This data is the land cover map of Qaidam Basin in 2010 and 2020 obtained by using Landsat 8 and sentinel-2 remote sensing images, and taking a large number of field sampling data points as classification samples and using random forest classification algorithm. After land classification with obvious comparison with satellite images, it is found that the classification accuracy is reliable, and different land cover can be basically accurately divided (kappa coefficient is greater than 0.9).
Landsat 8 and sentinel-2 remote sensing images.
The first step, indoor preparation, mainly includes consulting relevant data to understand some characteristics of Qaidam Basin, so as to prepare for our follow-up work.
The second step is field sampling. In order to improve the accuracy of classification results, we conducted a large number of field sampling in the study area. The sample coverage is wide and the number is sufficient.
The third step is to use the random forest classification method for classification. In the process of classification, we not only use the sample data, but also add some key features such as texture and slope.
The fourth step, classification post-processing, mainly includes the removal of small spots.
Step 5: result quality inspection.
This data is a data set obtained by using a large number of field samples and random forest classification. At the same time, we also add slope, texture and other features in classification. We strictly control errors in data processing. Finally, the classification results showed excellent, and the kappa coefficient could reach more than 0.9. Moreover, we also compared with satellite observation data and found that most land cover types were consistent.
| # | number | name | type |
| 1 | 2018YFC0406600 | National key R & D plan |
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.ncdc.Hydro.db2441.2022
10.12072/ncdc.Hydro.db2441.2022