Water quality monitoring and assessment of the Northern Baltic Sea using Earth Observation

This thesis concentrates on determining how accurate Earth Observation (EO), i.e. satellite instrument observation methods are for water quality estimation in the coastal waters of Finland, in the Northern Baltic Sea. Water quality refers to the characteristics of water that define its chemical, physical or biological properties. These are used to assess the ecological and chemical status of the waters based on various environmental regulations, such as the European Union (EU) Water Framework Directive (WFD). In water areas, where the quality is not good, such as in eutrophicated sea areas, the water quality parameters have high spatial and temporal variability. EO observations capability to cover this variability is beyond the possibilities of conventional station-wise water sampling. This is valuable in Finland with numerous coastal and lake water bodies. In this thesis, phytoplankton chlorophyll-a (chl-a) – as a direct indicator of eutrophication and one of the WFD's biological quality elements – is the key water quality parameter. Water quality parameters derived from one airborne and two satellite instruments algorithms are analysed against two types of field measurements: automated flow-through measurement systems and monitoring station water samples. The flow-through systems, either installed on smaller coastal boats or on board merchant ships (ferrybox), provide an efficient way for collecting a wide range of ground truth samples for EO algorithm development and validation. The routine water samples are collected at the station sites throughout the coastal waters of Finland and are used as the basis for the status assessment of water bodies. Finnish coastal waters represent an optically extreme and complex water type: absorption dominated humic waters, where EO algorithms often produce incorrect results. A Quality Grade (QG) method confirmed that the analysed EO model to determine chl-a concentrations is reliable for 62% (80 of the examined 129) coastal water bodies. Summarising results over the coastal monitoring sites confirm that the accuracy of the EO method is close to the determination accuracy of station sampling (0.6 μg l-1) near the chl-a target for WFD. The difference in statistical metrics calculated by EO and station sampling is 23% – well within the uncertainty limits of chl-a laboratory analyses. The statistical measures (mode and the geometric mean) are optimal for non-normally distributed EO observations for utilising them in the water bodies for assessment. Only in the optically specific estuaries (2.8 % of the examined area 7 out of 129 water bodies), the optical properties of the water can reach the extremes of their ranges and the chl-a estimation fails. This occurs particularly when the loading from the drainage basin is intensive (such as in the melting season in spring and after heavy rains). Moreover, EO algorithms for determining turbidity, absorption of coloured dissolved organic matter and Secchi disk depth were analysed and found to perform well. The work done with past EO instruments – medium resolution satellite and high resolution airborne - anticipated the efficient use of the EU Copernicus programme Sentinel-series instruments for the monitoring of the water areas of Finland.