DEVELOPING METHODS FOR WATER QUALITY MEASUREMENT : Using machine learning and remote sensing to predict absorbance with multispectral imaging

Water resources play an important role in society and fulfill various functions such as providing drinking water, supporting industrial production and enhancing the overall landscape. Water bodies, such as rivers and lakes, are particularly important in this context. However, as societies and economies develop, the demand for water increases significantly. This also leads to the release of domestic, agricultural and industrial wastewater, which often exceeds the self-purification capacity of water bodies. Consequently, rivers and lakes are getting more and more polluted, endangering the safety of drinking water and causing ecological damage, affecting human health and biodiversity. Water quality monitoring plays a crucial role in evaluating the state of water bodies. Traditional monitoring methods involve labor-intensive field sampling and expensive construction and maintenance of automatic stations. Although these methods provide accurate results, they are limited to specific sampling points and struggle to meet the demands of monitoring water quality across entire surfaces of rivers and lakes. This degree project aim at developing a method that can predict absorbance in water with the aim of remote sensing. Along with multispectral imaging and machine learning this work proves that this is possible. The result from multivariate analysis is an optimal model that can predict absorbance at 420 nm with RSQ of 0,996 and RMSE of 0,00081.