Now that NASA has released their data updated through July, we know that in that data set, this July was the hottest July on record with a temperature anomaly of 0.75 deg.C, i.e. it was 0.75 deg.C above “climatology” (which is what’s usual for the given month). It’s not the hottest temperature anomaly in the data set, however; that record still belongs to January 2007, at 0.96 deg.C above climatology.
Yet it does seem that this July, while not the hottest temperature anomaly on record, is the hottest month on record.
Every year, the global average temperature goes through an annual cycle — not just the temperature at a given location. In the northern hemisphere we tend to be hottest in July and coldest in January, but in the southern hemisphere the seasons are reversed, hottest in January and coldest in July. The seasons are definitely hemisphere-dependent.
But what about the global average? My first instinct, many years ago, was that earth would, overall, be hottest in January simply because we’re closer to the sun (at the perihelion of our orbit). But it turns out (as was quickly pointed out by a blog commenter) earth is actually hottest in July. That’s because when the southern hemisphere is tilted toward the sun in January, all that solar heat mainly strikes ocean, which dominates the southern hemisphere rather than land. The thermal inertia of the oceans is much greater than that of the land masses, so it heats up more slowly, and just doesn’t get that hot even at the peak of summer.
But in July, it’s the northern hemisphere that’s tilted toward the sun. The lower thermal inertia of land (mostly in the northern hemisphere rather than the southern) means it can heat up quickly, so the northern hemisphere reaches higher temperatures at its summer peak than the southern hemisphere does at its summer peak.
Why not translate those temperature anomalies into actual temperature estimates?
The catch is, that from thermometer records it’s hard to determine climatology — what the actual absolute temperature is, globally averaged. That’s why the major temperature data providers track earth’s temperature with anomalies; they can be determined more accurately, and give us just as good a picture of how temperature has changed (which is, after all, what we’re mainly interested in).
But there’s one source which seems to give pretty good estimates of absolute temperature — reanalysis data. I don’t think anyone would trust them to be precise enough to rank individual months, but they just might be our best source of data to estimate climatology.
There are two reanalysis products that are well-known and easy to get: an American version from NCAR/NCEP, and a European version called “era-interim.” Let’s use both to define climatology, then add the anomaly values from NASA to estimate actual absolute temperature.
We’ll start with NCAR/NCEP. Here’s what it gives for global temperature:
I first isolated only the complete years — those with all 12 months — so 2015 is out. Then I computed the average temperature for each month separately. That gives me the NCAR/NCEP climatology:
Indeed July is the hottest month on average, January is coldest. The difference is, according to these data, 3.28 deg.C.
Now we can add this estimate of climatology to NASA temperature anomaly data:
We can also “zoom in” to see which month was hottest:
It turns out that yes, this July is the hottest month on record. It’s not the hottest temperature anomaly, but it’s the hottest temperature. In 2nd place is July 2011, coming in 3rd is July 2009.
What about era-interim reanalysis data for climatology? Here’s the data (expressed in Kelvins rather than deg.C):
Here’s the climatology:
July is still the hottest month on average, January still coldest, and the difference using era-interim data is 3.48 K (equivalent to 3.48 deg.C)
Adding NASA anomalies gives this:
And again, we can zoom in to see which is hottest:
In this calculation also, this July is the hottest month on record. Not the hottest temperature anomaly, but the hottest temperature. Only this time, 2nd place is taken by August 2014, while July 2011 has dropped to 3rd and July 2009 to 4th.
The rivals are very close, and the uncertainties are big enough, that of course we can’t be sure which is really the hottest month ever. The leaders are in what is properly called a “statistical tie.” But the fact remains, that July 2015 takes the gold medal, according to both judges. It’s just another sign how much the earth is heating up. Which it is.
Global warming is real. It’s caused by humans. And it’s dangerous. Very dangerous.
Open Mind 
Too soon to call it, but August could bury July. Either way, these new AGW records don’t seem to be lasting very long.
I suggest find warmest 12 month period and next warmest non overlapping 12 month period. Does this avoid the statistical tie problem? If not try 24 month non overlapping periods ….
