Possible Error in OHC?

An objective analysis of the evidence for global warming suggests little if any anthropogenic effect, consistent with a direct radiative effect from increased CO2. It is also obvious that global temperature and ocean heat content should be related, so it’s somewhat surprising to see OHC rising so fast around 2002-3 when ocean temperature is relatively stable (upper line below).

fig3

Also given the known issues with this data set, it’s bizarre to see top bureaucrats and scientists like Wong/Steffen placing so much importance on the trajectory of OHC over a short run since 2000. This is reminiscent of the missplaced importance the Garnaut commission placed on the now discredited ‘worse than we thought’ finding of Rahmstorf and others, that he has admitted was an error.

Slide1

At the suggestion of cohenite, I plotted the distribution of changes in the OHC data (above), at lags of one, two, three and four years. These changes have been normalized to a mean of zero and the x-axis is the standard deviation. The changes in the OHC around 2002-3 stand out, particularly at lag 2 and 3, as 3-sigma events. That is, the probability of a change in OHC of this magnitude is around 0.001, or there is a high probability that this jump is an outlier, and due to some problem in the data.

See WUWT and Bob Tisdale for more discussion on this issue.

The above-the-fold figure illustrates the effect of treating one, and then two years changes as outliers in this type of data. Because of the high autocorrelation, removal of a step up affects all subsequent points, and they all get shifted down. The result is still a slightly increasing OHC, but far less alarming that the impression given previously.

Obviously this does not prove the data are in error. But people remove 3-sigma data points all the time, particularly when caution over possible instrumental errors is called for, as could be the case with the ARGO floats here. Does anyone else feel uncomfortable with so-called leaders incapacity to distinguish preliminary results, and using them for self-serving purposes?

  • Anonymous

    Where global measurements are involved, there is always a step where points data are converted to gridded data. This type of step is crucial in mining and has been studied deeply.

    When you add satellite derived data, more and more “unusual” errors accumulate, so that difference maps will show artefacts. e.g. what was originally a surface grid with parallel, straight sides of latitude and curved sides of longitude assumes a different shape as the satellite altitude changes from drag and reburns, especially when the satellite view is inclined normal to the path. The sides of the grid cell can bulge in longitude and assume a “different” area that needs quite complex correction to both size and position.

    It might be an interesting exercise to take watermelon slices, say 5 degrees apart in longitude, and compute their areas in the pole to equator quadrant, for different of the satellite and earth models. My hunch is that you will get different areas from different models at different times- especially when different satellites are used.

    This might seem a bit OT, but you have to go through this process when calculating a global ocean heat content and a global ocean temperature. There is good reason to suspect that different organisations working on these data sets will produce different rsult, for reasons arising more from geometry than climate.

    Currently, I’m sounding out some mining people for state of the art references to interpolation of point to grid and some surveying types on the godesy aspects of lat/long grids. In the fullness of time, I hope to have more to report.

  • sherro

    Where global measurements are involved, there is always a step where points data are converted to gridded data. This type of step is crucial in mining and has been studied deeply.When you add satellite derived data, more and more “unusual” errors accumulate, so that difference maps will show artefacts. e.g. what was originally a surface grid with parallel, straight sides of latitude and curved sides of longitude assumes a different shape as the satellite altitude changes from drag and reburns, especially when the satellite view is inclined normal to the path. The sides of the grid cell can bulge in longitude and assume a “different” area that needs quite complex correction to both size and position. It might be an interesting exercise to take watermelon slices, say 5 degrees apart in longitude, and compute their areas in the pole to equator quadrant, for different of the satellite and earth models. My hunch is that you will get different areas from different models at different times- especially when different satellites are used.This might seem a bit OT, but you have to go through this process when calculating a global ocean heat content and a global ocean temperature. There is good reason to suspect that different organisations working on these data sets will produce different rsult, for reasons arising more from geometry than climate.Currently, I'm sounding out some mining people for state of the art references to interpolation of point to grid and some surveying types on the godesy aspects of lat/long grids. In the fullness of time, I hope to have more to report.

  • http://bobtisdale.blogspot.com/ Bob Tisdale

    David, would the 1997/98 El Nino represent the same type of statistical outlier since 1955? If the rise in 2003 is a residual effect of that El Nino, actually of the 2-year La Nina that followed it, would it then be that unusual? If that makes sense, the only thing left to determine is why there would be that long of a lag between the end of that multiyear La Nina and the OHC shift.

