Content area
Full Text
HIGHLIGHTS
INTRODUCTION
Groundwater plays an imperative role in supplying water for agriculture, industry, and drinking needs. Meanwhile, modeling groundwater has substantial effects on the economy, notably on water resource management (Dagès et al. 2012). Recent variations in the hydrologic cycle due to changes in global climate, population growth, and agricultural area expansion have led to the overuse of groundwater resources in most parts of the world. Accordingly, the management of these resources is challenging due to problems caused by overuse, including reduced groundwater levels, diminished water quality, and increased harvest costs (Wada et al. 2010).
Undoubtedly, numerical models of groundwater are essential tools for groundwater resource management. In recent years, different researchers have applied numerical models to manage groundwater resources (e.g., Yihdego et al. 2016; Kamali & Niksokhan 2017). The main challenge researchers face in utilizing numerical models is the calibration of these models. In general, these models can be calibrated by trial and error and automatically. In recent decades, studies have proposed various methods based on inverse modeling for the automatic estimation of hydrological parameters (e.g., Dai & Samper 2006; Shang et al. 2016). Automated calibration methods are far faster than trial-and-error calibration and offer better outcomes due to their capability of searching more parameter space (Droogers et al. 2001). Further, different software applications have been developed for the automatic calibration of groundwater models, including PEST (Doherty et al. 1994), UCODE (Poeter & Hill 1998), and HydroPSO (Zambrano-Bigiarini & Rojas 2013). Although these applications accelerate the calibration process, they may obtain irrational values for parameters since they mainly aim to match observational and computational values regardless of physical realities (Delottier et al. 2017).
The findings of the applied models for groundwater modeling heavily rely on the input parameters of the model. On the other hand, the scarcity of field data, poor knowledge of the conceptual model, various modeling scenarios, and hydrogeological complexities cause uncertainties in the model (Wu & Zeng 2013). Thus, automated calibration methods based on uncertainty analysis have been of great interest to researchers in recent years. Interval estimates of parameter values are provided in these methods, unlike other optimization methods that generate a point and definite parameter estimates. Hydrologists have utilized uncertainty-based methods for several reasons such as errors in measuring...