The Land Use and Land Cover (LULC) segmentation represents a critical challenge in environmental monitoring and sustainable development. Traditional Convolutional Neural Networks (CNNs) excel at local pattern recognition but struggle with comprehensive spatial understanding, while transformer architectures capture global contexts at significant computational expense. This research introduces an innovative hybrid model that strategically combines the strengths of CNNs and vision transformers. We develop a segmentation approach that transcends existing methodological limitations by efficiently extracting local features through CNNs and leveraging transformers' ability to comprehend long-range dependencies. The proposed framework achieves a high accuracy of nearly 95 % and a mean Intersection over Union (IoU) of nearly 91% with reduced computational complexity, making advanced geospatial analysis more accessible. This approach advances technical capabilities and empowers researchers and policymakers with precise, timely insights into landscape dynamics, enabling more informed environmental decision-making across diverse geographical contexts.