    • Anonymous

      Bob, Its not as large a change. The 2002-2003 change is the only event that large. As to the second question, you are suggesting a lower rate of heat increase leads to a higher rate of heat increase later on. It doesn’t seem likely on thermodynamic grounds, as impulse changes would normally be dissipating. But its not impossible if the smaller impulses are accumulating ‘out of sight’. Hard to rule things out in a complex system.

      • http://bobtisdale.blogspot.com/ Bob Tisdale

        David, sorry, I should have clarified my speculations above…that the 2003 rise would assumedly result from a lagged temporary shift in cloud cover. The rebound of tropical Pacific OHC during the recharge phase of ENSO is likely the result of atmospheric circulation changes, primarily cloud cover. What then happened to Hadley and Walker circulation, wind stress, cloud cover, etc., globally when the 1998/99/00 La Nina relaxed and ENSO conditions returned to neutral? Sorry, just thinking out loud. A three-year lag does seem excessive and unlikely.

        • http://bobtisdale.blogspot.com/ Bob Tisdale

          Here are some OHC spaghetti graphs for further discussion about the OHC shift in 2003. The first is the OHC from Jan 1955 to Jun 2009 for the North and South Atlantic, North and South Pacific, and the Indian Ocean. As you can see the biggest shift in 2003 takes place in the South Atlantic.
          http://i38.tinypic.com/2q8xdu8.png

          And here’s the same comparison of the major ocean basins with the data limited to Jan 2000 to Jun 2009.
          http://i33.tinypic.com/w14u3n.png

          Using the South Atlantic for reference, here’s a comparison of Arctic and Southern Oceans from Jan 1955 to Jun 2009. Both have sizeable increases but they appear to be part of a multiyear rises.
          http://i37.tinypic.com/333v6zo.png

          And here’s the same comparison of the South Pacific, Arctic and Southern Oceans with the data limited to Jan 2000 to Jun 2009.
          http://i34.tinypic.com/311nspg.png

          So the 2003 shift in the South Atlantic OHC seems to be the outlier. For this part of the post, I’ll borrow two graphs from my post “North Atlantic Ocean Heat Content (0-700 Meters) Is Governed By Natural Variables”:
          http://bobtisdale.blogspot.com/2009/10/north-atlantic-ocean-heat-content-0-700.html

          The early 2000s rise in tropical South Atlantic OHC appears to lag the tropical North Atlantic by 1 to 2 years. Curious.
          http://i34.tinypic.com/3532k5l.png

          And the tropical North Atlantic seems to shift in response to the multiyear La Nina that followed the 1997/98 El Nino.
          http://i38.tinypic.com/2jcbgvb.png

          Now for two new graphs. The first is a comparison of the Low Latitude (25S-0) and Mid-To-High Latitude (60S-25S) South Atlantic from Jan 1955 to Jun 2009. The Mid-To-High Latitude OHC anomalies of the South Atlantic appear to mimic the tropical OHC anomalies, or vice versa, with the Mid-To-High Latitudes being noisier. Is the higher variability a result of the Southern Annual Mode, like the North Atlantic is impacted by the NAO? Dunno.
          http://i37.tinypic.com/vqmsrd.png

          And here’s the short-term version of the Low Latitude (25S-0) and Mid-To-High Latitude (60S-25S) South Atlantic OHC comparison.
          http://i37.tinypic.com/1z3bif9.png

          Is the 2003 OHC shift a natural effect? Is it a result of an overcorrection of ARGO biases? Dunno.

          I’ll try to take a closer look at the Southern Ocean and Arctic Ocean OHC data soon.

  • http://bobtisdale.blogspot.com/ Bob Tisdale

    David, would the 1997/98 El Nino represent the same type of statistical outlier since 1955? If the rise in 2003 is a residual effect of that El Nino, actually of the 2-year La Nina that followed it, would it then be that unusual? If that makes sense, the only thing left to determine is why there would be that long of a lag between the end of that multiyear La Nina and the OHC shift.

  • davids99us

    Bob, Its not as large a change. The 2002-2003 change is the only event that large. As to the second question, you are suggesting a lower rate of heat increase leads to a higher rate of heat increase later on. It doesn't seem likely on themodynamic grounds, as impulse changes would normally be dissipating. But its not impossible is the smaller impulses are accumulating 'out of sight'. Hard to rule things out in a complex system.

