Investigating Natural Resource Quantitative Retrieval Utilizing Sentinel-2A Satellite Imagery

This research utilized Sentinel-2A satellite imagery to develop a quantitative remote sensing inversion model for natural resources, employing a decision tree classification approach and various indices such as MNDVI. The land use of BD City in 2020 was extracted and evaluated for accuracy. The findings indicated an overall inversion accuracy of 95.65% and a Kappa coefficient of 0.9461. The average mapping accuracy was 93.13%, with an average user accuracy of 92.99%. The average omission error and commission error were 6.87% and 7.01%, respectively, demonstrating high classification precision. Indices like MNDWI, NDVI, MRENDVI, and PCA3 were instrumental in enhancing inversion accuracy. Forests and grasslands covered the largest area, approximately 62.56% of BD City's total area. The second largest was arable land, accounting for 24.32%. Artificial structures made up about 12.30%. Other land types were less than 1%. The distribution was consistent with the Third National Land Survey. The study's results are beneficial for natural resource management in BD City and provide valuable data extraction methods and experience, offering insights for resource monitoring in other regions.