The “statistical tie problem” really isn’t a problem. In a distribution which is rising annually at a rate which is a fraction of the standard error of the trend it is uncommon that any single year will be statistically significantly above each and every of all preceding years. That’s just a fact. Not a problem.
Your “solution” would work. In point of fact as a side effect it would also reduce autocorrelation greatly. But it would never convince deniers as they would just accuse you of fraudulently “adjusting” your analysis. They already do this when anyone points out how temperature rise by decade is quite obvious. Deniers look for anything they can to supposedly “DISprove” warming. Anything which does not is suspect, fraudulent, or adjusted.
Nice analysis-
A few irrelevant factoids, just because:
1) It’s true that the Earth is closer to the Sun during the SH summer, but because of Kepler’s law, the Earth is moving faster around the Sun during perihelion and so SH summer is shorter than the NH equivalent. This comes up when thinking about the precessional component of the Milankovitch problem. When the solar intensity is integrated over the summertime, precession-driven changes in duration and intensity nearly
balance (basically leaving obliquity to rule the story).
2) The NH is actually warmer in the annual-mean than the SH, not due to thermal inertia (which does indeed explain the global mean temperature being higher during the boreal warm season, as was the point of this post, but is offset by more NH cooling during the boreal cool season when you integrate over the seasonal cycle); rather the relatively warm NH is because of net northward ocean heat transport (across the equator). The atmosphere does only partly compensate with net southward transport, which actually works to position the ITCZ (tropical rainbelt) a few degrees north of the equator in the annual mean.
Don’t forget to compare apples to apples. Reanalyses give global SAT but Gistemp is a blended SST and land SAT index. The warmest SST month is August, and since sea covers 71 % of the globe, it would likely make August the warmest “blended” month.
Also, this August will likely be warmer than July measured in anomalies.
Actually, based on the numbers shown here I think Tamino has used global surface temperature rather than 2m SAT. It doesn’t make much difference to the shape of the annual cycle. The warmest month for SSTs may be August in observational data, but I suspect that’s influenced by geographical coverage of available observations and not necessarily a genuine reflection of global annual cycle. Getting a meaningful annual cycle for SSTs will always be tricky due to sea ice seasonally affecting ability to measure.
But really, it probably doesn’t matter much to the global average cycle – sea covers 71% of the globe but the amplitude of the annual cycle is about 20x as large for global land compared to global ocean. Therefore land temperature variations dominate the global land/ocean annual cycle.
Nice analysis!
Just one small quibble. To be fully consistent with GISS, the climatology should really be averaged only from 1951-1980. The result won’t change but the absolute values might.
There is also at least one absolute temperature estimate from the surface temperature datasets – Jones et al. (1999) see http://www.researchgate.net/publication/201169746_Surface_air_temperature_and_its_changes_over_the_past_150_years._Rev_Geophys
They also have July as warmest (15.9C) and January coolest (12.2C) in the global mean. Slightly larger range than your reanalysis results, but supporting the result.
See the absolute temperature annual cycles for global, NH and SH means here:
Did Jones et al 1999 also produce land-only estimates of absolute temperature?
We have monthly fields of absolute temperature estimates from the Jones et al. work on our website, see Absolute at the bottom of the data table here:
http://www.cru.uea.ac.uk/cru/data/temperature/
You could mask out the oceans to estimate the land, but beware that coastal grid cells will be a blend of land and marine temperatures.
New et al. 1999 produced a land-only gridded climatology of absolute temperature (and precip, cloud, etc) – the well-known CRU TS dataset. A global-mean of this would be a land-only absolute temperature estimate and this contributed to the absolute temperature fields of Jones et al 1999. But New et al. 1999 didn’t tackle Antarctica (which Jones et al. did), so while New et al. is land-only, you would need to extract Antarctica from Jones et al. Absolute file to get a truly global, land-only absolute temperature.
An adjustment to my earlier comment.
I checked with data from Knmi climate exporer. Global absolute land temps (according to various sources) are on average 0.4 C warmer in July than in August. ERSSTv4 absolute temps are 0.09 C warmer in August than in July. Making a 71/29% balanced blend of those results in a win for July by 0.05 C. With such narrow margin, August global temps can easily beat July, an anomaly of 0.80 is enough. Thus, August 2014 (+0.81) is likely the warmest month so far ( or a tie with July 2015), but will soon be replaced by August 2015, I believe..
