One of the most vivid predictions of global warming theory is a “hotspot” in the tropical upper troposphere, where increased tropical convection responding to warming sea surface temperatures (SSTs) is supposed to cause enhanced warming in the upper troposphere.

The trouble is that radiosonde (weather ballons) and satellites have failed to show evidence of a hotspot forming in recent decades. Instead, upper tropospheric warming approximately the same as surface warming has been observed.

It has been also been pointed out, with some justification, that our lower tropospheric temperature product really can’t be used to find the hotspot since it peaks too low in the troposphere, and our mid-troposphere product might have too much contamination from cooling in the lower stratosphere to detect the hotspot.

A recent paper by Sherwood and Nishant in Environmental Research letters presented a reanalysis of the radiosonde data and claims to find evidence of the hotspot. I’ve looked through the paper and find the statistical black box approach they used to be unconvincing. I’ll leave it to others to examine the details of their statistical adjustments, what what the physical reasons for those adjustments might be.

Instead, I want to introduce you to a new product that is made possible by the new methods we now use in Version 6 of our UAH datasets (links at the bottom).

Since we now have a tropopause (“TP”) product, we can combine that with our lower stratosphere (“LS”) product in such a way that we pretty well isolate the tropical upper tropospheric layer that is supposed to be warming the fastest.

The following plot of the satellite weighting functions shows that a simple linear combination of the TP and LS weighting functions (from MSU3/AMSU7 and MSU4/AMSU9, respectively) gives peak weight in the layer where the strongest warming is expected to occur, approximately 7-13 km in altitude:

If we apply the coefficients (1.4, -0.4) to the TP and LS products, the resulting “UT” (upper troposphere) product for the tropical oceans (20N-20S) produces monthly anomalies since 1979 as shown by the bright red line in the following plot (I have added offsets to all time series so their linear trend lines intersect zero at the beginning of 1979):

Note that the linear warming trend in the UT product (+0.07 C/decade, bright red trend line) is less than the HadSST3 sea surface temperature trend (light green, +0.10 C/decade) for the same 20N-20S latitude band, whereas theory would suggest it should be about twice as large (+0.20 C/decade).

And what is really striking in the above plot is how strong the climate models’ average warming trend over the tropical oceans is in the upper troposphere (+0.35 C/decade, dark red), which I calculate to be about 1.89 times the models’ average surface trend (+0.19 C/decade, dark green). This ratio of 1.89 is based upon the UT weighting function applied to the model average temperature trend profile from the surface to 100 mb (16 km) altitude.

So, what we see is that the models are off by about a factor of 2 on surface warming, but maybe by a factor of 5 (!) for upper tropospheric warming.

This is all preliminary, of course, since we still must submit our Version 6 paper for publication. So, make of it what you will.

But I am increasingly convinced that the hotspot really has gone missing. And the reason why (I still believe) is most likely related to water vapor feedback and precipitation processes, which largely govern the total heat budget of the free-troposphere (the layer above the turbulently mixed boundary layer).

I believe the missing hotspot is indirect evidence that upper tropospheric water vapor is not increasing, and so upper tropospheric water vapor (the most important layer for water vapor feedback) is not amplifying warming from increasing CO2. The fact that UT warming is indeed amplified — by about a factor of 2 — during El Nino events in the above plot might be related to the relatively short time scales involved, since convective heating and radiative cooling are far out of balance during short term variations, but are much closer to being balanced in the long-term with global warming.

The lack of positive water vapor feedback is an especially controversial assertion to make, given that (1) SSM/I satellite measurements of water vapor have indeed been increasing in lock-step with SST warming, and (2) probably a unanimous opinion in the IPCC climate community that water vapor feedback is positive.

But the SSM/I measurements are largely insensitive to the very low levels of upper tropospheric water vapor, so they can’t tell us anything about upper tropospheric vapor. And while lower-tropospherc water vapor is governed mostly by SST, upper tropospheric vapor is governed by precipitation processes, and we don’t even understand how those might change with warming, let alone have those physics included in climate models.

