Estimation of optically active freshwater substances using high resolution satellite sensors for Norway

With a changing climate altering the composition of substances in freshwater the need for a frequent and cost-effective monitoring method have arisen. To solve this issue remote sensing using high resolution satellite sensors were applied. Based on their high spatial, temporal and radiometric resolution the Sentinel-2 Multispectral Imager (MSI) and PlanetScope constellation were chosen. The three most optically active substances, namely, chlorophyll a (Chla), total suspended matter (TSM) and coloured dissolved organic matter (CDOM) were to be estimated. Due to the lack of in-situ measurements of CDOM the substance known as dissolved organic carbon (DOC) was used instead. Chla and TSM were only included to measure the effect it could have on the water leaving reflectance of CDOM/DOC. Norway was selected as a suitable study area since an increase in DOC had been observed. Changes in precipitation, higher temperatures and precipitation acidity are factors that are suspected to cause this increase. Based on promising results from previous studies an estimation approach using empirically derived band ratios were selected. With a total number of 2409 DOC in-situ measurements for the study area only 17 were suitable for regression modelling. A set of strict criteria were the main reason for dismissing such a large proportion of data. The results showed that Chla in the range from 0.001 – 0.1 mg/l had no effect on DOC. TSM concentrations above 2.84 mg/l was found to have a significant impact on the DOC regression model. Taking that into consideration good statistical results for estimating DOC using the green/red band ratio was achieved. Top of the atmosphere (TOA) radiance was preferred since both atmospheric correction algorithms Sen2Cor and C2RCC had certain disadvantages making them unsuitable for small sized lakes and rivers. PlanetScope’s products was also dismissed due to a low signal to noise ratio (SNR). The TOA DOC regression model had a statistical result with a correlation of -0.95, R 2 of 0.9, p-value of 5.8e-09 and a standard error of 1.5 mg/l. However, a limitation between the in-situ measurements and the spatial resolution of the MSI led to a reduced range for the TOA DOC model. These results show that the MSI has a sufficient potential for DOC estimation. PlanetScope is in an early development stage making it possibly a viable alternative in the future. This study also suggests what future work on this field should take into consideration to get good in-situ measurements for remote sensing use.