  • http://bobtisdale.blogspot.com/ Bob Tisdale

    David, sorry, I should have clarified my speculations above…that the 2003 rise would assumedly result from a lagged temporary shift in cloud cover. The rebound of tropical Pacific OHC during the recharge phase of ENSO is likely the result of atmospheric circulation changes, primarily cloud cover. What then happened to Hadley and Walker circulation, wind stress, cloud cover, etc., globally when the 1998/99/00 La Nina relaxed and ENSO conditions returned to neutral? Sorry, just thinking out loud. A three-year lag does seem excessive and unlikely.

  • http://bobtisdale.blogspot.com/ Bob Tisdale

    Here are some OHC spaghetti graphs for further discussion about the OHC shift in 2003. The first is the OHC from Jan 1955 to Jun 2009 for the North and South Atlantic, North and South Pacific, and the Indian Ocean. As you can see the biggest shift in 2003 takes place in the South Atlantic.http://i38.tinypic.com/2q8xdu8.pngAnd here’s the same comparison of the major ocean basins with the data limited to Jan 2000 to Jun 2009.http://i33.tinypic.com/w14u3n.pngUsing the South Atlantic for reference, here’s a comparison of Arctic and Southern Oceans from Jan 1955 to Jun 2009. Both have sizeable increases but they appear to be part of a multiyear rises. http://i37.tinypic.com/333v6zo.pngAnd here’s the same comparison of the South Pacific, Arctic and Southern Oceans with the data limited to Jan 2000 to Jun 2009.http://i34.tinypic.com/311nspg.pngSo the 2003 shift in the South Atlantic OHC seems to be the outlier. For this part of the post, I’ll borrow two graphs from my post “North Atlantic Ocean Heat Content (0-700 Meters) Is Governed By Natural Variables”:http://bobtisdale.blogspot.com/2009/10/north-at… The early 2000s rise in tropical South Atlantic OHC appears to lag the tropical North Atlantic by 1 to 2 years. Curious. http://i34.tinypic.com/3532k5l.pngAnd the tropical North Atlantic seems to shift in response to the multiyear La Nina that followed the 1997/98 El Nino. http://i38.tinypic.com/2jcbgvb.pngNow for two new graphs. The first is a comparison of the Low Latitude (25S-0) and Mid-To-High Latitude (60S-25S) South Atlantic from Jan 1955 to Jun 2009. The Mid-To-High Latitude OHC anomalies of the South Atlantic appear to mimic the tropical OHC anomalies, or vice versa, with the Mid-To-High Latitudes being noisier. Is the higher variability a result of the Southern Annual Mode, like the North Atlantic is impacted by the NAO? Dunno.http://i37.tinypic.com/vqmsrd.pngAnd here’s the short-term version of the Low Latitude (25S-0) and Mid-To-High Latitude (60S-25S) South Atlantic OHC comparison.http://i37.tinypic.com/1z3bif9.pngIs the 2003 OHC shift a natural effect? Is it a result of an overcorrection of ARGO biases? Dunno. I’ll try to take a closer look at the Southern Ocean and Arctic Ocean OHC data soon.

  • cohenite

    So, the 'artifact' is mainly a factor of something in the Atlantic, predominantly the North Atlantic, data?

  • cohenite

    So, the ‘artifact’ is mainly a factor of something in the Atlantic, predominantly the North Atlantic, data?

    • http://bobtisdale.blogspot.com/ Bob Tisdale

      Cohenite, the North Atlantic shows a long-term rise that appears to be caused in part by cycles in AMO/AMOC, but the South Atlantic has a more variability with a sudden upward shift in 2003.

  • http://bobtisdale.blogspot.com/ Bob Tisdale

    Cohenite, the North Atlantic shows a long-term rise that appears to be caused in part by cycles in AMO/AMOC, but the South Atlantic has a more variability with a sudden upward shift in 2003.

  • cohenite

    So, the 'artifact' is mainly a factor of something in the Atlantic, predominantly the North Atlantic, data?

  • http://bobtisdale.blogspot.com/ Bob Tisdale

    Cohenite, the North Atlantic shows a long-term rise that appears to be caused in part by cycles in AMO/AMOC, but the South Atlantic has a more variability with a sudden upward shift in 2003.

Bad Behavior has blocked 51434 access attempts in the last 7 days.