Instead, I suspect the models have been adjusted so that precipitation systems detrain more water vapor into the upper troposphere with warming, simply because that’s what we see on short time scales, say during El Nino events, and so the convective parameterizations in the models are adjusted to meet that expectation.

As part of a DOE contract we have, we will be examining 183 GHz measurements of upper tropospheric vapor, but those are available only since 1991 from the DMSP satellites, and late 1998 from the NOAA satellites. And from what I’ve read, it might not be possible to get meaningful trends from those data. So, at this point it’s not clear that we can get long term trends from water vapor…although there has been some tantalizing evidence of upper tropospheric drying since the 1950s in radiosonde data.

You can read more about the issues involved in determining water vapor feedback, and why I think it might not be amplyfing global warming, here.

For those interested in combining the TP and LS products themselves, the new Version 6 files (look for “beta2” in the filenames) are located here:

Lower Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tlt
Mid-Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tmt
Tropopause: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/ttp
Lower Stratosphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tls

The latest NASA MODIS satellite imagery from today shows that the huge Baiji oil complex continues to burn as the Islamic State torches the facilities there.

The MODIS thermal infrared sensors first indicated fires there on April 11, and by April 18 the black clouds of smoke had drifted almost 300 miles, well past Baghdad:

To give some idea of the size of the smoke cloud, here’s a wide angle view from April 18 stretching from Israel and the Mediterranean Sea to Baghdad (click image for full-size):

Here’s today’s imagery (May 19, 2015); the size and extent of the smoke cloud changes daily depending mostly on wind conditions:

From what I’ve read, even if Iraqi forces regain control of the refinery, the complex is so expansive that IS can render it largely unusable by continuing to attack critical portions of the complex.

As Subtropical Storm Ana churns off the southeast U.S. coast, the global atmosphere has exceeded 400 ppm carbon dioxide content for the first time in…well…who knows?

And also on tap for this month (May 25th, Memorial Day) is another milestone: 3,500 days since the last time a major hurricane (Cat 3 or stronger) struck the U.S., which was Hurricane Wilma in 2005.

Maybe we can all pause to remember the “good old days”, when hundreds or thousands of people died in major hurricanes. /sarc

You remember 2005, right? Hurricane Katrina? So many hurricanes that the National Hurricane Center ran out of names? The next year, Al Gore blamed it all on humanity’s carbon dioxide emissions in his movie, An Inconvenient Truth.

You might not remember that 2 years ago news reports also were reporting we hit record CO2, at 400 ppm. So why the latest report regarding 400 ppm? Well, because now we’ve exceeded 400 ppm, rather than just hitting 400 ppm.

The minor distinction illustrates an important fact: it takes a huge amount of CO2 emissions to raise the atmospheric CO2 concentration by even a tiny amount.

It took nearly a century to raise atmospheric CO2 concentrations from 3 parts per 10,000 to 4 parts per 10,000. That’s right, nearly a century to add 1 molecule of CO2 to every 10,000 molecules of atmosphere.

Most people aren’t aware that the atmospheric concentration would have gone up twice as fast if not for the fact that nature loves the stuff. No matter how fast we produce it with our cars and planes and power plants, nature sucks up half of it, like a starving dog that has just been fed dinner.

In fact, without CO2 life as we know it on Earth would not exist.

More CO2 has led to global greening. Increased agricultural productivity. It probably has contributed to recent warming, in my professional opinion, but that warming has been relatively benign, with no observable increase in severe weather.

Which brings me back to hurricanes. There is a huge amount of natural variability in global hurricane activity from year to year, and even decade to decade. For example, see Dr. Ryan Maue’s charts here.

This extreme variability would happen with or without humans, just like it happens in tornado activity. Yet, many people tend to anthropomorphize everything that happens in nature. Changes in nature are seen as an extension of changes in human behavior, specifically our use of fossil fuels. It really isn’t much different from medieval witches being blamed for bad things that happened.

Eventually a major hurricane will strike the U.S. again. Maybe it will be this year, maybe next year. No one knows.

