clustering, case study, data-driven methodology, unsupervised learning, water monitoring, water quality, Agriculture (General), S1-972, Engineering (General). Civil engineering (General), TA1-2040

Several factors contribute to the increase in irrigation demand: population growth, demand for higher value-added products, and the impacts of climate change, among others. High-quality water is essential for irrigation, so knowledge of water quality is critical. Additionally, water use in agriculture has been increasing in the last decades. Lack of water quality can cause drip clog, a lack of application uniformity, cross-contamination, and direct and indirect impacts on plants and soil. Currently, there is a need for more automated methods for evaluating and monitoring water quality for irrigation purposes, considering different aspects, from impacts on soil to impacts on irrigation systems. This work proposes a data-driven method to address this gap and implemented it in a case study in the PCJ river basin in Brazil. The methodology contains nine components and considers the main steps of the data lifecycle and the traditional machine learning workflow, allowing for automated knowledge extraction and providing important information for improving decision making. The case study illustrates the use of the methodology, highlighting its main advantages and challenges. Clustering different scenarios in three hydrological years (high, average, and lower streamflows) and considering different inputs (soil-related metrics, irrigation system-related metrics, and all metrics) helped generate new insights into the area that would not be easily obtained using traditional methods.