Category Archives: Climate models

Climate models, global climate change

Volcanoes: Ridley et al. (Bleg)

9 Comments

I’m reading “Total volcanic stratospheric aerosol optical depths and implications for global climate change
D. A Ridley1, S. Solomon2, J. E. Barnes3, V.D. Burlakov4, T. Deshler5, S.I. Dolgii4, A.B. Herber6, T. Nagai7, R. R. Neely III8, A.V. Nevzorov4, C. Ritter9, T. Sakai7, B. D. Santer10, M. Sato11, A. Schmidt12, O.Uchino7, J. P. Vernier13,14″

It’s an interesting paper. I have a big question and I’m hoping to that some readers might be able to point to resources. Briefly:

Is there a brief resource that lists the volcanic forcings used by groups who contributed projections to the AR5.

Specifically, I want to know which groups ‘froze’ volcanic forcings after 2000 and which did not. For groups that did not freeze volcanic forcings in 2000, I’d be interested in knowing when the year they did freeze them, and at what level those forcings were frozen.

RidleySaysFrozeThe reason I wish to know: Ridley seems to suggest “many” froze forcings. I don’t know what fraction ‘many climate model studies to date’ constitutes nor which models. Are the “many climate model[s]” the AOGCM’s used to create projections in the AR5? Or are they EMICs? Etc. Of the AOGCM runs, is ‘many’ 20% ‘froze’ or was it more like 80%?

My impression (which could be incorrect) is that Model E did not freeze volcanic aerosols in 2000. (NASA’s Model E page says,

“The simulations (using TCADI) denoted by p3001 are continuations from 1991 of the p3 simulations including further updates to the volcanic and solar forcings through to 2012. The volcanic forcings are from a more recent version of the Sato et al. data than used in the historicalExt runs described below.”

But perhaps I’m jumping to the conclusion that these were the runs used for “projections” in the AR5 and perhaps they did ‘freeze’ volcanic aerosols in 2000 and then use that to create projections. (Note: if they ‘froze’ in 2012, one might expect the ‘mean’ for their trends from 2012-2014 should be lower than the earths, because the volcanic aerosol cleared somewhat. Any discrepancy would be undetectable as it would be swamped by ‘weather noise’.)

I would like to know the year other modeling groups who contributed to the AR5 (or AR4) ‘froze’ volcanic aerosols.

BTW: I’m pretty sure it’s true that of those that included volcanic aerosols in the AR4, ‘many’ did freeze in 2000, though once again, I believe Model E did not. I think they froze later– around 2003 or so. That said: many models in the AR4 did not include volcanic aerosols at all; that makes it very difficult to speculate how the variation in aerosols in 2000 mattered– as one ought to two off-setting factors: (a) that those models were not including the tail end of the ‘recovery’ to Pinatubo — which might be small but still not zero– and (b)also did not include the ‘cooling’ effect of newer, but much smaller, eruptions. Nevertheless, it would be interesting to have a listing.

I know someone is going to be telling me I can hunt these down at PCMDI: I prefer not to slog away at PCMDI. What I want is for someone to tell me if someone else has already gone through and at least has information on a ‘yes/no’ froze in 2000 level and/or ‘which year froze’. Or if someone else wants to slog through PCMDI and find out the year any particular group ‘froze’ volcanic aerosols, that would be lovely. This is not a task that requires high technical skills: just time. So if you are curious, and have time, you could look. 🙂 )

The effort is interesting because it is important because the degree to which one ‘ought’ to believe the discrepancy between either AR4 or AR5 models and observations can be ‘explained’ by projections being based on ‘frozen in 2000’ aerosols depends on the fraction of modeling groups that froze volcanic aerosol forcings in their input files when creating runs to actually contributed to the AR4 or AR5. I have no idea what this fraction is.

If “many” froze them means roughly 20% did, then the discrepancy between model mean and runs is difficult to explain by the difference in ‘real’ and ‘modeled’ volcanic aerosols. In contrast, if 80% froze them, we can test that too. We can also dig through and see whether there is any systematic difference between 2000th century trends between models that did ‘freeze’ volcanic aerosols and those that did not. (Assuming some did and some did not, of course.)

