Observation B ias and O cean R eanalyses

Observation Bias and Ocean Reanalyses

James Carton, Gennady Chepurin, Semyon Grodsky, Tony Santorelli, and Benjamin S. Giese (TAMU)

Department of Atmospheric and Oceanic ScienceUniversity of Maryland

Session 3: Observational Data Recovery and Bias Correction Efforts

Motivation:

> Bias problems in ocean observation sets: Gouretski and Koltermann, 2007, Wijffels et al. (2008); Levitus et al. (2009); Ishii and Kimoto (2010), Gouretski and Reseghetti (2010), Giese et al. (2010)

> Problems interpretation of observations: Grodsky et al. (2008)

> Bias problems in current surface forcing data sets: Bromwich and Fogt (2004), Large and Yeager (2008)

> Model problems

1910-1919

HYDRO DATA

Wijffels et al. (2008)

BT History

1944 mechanical BT developed at WHOI

1968 Sippican gets contract to produce XBT

XBT depth = at – bt2

1980s Recorders change offering two drop-rate algorithms

1995 Production moved to MX

1996 wire coating changes; winding specifications were changed

1999 mesh added which affects the spooling of the first few meters.

Global heat content 0/700m

ORIGINAL

Discuss this later

Heat Content anal – obs

for 7 reanalyses

108Jm-2

A-O > 0

Fall Rate Formulae

Sippican Fall Rate:xbt depth = at – bt2

a=6.472 and b=0.00216

Hanawa (1994):a=6.691, b=0.00225

Hanawa data

Temp error interpreted as depth error

Temp Error

Levitus et al. 2010

Temp Error after

Levitus et al bias correction

Global heat content 0/700m

ORIGINAL

Corrected

Global heat content 0/700m

WIJFFELS ET AL. 2008, Ishii and Kimoto, 2010

LEVITUS ET AL., 2009

MEAN BIAS CORRECTION

When the observations are assimilated into SODA

Experiments

2.1.0 Hanawa

2.1.2 Levitus

2.1.4 Wijffels

<A-O>

Hanawa

Levitus et al

Wijffels et al

Too cold

Too warm

Too warm

When the observations are assimilated into SODA

1997/8 El Nino

Correcting the BTs causes a 30cm/s change in U(z=50m) and a 0.5C change in T(z=50m)

Surface drifters

Memoriam: Pearn P. Niiler 1937-2010

Surface Drifters:Velocity trend bias

Explanationsthe switch to mini-drogues? orunidentified drifters missing drogues?

Difference between MLT and SST

Surface windsPercentage of the ocean covered by the ICOADS wind observation

Wind Coverage100% means that 120 months of 2x2 deg binned data is available during a decade

Time correlation of anomalous zonal wind stress from ICOADS observations and the 20thC reanalysis for the (top) first and (bottom) second half of the 20th century

Running standard deviation of monthly anomalous zonal wind stress at the Vlissingen weather station (OBS) and from the 20thC Reanalysis (REAN). This suggests that REAN underestimates the magnitude of anomalous winds by ~50% (or 75% of variance is missing). But temporal correlation of observed and reanalysis wind anomalies is high, ~0.9.

1910-50 1951-00

TCORR 0.87 0.93

TREGR 0.53 0.51

Summary• Much recent attention has been devoted to identifying bias in the BT

archive, discussed in two recent workshops. Ocean reanalysis offers an approach complementary to XBT-CTD comparisons.

• SST is generally interpreted as MLT. In fact the two are systematically different in ways that need to be clarified.

• Surface drifters have an unrealistic trend. Unraveling the cause and removing the trend requires returning to the original data, an effort underway at AOML (Lumpkin). Possible cause: unidentified undrogued drifters.

• ICOADS comparisons to the 20CR offers a way to examine the accuracy of the surface winds.