Known Issues with AOD1B RL04

Trends in simulated ocean bottom pressure

OMCT simulations are intended to simulate in particular short-term variability of ocean bottom pressure in response to rapidly varying atmospheric conditions. In the long run, however, the model state is drifting more rapidly than, e.g., current state-of-the-art coupled atmosphere-ocean models that are prepared to reproduce climate variability over many centuries. Low frequency variability and trends in OMCT ocean bottom pressure are primarily related to ongoing warming and cooling of water masses at intermediate depths, and its secondary effects on the thermohaline circulation. They are therefore much less reliable than the high-frequency variability and should not be interpreted geophysically.


Figure 1: Ocean bottom pressure trends from OMCT RL04 as calculated from monthly mean bottom pressure coefficients (GAD-products) covering 2004 - 2010.

Discontinuities in atmospheric surface pressure in mountainous regions

The basis of the atmospheric part of the AOD products are operational analyses from ECMWF - that is, data from a numerical weather prediction model which is intended to provide best possible state estimates and corresponding medium-range forecasts to its users. Therefore, the ECMWF is upgraded periodically to incorporate improvements in the physical model, the numerics, the data assimilation scheme, and to accomodate new observing technologies. Those changes to the model consequently lead to inconsistencies in the time-series of model states, most easily realized from series of atmosheric surface pressure at high altitudes in mountainous regions. ECMWF model changes are usually performed twice a year, a complete history might be found here.


Figure 2: Surface pressure values from ECMWF operational analyses during 2007 for selected positions in Chile (top left), the Himalaya (top right), the Amazonas basin (middle left), western Antarctica (middle right), the South Pacific (lower left) and the Weddell-Sea (lower right).

Discontinuity in simulated ocean bottom pressure on May 26, 2009

During May 2009, the OMCT processing chain has been substantially modified in order to cope with major hardware changes at the German Climate Computing Centre. This made it necessary to replace a number of pre- and postprocessing software bits, which included also the library used to interpolate atmospheric forcing data to the OMCT compute grid.

Atmospheric surface pressure is sensitive to the local geopotential height and therefore has its strongest lateral gradients over steep topography. Differences due to a change of the method of interpolation are only relevant close to the coasts, and in particular at places where shelf-ice determines the lower boundary of the ECMWF model while OMCT still simulates ocean mass variability below the ice-body.

Since long-term changes in atmospheric pressure are quickly compensated by an adjusted sea-surface (inverse-barometric reaction of the ocean), the discontinuity is hardly detectable in simulated time-series of ocean bottom pressure ('oba'-coefficients in AOD1B, GAD for the monthly products). However, the water-column contribution to bottom pressure variability ('ocn'-coefficients in AOD1B, GAB for the monthly products) is reflecting the discontinuity. Since this water-column contribution is later added to the gravity effects of the three-dimensionally distributed atmospheric masses (to form the 'glo'-coefficents or the GAC-product), it is also mirrored additive-inversely by the final GSM gravity fields that where calculated by reducing the non-tidal atmosphere-ocean mass anomalies by using those 'glo'-coefficients.


Figure 3: OMCT RL04-based ocean bottom pressure anomalies averaged over the period before the discontinuity (Jan 1, 2009 until May 25, 2009), after subtracting the mean bottom pressure field calculated for 2009.

Bias in wind-stress calculation during Jan 2007, Jan 2008, and Jan 2009

Wind stresses required to force the ocean model are calculated from surface wind speeds, the air density in the surface boundary layer and a time-variable roughness coefficient, that in turn depends on the stability of the atmospheric boundary layer and the wind speed (Beljaars, 1997). Due to a scripting error, specific humidity required for the density calculation have been obtained from a level at tropopause altitude for Jan 1st - Jan 31st for a number of years (2007-2009), biasing wind stresses low for those months, which consequently lead to anomalous mean bottom pressure fields when compared to a model run unaffected by that bias as, e.g., OMCT RL05.


Figure 4: Differences between monthly bottom pressure anomlies simulated with OMCT RL04 and OMCT RL05 for January 2006 (top left), January 2007 (top right), January 2008 (middle left), January 2009 (middle right), January 2010 (lower left), January 2012 (lower right).

July 16, 2012, Dr. Henryk Dobslaw