I’d anticipated the 12 month average would begin a steady decline until the start of the next El Nino. But since that appears to be back in play, I’ll show the 12 month average in this months UAH post:

(Note– despite the apple/oranges element of comparing surface temps to UAH, the multi-model mean for A1B is shown. Both observations and projections are use the Jan 1980-Dec 1999 baseline.

Of course, there is something more important than climate change: Who took the quatloos. This month, Denny took the #1 spot and won quite a few Quatloos. Here are everyone else’s bets:

The net winnings for each member of the ensemble will be added to their accounts.

As summer grows old
my red hot peppers ripen
I need recipes!

This summer the Chicago area has been a bit hot and amazingly humid. I live in the burbs, and sneak veggies into spots planting hot peppers, tomatoes and basil along the side of the house. All are doing just fine. I always grow jalepenos, which we use to cook “dragon turds” (recipe will be made available on request. This year, I added these dastardly little peppers, based on an impulse buy at the garden center. I don’t remember what they are; maybe thai?

Anyway, they are extremely hot and too small to be used for dragon turds. All 6 starts lived and are covered in peppers. Because of my diet, I’ve been on a cooking / exercise binge. (That’s actually the explanation for less blogging here. Some have noticed I’m posting the more successful experiments at my former knitting blog which is now more of a cooking blog. I don’t post the experiments I consider failures. ) Anyway, if any of you have suggestions for this particular type of pepper, I’d give it a try.

Meanwhile, I do notice that Stephan, who is rooting heavily for ice recovery, is aching for us all to look at the dmi ice cover plots:

This ice plot shows ice cover not only exceeding the 2007 minimum but also 2008 and 2009.

Bear in mind, I don’t use DMI for bets. My thoughts in running bets are that by picking a metric, we reduce the contribution to confirmation bias caused by each of us repeatedly looking at whichever metric is closer doing what we “like” at the current moment. In that regard, we monitoring NH Ice, I happen to look at Jaxa.

JAXA shows ice extent above the 2007 minimum.

Still, this year has had some surprises: We entered the ice melt season with a relatively high ice extent as evidenced by the April high value of the red trace shown in the JAXA image above. We then experienced a rapid June ice-melt I described as devastating. It appeared we might be in serious risk of setting a new all time low for minimum ice. But now the risk seems less likely. But, who knows?

Those of you who have ideas, let us know. Those who have recipes for my peppers, please suggest. Recipes for labor day bashes are particularly welcome. Those who want to talk about other things: yammer away.

There are several interesting (to me) things to note:

• We are no longer flirting with “maximum anomaly for ‘month X’” with either metric.
• The best fit trendlines since 1980 remain below 0.2 c/decade shown in orange to help people get a rough sense of how warming is progressive to oft repeated about 0.2 C/decade mentioned in the AR4.
• This month, the recorded temperature anomaly lies itself below the trace corresponding to a trendline with a slope of 2 forced to a value of zero at the center of the time series.
• Computed ending in a period where surface temperatures appear to transition between “El Nino” and “La Nina”, the best fit trends since 1980 are comparable whether computed using HadCrut or GISS (0.162 C/decade vs 0.169 C/decade). So, it should be difficult to explain a great deal of slower-than-expected warming being due to ‘an artifact of the “Arctic hole” in the Hadley data’ — a theory both stefan and rasmus gave to explain differences in short term trends computed based on HadCRUT and GISS.

Twelve Month Averages
Many of us (meaning I) have been interested in both a) whether GISS or Hadley will break long term the long term average for surface temperature this year and b) whether (or by how much) the 12 month temperature anomaly would break through the “multi-model mean” value for the anomaly projected under the A1B scenario. The answer for Hadley can bee seen by examining the graph below:

There is pretty much no way short that HadCRUT NH/SH is going to set an all time record this year. We see that this month’s twelve month average is far short of the record set in 1998 and this month’s anomaly was lower than July 2009, causing a down tick in the 12 month average. Since temperature were rising during late summer and fall of 2009 and are now falling, it’s probable we have seen the maximum 12 month average for HadCRUT this year. Moreover, the peak temperature never managed to hit the level predicted for the multi-model mean. (Admittedly, this could be due to ‘an artifact of the “Arctic hole” in the Hadley data’, but … well… Let’s turn to the GISS 12 month averages shown below:

As previously discussed, GISS has set an all time record for highest 12 month average. Using the most recent values of GISTemp, the peak value appears in May 2010; the average then declined in June and July. Will the Jan-Dec. 2010 value be a more formal record high? I can’t say for sure– but if I had to bet, I’d say it will. Temperatures would have to decline at a pace exceeding the fondest dreams of the most hardened stone-cold cooler. I doubt there is even enough time for a Pinatubo type eruption on top of a growing El Nina to avoid a record for GISS.

