Interactive comment on “ Empirical correction of XBT fall rate and its impact on heat content analysis ”

Dear reviewer, Our XBT correction method builds on previous works, including the results of Gouretski and Koltermann (2007) , Wijffels et al (2008), Thadathil et al (2002) and Gouretski and Reseghetti (2010). We used a collocation method to compared XBT and CTD/OSD profiles. We highlight the need to separate deep and shallow XBT (Wijffels et al ,2008) and to take into account the water of the sea where the probe had been deployed (Thadathil et al ,2002). It is a combination of previous correction. We chose to calculate for each year between 1968 and 2007 a parabolic correction because it provides a good fit with the calculated depth bias. In contrast to Gouretski

Introduction  Expendable BathyThermograph (XBT) system does not measure directly the depth of the probe, it uses a fall rate to estimate it  Is there a correct depth equation for correcting temperature as a function of depth from XBT that could be applied to the global datasets ? The W08 correction is a reference for the treatment of XBT, but how does this correction vary with the method of comparison of XBT and CTD profiles ?  Correcting individually each type of XBT cannot be envisionned but can we refine the W08 correction including regional correction ?  What is the impact of such a correction on the calculation of the ocean heat content ? The W08 corrections have been applied to our collocated profiles

Median bias = original depth eq. (green) and corrected by W08 (red) function of depth on average over the study period.
Test of the W08 correction  The linear correction is not always performing well (with our collocation method) especially between 1975 and 1985. It provides too strong correction below 500m depth and a too small correction for surface layers.

Evolution of the median bias as a function of depth and years original depth eq. corrected by W08
A new correction Second order correction  Annual median depth correction computed using:


The difference between collocated profiles do not seem to indicate a linear function for depth correction, but rather a second order function with an offset,  Between the surface and 30m, the bias doesn't follow a parabolic behavior because of high variability noise due to the surface mixed layer.


Correlation between depth correction term and the deployment latitude.

Median XBT-CTD depth bias at 100m function of absolute latitude for XBTS (red) and XBTD (blue)
We can't distinguish XBTS to XBTD comparing depth correction at a given depth.

 Separation of XBT into 4 classes:
 XBTS and XBTD  Low and high latitudes (40°N⁄S)

Offset in meters calculated for XBTD deployed in high/low latitudes (blue/black) and XBTS deployed in high/low latitudes (red/green) function of years.
 Maximum of the offset between 1970 and 1985.


The correction reduces the median temperature bias.  Contribution of the offset is significant. median raw bias (top), corrected by a parabolic function (middle) and by a  parabolic function added an offset (below) funtion of depth and  A strong negative temperature bias is found in the western Pacific (from 0 to 60°N, West of 180°W) after the global correction.  It is predominantly located at 300m between 1970 and 1985.  A regional correction is available  These profiles (years 1968 to 1985) have been separated from the global dataset

Evolution of XBT-CTD median globally corrected bias for XBT deployed in western Pacific, function of depth and years
Heat content analysis

Evolution of [0-700] m ocean heat content calculated from only WOD05 XBT (black), corrected XBT (green),all data from WOD05 (red) and all corrected data (blue)
 The calculation of the ocean heat content confirms that on average XBT temperature data are now closer to CTD temperature data.


Using the same methodology, we corrected MBT (second order correction and an offset, latitude classes). Conclusion  According to W08, XBT are subject to a depth bias varying with the year of deployment.
 However, our collocation method reveals that this bias should be better corrected with a second order function added to an offset.


Behavior of XBTS and XBTD are quite different and depends on the latitude of deployment.


We confirm that the maximum of heat content during the 70's in early papers can be explained by the XBT bias.


We have now available a corrected database and we are now working on field reconstruction using a EOF method (DINEOF, Beckers et al , 2003).