The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLVIII-M-1-2023
https://doi.org/10.5194/isprs-archives-XLVIII-M-1-2023-361-2023
https://doi.org/10.5194/isprs-archives-XLVIII-M-1-2023-361-2023
21 Apr 2023
 | 21 Apr 2023

THE SPATIAL DISTRIBUTION OF SELECTED OPTICAL ACTIVE COMPONENTS IN THE GULF OF IZMIT USING BIVARIATE/MULTIVARIATE REGRESSION ANALYSIS

F. Sunar, A. Dervişoğlu, N. Yağmur, E. Aslan, and H. Atabay

Keywords: Water Quality, Gulf of Izmit, Sentinel-2, Bivariate/Multivariate Regression Models, Spatial Mapping

Abstract. Compared to traditional field-based (in-situ sampling) measurements, satellite-based remote sensing is an accurate, timely and cost-effective approach to monitor the dynamics of water bodies using images at different spatial and temporal scales. With satellite-based remote sensing techniques, qualitative measurements obtained directly from satellite images are integrated with in-situ measurements, enabling the establishment of spectral statistical relationships between satellite data and water quality physical indicators such as suspended solids, turbidity and chlorophyll-a. In this study, the spatial distribution of three water quality parameters (Chlorophyll-a (Chl-a), Secchi disk and Conductivity (EC)) which are optical active components (OAC) in the Gulf of Izmit were evaluated using in-situ water quality measurements, together with both field-spectroradiometer measurements and Sentinel-2 satellite imagery. In-situ water quality and field-spectroradiometer measurements were collected at the same date with the satellite overpass. Bivariate and multivariate regression models were established to analyse the correlation of in-situ water quality measurements with two different measurement datasets (i.e. satellite and spectroradiometer), and then the results were evaluated with two accuracy metrics Root Mean Square Error (RMSE) and Mean Absolute Error (MAE); and compared visually with the spatial distribution maps of the three water quality parameters generated by the ordinary Kriging interpolation method.