But what about the temperature compared to A1B projections? Well, at the top of El Nino, GISS barely pierced the “multi-model mean”. As we all know, temperatures go up; temperature go down. All other things being equal, if the projections are on track we might expect the temperature to lie above the projections at the top of El Nino and below them during La Nina. Of course, all other things may not be equal. It will be interesting to see whether we pierce the multi-model mean during the next El Nino.

Willis spends much of his article focusing on GISTemp’s UHI adjustments to the site between 1961-1980. I’ll cover this quickly, but its really not germane to the main question at hand, mainly because the IPCC does not use GISTemp. NASA’s primary adjustment to GHCN data (outside the U.S., where they use USHCN data) is their UHI correction procedure. Per Hansen et al 2010, their UHI adjustment procedure:

“…uses nearby rural stations to define the long-term trends while allowing the local urban station to define high frequency variations, nominally as described by Hansen et al. [1999], but with details as follows. If an urban station has at least three rural stations within 500 km, all of these are used for the adjustment with closer stations receiving greater weight, as described above.  If there are not three stations within 500 km, but there are three or more stations within 1000 km, these stations are used for the adjustment.  The mean temperature trend of the rural stations is computed as a two-segment broken line, as in Hansen et al. [1999] but the knee of the broken line is variable (rather than being fixed at 1950) chosen so as to minimize the difference between the urban and rural records.”

The net effect of this procedure globally is to slightly lower the trend vis-a-vis using all raw data. I imagine our friends at NASA would readily admit that this does not produce the most accurate estimate of actual temperatures for any given individual station, but will result in more accurate regional and global trends by reducing the impact of stations in urban areas.

As Willis points out, there is a single GHCN station for all of Nepal, located at Kathmandu Airport, that is comprised of three separate records (or dupes in GHCN’s parlance):

He errs, however, in asserting that the records do not have a single year of overlap. Dupes 0 and 1 overlap in 21 months, with most absolute temp values being exactly the same but with a few slightly off (all < 1 C). A histogram of the differences between Dupe 0 and Dupe 1 in months where they overlap is shown below:

Likewise, Dupe 1 and Dupe 2 overlap in 26 months, and in all cases are exactly the same. This, coupled with the general consistency of the record between Dupes, suggests that they can be effectively treated like a single continuous record.

While the Kathmandu station in GHCN has a near-continuous record between 1951 and 1980, it becomes much more fragmentary post-1980, with many gaps. The figure below shows the anomalies for the Katmandu station, with the gaps in recent years clearly visible:

There is a clear warming trend, especially in recent years, though the missing data makes it somewhat difficult to assess. Thankfully, GHCN is no longer the only game in town for climate data, and we can turn to GSOD to fill in the gaps. GSOD has 12 stations in Nepal, who together create a complete record from 1976 to present. Only 4 of these station records cover a significant period of time, and only one (Katmandu Airport, conveniently enough) has a fully complete record in recent years.

If we compare the GHCN record and the GSOD record for the Katmandu Airport, we find that the absolute temperatures are similar but not identical:

We can also look at the differences between the two:

The anomalies are also similar, though the GSOD record shows notably higher anomalies in recent years:

To return to the question at hand, over what period has Nepal warmed by 0.9 C per decade, we can calculate the trend to present for each month since 1951 for both the GHCN and GSOD records.

We find that, for GHCN, that the trend over the last 19 (or fewer) years is greater than or equal to 0.9 C per decade. For the GSOD record, the past 30 years have a trend that exceeds 0.9 C per decade. While the trend is smaller for the full series (0.17 C per decade from 1951-present for GHCN, 0.67 C per decade from 1976 to present for GSOD), it is clear that Nepal is warming faster than the global land average, especially over the past few decades.

In this case, and assuming that the authors drew upon the same temperature records, 0.9 C per decade seems to be a bit of an exaggeration. If we want to focus on the “modern warming period” of 1975 to present, the highest warming rate that could be reasonably claimed is a still-respectable 0.67 C per decade.

