Abstract's details
Adapting ocean prediction systems to observations
Event: 2023 Ocean Surface Topography Science Team Meeting
Session: Science III: Mesoscale and sub-mesoscale oceanography
Presentation type: Forum only
Adapting ocean prediction systems to observations
Authors: Gregg Jacobs1, Brent Bartels2, Joseph D’Addezio1, Chris DeHaan2
Affiliations:
1Naval Research Laboratory, Ocean Dynamics and Prediction, MS, USA
2 Peraton, MS, USA
Observing system coverage changes regularly. This includes the primary satellite observing systems returning sea surface height and sea surface temperature as well as in situ observing systems including profiling floats and uncrewed gliders and surface vehicles. Ocean forecast systems use these observations to compute corrections on a regular schedule. Typically, the data assimilation process does not consider the observation distribution and changes over time. During the S-MODE pilot campaign, we demonstrated a reduction in horizontal assimilation correlation scales in the vicinity of locally high resolution observations leads to predictive skill at smaller scales. However, the adaptive assimilation was fixed in time as the S-MODE deployments were consistently in an area over time.
To extend the process further, we developed an approach to estimate the scales resolved by the observations provided to the data assimilation process prior to each model forecast. This process occurs daily. Data sources are treated separately so that sea surface height and profile data provide one set of scales resolved around the globe while sea surface temperature provides a different set of scales. In the assimilation process, the scales estimated from the satellite sea surface height and in situ profiles are the control on the horizontal decorrelation scale in the data assimilation.
The regular scale estimation in the assimilation process was tested during the 2023 S-MODE campaign and subsequent SWOT cal / val in neighboring areas. Prediction experiments included all regular satellite observations as well as the in situ observations from the two campaigns. Two identical ocean forecast systems were run over the time. One system used the regular assimilation process, and the second used the horizontal decorrelation scale updated prior to the assimilation. Comparison to the in situ observations prior to assimilation shows the improved skill in the system adapting scales to the observations. The results provide demonstration that the automated estimation of scales leads to advancement in ocean prediction capability.
Back to the list of abstractAuthors: Gregg Jacobs1, Brent Bartels2, Joseph D’Addezio1, Chris DeHaan2
Affiliations:
1Naval Research Laboratory, Ocean Dynamics and Prediction, MS, USA
2 Peraton, MS, USA
Observing system coverage changes regularly. This includes the primary satellite observing systems returning sea surface height and sea surface temperature as well as in situ observing systems including profiling floats and uncrewed gliders and surface vehicles. Ocean forecast systems use these observations to compute corrections on a regular schedule. Typically, the data assimilation process does not consider the observation distribution and changes over time. During the S-MODE pilot campaign, we demonstrated a reduction in horizontal assimilation correlation scales in the vicinity of locally high resolution observations leads to predictive skill at smaller scales. However, the adaptive assimilation was fixed in time as the S-MODE deployments were consistently in an area over time.
To extend the process further, we developed an approach to estimate the scales resolved by the observations provided to the data assimilation process prior to each model forecast. This process occurs daily. Data sources are treated separately so that sea surface height and profile data provide one set of scales resolved around the globe while sea surface temperature provides a different set of scales. In the assimilation process, the scales estimated from the satellite sea surface height and in situ profiles are the control on the horizontal decorrelation scale in the data assimilation.
The regular scale estimation in the assimilation process was tested during the 2023 S-MODE campaign and subsequent SWOT cal / val in neighboring areas. Prediction experiments included all regular satellite observations as well as the in situ observations from the two campaigns. Two identical ocean forecast systems were run over the time. One system used the regular assimilation process, and the second used the horizontal decorrelation scale updated prior to the assimilation. Comparison to the in situ observations prior to assimilation shows the improved skill in the system adapting scales to the observations. The results provide demonstration that the automated estimation of scales leads to advancement in ocean prediction capability.