I have a few other comments that I could make on this paper. The main one that first comes to me is this:

When finding likely effect of forcings on trends in the 2000’s it strikes me as a bit ambitious to simply do everything relative to the volcanic forcing in 2000. The reason is that the atmosphere does have a response time. So, if one wants to find out the effect incorporating forcings in the lower stratosphere into the recent trends, one ought to know what those were in past years before deciding the effect of any increase in Stratospheric Aerosol Optical Depth (SAOD) during the 2000’s. It appears no AERONET data exists prior to 1995; I’ll take the researchers word that the data are tenuous prior to 2000. But this is a factor we should consider– and possibly one ought to widen the range of possible effect on temperature in that paper. (Perhaps doubling it.)

That said: it’s an interesting exercise, and worth being aware of. It could potentially explain the discrepancy between the model mean and the observed trend. (Which will mean that the discrepancy between a projected model mean trend of “about 0.2 C/decade” existed and still does as opposed to ‘did not exist’.)

Climate models, global climate change

Temperature: Compare to AR4.

173 Comments

I haven’t been following Climate so much. But I noticed ‘rumors’ that the hiatus was over… on twitter that. This motivated me to whip out my script and see how temperature have progressed. As the AR5 is now “officially” out, I thought it might be useful to compare the temperature to the previous prediction from the AR4, which is now so “yesterday”. Nevertheless, many of the AOGCM’s used to form the basis of the AR5 are more-or-less the same as those used in the AR4, so I thought some of you might enjoy seeing the observations shown next to the multi-model mean from the AR4.
TempsSince2000

The above shows annual average temperature along with the IPCC projection of annual average temperatures.

Recall: In the AR4, the IPCC projected warming to proceed at about 0.2 C/decade for the first two (or was it three?) decade of this century. Whatever it was, we are quite a large fraction of the way there. Also, in the AR4, like it or not, the IPCC authors elected to communicate uncertainty using ±1 σ error bars around a multi-model mean. To match them I show the observed temperature on a similar graphs. You can see the present temperature while certainly ‘up’ relative to recent temperatures is just brushing the lower -σ boundary for temperature itself. If we use a ‘red noise’ model for time series to compute uncertainty intervals for trends consistent with observations, the 0.2C/decade warming trend is well outside the 2σ uncertainty intervals consistent with the observations.

Of course, this is now old news: Even the IPCC has admitted a ‘hiatus’. Even if many won’t quite want to admit the models are somehow “wrong”, the peer reviewed literature is now permitting people to observe that the temperatures are not exactly following the mean trend. But it’s well worth comparing observations to projections now that those projections are (likely) to be deemed “old news”.

Anyway: I’m rather unconvinced ‘the hiatus’ is over. That said: it’s a bit difficult to say for sure because the definition of ‘hiatus’ is rather vague. It does seem to me we are going to need to see quite a bit of warming to overcome the doubts of those who think models are not well suited to predicting warming over periods as long as 20 or 30 years.

Climate models, Data Comparisons, Toy Physics

Observation vs Model – Bringing Heavy Armour into the War

332 Comments

As I have noted before, most of the AOGCMs exhibit a curvilinear  response in outgoing global flux with respect to average temperature change.  One of the consequences of this is that there is a well-reported apparent increase in the effective climate sensitivity with time and temperature in the models; in particular, the effective climate sensitivity required to match historical data over the instrument period in the GCMs is less than the climate sensitivity reported from long-duration GCM runs.   This is not a small effect, although it varies significantly between the different GCMs.   In the models I have tested, it accounts for about half of the total Equilibrium Climate Sensitivity (“ECS”) reported for those models.   (Equilibrium Climate Sensitivity is defined by the IPCC as the equilibrium temperature in degrees C after a doubling of CO2.)  In general, models which show a more pronounced curvature will have a larger ratio of reported ECS to the effective climate sensitivity required to match the model results over the instrument period, and vice versa.

