Ice Core CO2 and Temperature
By Ed Caryl
There has been some interest in the lag in ice cores between temperature as measured from oxygen isotope differences, and the CO2 content as measured directly in gas bubbles. The literature seems to agree that the lag is something between one millennium and 800 years. As I had downloaded the EPICA Dome C data for both, resolved to 100 years, I decided to play with the numbers.
The first thing discovered is that the CO2 data resolution for times older than 22,000 years was not 100 years, but a variable number much coarser than that, interpolated in 100 year intervals, The original data resolution is up to more than a thousand years for some stretches of time.
The other thing discovered in the charted data are three groupings in time: the ice age time up to 20,000 years ago, the melt interval from 20,000 to 10,000 years BP, and the recent 10,000 years. The following charts show what the whole 100,000 year time interval looks like. The first figure is an XY plot of temperature versus CO2 concentration for the last 100,000 years.
Figure 1a is the CO2 plotted against temperature with no delay. Figure 1b is a plot of best fit for delays between plus 1000 years and negative 2500 years. For best R2 value the delay is 900 years. For highest temperature response versus CO2, the steepest linear trend line, it is at 2200 years.
An 800 to 1000-year value is what most researchers are finding looking at ice core data. Look here, here, and here. But in the current era, the last 50 years, lag times are less than a year. Look here, here, and here.
But notice the grouping in Figure 1a. There seems to be a difference between colder and warmer times, and a large gap where the temperature was increasing rapidly. What do we see when these groups are isolated and measured?
Figure 2a is the CO2 plotted against temperature with no delay for the time period from 100,000 years BP to 20,000 years BP. Figure 2b is a plot of best fit for delays between plus 0 years and negative 2500 years. For best R-square value the delay is 2100 years. The highest temperature versus CO2 response (the steepness of the trend line) is 2200 years.
During the ice age, the data has low resolution leading to low R2 values, but the delay is longer than 900 years. It is now measured at 2100 and 2200 years for the R2 and temperature response values, a much longer delay. For the emergence interval when the ice is melting and temperature and CO2 are rising sharply, and the CO2 data resolution is much higher, we should get a clearer picture of the delay, and indeed we do.
Figures 3a and 3b are of the time period from 20,000 years BP to 10,000 years BP. During this interval the delay is much shorter, the R value much higher, and the delay is about 400 years. It is longer at 1000 years for the maximum temperature change for a CO2 change (the trend slope).
So now that it has warmed up, and again holding a more or less constant temperature and a more or less constant CO2 value, what is the delay during the last 10,000 years?
Figures 4a and 4b show a multiple number of delays, including zero delay, 300 years and 700 years for the last 10,000 years. It looks like multiple delays are happening in this interval.
In Figure 4a, there appear to be two groups of data points, above and below 265 ppm CO2. The dividing point happens to be at 5000 years ago. If we split the data at 5000 years BP, we get these two sets of plots.
Figure 5a and 5b are plots from 10,000 year BP to 5000 years BP. In this period the delay is 700 years and both methods agree.
Figures 6a and 6b are plots from 5000 year BP to the present. The delay is zero but the R2 values are very low. The relationship between temperature and CO2 appears to be very low.
I will summarize with two more plots. Figures 7a and 7b are summaries of the previous plots:
Figure 7a is a plot of delay versus the R2 value, and 7b is the delay versus the average CO2 concentration during each interval.
As the CO2 concentration goes up, the delay, and the R2 value goes down, until at the present time, and indeed for the last 5000 years since the CO2 concentration has reached 265 ppm, there has been no delay.
There are several possible interpretations for these observations.
First, there is a delay dependent on CO2 concentration. The difficulty with this interpretation is that no one has proposed a mechanism for such a delay. Indeed, few have even noticed that the delay changes over time.
The second possibility is that there is no delay, and that what we observe is simply due to gas rising in the firn before it closes off. This has been proposed, here, as at least a partial explanation. But this doesn’t explain the variation in the delay, and certainly does not explain why there is no delay at all for ice less than 5000 years old.
A third explanation might be that the delay is due to the thermohaline oceanic circulation. When the ice age is at maximum, the circulation slows and large temperature fluctuations (see Dansgaard–Oeschger events below) during the ice age get embedded in the circulation, reappearing as large CO2/temperature differences when sea water returns to the surface after one cycle. When it warms up, the circulation speeds up again. When it gets as warm as it is now, the constant ocean surface temperatures completely hide the delay as the temperature differences are lower, and short term temperature versus CO2 solubility in the surface ocean dominates.
A fourth explanation might be that the delay is related to Dansgaard–Oeschger (D-O) events. These occur at multiples of 1470 years, and a few missing ones in the record would account for the average delay measuring 2200 years with a rather broad peak. The third and fourth explanations might be combined. The equivalent Bond events in the Holocene are of a much lower magnitude, and may not appear in the last 5000 year record.
During the ice ages the atmosphere was much drier than during interglacials. We know this because we know that deserts and grasslands were much larger and rain forests smaller. There is much less evaporation over ice than over open water. The greenhouse effect would be lower due to both CO2 and water vapor reductions. During the Holocene, CO2 increased along with water vapor. The temperature response to both is logarithmic, the same added CO2 and H2O has less and less effect. However, the larger variable is H2O. It can vary from nearly zero over deserts to saturation over tropical seas. Of these two gases, water vapor in all forms is the control on temperature, not CO2. This is why we see temperature driving CO2, and not the other way around.
Nowhere here do we see any indication that CO2 is driving temperature.