ticketticket ticket ticket ticket

Summary of Abstract Submission

Abstract Submission No.ABS-03-0096
Title of AbstractCoastal storm surge and extreme sea level simulation using machine learning
AuthorsSreeraj P*
OrganisationINCOIS, Hyderabad
Hyderabad, Telengana, India
Pincode: 673620
Mobile: 9526644830
E-mail: s.puthiyadath-p@incois.gov.in
AbstractThere has been significant progress in understanding the mean sea level (MSL) change on both global and regional scales. Coastlines are also potentially prone to episodic high-intensity sea levels or extreme sea levels (ESLs) and coastal flooding; however, the understanding of these short-span (hours-days) events, driven by the combination of MSL, tides, storm surge, and waves is very limited. Only a few tide gauges of the global coastline have a long record of high frequency (minute-hours) sea and many tide gauges pose large data gaps. This limits a better understanding of storm surges along densely populated coastal belts. Historic storm surge estimation from the model-based studies and global reanalysis had considerable uncertainty due to the lack of proper validation and biases along the Indian Ocean coastlines. Here, we provided an AI/ML-based data-driven reconstruction (IOSSR) of high-frequency (hourly) storm surges along the Indian Ocean Coastline. Using the Long Short Term Memory (LSTM) neural network with predictors includes hourly MSLP gradient, 10m wind speed, and 2m air temperature, we reconstructed the high frequency (hourly) storm surge data along the TG locations in the Indian Ocean coastlines from 1979 to 2021. Validation of hourly surges and monthly extremes shows our data performing fairly well with a mean RMSE of less than 7 cm (<13% mean relative RMSE) and a mean correlation of 0.73. Simulation of extreme surges during cyclone events are outperforming presently available global surge reconstructions.
KeywordsExtreme sea level, storm surge,machine learning,coastal flooding,data reconstruction
For Awardsyes
Date Of Birth04-01-1995