Detection, monitoring and temporal analysis of cyanobacteria by remote sensing techniques for Lake Bafa and evaluation of relationship with loads reaching the lake

Cyanobacterial harmful algal blooms (CyanoHABs) seriously damaging effects on human and environmental health. Factors that ultimately lead to the formation of different types of algal blooms in inland waters have been investigated for years. Mounting evidence indicates that global climate change, uncontrolled urbanization, increased nutrient loading and irregular flow regimes of freshwater systems are contributing to the increased frequency, severity, extent, and broader geographic distribution of CyanoHABs. Monitoring and estimating pigment concentrations in water bodies has a critical role in the search for early intervention or prevention methods. Traditional monitoring techniques, although highly accurate, are vastly insufficient in terms of spatial and temporal coverage. Unlike the traditional method, remote sensing-based methods have excellent temporal and spatial coverage and are ideal for developing and data-deficient regions. Located in the Mediterranean region, which is the hot spot of climate change according to the United Nations Environment Programme (UNEP), Lake Bafa has been struggling with CyanoHAB and pollution problems for years. The lake is one of the most important wetlands of Turkey and is trying to be protected with hydrological projects. Lake Bafa is located in the Büyük Menderes River Basin (BMN), an important river basin in terms of water management. The BMN basin is a hydrologically highly regulated basin with dense dams and irrigation structures and is also under pressure from agricultural and industrial pollution. This thesis focuses on monitoring and investigating the cause of CyanoHABs in Bafa, a shallow alluvial barrier lake formed by the BMN. A Random Forest Chlorophyll-a (RFchl-a) pigment quantification model was calibrated and validated with a root mean square error (RMSE) of 18.10 and 14.25 µg/L and a normalized percent root mean square error (%NRMSE) of %15.8 and %29.2 respectively, from measured in-situ Chl-a values between 2013, 2014, 2018, and 2019 using Rrc RED, Blue/Green, Rrc SWIR and Floating Algae Index (FAI). A time series of remotely estimated RFchl-a was developed from 2013 to 2019 Landsat-8 OLI sensor data. CyanoHABs drivers related to environmental factors such as nutrient, sediment, hydrological and meteorological outputs were obtained from a detailed non-point source Soil and Water Assessment Tool (SWAT) model. It was calibrated for 2010-2013 and validated for 2014-2019 years. The results for the flow data ranged from R2 (coefficient determination) 0.64-0.92, NSE (Nash-Sutcliffe Efficiency) 0.63-82, and R2 0.57-0.97, NSE 0.54-0.92; for the sediment data, R2 0.55-0.82, NSE 0.33-0.67, and R2 0.66-0.81, NSE 0.10-0.79; for the nutrient data, R2 0.59-0.79, NSE -1.55 - 0.50, and R2 0.40-0.77, NSE -0.51-0.71 obtained for the calibration and the validation period consequently. The evaluation of the SWAT model and RFchl-a time series showed that the parameters affecting RFchl-a values are seasonal. A significant positive correlation (Pearson correlation coefficient) was observed between RFchl-a and the most significant parameters of evapotranspiration (0.603), soil water content (0.592), and reservoir volume (0.550) for the dry season, percolation (0.701), precipitation (0.699), and surface runoff (0.659) for the wet season. Outputs of this study revealed that combining Landsat-8 OLI data and the SWAT model has the potential to analyze the drivers of CyanoHABs spatio-temporal variability of inland water bodies in a basin scale.