Drought Prediction with Raw Satellite Imagery and Ensemble Supervised Machine Learning

Abstract

Drought is one of the biggest challenges that environmentalists face today because of rapidly evolving climate. The negative impacts of drought on the economy, humans and other living organisms endure long after the ending of drought, and with time its intensity also increases. One way to fight the adverse effects of drought is to perform drought prediction and so that appropriate decisions can be made accordingly. Drought prediction can be made considering vegetation and water level in any region therefore, in this research we are using satellite images to predict drought conditions and its various stages, like if it is about to come or it has passed. All these predictions will be helpful for authorities to make informed decisions. We are employing supervised machine learning nevertheless to obtain the best results. We are using boosting and bagging which is ensemble supervised machine learning techniques. The experiments performed proved that bagging is better than boosting classifiers and it is less computationally expensive; and boosting on the other hand is less accurate and computationally expensive.Contribution/Originality: This research performs drought prediction on tharparkar district using raw satellite imagery and ensemble machine learning techniques.

Evaluation of deep learning approaches for classification of drought stages using satellite imagery for Tharparker

“…The studies discussed above either use some index or remote sensing for prediction, but this study uses raw satellite images which create ease of use. The preliminary results of this study are presented, using machine learning and ensemble machine learning technique [19].…”

Prospective analysis of spatial heterogeneity influence on the concordance of remote sensing drought indices: a case of semi-arid agrosystems in Morocco (Moulouya and Tensift watersheds)

“…The unpredictable occurrence in time and space of drought sequences has serious socioeconomic and ecosystem consequences in regions heavily dependent on rainfall (Lui et al 2018;Sinha et al 2017;Cartwright et al 2020;Liu et al 2020;Liu et al 2016;Wu et al 2020;Jiao et al 2021). Nevertheless, the negative effects of drought sequences can be mitigated by early warning and drought monitoring actions (Enenkel et al 2015;Masinde, 2015;Didier et al 2017;Raza et al 2021;Jiao et al 2021;Chandrasekara et al 2021).…”