Use of Remote Sensing for Monitoring Irish Water Quality

The Water Framework Directive (WFD) is the main legislation regulating water management in Europe, with the principal purpose of protecting and enhancing ‘good ecological statuses’ of all waters. Therefore, there is a need for effective and large coverage monitoring of water bodies to support the decision making of such directives. Remote sensing offers a cost-effective way to meet such need and has been used globally to alleviate the limitation of conventional method of water quality monitoring. However, the use of remote sensing for water quality monitoring still has significant challenges (i.e. low correlation efficiency; not applicable in different environmental settings etc.,.) and needs to be explored further. Especially, the usefulness of remote sensing for small water bodies has rarely been explored in previous research. Therefore, this thesis is aiming to address the knowledge gaps concerning the use of remote sensing techniques for small water bodies, with the primary focus of Irish water bodies. The efficacy of remote sensing to detect Irish water bodies was initially evaluated. Systematic comparison was conducted between satellite-derived maps and the existing water maps. Results show that both Landsat-8 and Sentinel-2 have potential in assessing water quality across various water types in Ireland. The relationship between chlorophyll and Sentinel-2 remote sensing reflectance was then investigated. Models were built using different zoning classification methods and the results underscored the considerable impact the zoning methods have on accuracy. Three regional-scale methods (k-means clustering, river basin district, catchment) and a national scale method were tested. Results highlight the superior performance (higher R2 and lower RMSE) of the regional-scale approaches. K-means clustering emerged as the most stable with the highest accuracy. Additionally, the results show that regions with higher trophic state (eutrophic) waters consistently exhibited superior performances. Finally, to solve the problems of discontinuous monitoring hindered by single-sensor measurements, a multi-platform approach model was then built based on four optical satellites (Sentinel-2, Landsat-8, MODIS Terra and Aqua) spanning from 2016 to 2022. Results highlighted the advantages of using our multi-sensor model, which can provide 550% times the number of water quality estimates compared with in-situ measurement alone. Moreover, higher accuracies were also observed through the dataset, with average accuracies for oligotrophic, mesotrophic, and eutrophic of 79.7%, 70.3%, and 66.6%, found respectively. From the spatial analysis of our multi-platform approach model, we found that most of the eutrophic cases were from North Western, Neagh Bann River Basin Districts, this was consistent with the in-situ measurement spatial variation. Our multi-platform approach model discovered that while the average percentage of eutrophic cases across Ireland decreased from 25.50% in 2017 to 18.68% in 2021, there is an increasing trend from 2021 to July 2022.