The products during the flood season are sourced from GRACE land water storage and extreme precipitation. We used GRACE land water storage precipitation data, combined with high-frequency filtering, anomaly detection, and flood potential index methods, to successfully extract the global historical flood days from April 1, 2002 to August 31, 2016. We further compared and validated the data with Dartmouth Flood Observatory (DFO) data, Global Runoff Data Center (GRDC) flow data, news reports, and social media data. The results indicate that GRACE based flood days can cover 81% of flood events in the DFO database, 87% of flood events extracted by MODIS, and supplement many other flood events not recorded by DFO. In addition, compared with the flood events obtained from GRDC flow data, the detection probability of 261 watersheds is greater than or equal to 0.5, with a probability of 62%. These detection capabilities and results are both very good. We finally provide flood day products with a 1 ° spatial resolution covering the range of 60 ° S-60 ° N from April 1, 2002 to August 31, 2016. This study provides a data foundation for the mechanism analysis and attribution of global flood events.
Daily GRACE TWS: https://www.tugraz.at/institute/ifg/downloads/gravity-field-models/itsg-grace2018/ .
Precipitation data: This study used Global Precipitation Measurement (GPM) data to calculate extreme precipitation.
The Dartmouth Flood Observatory: DFO dataset records large-scale flood events from various news reports and government websites. It includes the start and end time of each flood, the country of occurrence, the approximate scope, the cause of the flood, and the degree of damage. It is a rare and useful product for studying global historical floods.
MODIS derived flood inundation data: The flood inundation data used in this study were collected from a total of 807 flood inundation data points recorded in the 60 ° S to 60 ° N latitude region between April 1, 2002 and August 31, 2016. This product uses atmospheric corrected Terra (MOD09GA/GQ) and Aqua 135 (MYD09GA/GQ) MODIS images and threshold analysis methods (including 3-day standard method, 2-day standard method, 3-day Otsu, and 2-day Otsu methods) as well as slope constraints (slopes greater than 5 ° are covered) to extract inundation at a spatial resolution of 250 m based on DFO recorded flood events. The extracted results were compared and validated with 30 meter resolution inundation data from Landsat 5, 7, and 8 images, and quality control analysis of flood maps was conducted.
GRDC Drainage Data: The Global Runoff Data Center is an international data center operated under the auspices of the World Meteorological Organization 145. The corresponding dataset is a flow product that records the average flow, country, longitude, latitude, and river name associated with each flood event. This study selected records from April 1, 2002 to August 31, 2016 as the validation dataset to validate the extracted floods. In order to facilitate comparison with GRACE derived flood days and eliminate the influence of random errors, we considered the spatial average of 150 from the HydroSHEDS Basin level 4 data, which includes flow measurement values, and converted it into flood events using the following statistical methods. These data can be obtained from https://www.bafg.de/GRDC/EN/Home/homepage_node.html get.
It mainly includes three parts: data preparation, flood day extraction, and result verification. The flood data extraction step uses daily precipitation and daily GRACE TWS data, while the flood validation step uses daily flow, DFO, MODIS derived flood inundation, and social media data. The second part of the flood extraction process is mainly based on using high-frequency TWS signals and flood potential indices to obtain preliminary possible flood days; Then, extreme precipitation constraints are used to obtain the final number of flood days. The third part of the flood verification process includes comparing the flood range extracted from DFO records and MODIS images, comparing it with the floods obtained from GRDC emission data, and finally comparing it with major flood events recorded on social media.
This study successfully extracted global flood days using GRACE TWS and extreme precipitation data from 60 ° S to 60 ° N latitude from April 1, 2002 to August 31, 2016. The results were compared with the flood events recorded by DFO in time and space, and the results showed that our detection not only identified 81% of the flood events recorded by DFO, but also supplemented a large number of flood events not recorded by DFO. To further verify the reliability of the derivative product, we compared it with flood events extracted from global GRDC emission data, and detected a probability of over 0.5 with a probability of 62%.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB6523.2024
10.5281/zenodo.6831384