Science a GoGo's Home Page Forums General Discussion Climate Change Forum Differences in Global Temperature Estimates

First off, this is a completely honest question! I'm just trying to clear something up for myself. It's too bad I have to start a thread off with these disclaimers, but regardless....

I went and had a look at the monthly estimates of global temperature. Looked at two sources. The Hadley Institute and NOAA. At first I just wanted to see how widespread our (southern Ontario) cool spring was, but then I wondered if the separate datasets lined up with each other.

I quickly realized, they weren't going to line up perfectly with one another. The temperatures were reported as anomalies from the long term average, which was different for each dataset. The Hadley Institute uses the 1961-1990 as the average, NOAA uses 1901-2000 as average. Because of this I expected NOAA to have larger anomalies than Hadley (since NOAA is using the cooler average) - but I figured it would be a straightforward upward translation (both would show the same fluctuations and trends).

But not so. The Hadley Institute shows almost no increase in global temperature over the last 6 and a bit years. NOAA on the other hand shows about a 0.1 C rise over the same time. I've thrown a quick graph together showing this below (hope it works).





I haven't done anything to the data, pulled it directly from
http://lwf.ncdc.noaa.gov/oa/climate/research/anomalies/anomalies.html
http://hadobs.metoffice.com/hadcrut3/diagnostics/global/nh%2Bsh/index.html


Am I missing something? If not, is there any insight into this difference?

From what I can tell - it can be due to NOAA trying to "fill in" gaps in the climate network by kriging, whereas Hadley excludes these areas.

Any ideas anbody?

I'm not sure about some of your questions, or what you see as problematic; but let me just give my impression of the graph.

Both lines seem to be measuring the same parameter -moving in lockstep (roughly). The NOAA data seems to have a bit more resolution and range in it (that might be a clue), but they look very similar.

So there is a <0.1 degree drift over 6 years between two different sources. Is this that much of a problem? These are referred to as "Estimates." I'm impressed that the difference isn't larger.

What about the ~0.5 degree increase, over whatever "normal" baseline is used, that both lines reflect (~constant over the 6 years)?

Maybe I'm not understanding the purpose of the graph. You say the Hadley data shows "almost no increase" over 6 years.
Do you mean- except for the 'plus' ~0.5 degree anomoly?

Let me know what....

~~Samwik

I guess I wasn't clear enough - sorry about that.

My concern is, both NOAA and Hadley use the same base data. Maximum and minimum monthly temperatures from the global climate station network. If they're using the same data, they should be showing the same thing.

Yes, NOAA does show a higher anomaly than the Hadley. This is because of the cooler "average" temperature NOAA is using to calculate the anomalies, than what Hadley is using. I don't see that as an issue.

What is concerning to me, is the trend differs for each plot. Hadley shows almost no rise in temperature over the past 6 years, NOAA shows a very significant rise. Again, these plots are based on the same data! How can that be?

Either NOAA or Hadley are doing something to the data that is causing this discrepancy. Which one is right? Which one is wrong? Which one is influencing policy?

Without looking at either website, here's my 2 cents:

If what you are graphing are deviations from the norm, the norm must be lower for the larger data set (NOAA) than the smaller (Hadley) data set.

soilguy - Look at the trends (thick black lines), not the absolute differences between the sets. I'm not concerned with the differences caused by the different normal periods.

Differences between the norms will not effect the trends. The only thing that will effect the trends are the monthly estimates of global temperature. NOAA shows an increasing temperature of about 0.08-0.09 over the 6 year period. Hadley doesn't show any significant increase in temperature.

To clarify - the trends are the critical thing here......one is trending upwards, one is stable.

How can this be, when both are using the exact same data?

"Differences between the norms will not effect the trends." -Canuk

Why not?
I'd think if they use a different baseline (norms?), to compare the same readings to, that they'd get different "answers."

...and, "the trends are the critical thing here" -Canuk

Why?
Isn't the y-axis showing a ~0.5 degree "anomoly" from some supposed baseline? Isn't that the critical point? I see the graph as a 6 year snapshot of a much longer line that may trend up or down by 10-20% (~0.1degree), and over decades may trend up or down "overall" (again, limited by the time range).