Kyle Armour et al have produced a paper, Armour 2012 , which offers a simple, elegant and coherent explanation for this phenomenon.   It comes down to geography.

Continue reading Observation vs Model – Bringing Heavy Armour into the War →

Climate models, Statistics

Uncertainty: Let’s just call it an acquaintance!

69 Comments

Recently, I became aware of the oddly titled Uncertainty is Not Your Friend by Stephan Lewandowsky posted at Planet3.0org back in June. I vaguely recalled reading this and thinking it a rather strange mixture of mangled text, graphics and math. Ben Pile commented on the paper blog post. But at the time I didn’t consider the pastiche of misinformation mixed with some not-necessarily wrong observations was worth commenting on. Today I’ll comment on a few items so as to clarify which statements in that specific article appear totally false, which are worded so strangely as to be impossible to parse and note things that happen to be true.

I will start by saying this: Of course it’s true that uncertainty is “not your friend”. I will add that “A bird in the hand is worth two in the bush” and “Better the devil you know.”.

I’ll also say that generally speaking, when one accounts for uncertainty of anything we might call X, it is often the case that the cost of our best estimate of X will be lower than our best estimate of the cost of X. This appear to be the case that for climate change. (If we use the mean or expected value to represent our “best estimate” of anything, the previous statement can be expressed using mathematical symbols. If $latex E[X] $ is the expected or mean value of $latex X $ and $latex C(T) $ is the cost associated with a temperature rise of $latex T $ , then often $latex C(E[T]) < E( C[T]) $. I don't know if the property $latex C(E[T]) < E( C[T]) $ means uncertainty is "our friend". I would suggest uncertainty would not be our friend if $latex C(E[T]) > E( C[T]) $ either. I’ll defer discussing costs for now.

I want to focus on certain specific verbiage in Lewandowsky’s first post which seems to focus on the probability that temperature would change more or less than “anticipated”. Specifically, in his first of three posts he writes things like

So uncertainty doesn’t just mean that things could be worse than anticipated—in the case of climate, chances are that things will be worse rather than better than anticipated.

Uncertainty in climate evolution means things are likely to be worse, rather than better, than anticipated.

In paper 1, he the “thing” he focuses on is the change in temperature. But I had to scratch my head for a while because the following question popped into my head: “Does that depend on what is ‘anticipated’?”

I continued reading and hoped the examples might reveal which properties of the uncertain temperature defined the changes that are “anticipated”. Based on the example, it seemed that — possibly- the value that assumed to be ‘anticipated’ (by someone) is the mean value of the temperature change. At least, it seem to clarify his points, Lewandowsky used probability distributions for climate sensitivity that all shared the same mean temperature. To wit:

To make my point, I ensured that the mean of the four distributions is identical (around 3, with a tiny amount of deviation introduced by the simulation). However, the standard deviations (spread) of the distributions differ considerably, from .49 in the top left to 2.6 in the bottom right. The spread of each distribution characterizes the extent of uncertainty surrounding the mean estimate of 3 degrees.

Reading this, it seems whatever Lewandowsky means, he is comparing probabilities computed when we assume a) similar functional forms for the probability distribution function, b) identical mean values for the change in temperature and c) an increase in variance about the mean.

So, I thought I’d just plow through the math, making changes where I thought suitable. I noticed Lewandowsky drew heavily from “Roe, G. H. & Baker, M. B. (2007).” (RB) but substituted his own probability density function for theirs which is based on a normal distribution for the “climate feedback parameter” $latex f $. I thought I’d follow RB more closely, but introduced a tweek motivated by problems that arise in their probability distribution function as their dimensionless climate feedback parameter approaches $latex f $ approaches or exceeds 1. (Some of these problems were identified by Zaliapin, I. & Ghil, M. 2010). For the purposes of the discussion here I think it is sufficient to resolve the problem by simply noting that empirical evidence suggests $latex f<1 $. To reflect this known requirement, I replaced the normal distribution for $latex 1-f $ used by RB with an Inverse Gaussian Distribution with mean freed back $latex \overline{f}= E[f] $ and standard deviation $latex \sigma_f $ Then, to create my examples, I used the numerical values of $latex \overline{f}= E[f] = 0.62 $ and $latex \sigma_f =0.13 $. These values originate from a “suite of GCM simulations” and correspond to trace “A” in RB2007 figure 3. The resulting probability density function for the temperature rise shares the skewness of those in RB and the log-normal distribution used by Lewandowsky

