Quasi-Coupled Wind-Wave Data Assimilation in Extreme Events

Abstract

Extreme weather events are costly to society - causing loss of life, property damage, and disruption of local economies. In the US a,lone, economic damages associated with extreme weather have exceeded 1.8 trillion dollars since 1980 (NCEI 2021). Improved understan,ding and predictability of these events is critical for many aspects of American life, including safety on land and at sea, mitigati,on of property damage in coastal areas, and the provision of accurate early warning forecasts to the public. Historically, the forec,ast skill of numerical models for extreme ocean weather, and ocean waves in particular, has been strongly constrained by the sparse, availability of high quality observational data to help guide the operational models. Only recently, with the introduction and cont,inued growth of global distributed drifting wave-sensor networks, has this oceanic data gap started to become smaller. However, curr,ent operational wave forecast systems are not ready to take advantage of these distributed observations, and new advances in Data As,similation (DA) strategies are needed to fully realize the potential provided by these networks. The objective of this project is to, demonstrate that the assimilation of dense distributed drifting sensor networks has the potential to result in dramatic improvement, in our ability to forecast extreme wave events. We intend to develop a wave DA system based on the Local Ensemble Transform Kalman,Filter that can effectively and efficiently use spectral wave observations to provide improved estimates of the wave state. We antic,ipate that such a wave-DA system alone will tremendously improve forecast skill for energetic swell waves However, improved forecast,s of locally generated waves will require updating the atmospheric forcing as well, and we will therefore develop an improved surfac,e wind estimate through post-processing of ensemble surface wind forecasts. We hypothesize that observed errors in the ensemble mean, wave model contains information on which (linear combination of) atmospheric forecasts is best representative of current and future, atmospheric conditions. We will train a deep convolutional neural network to estimate the best weighted ensemble mean based on obse,rved wave errors. This data-driven quasi-coupled forecast in turn will be used as forcing to produce an improved medium-range wave f,orecast.This project will result in the complete development, testing and validation of the first quasi-coupled wave-wind DA system,that ingests large numbers of heterogeneous drifting sensors to immediately improve both surface wind and wave forecasts in extreme,conditions. We will use data collected from aerial deployments of sensors in front of hurricanes and available from the Sofar sensor, network. The assimilation strategies will be implemented using operational models for direct validation against other operational p,roducts, specifically for extreme events.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 13, 2022
Source ID
N000142212394

Entities

People

  • Pieter Smit

Organizations

  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Environmental science

Readers

  • Distributed Systems and Data Platform Development
  • Economics
  • Ocean-Atmosphere Mesoscale Modeling, Data Assimilation, and Flux Boundary Layers

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference