The outputs of the Urban Flash Flooding in England project are now available via the @climatenodeuk website. The project used #NLP to analyse 17000+ articles from ~300 websites to detect events and map flood impacts for 56 key dates since 2010.
https://www.climatenode.org/maps/about_UFF_maps.html
Why do this? We have little historic data on surface water flooding. As far as I'm aware, this is the first attempt to create a detailed geospatial dataset for these events. Such data has potential applications in validating flood forecasts and models, as well as planning.
The outputs are:
• A set of 56 downloadable interactive maps of flooding impacts on each of the key dates identified
• geoJSON files containing the data for all maps, and each map individually
• A csv file providing details of all of the urban flash flood events detected, including those which weren't mapped due to a lack of geographical detail
Data is free and publicly available
Users can explore individual events, e.g. this is the map for the severe flash flood event which happened in London on 25th July 2021
https://www.climatenode.org/maps/UFFE_map_20210725.html
Please have an explore if of interest and feel free to drop me any questions. I would be really interested to hear from anyone who is likely to use this data to learn more about how they will use it.
The project used two distinct NLP techniques. Text classification was used to detect potential urban flash flood events from articles and distinguish them from other types of flooding. Named Entity Recognition was used to extract the names of places, streets, buildings, etc.