Update

Per a commenter at WUWT, we find a link to Shrestha et al:

Shrestha, Arun B.; Wake, Cameron P.; Mayewski, Paul A.; Dibb, Jack E.. Maximum Temperature Trends in the Himalaya and Its Vicinity: An Analysis Based on Temperature Records from Nepal for the Period 1971–94. Journal of Climate, 9/1/99, Vol. 12 Issue 9 pp:2775-2786

Analyses of maximum temperature data from 49 stations in Nepal for the period 1971–94 reveal warming trends after 1977 ranging from 0.068 to 0.128C yr21 in most of the Middle Mountain and Himalayan regions, while the Siwalik and Terai (southern plains) regions show warming trends less than 0.038C yr21. The subset of records (14 stations) extending back to the early 1960s suggests that the recent warming trends were preceded by similar widespread cooling trends. Distributions of seasonal and annual temperature trends show high rates of warming in the high-elevation regions of the country (Middle Mountains and Himalaya), while low warming or even cooling trends were found in the southern regions. This is attributed to the sensitivity of mountainous regions to climate changes. The seasonal temperature trends and spatial distribution of temperature trends also highlight the influence of monsoon circulation.

The Kathmandu record, the longest in Nepal (1921–94), shows features similar to temperature trends in the Northern Hemisphere, suggesting links between regional trends and global scale phenomena. However, the magnitudes of trends are much enhanced in the Kathmandu as well as in the all-Nepal records. The authors’ analyses suggest that contributions of urbanization and local land use/cover changes to the all-Nepal record are minimal and that the all-Nepal record provides an accurate record of temperature variations across the entire region.

It looks like they had access to a significantly larger set of stations than is available in either GHCN or GSOD, and are studying the period from 1977-1994 for different regions of the country. This analysis clearly has nothing to do with the GISS UHI adjustment algorithm, Willis’ protestations notwithstanding.

I downloaded today, and I’ve been reading discussions at various blogs. I’m rather glad I waited to post as the comments at various blogs contain a number of obviously wrong knee-jerk reactions.

The finding that many will aggravate “some” and please “others” is this:

Over the interval 1979 to 2009, model-projected temperature trends are two to four times larger than observed trends in both the lower and mid-troposphere and the differences are statistically significant at the 99% level.

As many of us are aware, a 17 author paper lead by Santer pubished in 2008, mysteriously cropped their analysis in 2000, shortening the record and coming up with a “failed to reject” conclusion.

I’m looking at more details of the method in MMH because I’m particularly interested in how to deal with the circled term in the equation below:

I chuckled when I read MMH’s discussion of the circled term below (which is well understood and know to be required):

While detrended climate model projections may be uncorrelated with observations, the assumption of no covariance among trend coefficients implies models have no low frequency correspondence with observations in response to observed forcings, which seems overly pessimistic.

“Overly pessimistic” would fall in the category of “understatement”.

In fact, because deviation from linearity in temperature from 1979-2009 called various things like “Eruption of Mt. Pinatubo”, or “Non-linear variations in aerosol loading”, and these deviations are thought to be shared in both the observations and the model runs, assuming the circled term is ‘zero’ amounts to assuming the models runs do not exhibit any signature from these runs. As quite a few readers are aware, I’ve been trying to look at difference between model runs and observations precisely because I think that term is not zero. Finding a cleaner way to deal with it would be great.

Anyway, congrats guys!

I’m game. Since this is a game, I’m not going to provide a lot of caveats for my method.

I’ve provided my “best” naive prediction for future values: Shown in brown. These are based on simple linear extrapolation. No assumption the 52 values are 52 weeks or anything like that. I’m assuming being close to 52 weeks is a coincidence.

(Update: When I updated, I added someone else’s naive projection from Roger’s blog. It was made using “minitab”.)

I’ve also curves to bound my ±1 standard deviation uncertainties for actual values– those are indicated by the dashed orange traces. That is, in the future, my “best” guess for the value (stock price? height of corn in Illinois? ) is in brown, and I expect there roughly a 2/3rd chance the individual values will fall inside the dashed orange lines. (I also expect that if I acquire loads of data, the future trend has a pretty good chance of falling between the purple lines.)