However, if we are to use the “method of eyeball” to compare features like the mode, mean, median, it is more convenient to examine the cumulative distribution plot. Below, I show the cumulative distribution for the parameter values I picked:

Notice that I’ve highlighted the “mean value” in red. On the right hand side, I’ve indicated the temperatures that bound 90% of the range we might anticipate. So, based on this distribution, we might say that that we anticipate that the temperature will likely fall between 1.93 C ad 5.76 C. The expected value is 3.53C. We think there is a 5% chance the temperature will fall below 1.93 C and a 5% chance it will fall above 5.76 C.

What if the uncertainty was higher?
Now, suppose that someone suggests that we should explore what happens to our probability estimates if we:

  1. Use the same functional form for the probability distribution function
  2. Hold the expected value of temperature constant (as Lewandowsky did) but
  3. Increase the uncertainty. That is: double the standard deviation about the expected value.

When we do that we obtain the following cumulative distribution function:

Recall that in the previous it was decided that the “anticipated” temperature range was (1.93C to 5.76 C); there was a 5% chance chance temperature would fall below this range and a 95% chance they would fall above this range. Examining the new distribution

  1. The probability that temperature will fall outside the (1.93C to 5.76 C) range is now estimated to be 31% which is larger than 10%. This outcome is not surprising: uncertainty is defined this way.
  2. The probability the temperature will fall below the anticipated lower bounds has increased to 19.1%.
  3. The probability the temperature will fall above the anticipated upper bound is 12.1%.
  4. 19.1% > 12.1% so the probability that temperature will fall below the range currently anticipated will rise more than the probability it will rise above the range currently anticipated.

Based on this, I would suggested that based on what people might understand by the words “anticipated” and “likely” (meaning more probable) and the model for probability distribution I described, the following claim appears to be backwards:

Uncertainty in climate evolution means things are likely to be worse, rather than better, than anticipated.

The figure above shows that assuming that if, as we gain understanding, the mean value of the predicted temperature rise remains constant but our uncertainty in our ability to predict temperature increases, then we will find the probability that temperature will fall below the current estimated lower bounds increase more than the probability that temperatures will exceed the current estimated upper bounds.

For those who might think finding the increasing uncertainty means that the probability of lower temperature rises rises more rapidly than the probability of higher temperatures means “uncertainty is our friend”: That’s not the case. Uncertainty is uncertainty. It’s really never a “friend”. We can better understand why uncertainty is not a friendly thing when we turn to discussing costs. At that time we can discuss whether its unfriendly nature means what Lewandowsky thinks it means.

References
“Roe, G. H. & Baker, M. B. (2007). Why Is Climate Sensitivity So Unpredictable?Science, 318, 629-632.”

Zaliapin, I. & Ghil, M. (2010). Another look at climate sensitivity. Nonlinear Processes in Geophysics, 17, 113-122

Climate models, Data Comparisons

Ocean Heat Uptake Efficiency, Chicken-laying Eggs and Infinite Silliness

75 Comments

In a recent post here, I discussed the nonlinearity of the Earth’s radiative flux response to temperature change exhibited in most of the GCM results, and the application of an incompatible linearised model to those results for the purpose of analyzing climate feedbacks.

The conversation in the comments section, perhaps inevitably,  got onto the subject of “ocean heat uptake efficiency”, and more specifically onto the 2008 paper by Gregory and Forster, (“GF08”).  The assumption underpinning ocean heat uptake efficiency is that you can characterize the net flux imbalance as a simple linear function of temperature change,  with a constant of proportionality, ĂŽÂş.  I expressed the view in the comments section that the concept was inelegant, ultimately unnecessary and founded on a mathematical fallacy.  With Lucia’s permission, I would like to try to defend this comment, and take the opportunity to address some of the questions which arose.

