Exploring the link between drought‐related terms and public interests: Global insights from LSTM‐based predictions and Google Trends analysis

Abstract

Effective drought monitoring is of paramount importance in hydrology. It aids in mitigating the detrimental effects of water scarcity, facilitates sustainable resource management, and informs policy decisions. Therefore, it is crucial to comprehensively comprehend the dynamics and trends of drought‐related phenomena. This study aims to explore the relationship between six low water quantity terms including drought, water crisis, water scarcity, water shortage, water stress, and water insecurity and some socio‐economic, geographic, and demographic parameters for different regions of the world and to predict the future trend of public interest in the mentioned terms using Long Short‐Term Memory neural network (LSTM) models. Google Trend data analysis was used to examine the public interest in these terms from 2017 to 2022. The LSTM models were trained using historical data on the studied terms, and their performance was evaluated using Root Mean Square Error (RMSE) indicator. The Google Trend data analysis showed that public interest in water shortage and water insecurity increased significantly from 2017 to 2022. The LSTM models showed promising results for predicting future trends in the mentioned water related issues, with RMSE scores (dimensionless) ranging from 0.04 to 0.43. The most significant socio‐economic, geographic and demographic parameters were found to be population, life expectancy, access to drinking water, and access to Internet while there was no correlation between precipitation and searched terms. The results suggest that LSTM models can be an effective tool for forecasting water related issues and emphasizes the importance of socio‐economic, geographic and demographic parameters in the public search behaviour around the world. The study also highlights the increasing public awareness of water related issues and the need for sustainable water management practices, particularly in regions with high water shortage and insecurity. The LSTM‐based prediction models in this study have practical applications in early warning systems for droughts, community education on water conservation, prioritizing vulnerable areas, assessing public perception of climate change's relation to droughts, and evaluating water policies. Further research is needed to improve the accuracy of these models incorporating the effective parameters and to develop effective strategies for managing water resources in regions facing water scarcity.

Quantitative study of rainfall lag effects and integration of machine learning methods for groundwater level prediction modelling

“…Nowadays, AI methods have been successfully employed for GWL modelling and prediction in aquifers of different geological and climatic regions (Daneshvar Vousoughi, 2022; Khan et al, 2023; Naghipour et al, 2023; Rajaee et al, 2019). In this field, long short‐term memory (LSTM) has found extensive application in this domain (Shahabi‐Haghighi & Hamidifar, 2023). By incorporating gates and well‐defined memory units, LSTM enhances the recurrent neural network model and resolves the problems of gradient vanishing and explosion encountered in handling long time series within the recursive neural network model (da Silva et al, 2013; Kratzert et al, 2018; Zhang et al, 2018).…”