This dataset is the first vectorized dataset of silt soil formed by inspection dams on CLP, combining high-resolution and easily accessible Google Earth images with object-based classification methods. This not only provides basic information for accurately assessing the ecosystem service functions of water retaining dams, but also helps to interpret the current changes in sediment transport in the Yellow River and plan future soil and water conservation projects.
The vectorized dataset format for dam sites is shapefile (. shp), where each record is depicted as a face and includes subsequent attributes such as longitude, latitude, dam site area, dam site perimeter, sediment quantity, and sediment quality. In the attribute table, the fields Area (unit: m2) and Shape_Leng (unit: m) represent the area and perimeter of the dam surface, calculated in WGS1984-EASE-GrideGlobal coordinates; The field volume (unit: m2) and mass (unit: m) represent the sediment retention and mass of the inspection dam.
This dataset is generated based on all available images of the Loess Plateau in China from May 2016 to 2020 on Google Earth.
In this study, we combined high-resolution and easily accessible Google Earth images and used object-based classification methods to create the first vectorized dataset of the Central Plains River Valley Barrage. We first investigated and analyzed the main features of the dam, and obtained Google Earth images with a resolution of 0.3-1.0 meters for the optimal extraction period. Then, we obtained a rough layer of the dam using methods such as multi-scale segmentation, threshold classification, and river network overlay. Finally, utilizing self-developed human-computer interaction programs, combined with auxiliary data, visual interpretation, and expert knowledge, the classification accuracy of the dam was improved.
The dam strata of these two regions were summarized and their accuracy was verified through field investigations and validation samples obtained from Google Earth. The accuracy of the dataset was verified through 1947 collected experimental samples, with producer accuracy and user accuracy of 88.9% and 99.5%, respectively.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB6470.2024
10.5281/zenodo.7857443