Detecting abandoned farmland is crucial for effective agricultural land management. By leveraging satellite imagery, and using temporal and textural features, machine learning studies have identified such lands. However, demographic and locational factors, though recognized as key drivers of farmland abandonment, have not been used as classification attributes. Our research employs Support Vector Machines and Random Forests, integrating these socio-environmental factors, achieves accuracies of 89.2 % and 88.5 % for overall and user classifications, respectively. We discovered that factors such as parcel size, slope, and road proximity are more critical than other conventional features in identifying abandoned farmland, recommending further investigation into these socio-environmental factors to uncover more impactful classification attributes. Additionally, our findings contrast with other studies by demonstrating that both higher and lower harmonic components of NDVI and NDWI enhance classification accuracy. This research could greatly assist in agricultural land surveys and in developing policies for sustainable farmland preservation.