Mapping Paddy Lands using Remote Sensing Techniques and Land Use Land Cover Change Detection Over a Decade in Gampaha District, Sri Lanka

Rice production constitutes a critical component of food security across numerous Asian nations. However, rapid urbanization has significantly constrained the availability of agricultural land resources. This study evaluated the efficacy of remote sensing (RS) techniques for generating comprehensive paddy land maps across three distinct phenological stages—initial establishment, development, and harvesting—while simultaneously detecting land use land cover (LULC) transformations in Gampaha district, Sri Lanka, during the periods 2008/2009 and 2018/2019. Three remote sensing methodologies were systematically assessed: (1) supervised classification, (2) Tasselled Cap Transformation integrated with Normalized Difference Vegetation Index (NDVI), and (3) Land Surface Water Index (LSWI) combined with NDVI analysis. LULC change detection was conducted utilizing Landsat 5 and Landsat 8 satellite imagery processed through ArcGIS 10.8 software. Classification accuracy assessments revealed overall accuracies of 85.5% and 89.5% for supervised classification, 84.0% and 88.5% for Tasselled Cap-NDVI methodology, and 85.5% and 53.0% for LSWI-NDVI approach across the 2008/2009 and 2018/2019 periods, respectively. Results demonstrated that supervised classification proved most effective for identifying paddy fields during the harvesting stage, while Tasselled Cap-NDVI methodology exhibited superior performance during the development stage, and LSWI-NDVI analysis showed optimal results for initial stage detection. Temporal analysis revealed substantial LULC transformations during the study period, with built-up areas experiencing a 126.4% increase, concurrent with a dramatic 76.5% reduction in paddy cultivation areas. These findings underscore the severe pressure of urban expansion on agricultural land resources in the region.