• Data Accuracy & Methodology Uncertainties

Refining uncertainty estimates is a critical component of ongoing heat transport research at AOML. Uncertainties in the heat transport estimates may arise due to methodology, and due to data accuracy and representation issues. Uncertainties can include:

1. The accuracy of the XBTs
2. The accuracy of the salinity estimation for each XBT temperature
3. The accuracy and representativeness of the deep climatology (if that is used for the estimation of heat transport below the depth of the XBTs).
4. The accuracy and representativeness of wind fields used to estimate Ekman transports (one component of the ageostrophic motions).
5. The accuracy of geostrophy
6. The uncertainty in reference level using geostrophy (e.g. how well the boundary current transports and barotropic velocity is known).




Back to Top of Page...

In order to test the validity of the methodology in the North Atlantic, historical data along the 24N line are used in the simple geostrophic method to compare heat transports estimates. The four historical sections can be systematically reduced to temperature data in the upper 850 m to simulate XBT data and the resultant heat transport suggests what sort of errors can be found from using real XBT transects. These errors estimates are ongoing and more information will be posted here shortly.

Back to Top of Page...

In order to test the validity of the methodology in the South Atlantic, data from the WOCE A10 line (Siedler et al., 1996) is used in the simple geostrophic method to compare heat transports estimates. A10 data can be systematically reduced to temperature data in the upper 850 m to simulate XBT data and the resultant heat transport suggests what sort of uncertainties can be found from using real XBT transects. The POCM model heat transport is also analyzed to examine the possible errors from uncertainty in ageostrophic motions.

We found that the largest errors were introduced by using the Levitus Climatology to simulate the subsurface density field (about 0.2 PW) and the uncertainty in the absolute flow field (0.2 PW). A summary of the various uncertainties follows:

1. Simple method (vs. inverse solution with transport constraints) = 0.01 PW
2. Salinity estimation = 0.04 PW
3. Subsurface density field estimation = 0.21 PW (possibly a bias, not random error)
4. Water depth estimation = 0.01 PW
5. Absolute transport uncertainty = 0.27 PW (determined from POCM), absolute transport uncertainty = 0.2 PW (determined using different reference surfaces)
6. Ekman transport uncertainty = 0.1 PW (difference between wind field products)
7. Undersampled Argentine shelf transport = 0.01 PW
8. Undersampled South African shelf transport = 0.01 PW

Back to Top of Page...

Several projects are underway or are planned that will attempt to constrain these uncertainty estimates further, to provide overall uncertainties in the heat transport values presented:

1. An algorithm to determine salinity as a function of temperature, longitude, depth and season for the region 30°S to 40°S using the methodology published by Thacker (personal communication). Regional functions of temperature, latitude, longitude, and day-of-year fitted to salinity from CTD and ARGO data will provide more accurate estimates of salinity and thus density from XBT data. (Go to Project...)
   A summary of differences in results from the Thacker Salinity estimation method and Levitus' climatology, together with their differing impacts on overall Heat Transport estimates, are both summarized in the tables at bottom of this page...
2. Malvinas Current transport estimates using an approach that included altimeter and current meter observations (Vivier et al, 2001). The Brazil Current transport will be computed using a combination of altimetry and historical hydrographic data within a two-layer reduced gravity scheme (Wainer et al, 2000). (Go to Project...)
3. Boundary current transport estimates for the North Atlantic to include the time variable Florida Current transport observations funded by NOAA/OCO. (Go to Project...)
4. A comprehensive study of the wind products - climatology, reanalysis and satellite - performed to determine which one is the most appropriate for the region.
5. The methodology will be tested using a box inverse model approach that will generate formal error bars on the transports based on the accuracy of the various observations, improve the transports by incorporating the air-sea fluxes of the Atlantic Ocean, and potentially improve operational air-sea flux estimates in the undersampled Southern Ocean.

Reasonable estimates of the heat transport uncertainty will be derived (from models and CTD/ARGO data), that includes updated estimates from the improved methodology so that recommendations for future observing system improvements can be made, e.g., XBTs that extend to 1000m might improve our heat transport estimate by some amount. The ultimate goal of these uncertainty estimates is to quantify our confidence in the heat transport results and provide the scientific basis to improve their accuracy.

Back to Top of Page...

Table 1:

The following table shows estimates of total meridional heat transport in PW^(†) along the AX18 line. Heat transport estimates are shown using two different estimation methods, as outlined in the Methodology section of the AX18 FAQ:

1. Using the Levitus Salinity climatology (S(T,lat,lon)); and,
2. Using the Thacker et al Salinity estimation procedure (S(p,T,lat,lon))

(†): _{1 PW = 1 petawatt = 10}15 _Watts

The effect of these two different methods of salinity estimation leads to a ~0.1 PW difference in heat transport estimates. That suggests an additional uncertainty in heat transport due to the uncertainty in salinity. Improvements to the heat transport estimates may be made through observations of salinity either with expendable CTDs or profiling ARGO floats.

Back to Top of Page...