Chlorophyll-a (Chl-a) serves as a primary indicator of phytoplankton abundance and plays a central role in assessing coastal water quality and eutrophication processes. As an optically active substance, its concentration modulates water-leaving radiance, enabling retrieval from multispectral satellite observations. Conventional remote sensing models typically rely solely on spectral information derived from reflectance bands. This study extends this approach by integrating spectral information with spatial context using Geographically Weighted Regression (GWR). Nevertheless, standard GWR faces practical constraints such as high computational cost and challenges in conducting unbiased performance evaluations across heterogeneous areas. To address these limitations, we introduce a Station-Centric GWR (SCGWR) framework. SCGWR constructs separate locally calibrated regression models centred on each <em>in-situ</em> sampling station, thereby capturing site-specific relationships between Sentinel-2 reflectance bands and measured Chl-a concentrations. The performance of SCGWR was evaluated in the Gulf of İzmir using Sentinel-2A surface reflectance data together with synchronous <em>in-situ</em> Chl-a observations. Comparative analysis with Multiple Linear Regression (MLR) showed that SCGWR provides improved prediction accuracy while better representing local spatial variability. In particular, SCGWR produced a lower test RMSE (6.26) compared to MLR (6.95), together with higher correlation and concordance coefficients. Visual inspection of the predicted maps further indicated that SCGWR generates spatially coherent Chl-a patterns that correspond more closely with<em> in-situ</em> measurements. These results demonstrate the potential of the SCGWR framework as a reliable spatial modelling approach for improving satellite-based monitoring of coastal eutrophication and water quality dynamics.