How did I concoct this? First, I fit a least squares trend and found the line that bests fits the data Roger provided. This line has a slope of 1.21 units/time and the intercept at time=0 is 78 units. I checked to see if the trend was statistically significant: It was. Then, I checked the autocorrelation of residuals: It was 5%. So, I neglected auto-correlation and proceeded as though it’s not real. (It might be real, who knows? I’m being naive here. )

So, based on this, I do anticipate whatever we are examining has been increasing over time. Though I can’t be sure about the future, my naive forcast is this will persist into the future. I made my “best” estimate of the next 12 years of data based on the linear fit with no auto-correlation in residuals: These predictions are shown in brown.

If this is stock prices (or anything), before I gamble, I’m going to want an idea of risk. So, even though Roger asked for a naive, I added uncertainty intervals. Of course, I’m also going to give naive uncertainty intervals; I’ll compute them using the same assumptions I applied to the least squares fit.

Since I have a finite amount of data, the trend of 1.20 units/time can’t be exact, right? Excel tells me that using the assumptions above, the standard error in this trend is ±0.15 units/time.

But, I also know the intercept of 78 units is not exact. So, I shifted the time and found the standard error in the best fit curve running through the central time point: ±2.22 units. I used these to find my uncertainty in the expected value based on the fit. That’s shown in purple.

The purple line indicates is my uncertainty in the best curve that describes the data. It does not yet describe the “noise” or variability not explained by the trend. Even the most naive person can see there is a lot of noise around the best fit trend. I estimate the “noise” around my fit from the residuals from the straight line and get: ±16.2 units. I added these to the uncertainty in the trend (i.e. the purple lines) and show the bounds for my naively estimated uncertainty with a dashed orange trace. When gauging what might happen in the future, I expect that, more likely that not, the future values will fall within the dashed orange lines rather than splat on the brown crosses.

Are these uncertainty intervals adequate? Not entirely. After all, I neglected autocorrelation in the residuals, non linearity and all sorts of things. Autocorrelation in the residuals is 5%. I set it to zero, but it may be real. If it is, then I should have added that possibility to computation of my uncertainty intervals. Also, maybe the trend is non-linear, or maybe something important changed after the final data point.

But Roger wants a naive estimate, and this is mine: Naive, but accounting for a feature that is strongly indicated by available: This is the statistically significant positive trend.

Update: Julio noticed I lost the final point in Roger’s data set when I cut and pasted his series of numbers. I updated my image to show the uptick. Here’s the link to the first image.

The monthly anomalies since 1980 are shown below with July anomalies circled:

Now for the important stuff: Who took home the quatloos?

Our winner are: In first MikeP who netted 34.904 quatloos; in second Dolormin took home 27.324 quatloos! Use your quatloos wisely.

The net winnings for each member of the ensemble will be added to their accounts.

It appears RSS also reported their July anomaly, I’ll post plots tomorrow. Then, we can wait for the slower reporting surface records.

The cut-off is set at “8/15/2010″. I still haven’t quite figured out whose time zone this is– I think the clock is local to Dreamhost, which is in California.

The bets on UAH are now pre-scheduled to appear on the 15th of each month. I’m assuming most who bet know the drill. If you don’t, ask in comments.

Anyway: The downside, this process may itself is CPU intensive and so may also be memory intensive. So, things could be much worse today. Upside, once every single post is cached, both CPU and memory use may drop.

You can read about another bloggers experience here. The conclusion of the blogger (who also writes lots of nice plugins) is:

See that nice dip in the graph for this week? I started to preload the cache used by WP Super Cache last Sunday and it’s made a noticeable difference in the load on my server here. The big spike is the preloading process.

There should be no issues with stored cache files. That’s one thing Dreamhost is very generous on. But, if you have trouble commenting today, it may be due to the pre-load plugin operating.

I’ll also be exploring paged comments plugins. But I suspect running php to run a script to create a page for all the ‘bots may be a larger issue.

Here are the more detailed results:

The net winnings for each member of the ensemble will be added to their accounts.

Of course, announcing the winner also give us a result to review the status of the sea ice. I rarely say anything that isn’t rather obvious about NH Ice. The three most obvious points are:

• As usual, there is less NH Sea Ice present in July than in June.
• Unlike June, this is not the all time record low for the month; however it’s close!
• The general trend for July sea ice continues to be down.

Here is a graph of trends which also highlights the July average relative to past averages:

Unfortunately for me, the error logs show which processes were interrupted, but these are not necessarily the process that sucked up too much memory. In any case, the process that is interrupted tends to be the version of php being run. This is pretty unenlightening, since the problem almost certainly lies with a particular plugin not php overall.

