As important parts of the "Belt and Road" construction, Central and West Asia and Northwest China are facing severe challenges posed by flood disasters to regional sustainable development. In view of the problems of single existing flood disaster information in the region and insufficient analysis and extraction, this study is based on radar images of typical floods in Central, West Asia and Northwest China during the "Belt and Road" period from 2017 to 2022. It extracts water body information before and after the disaster through object-oriented method, uses change monitoring to determine the flood inundation scope, and combines the results with optical images to verify the results. Finally, flood remote sensing images and interpretation datasets are obtained to make up for the incomplete existing flood disaster information. Inadequate details. This dataset includes historical flood disaster datasets and flood interpretation datasets: (1) The historical flood disaster datasets include typical historical flood disaster data in Central, West Asia and Northwest China of the Belt and Road Initiative, corresponding Excel data, and DEM data. (2) The flood interpretation data set includes SAR image data, optical image data, and interpretation and annotation data covering the flood inundation area. The overall flood distribution range is 22°42 '25 "-47°22' 52" N, 39°21 '54 "-101°17' 24" E, including 10 typical flood disaster samples. By introducing Landsat optical image-assisted verification and manual interpretation feedback optimization, the accuracy of flood inundation range extraction results has been effectively improved, and the proportion of correctly classified samples has stabilized to more than 85%, effectively reducing information omissions and misjudgments, and verifying the data set. Reliability and integrity. The data source of this dataset is clear and the data quality is strictly controlled. It can provide rich sample data for hydrological research, disaster loss and disaster risk assessment in relevant areas of Central, West Asia and Northwest China, and serve as data support for flood monitoring work.
| collect time | 2017/11/21 - 2022/01/19 |
|---|---|
| collect place | Belt and Road regions |
| data size | 12.7 GiB |
| data format | *.shp,*.tif,*.img,*.xls |
| Data spatial resolution (/ M) | 5m×20m |
| Coordinate system | WGS84 |
The data set includes a flood point data set and a flood interpretation data set. The flood point data set includes flood point data, corresponding Excel data, and DEM data. The flood interpretation data set includes SAR image data, optical image data, and interpretation annotation data. (1) Flood point data set: The original data of flood points comes from the global disaster data platform. According to key information such as flood disaster occurrence time and death toll, the point data is filtered, and finally 10 typical flood disaster points are selected to participate in sample production. The attribute table of each piece of data includes information such as disaster occurrence time, latitude and longitude location, and death toll. digital elevation model (Digital Elevation Model (DEM) data is derived from the SRTMDEM released by NASA;(2) Flood interpretation dataset: The sample data is based on Sentinel-1 radar images and adopts interference wide format Single-view complex number of 20 scenes elevated track under (Interferometric Wide, IW)(Single Look Complex (SLC) image, polarization method is VV, spatial resolution is 5m×20m, image time range covers 2017-2022, data source is ASF Data Search website, high spatial resolution Landsat optical remote sensing data obtained from the U.S. Geological Survey Data Center (USGS), based on the above data, In this study, ENVI and ArcGIS software were used to interpret and label flood disaster points.
Data processing is mainly divided into two parts: (1) Flood point data set: The original data of flood points comes from the global disaster data platform. Data with coordinate information are selected and organized into Excel files. Based on key information such as flood disaster occurrence time and death toll, 10 typical flood disaster points are selected, and the attribute table of each data includes disaster occurrence time, latitude and longitude location, death toll and other information;(2) Flood interpretation data set: Based on Sentinel-1 radar image data, ENVI software is used to extract water bodies from pre-disaster and post-disaster images respectively using object-oriented methods. Use ArcGIS software to monitor changes, eliminate areas with insignificant changes before and after the disaster, and obtain the flood inundation scope. Subsequently, the results were verified by combining optical remote sensing images corresponding to the flood range, and interpretation data sets of 10 flood disasters were finally produced.
This research strictly follows the designed technical route for product production. Data quality control mainly starts from two aspects: data source quality and interpretation data verification: (1) Data source quality control: The location information of flood points comes from the global disaster data platform. The data is accurate and reliable, and is put into the production of data sets after strict screening. The images before and after the flood disaster are all from the NASA official website. The same set of disaster data strictly follows the same imaging geometric parameters and sensors. The data standards of the configuration are selected to avoid the impact of different parameters on the interpretation results;(2) Interpretation data verification: In flood area detection, optical remote sensing can provide rich spectral information and high image resolution, while radar remote sensing has all-weather flood area detection capabilities. Taking into account the rich spectral information of optical images, based on the interpretation results obtained based on the object-oriented method and change monitoring, the optical images after the flood disaster were selected for result verification. The optical images were greatly affected by cloud cover. In this study, the image with the least cloud cover was obtained by screening, and the image with the time closest to the disaster occurred was selected while meeting the cloud cover conditions. After processing, the interpretation data was tested to verify the reliability of the results. To sum up, this study ensured the quality of data sources through layer screening, relied on a large number of documents as a theoretical basis, determined scientific research methods, selected optical images with high image resolution to verify and analyze the research results, and improved the extraction accuracy., providing important data support for flood disaster monitoring.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
| # | title | file size |
|---|---|---|
| 1 | Data.zip | 12.7 GiB |
Belt and Road SAR flood object-oriented method change monitoring interpretation dataset
laAiDX
P2dOKZ4I
©Copyright 2005-. Northwest Institute of Eco-Environment and Resources, CAS.
Donggang West Road 320, Lanzhou, Gansu, China (730000)

