series of stream flow of a forested watershed (River Rimbaud) in the French Alps (area = 1.478 km2) (1966–2006). Non-stationarity in the recorded stream flow occurred as a result of a severe wild fire in 1990. Two daily models (AWBM and SimHyd) were initially calibrated for each of three distinct phases in relation to the well documented land disturbance. At the daily and monthly time scales, both models performed satisfactorily with the Nash–Sutcliffe coefficient of efficiency (NSE) varying from 0.77 to 0.92. When aggregated to the annual time scale, both models underestimated the flow by about 22% with a reduced NSE at about 0.71. Exploratory data analysis was undertaken to relate daily peak hourly rainfall intensity to the discrepancy between the observed and modelled daily runoff amount. Preliminary results show that the effect of peak hourly rainfall intensity on runoff prediction is insignificant, and model performance is unlikely to improve when peak daily precipitation is included. Trend analysis indicated that the large decrease of precipitation when daily precipitation amount exceeded 10–20 mm may have contributed greatly to the decrease in stream flow of this forested watershed.

Highlights

Most concceptual hydrological models such as AWBM and SimHyd for streamflow estimation require, as input, continuous daily precipitation and potential evapotranspiration
AWBM and SimHyd were found to be adequate in decribing the hydrology of the Rimbaud watershed using the workshop protocal, and the effect of the wild fire in the watershed on stream flow was found to be secondary relative to the declining precipitation and streamflow during the study period from 1968 to 2006 (Yu and Zhu, 2014)
Annual flows from the three distinct phases were assembled to illustrate model performance when the flow was aggregated at the annual time scale

Summary

Introduction

Most concceptual hydrological models such as AWBM and SimHyd for streamflow estimation require, as input, continuous daily precipitation and potential evapotranspiration. As part of a workshop on modelling non-stationary time-series of streamflow, attempt was made to compare the performance of AWBM and SimHyd when calibrated using daily data for two forested watersheds in France and the USA (Yu and Zhu, 2014). Of the 14 watersheds for the workshop, sub-daily precipiation data at hourly intervals were available for the Rimbaud watershed for the study period from 1966 to 2006 The objectives of this preliminary study were (1) to assess model performance for distinct phases of the watershed: (a) pre-fire, (b) fire-affected; (c) post-fire; (2) to relate model residuals to peak rainfall intensity on a daily basis; (3) to relate the decline in stream flow to changes in precipitation, potential evapotranspiration and rainfall intensities. Linear regression was used extensively for exploratory data anlysis of the relationship between rainfall intensity and model performance, and of the time series of precipitation, potential evapotranspiration, and rainfall intensities

Model calibration

Parameter values

Effect of rainfall intensity

Conclusions