Some thoughts:
1. Unless my conversion between Kelvin and Celsius is wrong it looks like the climatology from ERA-Interim is a full degree warmer than that from NCEP/NCAR. Partly this can be explained by the different data periods (ERA-Interim starts ~1980, NCEP/NCAR ~1950), but the residual perhaps gives an indication as to how accurate the climatologies are as an estimate for absolute global temperatures.
2. “My first instinct, many years ago, was that earth would, overall, be hottest in January simply because we’re closer to the sun (at the perihelion of our orbit). But it turns out (as was quickly pointed out by a blog commenter) earth is actually hottest in July. That’s because when the southern hemisphere is tilted toward the sun in January, all that solar heat mainly strikes ocean, which dominates the southern hemisphere rather than land. The thermal inertia of the oceans is much greater than that of the land masses, so it heats up more slowly, and just doesn’t get that hot even at the peak of summer.”
For all of your discussion you are talking about global average surface temperature, but immediately the question arises in my mind as to whether the Earth system as a whole – ie including the oceans, cryosphere and latent heat – holds more energy in January than in July. In my view this hinges on whether there is a hemispheric imbalance in albedo which is larger than the energy difference caused by being at perihelion.
From what little I know about this, I know that the albedo of the ocean is complicated, because it is related to the zenith angle. When sunlight strikes the ocean surface at an angle more will be reflected. This is then complicated by the presence of waves.
For all that, the Earth’s albedo should be relatively well observed.
[Response: There is some albedo discussion here.]
I understand it is the trend, and that these short time periods do not mean much of anything, but I wonder what the recent temps have done to the accuracy of the IPCC models, such as in
Response: Look here:
http://www.theguardian.com/environment/climate-consensus-97-per-cent/2015/aug/10/2015-global-temperatures-right-in-line-with-climate-model-predictions
]
Is this what you are asking about?
Thank you.
But it was a cold-ish July in the USA so none of this matters.
I see the sarcasm, but it is clear you are not on the US west coast. We’ve been setting records here.
FWIW, it was also a warm July throughout most of the Southeast, and also parts of New England. Overall, middle of the pack, not really ‘coldish’ except by contrast:
http://www.ncdc.noaa.gov/temp-and-precip/us-maps/1/201507?products%5B%5D=divisionaltavgrank
We are getting those “hottest day / month / year” now every now and then. So actually these are non-news. (Just like the bomb blasts in middle east, btw.) News would be, that non-news like those were no longer useful, because every flatearther now got it at last…
I learned a lot from this post. Thanks Tamino!
Phil Jones has reminded me that he re-visited his earlier (1999) work on absolute temperatures with a new (open access) paper in 2013:
Jones, P. D., and C. Harpham (2013), Estimation of the absolute surface air temperature of the Earth, J. Geophys. Res. Atmos., 118, 3213–3217, doi:10.1002/jgrd.50359.
http://onlinelibrary.wiley.com/wol1/doi/10.1002/jgrd.50359/full
Here’s their concluding paragraph:
In this paper, we have discussed the issue of the absolute surface temperature of the Earth. The difference between the value developed for 1961–1990 by Jones et al. [1999] and that from ERA-Interim are within 0.55°C, a value that is marginally larger than the 0.5°C uncertainty estimate given by Jones et al. [1999] for their climatology. The two are easily within the uncertainty estimate if the 0.2–0.3°C cold bias in ERA-Interim is acknowledged. The absolute surface temperature of the world is likely to be between 13.7 and 14.0°C for the 1961–1990 period and 13.9 and 14.2°C for 1981–2010. The spatial detail reveals that most of this difference comes from Antarctica and to a lesser extent Greenland and the immediate coastal areas around these two landmasses. There are also large differences along the coastlines of northern Eurasia particularly in DJF. These differences are suggestive of issues over the two landmasses and their adjacent sea-ice areas, which for large parts of Antarctica makes ERA-Interim up to 10°C cooler. High-elevation areas of Antarctica are much warmer (5–6°C) than the two sites with long records. ERA-Interim, therefore, has markedly reduced temperature gradients between the interior and coastal sites than evident at the limited number of sites in eastern Antarctica.