Isn't this focusing on a discrepancy that is small compared to the main point of the graph? Let me repeat from previous post:
"You say the Hadley data shows 'almost no increase' over 6 years.
Do you mean- except for the 'plus' ~0.5 degree anomoly?" -samwik

~~Samwik

In terms of the individual anomalies you'd be correct. That's not the issue here - as I said, I expected there to be a difference between the anomalies. Different norms, different anomalies.
But I'm talking about the trend of the anomalies. The difference between the two datasets should remain the same (regardless of the normal selected) for every time interval within that 6 years - provided that they are calculating similar global temperatures.
Now, what the graph is showing, is that the difference between NOAA and Hadley is NOT staying the same, but is actually increasing. This is despite using the same orginal data.




No - that's not the critical point here. The critical point is NOAA is showing global temperatures continuing to increase over the past 6 years, while Hadley has shown temperatures to flatten.
So with NOAA - the past warming is continuing. With Hadley, the past warming has greatly slowed or stopped.



Again - the issue is that here we have two respected, leading edge institutions, and they can't agree if temperature is increasing or has stabilized over the past 6 years.



This is not a small issue. NOAA is showing temperature to be increasing at phenomenal rate. 0.8 C over 6 years. That's huge. It's a warming rate of 13.3 C/100 years (obviously this rate can't continue, I just scaled it up for illustrative purposes). Is there this much uncertainty in our estimates of global average temperature?



Once again - I'm not concerned about the +0.5 degree anomaly. I'm talking about the fact that the start of the linear trend line for Hadley is at 0.44 C, and ends at 0.45 C (a difference of 0.01 C), and the start of the linear trend line for NOAA is at 0.51 C, and ends at 0.59 C (a difference of 0.8C).
0.8 is much higher than 0.1, which means the NOAA rate of warming over the past 6 years is 8 times what Hadley is showing.


They could present global temperatures as absolute values, and get away from the complicating factor of differing means, but the end result would be the exact same thing. Over the past 6 years global temperatures estimated by NOAA are increasing much, much faster than global temperatures estimated by Hadley. And they're using the same climate data!

"...which means the NOAA rate of warming over the past 6 years is 8 times what Hadley is showing." -Canuk

If this were true over a 100 yr. (even multi-decadal) time span, I would agree with you that "This is not a small issue."

But the time scale of this graph, compared with the differing resolution (variance on y-axis), makes it meaningless to draw a conclusion about a "trend" or "rate." IMHO

What does it look like on a larger time scale?

~SA

I'll try once more.....

My concern is, that you have two of the leading meterological instutions in the world, who are taking the exact same data from from the exact same climate station network, and coming up with different results.
It doesn't matter if that's only over 1 year or 100. We're not looking at whether the trends are statistically signifcant, we're looking at whether two different techniques for estimating global temperature produce similar results (aka similar trends). Apparently they don't.

I don't know what to say. Your point is "...exact same data from from the exact same climate station network, and coming up with different results." -Canuk

Well sure, that sounds like something must be wrong.
If they'd used the same techniques to analyze the data, then I could see questioning the differing results (even if not very different).

"...we're looking at whether two different techniques for estimating global temperature produce similar results (aka similar trends). Apparently they don't." -Canuk

different techniques?
for estimating?
produce similar results? (or not)

"I'm impressed that the difference isn't larger." -Samwik (1st post)-- Y'know, because they're different techniques, and they're estimates.

"We're not looking at whether the trends are statistically signifcant" -Canuk

But we should look. I don't feel they are signifcantly different, so to me the different techniques DO produce similar results.

Even if they are signifcantly different, then I'd just chalk it up to the difference in techniques. If that is the case, then maybe someone should ask if one techniques is better or not.

I think I do get what you're pointing out as a problem; but for the reasons posed above I don't see it as a problem. Are you familiar with statistical analysis? Maybe you can fill me in on some things. I only took one stat. class back in the late 70's (400 level).

Does using "different techniques" seem unscientific to you, or invalidate the results in some way?
I'd be more curious about the estimates. What kind of estimates and where in the process do they come into play?

Let me ask you this. Regarding the drift that you point out in the NOAA results: If you extrapolate (from this small snapshot) backwards in time, wouldn't it lead to an absurdly low temperature after some decades? Do you think that is how their technique is supposed to work?
Similarily, with Hadley, the result (extrapolated from this small snapshot) might not extrapolate back to actual temperatures. Do you see why that wouldn't invalidate the "technique?"

I know we're starting to repeat ourselves here (sorry), but what about the questions at the end (here, above)?


~Samwik