But you can be sure that when the current drought in U.S. major hurricane strikes ends, that, too, will be blamed on humans.

Yesterday I presented this time series of climate data and asked if anyone could determine any physical causes based upon it’s character:

I like the example because it shows realistic variability compared to, say, global average temperature variations.

I created it with a very simple function that actually has some basis in how the real climate system operates. Lance Wallace came closest to the right explanation. Bob Tisdale even gave a prediction of how politicians and environmentalists would have used it to call for energy policy changes. 🙂

But the time series, with its multi-decadal warming trend, was created entirely from a monthly series of random numbers. It’s what I call a “constrained random walk”.

How Does this Relate to the Real Climate System?
If you had random monthly cloud variations over the earth, it would cause a monthly random climate forcing as more or less sunlight was absorbed by the system. That effect is cumulative, since the heat is stored by the land and the ocean. So, every month’s value is just the previous month’s value plus a new random number (I used +/-0.5 as the range of random numbers in Excel). BUT…this would just produce a random walk, which almost always wanders away from the average state over time. This is in contrast to the real climate system, which has net negative radiative feedback (the more it warms, the more infrared energy it loses to space, restoring the system to an average state).

You can mimic this negative feedback by just subtracting off 10% of the previous month’s value from the next month’s value. In other words, instead of each month being the previous month’s value plus a random number (which would produce a random walk), use 0.9 times the previous month’s value instead. This is actually an approximation to the time-dependent energy budget equation in a 1D global energy balance model.

The reason for this example is to show that relatively rapid (monthly) forcing in the form of just random cloud variations can cause low-frequency climate variability…even multi-decadal temperature trends. You don’t need variations in solar activity. The reason why is the climate system’s “memory” — its ability to store energy. Certain preferred time scales of temperature variability tend to show up because of certain characteristics of the system — the depth of the ocean mixed layer, the time it takes the tropical atmosphere to overturn, etc.

This kind of variability is contained, to a lesser or greater extent, in all of the IPCC climate models. The cloud variations aren’t really “random” because they have physical causes, but they can seem random because the causes are myriad and complex. This is also the type of simple climate model forcing we used in our papers demonstrating how cloud feedbacks in the climate system have likely been misinterpreted, because researchers tend to assume cloud variations are caused by temperature variations while ignoring causation in the opposite direction.

Thanks to everyone for offering their ideas. I hope you are beginning to appreciate how some of the structure we see in global temperature variations might simply be just nature flipping a coin.

A little over a year ago I posted a climate riddle of sorts: a time series that showed warming, then a “pause”, then warming again, etc.

The point of the exercise was to demonstrate how a natural climate cycle sumperimposed upon a linear warming trend can cause what we have seen in global temperatures. I wasn’t necessarily advocating that’s what’s going on…although it would be at the top of my list of educated guesses.

Anyway, an ongoing e-mail discussion I’ve been having with Lubos Motl about a possible 3.7 year cycle in the satellite-based global temperatures led me to my second Magical Mystery Climate index riddle:

My question is this: What, if anything, can you infer about the physical cause(s) of this time series?

All I’ll tell you is that (1) it is “temperature-related”, and (2) the data aren’t “real”…the year labels are only meant to indicate a monthly time series a little over 36 years in length, like the satellite record.

If you feel so inclined, here are the data contained in the graph.

I will post the answer tomorrow.

I was watching a Ford commercial last night that highlighted their “EcoBoost” engine technology, which mostly involves turbocharging (nothing new) which allows higher efficiency, and thus greater power output with smaller engine displacements.

That “ecoboost” term sounded familiar, so I went and looked on my washing machine, and found this:

I have no idea what the setting does. I’m pretty sure my washer isn’t turbocharged. And it can’t mean “less water” because the washer already fails to wash my clothes as it is.

I have to wonder how many marketing meetings are now dominated by discussion of how to work “eco” into new (or existing) products. Everyone wants to Save The Earth™, so if we can do that while we are buying more stuff, so much the better.

