Estimating Parameters in Biological Ocean Models Using an Emulator Approach

• Activity Number: 14
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Michael Dowd*+
• Companies: Dalhousie University
• Keywords: dynamic system ; data assimilation ; emulator ; parameter estimation ; oceanography

Abstract

Dynamical models for ocean biology contain many unknown or uncertain parameters. Estimation of these parameters using available data is important for scientific progress, and improvement in predictive skill. These dynamical models are based on numerical ocean circulation models, and are computationally expensive to run. Here, we consider emulators to facilitate data assimilation (i.e., state and parameter estimation). We present work wherein a simple emulator approach, based on the so-called polynomial chaos expansion, has been used to estimate two key biological parameters (the phytoplankton carbon/chlorophyll ratio, and zooplankton grazing rate), and also achieve estimates of uncertainty in the biological state. The dynamic model was based on the Regional Ocean Modelling System (ROMS), with the data being daily observations of the sea surface chlorophyll in the northwest Atlantic Ocean from the SeaWiFS satellite. Estimation was based on an image distance metric that compares model surface fields to the satellite imagery. Results for time-varying parameter estimates are presented, and prospects for the use of emulators in ocean modelling is discussed.