Comparison of Landsat 8, Sentinel-2 and spectral indices combinations for Google Earth Engine-based land use mapping in the Johor River Basin, Malaysia
Accurate land use information is the basis for scientific research related to carbon cycle analysis, hydro-climatic modelling, soil degradation assessment, etc. It is also an indispensable basic information for local land management departments in land use planning and management. With development i...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Penerbit Universiti Kebangsaan Malaysia
2021
|
| Online Access: | http://journalarticle.ukm.my/18101/1/47689-164856-1-PB.pdf http://journalarticle.ukm.my/18101/ https://ejournal.ukm.my/gmjss/issue/view/1418 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate land use information is the basis for scientific research related to carbon cycle analysis, hydro-climatic modelling, soil degradation assessment, etc. It is also an indispensable basic information for local land management departments in land use planning and management. With development in big data and internet network, Google Earth Engine (GEE), a cloud-based computing platform allows users to perform satellite images processing more efficiently. This study aims to improve the land use mapping in a tropical region based on the GEE platform. Seven satellite images and indices combinations include Landsat 8 (C1), Sentinel-2 (C2), Landsat 8+Sentinel-2 (C3), Landsat 8+Indices (C4), Sentinel-2+Indices (C5), Landsat 8+Sentinel-2+Indices (C6), Normalized Difference Vegetation Index (NDVI)+Normalized Difference Water Index (NDWI)+Enhanced Vegetation Index (EVI)+Elevation (C7) were developed to evaluate the best combination for land use mapping in the Johor River Basin (JRB), Malaysia. The Random Forest (RF) algorithm was used to classify the land use land cover (LULC) with 222 training samples and 78 verification samples obtained through the Google Earth Pro higher resolution satellite images and field samplings. The results show that the overall accuracies of all the seven combinations are mostly more than 75%, ranging from 72% (C1) to 86% (C6). The findings show that adding of additional indices information before the land use classification helps to increase the overall accuracy significantly. For instance, the overall accuracy of the C6 combination is 14% higher than that of the Landsat 8 image solely. This study can act as a reference to effectively improve the land use mapping in cloud-prone tropical regions. |
|---|