Remote Sensing and Machine Learning Applications for the Assessment of Urban Water Stress: A Review

Water stress is a critical factor and depends on the balance between water demands and supplies at any given time and locations. The continuous expansion of urban areas attributed to rural immigration to urban centers and population growth, exacerbated by a changing climate, has broken the balance between water supplies and demands, making cities more water insecure. The use of remote sensing (RS) products and machine learning (ML) analytics for the assessment of water stress areas has increased in the past twenty years. This paper reviews scientific and technological evidence published in the past years in the intersection of RS, ML, and water stress. We explore how RS and ML are shaping the current and future needs for research and innovation for water stress assessment. This review focuses on the contrasting sources of water stress various water stress when water surpluses and deficits are present, and how indicators have incorporated the use of RS and ML to identify temporal and geospatial attributions, scales, and the metrics. It has been found that metrics such as rainfall, population, runoff, drainage network have been diversely and extensively are used in different case studies which plays a major role. For the water quality assessments, the parameters of salinity, pH, dissolved oxygen, suspended solids, and ammoniacal nitrogen, sediment load has been utilized. ML techniques such as ANN, XGBoost, SVM, CNN, ANFIS, RF have been implemented in the previous literature. These techniques have been found useful with the requirement of exploitation of other methods as well. The use of knowledge graphs can be promoted which can help in the integration of the various parameters and help in defining water scarcity as one entity. With the development and improving upon the urban water stress management can help in micro level planning at the city level to adapt to the water scarcity.