Evaluating the Large Scale Mean and Variability

It is important to check that the average properties of the ocean forecasting systems are providing a good representation of the ocean climate. This is usually done by comparing multi-annual averages to climatologies generated from observational data-sets.

One example of this is a comparison between a mean dynamic topography (MDT, such as that of Rio et al. 2005; or Maximenko and Niiler 2005), with the model's average sea surface height field. This provides a useful guide as to the ability of the model to represent the large scale ocean circulation (see for example Metzger et al. 2008).

Temperature and salinity can also be assessed using a suitable climatological data-set. In Fig. 22.2, the annual mean temperature anomalies from the World Ocean

1000

2000

4000

latitude (deg)

latitude (deg)

latitude (deg)

latitude (deg)

1000

2000

d- 3000

4000

0 50

latitude (deg)

0 50

latitude (deg)

0 50

latitude (deg)

0 50

latitude (deg)

1000

2000

d- 3000

4000

100 200 300

longitude (deg)

100 200 300

longitude (deg)

100 200 300

longitude (deg)

Fig. 22.2 Annual mean temperature anomalies from WOA05 climatology for 2008 from FOAM both without (a, c, e) and with (b, d, f) data assimilation. a, b Show a cross-section in the Indian Ocean along 90°E. c, d Show a cross-section in the Atlantic Ocean along 30°W. e, f Show a cross-section along the equator

1000

2000

4000

1000

2000

d- 3000

4000

1000

2000

2000

d- 3000

d- 3000

4000

4000

Atlas 2005 (Locarnini et al. 2006) are shown as cross-sections from two hindcasts of the V° resolution global FOAM system, one without assimilation and one with. This shows that the data assimilation is able to reduce the drifts of the model away from climatology. One has to be careful when performing these comparisons that any inter-annual signal is not contaminating the results. For example in Fig. 22.2f, there is a clear La Nina signal, where the model is representing the true deviation from climatology.

The variability in the model and observations can also be assessed. For instance, sea surface height (SSH) can be used to measure the amount of mesoscale activity. This can be estimated from observations provided by satellite altimeters, and also from ocean models. Figure 22.3 shows an example of this from the GLORYS reanalysis produced by Mercator using the V° resolution NEMO model with data assimilation. The standard deviations of the data are shown next to the standard deviation of the model fields from a 6 year period. Here, the data used to calculate the observed variability are being assimilated in the reanalysis product so this test is only useful to check that the assimilation of data is working correctly. The model analyses are reproducing the observed variability very well, including the western

Longitude

Longitude a

Fig. 22.3 Standard deviation of SSH for the period

2002-2007 from a Aviso data a

Fig. 22.3 Standard deviation of SSH for the period

2002-2007 from a Aviso data

Longitude

0.28

0.26

0.24

0.22

0.18

0.16

0.14

0.12

0.08

0.06

0.04

0.02

and b from the GLORYS reanalysis product

Longitude boundary currents which are difficult areas to accurately represent the mesoscale variability with resolution. The only regions where the model variability is significantly different to the observed are in the Zapiola Rise region and in parts of the South Pacific.

Both the average and variability comparisons described above are useful as a first-order check on the ability of the model to represent the large-scale ocean features, and can give confidence that the model is behaving as expected. However, they do not give information about the accuracy or skill of the model and so are of limited use to most users. More detailed investigations are required for this, and are described below.

0 0

Post a comment