Friday, 9 October 2015
The Met Office ocean analysis
by Laura Jackson, Chris Roberts and Drew Peterson, Met Office
At the Met Office we've been examining how well our different products (ocean analysis, decadal prediction and various density and temperature metrics) compare with the observed AMOC variability. In particular we've found that an ocean analysis developed from our Forecast Ocean Assimilation Model (FOAM) and used in our seasonal forecasting model (GloSea) is exceptionally good at reproducing the recent AMOC changes.
The figure shows the period from April 2004, when the RAPID data starts. Each panel shows the existing observed data from RAPID (red) alongside values from the GloSea analysis (blue). As well as the actual AMOC strength we have also calculated the AMOC in the same way as RAPID by calculating the components from the Florida Straits flow (the northwards flow between mainland USA and the Bahamas), Ekman flow (driven by the wind) and Upper Mid Ocean (flow throughout the remainder of the Atlantic from geostrophic interior flow and velocities in the western boundary region).
The Ekman flow (not shown) is identical since observational winds are used to drive the analysis. The Florida Straits (FS) and Upper mid ocean (UMO) components don't match individually as well as their sum, suggesting that although the FS and UMO components are not perfect, the analysis as a whole is doing a good job of constraining the combined sum of transports through the FS and Antilles currents. It is possible that a higher resolution might improve the representation of flows in the shelf region.
Since our model compares so well with the observations, we wanted to use it to predict what the observations being collected at the moment might show, and to enter the survey for the MOC strength from October 2014-September 2015. Unfortunately our analysis finishes in May 2015, so we have tried to estimate the last few months of data (see below for method).
The result of this analysis suggests an increase of the AMOC from Oct 2014-Sept 2015 from the 2009-2014 average by 0.7 Sv, which is made up of an Ekman increase of 0.6 Sv and an increase of 0.1Sv from the UMO + FS. It will be interesting to see how the new RAPID data compares!
For the Ekman component, we do not have wind stress for the full period, however we found that the Ekman component had a good correlation with the NCEP zonal wind strength at the same latitude. Hence we used a simple linear regression model with the NCEP wind strength which we have for the missing months. For other components we used the seasonal climatology plus the mean anomaly from the previous 12 months (ie assuming persistence of the anomaly from seasonal climatology). These last 4 months are shown in each panel in green.