NOAA is now reporting, “July 2015 was warmest month ever recorded for the globe.”
http://www.ncdc.noaa.gov/sotc/summary-info/global/201507
And apparently a few people have noticed.
https://news.google.com/news/story?ncl=d0adA_mPv5gEUNMhiXU0xO8IXpgsM&q=warmest+month+on+record&lr=English&hl=en&sa=X&ved=0CC0QqgIwAGoVChMIrpKTvJK4xwIVBqHbCh2aCAW5
Jeez! You beat me to it. But yes, they essentially duplicated Tamino’s analysis:
The appearance of a new hottest month in absolute terms is of course something we would expect to see in a warming climate but the annual cycle makes it rarer than a record warmest anomaly, that is assuming the months aren’t warming at dramatically different rates.
And of course the different months aren’t warming at the same rate and the numbers do look just a little dramatic.
Regressions through GISS, NOAA & HadCRUT since 1980 shows globally the boreal winter (DJF) warming slower than the annual average & autumns (SON) warming quicker – winter warming at 75%the rate of the annual average and autumn warming at 120% (this last with a bit more variation between GISS 127%, NOAA 119% & HadCRUT 114%).
But the recent monthly anomaly figures are quite exceptional. While NOAA July 2015 rated only 12th hottest anomaly, this calendar year so far we have had four top10 NOAA anomalies and all seven are top 20 anomalies. Indeed, the last 12 months have only seen one month not in the NOAA top 20.
I wonder if there is perhaps some merit in somehow using such clustering of these rankings of global temperature data. Perhaps Joe Public would find such statistics an easier concept to grapple with than fractions of a degree Celsius.
Thus the normal account – being told that the last 12-months average global temperature anomaly (NOAA) was a record anomaly at +0.81ºC, mucho scorchio, but likely a meaningless figure for Joe Public. The comparison with previous highs helps, previous rolling 12-month average records being +0.72ºC in January 2010 and before that +0.68ºC in 2006 & +0.67ºC in 1998. (And these previous ‘records’ were the maximums – we haven’t necessarily got to the 2015 maximum yet.)
But would it be easier for Joe Public to grasp if the last 12-month global temperatures were described as being “average-ranked” 14th. (Note that such an “average-rank” if averaged properly cannot drop below 6th for a 12-month period, so 14th (averaged properly 14.3) is not far from ‘the axis’. A more complex method could perhaps account for that.)
And the comparison with the previous records shows them now “average-ranked” 2010 – 45th, 2006 – 67th and 1998 – 76th.
But then, like a golf score cannot be lower than 18, if individual months are compared, on the NOAA record nine of these last 12 months have been the hottest of their month on record with only January (2nd), April (3rd) and November (7th) preventing a perfect score. Mucho mucho scorchio!!!
Al, maybe it would help if you would define “average rank”. Frankly, I don’t understand you yet.
Reblogged this on esthersvictory weblog.
Let me elaborate on what I don’t get.
I’d have thought that, if:
…then,
jan = 2
feb = 1
march = 1
april = 3
june = 1
july = 1
aug = 1
sept = 1
oct = 1
nov = 7
dec = 1
Total, 20, for a mean of 1.67. Where’s the 14.3 coming from?
Doc Snow,
The “average rank” 14th was calculated from the 12 month’s-worth of anomaly rankings within the entire record that total to 172:-
Last 12 months NOAA anomalies.
13.5 … 2014 8 …. +0.8°C
15.5 … 2014 9 …. +0.79°C
17 … 2014 10 … +0.79°C
61 … 2014 11 … +0.69°C
8 …. 2014 12 … +0.84°C
11 … 2015 1 …. +0.82°C
2.5 …. 2015 2 …. +0.89°C
1 …. 2015 3 …. +0.9°C
20 … 2015 4 …. +0.78°C
6 …. 2015 5 …. +0.86°C
4.5 …. 2015 6 …. +0.87°C
12 … 2015 7 …. +0.81°C
Thus the lowest “average rank” cannot drop below 6th for a 12-month period in such a scheme.
The ranking of each month individually that presently yields 1.67 is an alternative assessment.