Data-Driven Modeling and the Influence of Objective Function Selection on Model Performance in Limited Data Regions.

Thelma Dede Baddoo,Zhijia Li,Yiqing Guan,Isaac Kwesi Nooni,Kenneth Rodolphe Chabi Boni

doi:10.3390/ijerph17114132

Abstract

The identification of unit hydrographs and component flows from rainfall, evapotranspiration and streamflow data (IHACRES) model has been proven to be an efficient yet basic model to simulate rainfall–runoff processes due to the difficulty in obtaining the comprehensive data required by physical models, especially in data-scarce, semi-arid regions. The success of a calibration process is tremendously dependent on the objective function chosen. However, objective functions have been applied largely in over daily and monthly scales and seldom over sub-daily scales. This study, therefore, implements the IHACRES model using ‘hydromad’ in R to simulate flood events with data limitations in Zhidan, a semi-arid catchment in China. We apply objective function constraints by time aggregating the commonly used Nash–Sutcliffe efficiency into daily and hourly scales to investigate the influence of objective function constraints on the model performance and the general capability of the IHACRES model to simulate flood events in the study watershed. The results of the study demonstrated the advantage of the finer time-scaled hourly objective function over its daily counterpart in simulating runoff for the selected flood events. The results also indicated that the IHACRES model performed extremely well in the Zhidan watershed, presenting the feasibility of the use of the IHACRES model to simulate flood events in data scarce, semi-arid regions.

Highlights

Arid and semi-arid regions are recurrently prone to be dominated by extreme rainfall events with a high degree of spatial variability, usually resulting in rapid response profiles [1]
The fact that the comprehensive hydrological data required by conceptual or physically based models are rarely available in most watersheds has led to the rapid development of data-driven models (DDMs) in hydrology and environmental sciences
This study, applies the identification of unit hydrographs and component flows from rainfall, evapotranspiration and streamflow data (IHACRES) model, a hybrid conceptual-data-driven rainfall–runoff model to simulate a series of flood events with limited data over some years representing changes in land-use and landcover settings in Zhidan, a semi-arid catchment in northern China, applying objective function constraints in the form of diversifying the Nash–Sutcliffe efficiency (R2 in IHACRES)

Summary

Introduction

Arid and semi-arid regions are recurrently prone to be dominated by extreme rainfall events with a high degree of spatial variability, usually resulting in rapid response profiles [1]. Accurate modeling of these events would facilitate proper flood forecasting in these areas to save lives and property. Flood modeling remains one of the most challenging and important tasks of operational hydrology due to spatial and temporal variations in rainfall distribution and the tremendously complex and highly nonlinear nature of the rainfall–runoff relationship [2]. Res. Public Health 2020, 17, 4132; doi:10.3390/ijerph17114132 www.mdpi.com/journal/ijerph

Methods

Results

Discussion

Conclusion