ESTIMATION OF CHLOROPHYLL-A CONCENTRATIONS IN WETLANDS USING REMOTE SENSING.

Sammanfattning

Remote sensing techniques with an integration of machine learning algorithms is increasinglybeing tested as a method for water quality monitoring, particularly for the estimation ofChlorophyll-a concentrations. In this study, similar methods are used to estimate Chlorophyllafrom small-scale wetlands with open water in Sweden, where in-situ Chlorophyll-a field datahas been collected. Images captured by the drone and Sentinel 2A images were used to obtainwavelength bands (Blue, Green, Red, Red-edge and Near-infrared) data which was used asinput in the Random Forest models. This study compared the performance of random forestmodels developed using Drone imagery and Sentinel-2A multispectral imagery forChlorophyll-a estimation in two wetlands in Sweden. The random forest model clearlypredicted Chlorophyll-a from both Drone imagery (R²=0.845 training) and Sentinel 2Aimagery (R²= 0.903 training). Sentinel-2A model achieved higher training R² value than Droneimagery model, indicating stronger predictive performance. The cross-validation results forboth indicated substantially lower predictive ability (R² ≈ 0.02–0.05) indicating overfitting.Variable importance analysis revealed substantial differences between the two datasets. In theDrone Imagery model, the Red-edge band variable was identified as the most importantpredictor for Chlorophyll-a while Near-infrared variable was the most important for Sentinel2Amodel. Overall, the findings show that effectiveness of retrieving Chlorophyll-a by remotesensing using Random Forest models is possible, however it depends on the interactionbetween spatial resolution, spectral characteristics, and model structure.