Since I have no particular skills in identifying the problem, I’ve been turning off plugins. Some of you will notice that editing functionality for comments has been turned off. Some other plugins are also turned off, but most of you will not notice. (The plugin that insert ads is off. A plugins that makes the archives look pretty is off. The plugin that caches things to reduce CPU was turned off last night, and turned back on. I thought it might suck memory while sparing CPU. Well, turning it off definitely makes the blog use too much CPU– around 6 am server errors were nearly constant.)

Dreamhost staff suggested merely having unused plugins in the plugin folder can soak up memory, so I moved nearly all unused plugins out of the folder. (I had a shockingly large number of plugins in there.) If you experience no server errors today, we’ll have found the problem.

In the meantime, if any of you happen to have any expertise in optimizing WP installations, let me know. Thanks in advance.

But we do find that there has been a consistent pattern of failing to display the proper degree of openness, both on the part of the CRU scientists and on the part of the UEA, who failed to recognise not only the significance of statutory requirements but also the risk to the reputation of the University and, indeed, to the credibility of UK climate science.

On the allegation of withholding station identifiers we find that CRU should have made available an unambiguous list of the stations used in each of the versions of the Climatic Research Unit Land Temperature Record (CRUTEM) at the time of publication. We find that CRU‟s responses to reasonable requests for information were unhelpful and defensive.

On the allegations in relation to withholding data, in particular concerning the small sample size of the tree ring data from the Yamal peninsula, CRU did not withhold the underlying raw data (having correctly directed the single request to the owners). But it is evidently true that access to the raw data was not simple until it was archived in 2009 and that this delay can rightly be criticized on general principles. In the interests of transparency, we believe that CRU should have ensured that the data they did not own, but on which their publications relied, was archived in a more timely way.

On the allegation that CRU does not appear to have acted in a way consistent with the spirit and intent of the FoIA or EIR, we find that there was unhelpfulness in responding to requests and evidence that e-mails might have been deleted in order to make them unavailable should a subsequent request be made for them. University senior management should have accepted more responsibility for implementing the required processes for FoIA and EIR compliance.

Given the significance of the work of CRU, UEA management failed to recognise in their risk management the potential for damage to the University‟s reputation fuelled by the controversy over data access.

Under “broader Issues”:

Openness and Reputation. An important feature of the blogosphere is the extent to which it demands openness and access to data. A failure to recognise this and to act appropriately, can lead to immense reputational damage by feeding allegations of cover up. Being part of a like minded group may provide no defence. Like it or not, this indicates a transformation in the way science has to be conducted in this century.

Role of Research Sponsors. One of the issues facing the Review was the release of data. At various points in the report we have commented on the formal requirements for this. We consider that it would make for clarity for researchers if funders were to be completely clear upfront in their requirements for the release of data (as well as its archiving, curation etc).

I think it is fair to say that Muir Russell bent over backwards to interpret things as favorably as possible for scientists at CRU. Nevertheless, they slammed them for their consistent lack of openness toward data finding it indefensible, and advised that this sort of behavior feeds allegations of cover ups and risks the reputation of the scientists.

Refuse to share data and reputational damage will occur: This is a simple fact. It would be darn shame if scientists at UCAR and NCAR damage their own reputations by imitating the behavior for which CRU was justly criticized.

Even if it turns out NCAR is not subject to FOI, with some luck, NSF– the funding agency– will take Muir Russell’s recommendations to heart andstep in and save some scientists from further damaging the reputation of all climate scientists by indulging in what appears to be their own worst inclinations: not sharing data with people who might disagree with their findings.

As in previous posts, I’m watching the 12 year monthly averages to detect records. The HadCrut 12 month average is illustrated in blue; a “what if” 12 months average computed by assuming the current temperature freezes is indicated in green.

The 12 month HadCrut average has not broken the previous record for all time high 12-month temperature anomaly, and it appears unlikely to do so. Moreover, it’s looking more and more likely that the 12 month average temperature anomaly at the top of this El Nino will not pierce the multi-model mean projection based on IPCC models projected using the SRES.

