Accuracy analysis of mapping land use and occupation using Sentinel-2 and CBERS-4 images in the surroundings of a reservoirs
Detecting changes in land cover helps policymakers understand the dynamics of environmental changes to ensure sustainable development in the Caatinga biome. Thus, the identification of spatial characteristics by Remote Sensing has emerged as an important aspect of research, and, therefore, adequate...
| Autores: | , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2024 |
| País: | Brasil |
| Institución: | Universidade Federal de Santa Maria (UFSM) |
| Repositorio: | Revista Ciência e Natura (Online) |
| Idioma: | portugués |
| OAI Identifier: | oai:ojs.pkp.sfu.ca:article/84730 |
| Acceso en línea: | https://periodicos.ufsm.br/cienciaenatura/article/view/84730 |
| Access Level: | acceso abierto |
| Palabra clave: | Sensoriamento Remoto Aprendizagem de Máquinas Reservatório Pedro Mauro Junior Remote Sensing Machine Learning Pedro Mauro Junior Reservoir |
| Sumario: | Detecting changes in land cover helps policymakers understand the dynamics of environmental changes to ensure sustainable development in the Caatinga biome. Thus, the identification of spatial characteristics by Remote Sensing has emerged as an important aspect of research, and, therefore, adequate and efficient methodology for mapping the necessary land cover is a preponderant factor. In this study, data from the Sentinel-2 and CBERS-4 satellites captured by the MultiSpectral Instrument (MSI) and Panchromatic and Multispectral Camera (PAN) sensors, respectively, were used for classification and accuracy analysis for five land cover classes around dams located in the municipality of Belo Jardim, Pernambuco. The KNN algorithm (K-th nearest neighbor) with a value of k=1 was used for image training and classification. Recent high-resolution images from the European program WorldCover were used as a spatial and thematic reference image. After the Contingency Matrix analysis between the land cover maps and the reference data, an overall accuracy of 57.4% was obtained for the MSI and 54.5% for the PAN product. The results obtained showed that the MSI presented more satisfactory land cover maps than the PAN data. On the other hand, for the shrubby vegetation class, the PAN product presented an r of 0.5, while the MSI had an r of 0.47. Spatial and spectral characteristics of the images were the main causes of the variability found in the thematic accuracy coefficients. |
|---|