So, where did all this ecobabble come from? Well, as I recall the first ecoword was “ecology”, which from the Greek root words means “the study of annoying stuff”.

We now have eco-friendly eco-schools with eco-learning for eco-kids. Eco-cars, eco-news, eco-warriors, eco-awards. The list goes on eco nauseum.

The eco-trend does not seem to be nearing its eco-end, either. According to Google Trends, the term “eco” has been at an eco-high for several eco-years now.

The annoying part is that little if any eco-good is done with any eco-product, I suspect. History has shown that if we become less wasteful of some commodity, we will find a way to use more of it. As car engines become more fuel-efficient, we buy cars with bigger engines or we take longer drives.

Money we save on one thing ends up getting spent on something else, which inevitably uses more resources.

British company EasyJet has unveiled a new ecoJet technology to improve the energy efficiency of jet travel. I suppose if rocket engines become sufficiently efficient, we will all be taking eco-tourism trips into low Earth orbit.

Just think of how much energy we will be saving then!

NOTE: This is the first montly update with our new Version 6.0 dataset. Differences versus the old Version 5.6 dataset are discussed here.

The Version 6.0 global average lower tropospheric temperature (LT) anomaly for April, 2015 is +0.07 deg. C, down a little from the March, 2015 value of +0.14 deg. C (click for full size version):

The global, hemispheric, and tropical LT anomalies from the 30-year (1981-2010) average for the last 4 months for the old Version 5.6 and the new Version 6.0 are:

YR MON GLOBAL NH SH TROPICS
v5.6
2015 1 +0.351 +0.553 +0.150 +0.126
2015 2 +0.296 +0.433 +0.160 +0.015
2015 3 +0.257 +0.409 +0.105 +0.083
2015 4 +0.162 +0.337 -0.013 +0.074
v6.0
2015 1 +0.261 +0.379 +0.143 +0.119
2015 2 +0.157 +0.263 +0.050 -0.074
2015 3 +0.139 +0.232 +0.046 +0.022
2015 4 +0.065 +0.154 -0.024 +0.074

The global image for April, 2015 should be available in the next several days here.

The new Version 6 files, updated shortly, are located here:

Lower Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tlt
Mid-Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tmt
Tropopause: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/ttp
Lower Stratosphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tls

A secretive government can lie about many things, but it can’t hide its landscape from Earth orbiting satellites.

Most people are familiar with the nighttime satellite imagery revealing virtually no lights on in North Korea, presumably due to its extreme poverty. It’s always Earth Hour there.

But MODIS satellite imagery from yesterday shows that North Korea is cutting down its trees at an alarming rate, while South Korea shows about the same level of greenness compared to two years ago:

In contrast to PBS’s article on North Korea’s environmental collapse, which makes it sound like a case of simple neglectfulness or poor land management, North Koreans are just trying to stay alive. The poorest countries of the world have the worst environmental records as the land is denuded for firewood.

To get some sense of the North Korean mindset, read this candid, sad, yet humorous Tim Urban article, 20 Things I Learned While I Was In North Korea.

Now you’ll have to excuse me while I go change all of my computer passwords since one thing the North Koreans are good at is hacking the computers of people they don’t like.

by Roy W. Spencer, John R. Christy, and William D. Braswell

(a PDF version of this post is available here. Monthly updates will use Version 6 starting with the April update.)

Abstract

Version 6 of the UAH MSU/AMSU global satellite temperature dataset is by far the most extensive revision of the procedures and computer code we have ever produced in over 25 years of global temperature monitoring. The two most significant changes from an end-user perspective are (1) a decrease in the global-average lower tropospheric (LT) temperature trend from +0.140 C/decade to +0.114 C/decade (Dec. ’78 through Mar. ’15); and (2) the geographic distribution of the LT trends, including higher spatial resolution. We describe the major changes in processing strategy, including a new method for monthly gridpoint averaging; a new multi-channel (rather than multi-angle) method for computing the lower tropospheric (LT) temperature product; and a new empirical method for diurnal drift correction. We also show results for the mid-troposphere (“MT”, from MSU2/AMSU5), tropopause (“TP”, from MSU3/AMSU7), and lower stratosphere (“LS”, from MSU4/AMSU9). The 0.026 C/decade reduction in the global LT trend is due to lesser sensitivity of the new LT to land surface skin temperature (est. 0.010 C/decade), with the remainder of the reduction (0.016 C/decade) due to the new diurnal drift adjustment, the more robust method of LT calculation, and other changes in processing procedures.

