Machine Learning for Planning Water Supply Infrastructure in the Face of Climate Change

The evaluation of risks and assessment of developments in drinking water supply under the influence of climate change requires reliable predictions of future per capita drinking water demand and water supply over longer periods of time. It is obvious that the consumption of drinking water and water supply is strongly dependent on various environmental aspects, such as temperature, precipitation, number of summers and hot days that are directly influenced by climate change. There is currently no established mathematical model that can make reliable predictions based on the limited available measurement data and sociological data surveys. Therefore, the declared goal of our research project is to develop a machine learning (ML) strategy that enables the generation of ML models that allow reliable predictions of water demand and water supply in the context of climate change.

 

One of the special features of this challenge for ML strategies is the relatively small number of training samples per individual local water supplier. This is unavoidable as a significant number of local water suppliers only record consumption and supply on a monthly basis and often do not have continuous monitoring of their own database. This local data scarcity requires ML models specialized in data-poor applications.

 

A relatively high number of local training datasets to be used for this task makes the automation of the training process particularly attractive. For some years now, the concept of automated ML (AutoML) has enjoyed increasing popularity. AutoML focuses on the automatic selection of models, the selection of model components, the hyperparameter tuning of models and the evaluation of models. We intend to develop an AutoML strategy for data-poor applications that focuses on predicting water demand and water supply as a function of climate parameters. Using the ML models produced by AutoML, it would be possible to make individual and ensemble predictions of drinking water consumption and drinking water supply at each of the local water suppliers. Data from water suppliers throughout southern Germany will be used to train the models. The result of this work would make it possible to provide each local water supplier with an early warning system with the aim of defining specific measures for better management of water resource use in good time.