Subduction Temperature Errors 2014/04/29 N. Galbraith Temperature records were removed from the UOP ‘archives’ web area because of errors, first brought to our attention via email: I am interested in using the subduction array data, but I have noticed some peculiarities in the data and I was hoping that you might be able to help me. At some moorings and at some times (sufficiently often to show up on a three month average) the water temperature is higher at 10m and 30m depth than it is at 1m depth. It is occasionally also higher at 50m depth. I have looked in the single publication referenced at the bottom of the page http://uop.whoi.edu/uopdata/subduction/subduction.html that I have access to, but no mention is made of any such feature. Do you know how or where I might go to access the WHOI technical notes, or who I might contact to obtain further information about this and other subduction array data issues? Many thanks in advance for your help, John Kennedy John Kennedy Climate Monitoring and Research Scientist Met Office FitzRoy Road Exeter EX1 3PB United Kingdom Tel: +44 (0)1392 885105 Fax: +44 (0)1392 885681 E-mail: john.kennedy@metoffice.gov.uk http://www.metoffice.gov.uk Global climate data sets are available from http://www.hadobs.org 1) There are times when we might expect a temperature inversion: - A colder but presumably fresher layer was at time seen to slide in under the surface mixed layer (see xbt profiles 1-9), so as you went deeper thru that layer and came out the bottom you hit warmer but presumably saltier water so that the temp profile has inversions but I'd expect that density did not. - Rain is falling. Rain falls at the local wet bulb, usually colder than SST, forms shallow fresh lenses for a day or so depending on the wind mixing. The third profile from the left on the xbt profiles 180-189 might be evidence of such a fresh lense. Other profiles in this plot look to have a slight increase with depth of surface mixed layer temp. 2) the moorings lacked salinity sensors Ideally, I would have done a quality control step on the moored temperature data by computing density and looking to see if the density profile was stable. This is what we do now on more recent deployments where we accept as real salinity compensated temperature inversions, or, in high time resolution data, evidence of overturning. However, subduction was an early long-term mooring experiment and we only had temp sensors that would last. So no way to do the stability check, and against the time desribed in 1) above when there are real inversions, we found no good way to go after detecting and fixing false temp offsets. 3) a mix of instruments This experiment drew down the resources at my lab at WHOI and at the folks we partnered with at Scripps Inst of Oceanography (SIO). So a mooring like Sub 3 NW had a Vector Averaging Wind Recorder for SST, a WHOI Brancker temp logger at 1 m, an SIO temp logger at 10 m, then WHOI Brancker temp loggers at 30, 50, 60, 70.... Given this, I worry about the calibration of the NW3 10m temp vs the NW3 30 m temp as two different labs, two different processing schemes were involved. Then there is the issue of clock drift. These instruments had a sampling rate of 15 min typically. We checked the crystal clocks in the processing and tried to do linear interpolations to get to corrected time bases, but I suspect after 8 months we might have a skew of 1 or 2 sample intervals, which in the presence of internal wave variability might, depending on who lags or leads, give the type of ± temp difs at high frequency in your plot. If one or both of the instruments was at or near the base of the mixed layer, I'd expect something like this as the based moved up and down. 4) the 1 m vs 10 m I am most concerned about the cooler 1 m temp on your other plot. It looks to have the same cool bias before and after the diurnal peak. I do not think this is a geophysical signal. 5) strategy The issues above led to not implement a correction procedure on the archived data. From XBTS and CTDs there are at time temp inversions; we did not have salinity and could not do a stability check. If I was to work on this data, I would do the following: 1) look for times of strong wind (so there is good mixing) and take the temps from within the mixed layer. 2) Assume there is strong enough stirring that the mixed layer should be isothermal as well as isopycnal. 3) Identify there persistent biases between temps in the "mixed layer" and correct for them (to what? a bit arbitrary here to choose which is correct, but I'd add the SST from the met record in to include that and might correct to that as it was deeper than any skin temp). 4) check and see if a bit of bias correction helps the fast time repsonse issue. 5) if the fast time sampling temp data still shows plus and minus difs, one last effort can be made to verify time base issues. I do a coherence and phase calculation, looking to see if as you go to high frequencies the phase tails off to one side of 0 or the other, then I shift one time series with respect to the other by one sample interval and rerun the coherence/phase; sometimes by introducing a time shift you bring the phase back to zero. John, I hope this helps. Basically, there are limits to the data. Our group site http://uop.whoi.edu shows other moored data sets that we would be happy to share. In the years since the Subduction experiment, instrumentation has improved. The moored temp loggers we use now keep their calibrations to about 0.005 deg C and their clocks are good to a few minutes a year. And we have moored salinity now too. Let me know if we can help with other data. Best regards, Bob Weller 2014/04/29 Update I did a quick look at the temperature files on the web site, listing the mean value for each bin and then plotting those where the difference between the means showed an overall inversion. Some of these were caused just by a shallow instrument ending early and then the temps warming up, so a deeper bin had a higher mean. cd /Volumes/Buoy2_HD/Users/Shared/archive/work/Subduction/fromOcelotWeb infiles=dir('*.epic'); for ii=1:length(infiles) infile=infiles(ii).name; clear mday temp depth meta [mday, temp, depth, meta ]=get_epic_data(infile); fprintf('\n%s %.1f rate(minutes)\n', infile, median(diff(mday)*24*60)); for dd=1:length(depth) gv=find(temp(dd,:) < 100); fprintf('%.1f %.2f %.2f\n', depth(dd),mean(temp(dd,gv)),median(temp(dd,gv))); end end found inversions in subc2t.epic: 50.0 21.23 21.10 60.0 21.44 21.29 subne3t.epic - this one has a bad bin, shows many problems 10.0 19.87 19.15 30.0 20.11 19.46 subnw2t.epic 130.0 17.90 17.88 150.0 17.96 17.93 subnw3t.epic 1.0 20.29 19.72 10.0 20.31 19.75 Sub 1 NE Depth (m) .5 10 30 50 60 70 80 90 100 110 130 150 200 300 400 580 750 1500 Model VAWR VMCM VMCM VMCM TPOD VMCM TPOD VMCM TPOD VMCM TPOD VMCM VMCM TPOD TPOD TPOD TPOD TPOD SN 704 041 021 039 3274 032 3265 022 3288 030 3269 028 018 3300 3305 3268 3286 3293