[Fig 1]

As we discovered back during The Great GISTemp Mystery, the land temperature record published by GISTemp is not actually an estimate of global land temperatures. Rather, its an an approximation of global temperatures using only land stations. It does not use a land mask, and has strong zonal weightings (90°N to 23.6°N, 23.6°N to 23.6°S and 23.6°S to 90°S with weightings 0.3, 0.4 and 0.3, respectively) whose net result is to considerably lower the resulting temp record vis-a-vis a true land-only record. This is shown in Figure 1, where we can see the GISTemp land record running considerably below a standard land-masked and unweighted reconstruction using the GISS Step 0 data.

The claim that GISTemp would use such an odd weighting scheme seemed rather far fetched to me, since I’ve spent so much time working on reconstructing GISTemp using the underlying data. Indeed, the very fact that we can easily replicate Hadley-area GISTemp by using a standard 29% land / 71% ocean weighting of GISS Step 0 and HadISST/Reynolds suggests that GISTemp isn’t doing anything too screwy. The difference between standard GISTemp and GISTemp run using the Hadley area is entirely interpolation, and isn’t nearly large enough to account for a land/ocean ratio reversal.

[Fig 2]

Indeed, the only thing that can account for the odd 70% / 30% land/ocean ratio is mistakenly combining the GISTemp land series with an ocean series, and failing to take into account that GISTemp land series is not a true land record. Figure 3, below, shows how a 70 / 30 combination does replicate GISTemp land/ocean fairly well. It also shows, however, that using the correct GISS Step 0 series only requires a 35% land 65% ocean ratio to fully match the interpolated version:

[Fig 3]

We see that GISTemp’s interpolation adds an implied 6% to the global land area, rather than the 41% originally claimed. Even this is somewhat misleading though, since the area mainly covered by interpolation (e.g. the Arctic) is warming much faster than the average land temperatures, so adding a little bit of coverage there has an out-sized effect on the trend. Regardless, this exercise shows that folks should be a tad more careful before accusing our friends at NASA of deliberate manipulation, though it also suggests that GISS should probably alter the descriptive text on their website to make it clearer what exactly their land record represents.

As a small bonus, here is what you would get if you actually did a 70% land 30% ocean weighting using the correct land record:

[Fig 4]

Update

A more careful reading of Bob Tisdale’s post shows that he indeed pointed out this issue, remarking that “Note how the GISTEMP LST data extends out over the oceans. This is not the case for their combined product, because GISS masks the LST data over the oceans in its combined product. So in order to properly create a weighted average of GISTEMP land and sea surface temperature data with 1200km radius smoothing, the land surface data where it extends out over the oceans would first need to be masked.”

GHCN adjusted data isn’t actually used in many places; GISTemp and HadCRUT take in the raw data and do their own adjustments. We had assumed, however, that the NCDC land record used adjusted rather than raw data. A look at how reconstructions turn out using raw vs. adjusted data compare to NCDC suggests otherwise.

(click images to embiggen)

Updated! See below.

The v2.mean_adj file is curious in a number of ways. For one, it contains considerably fewer station records than v2.mean, particularly in recent years.

What is really odd, though, is that temperature reconstruction using unadjusted data turn out to be much closer to the NCDC land record than reconstructions using adjusted data:

Reconstruction with adjusted data, Zeke’s model and Mosh’s model

Using adjusted data, we see large divergences post-1988 between our reconstructions and the NCDC land record.

Reconstruction with raw data, Zeke’s model and Mosh’s model

Whereas these divergences mostly vanish in the unadjusted data.

So does NCDC land record use v2.mean or v2.mean_adj?

Update

Commenter Torn8o suggested that v2.mean_adj might contain only those specific values that have been changed from v2.mean, and that to really analyze the adjusted data we would need to combine v2.mean_adj with v2.mean (excluding any v2.mean values that have been adjusted).

If we do this, its much harder to judge post-1960 if NCDC land temps are more in line with the raw or adjusted data.

For the whole century, however, NCDC land temps still seem closer to v2.mean than v2.mean_adj, though its not beyond the realm of possibility that this is an artifact of methodological choices in model construction.

In comments BenjaminG noted:

It appears the last couple days of data, July 5th and 6th, 2010 have set new all-time record high absolute temperatures, at 254.32K.

I’m not reading precisely those numbers for AQUA-Channel 5 for July 5th and 6th, I read 254.317K and 254.301K, but I am reading 254.333K for July 10, which is a record all time high.

