Using altimeter and Argo data to estimate biases
Using altimeter and Argo data to estimate biases in XBT fall rate equations Josh K. Willis joshua. k. willis@jpl. nasa. gov Jet Propulsion Laboratory
Overview I. Ocean “cooling” – why overall accuracy in the XBT network is important II. Argo and altimeter data as calibration tools III. Time evolution of recent XBT biases IV. Remaining XBT errors
Upper-Ocean “cooling” Cooling!!! 2003 to 2005 cooling: -1. 0 ± 0. 33 W/m 2 (Averaged over Earth’s surface) from Lyman et al. (GRL, 2006)
XBT Data – Signals and Errors Eddies • Big signals in “isotherm displacement” • Lots of averaging required for large-scale • Average signals susceptible to systematic error
WHOI float biases
A correction to “recent cooling” Ocean Heat Content from 2004 to 2006 Removing the bad float data reduces the cooling but does not completely eliminate it. From Willis et al. , GRL, in prep.
A correction to “recent cooling” Ocean Heat Content from 2004 to 2006 Another bias: XBTs are biased warm, which also causes spurious cooling. From Willis et al. , GRL, in prep.
XBT bias & fall-rate errors – pair analysis Comparison of Isotherm Displacements XBT/Argo pairs ~12, 000 Argo/CTD pairs ~2, 000 Isotherm Displacement: Tclim – T Dz = d. Tclim /dz From Willis et al. , GRL, in prep.
The “pseudo pair” Coefficient of regression, r between SSH anomaly and T(400 m) • Much of upper ocean T variaibility is contained in SSH anomaly • Use AVISO SSH maps to make “pseudo” temperature anomalies: Tpseudo = a(z) * SSHA From Wijffels et al. , manuscript in prep.
“Pseudo-pair” comparison Comparison of Sippican Deep Blue probes with nearby Argo pairs (2004 – 2006), ~12, 000 • Pseudo-pairs give same bias, but have narrower distribution • More comprehensive means of test XBT bias because of SSH data availability From Wijffels et al. , manuscript in prep.
“Pseudo-pair” analysis of other data (a. k. a. sanity check) CTD data Argo profile data • CTD data show no significant bias during any time period • Argo floats show little bias except for WHOI/FSI floats From Wijffels et al. , accepted
Time dependence of XBT bias Time dependence of bias in Sippican Deep Blue XBT probes • Bias increases over time • Hi bias in later years may reflect “double” application of Hanawa et al. (1995) correction From Wijffels et al. , manuscript in prep.
Stretching factor by probe type
Revised ocean heat content estimate
Remaining Errors • Skewness: significant • difference between mean and median Could be a sign of wire stretch
Remaining Errors Gaussian Tail for positive depth error is much bigger
- Slides: 16