 

Continue reading Ocean Heat Uptake Efficiency, Chicken-laying Eggs and Infinite Silliness →

Climate models, Data Comparisons

The Arbitrariness of the IPCC Feedback Calculations

251 Comments

Introduction

Professor Isaac Held recently published a blog article  here with the title “The arbitrariness in feedback analyses”.   The title refers to the arbitrary selection of a reference response against which to assess feedbacks.   I agree with much of  Professor Held’s paper  on the issue of the arbitrariness of the starting point, which is a definitional issue, but I also believe that the paper leaves a very large elephant under the table.

I wish here to discuss the elephant.  The elephant is the application of a linearised response function to the nonlinear response observed in the GCMs when a large forcing is applied, and which gives rise to a quite different source of arbitrariness in the feedback analyses  reported by the IPCC .

The IPCC feedbacks calculated for each GCM are highly case-sensitive as well as being sensitive to the choice of the run-time over which they are estimated.  In particular, these values, which are based on large forcings applied in future scenarios, cannot be reconciled with the feedbacks effective (in the same GCM) over the historic instrumental temperature period.

Not only do the calculated values  have  little explanatory value outside the specific cases examined, but, even within those specific cases, the assumption that all individual feedbacks are linear with temperature, and therefore additive, makes the relative attribution between feedbacks within any model, quite fundamentally flawed.

Continue reading The Arbitrariness of the IPCC Feedback Calculations →

Climate models, Data Comparisons, Statistics

More Blue Suede Shoes and the Definitive Source of the Unit Root

35 Comments

Introduction

In a previous article here, I showed  that it was possible to invert the instrumental temperature series into the flux domain, and that the incremental flux series could then be decomposed into three bandwidths yielding a Very Low Frequency component (the “SURE”), a Low Frequency component (represented here by just two cycles of 60.8 years and 21.3 years approximately) and a High Frequency component.

The SURE curve – which controls the  average trajectory of the temperature series – is seen to be a relatively smooth curve, tailing over at late time, when expressed as incremental flux.  The signal shows no obvious evidence of any CO2 effect.  I concluded in my previous article that it is impossible to quantify the effects of AGW from the temperature series alone, notwithstanding that various authors have claimed so to do.

I also stated that this approach – that of inverting the temperature signal to net flux – explains why statistical tests on the temperature series “fail to reject” a unit root in the temperature series.  I will show in this article where the unit root comes from, but before I do so, I would like to clear up three questions which were left outstanding from the first article.

Continue reading More Blue Suede Shoes and the Definitive Source of the Unit Root →

Climate models

Gedanken: Relates to Discussion of Condensation in GCMs

87 Comments

Some of my readers, JeffId, Gavin and others are discussing the possibility that GCM’s do not properly account for condensation on local pressure. I am agnostic on this (I’m trying to think about it.) However, plowing through the comments at The Air Vent, I came across an argument between a number of people (Carrick, SteveF and others). I think the question that triggered the argument is the following one:

94. Gavin said

Consider a thought experiment. Take a closed container filled with super-saturated air and place it on a scale. Will you be able to detect the moment of condensation by monitoring it’s weight (i.e. the pressure on the scale)?