It’s worth putting this in context: As you can see, the earth’s average surface temperature exhibits an annual cycle which hits a maximum near the end of July and beginning of August; the maximum of the trace called “average” at the AMSU page hits a maximum of 255.816K July 30. This means that when El Nino warmth persists in the troposphere in July, we tend to see very high absolute temperatures at that time. Because temperature tend to lag the end of El Nino, this record may be partly owing to El Nino. That said, El Nino is dying, so you can’t explain the record purely based on El Nino or July. It’s a record, and interesting for that reason.

That said, I tend to be more interested in anomalies. In January, I was interested in watching to see if we hit an all time daily high anomaly for channel 5. I thought it might happen, but it didn’t on the old channel 5 data available at the AMSU page. Then, Roy switched to displaying the AQUA channel 5 data, and only recently provided the baseline to permit me to compute anomalies.

Now that a baseline is available, I subtracted the “average” from the data for individual years 2002-2010 and also from the min and max to create a daily anomaly. These are, admittedly noisy, but it’s still interesting to see whether the record for the all time high anomaly is broken. The anomalies are displayed below; 2010 is illustrated in red, with July temperatures highlighted with a narrower black trace:

As you can see, if viewed as daily anomalies, the July temperatures do not come close to achieving all time record highs. You can’t tell when the high occurred based on the level of detail available at the AMSU page. That record is 0.8023C and occurred on April 3 of some year. I’d guess 1998.

Anyway: In terms of anomalies, the AQUA-Channel 5 has not hit an all time record high this year. It has hit an all time record high if viewed in absolute temperatures.

Update: Zeke was trying to recollect the final appearance of Channel 5 AMSU anomalies before the switch to AQUA. I still have that:

I looked up “defalcations”, and found several definitions. This is Wikipedia’s definition of defalcation:

A defalcation is an amount of funds misappropriated by a person trusted with its charge; also, the act of misappropriation, or an instance thereof. The term is more specifically used by the United States Bankruptcy Code to describe a category of bad acts that taint a particular debt such that it cannot be discharged in bankruptcy.[1]

Defalcation occurs when a debtor commits a bad act while acting in a fiduciary capacity.[2] The classic example of defalcation is when a trustee recklessly invests trust funds and loses the money. If the beneficiary successfully wins a judgment against the trustee, and the trustee files for bankruptcy, the debt (the judgment) cannot be discharged in bankruptcy because the debt was the result of a defalcation.

Defalcation only applies when a debtor is acting in a fiduciary capacity. To constitute a defalcation, the conduct involves a degree of culpability that is greater than negligence, but the act does not need to rise to the level of a “fraud” under common law. Defalcation requires a showing of conscious behavior or extreme recklessness.[3]

The term is used in legal proceedings other than bankruptcy to refer more generally to embezzlement; it is often used in the context of the title insurance business. A title agent who misuses funds intended to be used to close insured transactions is said to be involved in a defalcation. Many title insurers have their own “defalcation units.”

I’ve looked up several other definitions. The word appears to have a very precise meaning, and to amount to an accusation of something rather bad. I see no evidence that John Abraham is committed anything that matches the dictionary definition of a “defalcation”. I also didn’t notice anything that would amount to a “defalcation” in Monckton’s 466 questions posted previously.

Have I missed something? Does someone know a definition of “defalcation” that might match the actual allegations published by Monckton in his pdf response to Abraham?

DDoes anything

Because many readers are interested in whether or not we reached a record 12 month average during this El Nina, the 12 month averages are shown in blue below. The 12 month running average for the multi-model mean projection driven by the A1B SRES is shown for reference:

We can see that:

1. The 12 month average NOAA/NCDC anomaly has not yet set a record.
2. A “what if” calculation was computed by assuming the current temperature will “freeze” at the current value is shown in green. This suggests the 12 month average would continue to rise as upcoming temperatures continue to exceed those from last year. Temperature sticking at current values is unlikely, but this does suggest NOAA that unless NOAA temperature decline soon, the NOAA record for the all time high 12 month average will be broken in the next few months.
3. The “what if” calculation indicates temperatures must freeze at current levels for the observations to achieve the multi-model mean value for projections. So, soon after the top of the most recent El Ninos, the observations appear to be just grazing the model mean. The model mean averages over a sufficient number of runs to expect little effect of El Nino, and if the model mean is unbiased we might expect the observations to fall above the model mean soon after the top of El Ninos and below soon after the top of La Ninas.

We’re still waiting for Hadley.

http://wattsupwiththat.com/2010/07/13/calculating-global-temperature/