If anyone can explain to me why I am getting this i’d be very appreciative. The fact Spannagel record appears to have been tuned to the 14C concentration means almost certainly that the dating is off! Unless i can get hold of an independent age model for Spannagel, I don’t think it will be useful in the slightest.

Apologies for repeating the post in previous threads, but the use of this is a proxy in reconstructions is alarming considering the number of caveats with this record.

]]>After more closely looking at the temperature and precipitation data extracted from the GHCN (daily Wudu), GISS (monthly Wudu) and TuTiempo.net, I found that the years 1968 and 1999 (for the time period 1952-2003) should be excluded due to my inability to locate complete data at either of the three links below. Note that the year 1974 is excluded due to the lack of O18 data for that year in the Zhang paper.

I had reported the data as the Zhang authors did in their paper using a 5-point average (average of the point of interest plus the 2 above and 2 below). I did not adjust the slope standard deviations as I was unsure how to properly do that with the 5-point averages used. As a matter of convenience, I went back and determined the lag 1 correlation of the regression residuals and adjusted the slope standard deviations using the techniques used in Santer et al. (2008). I show all the pertinent regression values and the adjusted slope standard deviations for each regression below. The Wudu precipitation data was used in units of meters of rainfall, the Wudu temperature was used in degrees C and the NH temperatures were anomalies in degrees C.

The adjusted standard deviations show that, in all cases except one (including the O18 versus NH temperature over the 1880-2003 time period), the regression slopes are not statistically different than 0 (p equal or less than 0.05). The only regression that did not include 0 in its 5%-95% CI was the O18 versus Wudu temperature for 1952-2003 and that CI almost touched on 0.

If the lag 1 residual adjustment for the slope standard deviation is properly applied here and with the dating uncertainty in mind as noted by the Zhang authors, my simple-minded analysis would indicate that going beyond a qualitative statement about the relationships of temperature and precipitation to the O18 values in the Zhang paper would be a stretch.

]]>After reading a few papers on the dating processes involved here, I have a better feel of why the uncertainties exist. That brings me to reason for questioning in the first place: What is the more exact nature of that uncertainty? The discussion at the Fleitmann 2007 thread has touched on the issue of wiggle matching and whether that is appropriate given the gensis of the dating errors. It would appear that the author, Fleitmann, has consulted statisticians on these matters without a solution. I would simply like to see a more complete explanation of the errors or be pointed to a link that can afford one.

]]>The problem associated with assigning dates to individual isotope values stems from several problems. 1) Different samples are used for the oxygen isotopes and dating; 2) the size of samples used for dating is very much larger than that for the isotope measurement, and 3) it is impossible to correlate or interpolate the position of samples taken for isotopes with those taken for dates.

Paul, I can comprehend 1 and 2, but I am still attempting to get my mind around 3. The isotope measurements are listed along with an individual year in the Zhang paper, and I, therefore, have the idea that someone has attempted to make an interpolation to an individual year. A larger sample for dating than isotope measurement and yet the dating for that sample can be within plus/minus 1 year.

It’s not that I have any reason (or the background) to doubt any of these explanations or published precisions. I just wish I could visualize it better.

]]>Kenneth, I would think that the age plus precision in the second column of your table is the calculated age for a particular sample submitted for dating. U series can produce ages with very high precisions. The Minnesota lab is amongst the best in the world.

The problem associated with assigning dates to individual isotope values stems from several problems. 1) Different samples are used for the oxygen isotopes and dating; 2) the size of samples used for dating is very much larger than that for the isotope measurement, and 3) it is impossible to correlate or interpolate the position of samples taken for isotopes with those taken for dates.

]]>The Zhang paper does reference results using climate model simulations that might be of interest to analyze, but the SI for that paper did not have the temperature data used in the paper nor a reference to where it might be located and linked.

I have downloaded the daily temperature and precipitation for Wudu per Steve M’s instructions and would now like to do comparisons of the temperature and precipitation series.

My problems are with understanding the 15 year uncertainty in assigning particular O18 data that was noted in the Zhang paper comment and excerpted below:

Errors are small, ±1 to ±5 years (2σ), due to high uranium concentrations

(from 5.8 to 10.8 ppm) and low initial thorium contents. The error in assigning a date to a

particular oxygen isotope value is about 15 years because of the uncertainty in relating

the position of the sub-sample used for dating to the position of the sub-samples used for oxygen isotope analysis.

The table below, excerpted in part from the Zhang SI, list in the last column some ages with 2 sigma limits that appear to me to indicate that date resolution is rather fine. On the other hand, when the Zhang authors say that the assignment of a particular date can have an uncertainty of 15 years, I am not certain what this means. Is the 15 years a 2 sigma range? Would not an assignment of a date to a given O18 measure, correct or not, in turn temporally locate the adjacent assignments through the resolution of the dating or are the adjacent date assignments made strictly by distance from the measured data? Could a whole segment of adjacent dates be misplaced by the same amount or could the dates within the segment be “off” by various amounts?

Sample——————- Age

WX42B-2-1—————- 1990 ±1

WX42B-2-2—————- 1990 ±1

WX42B-0 —————— 1970 ±1

WX42B-2-2 ————— 1943 ±1

WX42B-1——————-1897 ±1

WX42B-2-3 ————— 1868 ±1

WX42B-2 ——————1709 ±1

WX42B2-4—————– 1602 ±2

WX42B-3 —————— 1424 ±1

WX42B2-5—————– 1333 ±2

WX42B-4 —————— 1215 ±2

WX42B2-6 —————- 1104 ±2

WX42B-5 —————— 971 ±3

WX42B2-7 —————- 897 ±3

WX42B-7—————— 752 ±2

WX42B-8 —————– 571 ±3

WX42B-9 —————– 489 ±5

WX42B2-9—————- 427 ±3

WX42B-10 ————— 348 ±4

WX42B-11 ————— 192 ±4

I have absolutely no desire to publish a letter on this matter or most others. If anyone wanted to work it up into a letter they are welcome to any material I have. I doubt very much that the point I am attempting to make here would be deemed important by the editors.

]]>The Wudu temperatures in the regressions are in degrees C and for the NH temperatures anomalies in 100ths of a degree C were used.

It’s a little confusing to use these different units in your regressions, since it causes an apparent change of two orders of magnitude between the response to Wudu temperature and NH temperature. NH temperature should have a somewhat muted effect, but this is pretty extreme!

The only reason that NH temperatures are sometimes measured in .01dC units is that using 80-column 1960s-vintage punched card computer technology, if you didn’t drop the decimal point, you lost a valuable digit of precision. Today that’s no longer relevant, but bureaucracies are slow to change their ways…

On this blog it doesn’t much matter, but when you write up your letter to Science and accompanying SI refuting Zhang et al’s orientation, you should use consistent units of dC.

]]>Slope it is and slope it should have been.

]]>