I bet you all think you know the answer to this question. Right? I’m going to answer this question in agonizing nit-picky detail. Continue reading Gedanken: Relates to Discussion of Condensation in GCMs →

Climate models

Comparison of a trend of 0.2C/decade to NOAA: Since 2001 ( For Nathan )

122 Comments

In comments yesterday, Nathan seemed to indicate he wanted to see the result of some sort of test with data beginning in January 2001. Here’s a graph just for Nathan:

Above is a graph showing how the trend of 0.2 C/decade fits into ±95% uncertainty intervals computed for NOAA observation. The uncertainty intervals are estimated assuming the residuals to the linear fit have an ARMA(1,1) structure, and part of the ‘weather noise’ is explained by the MEI index. (I have documented the methodology for estimating uncertainty intervals using ARMA(1,1) only partially at the blog. The method of estimating the uncertainty intervals uses a “number of effective data points” method that is asymptotically correct as we acquire an infinite number of samples. In my applications, the coefficients required to estimate the effective number of data points are based on the first three lagged correlations in the correlogram for the residuals and monte-carlo tests indicate my method returns uncertainty intervals that are slightly too large.)

The graph shown is one of the “default” graphs in the spreadsheet I update when new monthly temperature anomalies are published. I alternate showing new data along with trends 1980 or 2001, with no particular pattern. Sometimes I show a graph with both trends. If I noticed that the 2001 graph flipped form “rejecting” 0.2 C/decade (as it has been for many months) to “accepting”, I would consider that news and show highlight that graph. Similarly, if 1980 flipped from “accepting” to “rejecting”, I’d highlight that one. Otherwise, it’s a toss up; earlier this month I happened to pick the 1980 graph when presenting GISSTemp for this month.

As long as I’m showing the 2001 graph, I’ll discuss what this communicates. When the analysis using the assumptions discussed briefly above is performed, a trend equal to exactly 0.2 C/decade falls outside the ±95% confidence intervals for the trend consistent with the NOAA/NCDC data observations. This means that based on the assumptions of this analysis, if we select a confidence level of 95%, we should treat the assumption that the trend is 0.2 C/decade as false.

In contrast, because, for the purpose of analysis, we’d assumed the trend of 0.2 C/decade projections is true, if that trend fell inside the uncertainty intervals, we’d continue to assume it was true.

Note that currently, if we create equivalent graphs for GISSTemp or Hadley, the analysis based on HadCRUT returns the result similar to NOAA shown above; while the analysis using GISSTemp results in a ‘fail to reject’ conclusion.

In either event, before we’d performed the analysis, we would know that we would ultimately accept a chance of making the wrong decision, with errors falling in two possible classes:

  1. False positive error. (i.e. ‘Type I’ or α error.) Contingent on our statistical model being correct, (i.e. the residuals are ARMA(1,1), etc.), the chance of decreeing 0.2 C/decade false when it is, in fact, true would be α=(1-95%)=5%. The way these tests are constructed, false positive error remains constant at the chosen level of α no matter how much data we collect. I’ve set up my test to give false positives at a rate of 5%.
  2. False negative error. (i.e. ‘Type II’ or β error.) Once gain, contingent on our statistical model being correct, we have some chance of failing to reject the null hypothesis. In this case, that is saying the trend of 0.2C/decade is correct when it is incorrect. The rate of β error can never truly be known because it is a function of the the magnitude of the actual trend– which cannot be known. However, in the limit of zero data, this rate is 95% and declines as we obtain more data. So, this type of error– accepting the null hypothesis, is the error we generally worry about when we have very little data.

Other sources of making the wrong diagnosis is using an incorrect statistical model. This is not directly related to using a short time period. There is, however, an indirect effect because lack of data can make it difficult to perform fiduciary tests to determine whether or statistical model is correct. For example, in the current case, if we have insufficient information, but had assumed the residuals are AR(1) as opposed to ARMA(1,1), white or any other statistical model, tests to show our assumptions are incorrect will have little power.

In any case: What the result illustrated in the graph above indicates is that, assuming the statistical model is true, and assuming 2001 is a good start year, we should reject the assumption that the trend of 0.2C/decade falls inside the range consistent with data.

Have I shown results of other tests starting in 2001 recently?

Since I’m not entirely sure why Nathan seemed to suggest I stopped doing something starting specifically in January, 2001, I think it’s worth showing that I have been recently presented results of analysis using data that begin specifically in 2001

That’s easy enough to show Nathan that I am in habit of including what results of statistical tests if we begin in Jan 2001 along with results starting in other years. For example, in January of this year, I posted various graphs comparing the multi-model mean projections computed from 22 models used by the IPCC to create projections published in the AR4 to observations. That method used in that post permitted me to include the uncertainty arising from both the ‘weather noise’ and spread in biases for that collection of 22 models. The results for Hadley and GISTemp were posted here and here respectively.

