Optimization of Calibration Parameters for an Event Based Watershed Model Using Genetic Algorithm

Abstract

In this study, an event based rainfall runoff model has been integrated with Single objective Genetic Algorithm (SGA) and Multi-objective Genetic Algorithm (MGA) for optimization of calibration parameters (i.e. saturated hydraulic conductivity (K s ), average capillary suction at the wetting front (S av ), initial water content (θ i ) and saturated water content (θ s )). The integrated model has been applied for Harsul watershed located in India, and Walnut Gulch experimental watershed located in Arizona, USA. Nash-Sutcliffe Efficiency (NSE) and correlation coefficient (r) between observed and simulated runoff have been used to test the performance of runoff models. The SGA and MGA integrated runoff model performance is also compared with the performance of the Hydrologic Engineering Center-Hydrologic Modeling System (HEC_HMS) model. Range of NSE values for study watersheds with integrated MGA, integrated SGA, HEC_HMS and for the event based rainfall runoff models are [−0.61 to 0.79], [−0.5 to 0.74], [−3.37 to 0.82] and [−5.78 to 0.53] respectively. Range of correlation coefficient values for study watersheds with integrated MGA, integrated SGA, HEC_HMS and for the event based rainfall runoff models are [0.18 to 0.95], [−0.55 to 0.90], [−0.18 to 0.97] and [−0.12 to 0.86] respectively. From the results, it is evident that the integrated model is giving the best calibrated parameters as compared to manual calibration methods. Genetic Algorithm (GA) integrated runoff models can be used to simulate the flow parameters of data sparse watersheds.

Exploiting the use of physical information for the calibration of a lumped hydrological model

“…In all cases, parameter optimization is carried out using an automatic calibration performed by a genetic algorithm (GA) tool (Goldberg, ) using the root mean square error (RMSE) as objective function or a specifically developed error function, called E snow , for the snow‐pack module (see Equation ). In recent decades, GAs have been widely used for the automatic calibration of hydrological models (e.g., Boisvert, El‐Jabi, St‐Hilaire, & El Adlouni, ; Boyle et al, ; Dumedah, Berg, Wineberg, & Collier, ; Franchini & Galeati, ; Kaini, Artita, & Nicklow, ; Reshma, Reddy, Pratap, Ahmedi, & Agilan, ; Wang, , ).…”

Modelling nonlinear trend for developing non‐stationary rainfall intensity–duration–frequency curve

“…For multi-modal objective functions, global search methods have been developed that are especially designed for locating the global optimum and not being trapped in local optima (Madsen, 2003). The generic algorithm (GA) is one of the efficient global search methods (Deb et al, 2002;Rajasekaran and Vijayalakshmi, 2003;Reshma et al, 2015). We have used MOGA to estimate the values of a, b and c and the general steps involved in estimating (optimizing) three variable's value using MOGA are as follows:…”

An automatic parameter calibration method for the SWAT model in runoff simulation

“…The unique crossover, mutation, and other genetic operators in GA enable it to exhibit better global optimization capability than PSO (Yi-wu, Qing-yin, & Xian-cheng, 2010). However, the performance of the GA integrated model may be improved by increasing number of calibrations, which results in the time consumption (Reshma, Reddy, Pratap, Ahmedi, & Agilan, 2015). Consequently, the GA-PSO method, by introducing genetic ideas (such as crossover and mutation) into PSO, aims to integrate the advantages of the two algorithms to improve a single algorithm, that is, to achieve rapid convergence and ensure global optimization capabilities (Garg, 2015;Jatana & Suri, 2019;Lim, Ponnambalam, & Izui, 2017;Sheikhalishahi, Ebrahimipour, Shiri, Zaman, & Jeihoonian, 2013;Yuming & Renjin, 2014;Zhang, Liu, Wu, Cai, & Ma, 2016).…”