110 likes | 212 Vues
This study by Zhang, Wilkin, Levin, and Arango explores the use of sensitivity analysis with ADROMS to improve ocean forecast and measurement along the New Jersey coast. The study focuses on adjoint sensitivity analysis to understand the causes of certain events in the area. Results show physically reasonable outcomes in idealized scenarios. Users are advised to carefully consider model linearization.
E N D
Sensitivity Analysis of SST along NJ coast with ADROMS Weifeng (Gordon) Zhang John Wilkin Julia Levin Hernan Arango Institute of Marine and Coastal Sciences, Rutgers University Oct. 2005
Background • Plenty of measurement in New Jersey shore area; • Ocean forecast and hind-cast with forward ROMS has been running; • To use data assimilation in these area to improve the forecast and measurement; • Adjoint sensitivity analysis with ADROMS is applied first to test the model based on our understanding of the physics in this area. • Adjoint sensitivity analysis: What causes the event?
Forward Model–Background for Adjoint model Three idealized cases (steady low river discharge, no tide, no surface exchange): No wind Southwestward wind Northeastward wind
Adjoint sensitivity set-up For upwelling event: Background state: 5 days bi-hourly forward model output. Event time: Adjoint forcing: Adjoint variable: FORWARD_MIXING: Adjoint model reads mixing coefficients from forward output
Summary • ADROMS gives physically reasonable results for sensitivity analysis on idealized application in NJ shore area • Users need to think about the definition of the measure of certain event, J • The validity of linearization to the nonlinear model has to be considered carefully and proved first ______________________________________________________ Great appreciation goes to John Wilkin, Hernan Arango, Julia Levin, Andrew Moore and B-J Choi!