Both graphs contain results of analysis initiated in decadal years since 1950, supplemented with the year 2001.

Some new readers might wonder why I always include 2001 in particular. I always include 2001 because the emissions scenarios or SRES used to drive climate models was first released in Nov. 2000 (a noted here). So 2001 is, in some sense, the first year presenting a pure comparison of observations to projections driven by the SRES. (On can argue whether purity matters, but I think it’s worth noting the distinction between testing hindcasts and forecasts.)

I’m aware this post may prompt Nathan to clarify what precisely he thinks I stopped doing. One thing I do not believe I have stopped doing is comparing observations to IPCC projections with the analysis using data beginning in January 2001 and reporting whether the statistical test guides us to deem the projections ‘true’ or ‘false’. Quite often, if we chose a confidence level of 95%, the result of a statistical analysis guides us to conclude a trend that we can connect to information reported in the IPCC AR4 is found to be is false. (For example, the trend of 0.2C/century is diagnosed as false above.) Occasionally, we the analysis does not guide us to conclude a trend is not false. By convention, if that trend was the null hypothesis, we continue to treat it as true– because, by convention, we treat a null hypothesis as true even if when we have absolutely no data.

It seems to me that I show graphs and presents result of analyses using data starting in Jan. 2001 rather consistently and plan to continue to do so. I will, of course, also continue to show what happens if we apply the same analysis using longer time spans, or what happens if we change the assumptions in our statistical model and also sometimes publish blog posts that don’t happen to include a graph or a trend computed starting in 2001.

Update

I goofed up and showed a NOAA graph but discussed GISSTemp. I’ll update by adding the GISSTemp, which may well show trends inside the ±95% uncertainty intervals as that’s what we get when I look at the full model with the spread as see if you click the link discussing that more detailed analysis.

Figure 2: GISSTemp since 2001.
Climate models

HadCrut Compared to IPCC Simulations Ending Dec. 2009.

39 Comments

In a previous post, I reported trend analysis of the difference between GISSTemp observations and the model mean projection for surface temperatures from simulation from IPCC models extended into the 21st century using the A1B scenario tell us we should reject the hypothesis that the model mean simulation agrees with the observations reported by GISS. Specifically, trend analysis beginning with start dates prior to 1980 all indicate we should reject the multi-model mean trend as reproducing the observed trend.

In that post, I also said I would apply the same analysis to Hadley when the December 2009 data arrived. Below, I have a figure showing mean trends for the difference between observations and the multi-model mean and uncertainty intervals based on HadCrut NH&SH through Dec. 2009.


Because the trend is fit to the difference between observations and the multi-model mean simulation, we expect a trend of 0C/century if the multi-model mean simulation faithfully reproduced the mean trend. (The IPCC projections in the AR4 were based on this assumption.)

Examining the figure above, we see the notion the multi-model mean of simulations for the earth’s surface trend faithfully reproduce the HadCRUT observations should be rejected at a confidence of 95% if we elect to do the analysis with start years beginning in 1950, 1960, 1970, 1980, and 2000 or 2001. The uncertainty intervals are based on the assumption the residuals from a straight line can be described by an ARMA(1,1) noise model, and the influence of El Nino has been accounted for using the MEI index.

Some readers may be curious as to whether this result differs from the similar analysis using data through Nov. 2009. Of course, they don’t differ much, but interestingly enough, the trend computed from1980 just kicked over into “rejection” territory; it was in “fail to reject” before the December anomaly was reported. (For earlier results see previous post. Of course, this is a mostly rhetorical issue, since in both cases, the upper range of the uncertainty intervals just grazes the key “0 C/decade” value. Continue reading HadCrut Compared to IPCC Simulations Ending Dec. 2009. →