Genetic programming based monthly groundwater level forecast models with uncertainty quantification

Abstract

Modeling hydrogeologic processes facilitates in accurate prediction/forecasting of groundwater level variations. Still, the uncertainty in model prediction is a major concern that requires detailed investigation. There could be several factors which introduce uncertainty such as inherent assumption, various levels of model complexity and simplicity. In general, model inputs, parameters and structure are the major sources of uncertainty while quantifying model prediction uncertainty. In this study, a genetic programming (GP) based models have been employed for forecasting groundwater level variation along with prediction uncertainty quantification. Though various sources induce uncertainty in the model prediction, the input uncertainty quantification has received little attention. Hence, the input uncertainty has been considered for the analysis in this study. The proposed method is demonstrated using measured monthly values of rainfall and corresponding groundwater level data of Amarawathi basin, India. It is observed that the prediction along with uncertainty quantification improves the confidence level of models while making decisions, in particular for effective planning and management of groundwater resources.

A Transdisciplinary Review of Deep Learning Research and Its Relevance for Water Resources Scientists

“…Savic et al (1999) used GP to create a rainfall-runoff model whose formula only relied on rainfall from previous days. Researchers used GP to create models for monthly groundwater levels from several wells, where GP showed good performance but prescribed complex functions of rainfall forcing and groundwater level in past time steps (Kasiviswanathan et al, 2016). When GP was used to predict saturated hydraulic conductivity (K s ), it generated a simple, regression-like equation (Fallah-Mehdipour et al, 2013).…”

Ensemble modelling framework for groundwater level prediction in urban areas of India

“…Similarly, support vector machine (SVM) a relatively newer technique has also been used for the groundwater level prediction in various site conditions (Yoon et al, 2011;He et al, 2014;Gong et al, 2016;Zhou et al, 2017). More recently, fuzzy theory and genetic programming (GP) have also been used to study the groundwater levels (Kurtulus and Razack, 2010;Güler et al, 2012;Shiri and Kisi, 2011;Fallah-Mehdipour et al, 2013;Kasiviswanathan et al, 2016). Further, latest techniques like extreme learning machine (ELM) which are much simple in design and application then ANN or SVM have also been used in groundwater modelling studies (Yadav and Eliza, 2017;Alizamir et al, 2018).…”

“…It was pointed out in many studies that the generalization ability of machine learning based models are significantly influenced by selection of appropriate input variable (Maier and Dandy, 2000;Galelli et al, 2014;Quilty et al, 2016;Sahoo et al, 2017). The most obvious input variables in groundwater level predictions studies are rainfall, evaporation, temperature and pumping patterns (Yoon et al, 2011;Singh et al, 2014;Mohanty et al, 2015;Kasiviswanathan, 2016;Chang et al, 2016;Barzegar et al, 2017;Wunsch, 2018). Further, groundwater levels are also partially controlled by interannual to multidecadal climate variability (Kuss and Gurdak, 2014;Sahoo et al, 2017;Velasco et al, 2017).…”

COVID-19 Lockdown Disruptions on Water Resources, Wastewater, and Agriculture in India

“…Such research studies that closely monitor different economic segments (i.e., high, middle, and low income) in the urban settlements for the possible changes in supply-demand patterns during the post-COVID-19 period will provide additional insights. Thus, creating a comprehensive database for domestic and commercial water use with improved monitoring devices, various machine learning algorithms (Kasiviswanathan et al, 2016;Sun and Scanlon, 2019) and empirical models (Huang et al, 1998) can be exploited to extract the useful information for managing the urban water requirement.…”