1. Introduction & Some Results

After three years of work, we have (hopefully) finished our Version 6.0 reanalysis of the global MSU/AMSU data. Many procedures have been modified or entirely reworked, and most of the software has been rewritten from scratch. (Please, before you ask a question, read the following to see if your question has already been answered.)

The MSU and AMSU instruments measure the thermal microwave emission from atmospheric oxygen in the 50-60 GHz oxygen absorption complex, and the resulting calibrated brightness temperatures (Tb) are nearly equivalent to thermometric temperature, specifically a vertically-weighted average of atmospheric temperature with the vertical weighting represented by “weighting functions”.

One might ask, Why do the satellite data have to be adjusted at all? If we had satellite instruments that (1) had rock-stable calibration, (2) lasted for many decades without any channel failures, and (3) were carried on satellites whose orbits did not change over time, then the satellite data could be processed without adjustment. But none of these things are true. Since 1979 we have had 15 satellites that lasted various lengths of time, having slightly different calibration (requiring intercalibration between satellites), some of which drifted in their calibration, slightly different channel frequencies (and thus weighting functions), and generally on satellite platforms whose orbits drift and thus observe at somewhat different local times of day in different years. All data adjustments required to correct for these changes involve decisions regarding methodology, and different methodologies will lead to somewhat different results. This is the unavoidable situation when dealing with less than perfect data.

After 25 years of producing the UAH datasets, the reasons for reprocessing are many. For example, years ago we could use certain AMSU-carrying satellites which minimized the effect of diurnal drift, which we did not explicitly correct for. That is no longer possible, and an explicit correction for diurnal drift is now necessary. The correction for diurnal drift is difficult to do well, and we have been committed to it being empirically–based, partly to provide an alternative to the RSS satellite dataset which uses a climate model for the diurnal drift adjustment.

The following plot (Fig. 1) shows the variety of satellites making up the satellite temperature record and their local solar time of observation as the satellites pass northbound across the Equator (ascending node).

Also, while the traditional methodology for the calculation of the lower tropospheric temperature product (LT) has been sufficient for global and hemispheric average calculation, it is not well suited to gridpoint trend calculations in an era when regional — rather than just global — climate change is becoming of more interest. We have devised a new method for computing LT involving a multi-channel retrieval, rather than a multi-angle retrieval.

The MSU instrument scan geometry in Fig. 2 illustrates how the old LT calculation required data from different scan positions, each of which has a different weighting function (see Fig. 2 inset). Thus, only one LT “retrieval” was possible from a scan line of data. The new method uses multiple channels to allow computation of LT from a single geographic location.

The LT retrieval must be done in a harmonious way with the diurnal drift adjustment, necessitating a new way of sampling and averaging the satellite data. To meet that need, we have developed a new method for computing monthly gridpoint averages from the satellite data which involves computing averages of all view angles separately as a pre-processing step. Then, quadratic functions are statistically fit to these averages as a function of Earth-incidence angle, and all further processing is based upon the functional fits rather than the raw angle-dependent averages.

Finally, much of the previous software has been a hodgepodge of code snippets written by different scientists, run in stepwise fashion during every monthly update, some of it over 25 years old, and we wanted a single programmer to write a unified, streamlined code (approx. 9,000 lines of FORTRAN) that could be run in one execution if possible.

Before addressing details of how the new Version 6 processing is different from the old (Version 5.6) processing, let’s examine some results. First let’s look at time series (Fig. 3) of the global average lower tropospheric temperature (LT), and how it compares to the old (Version 5.6) LT:

Note that in the early part of the record, Version 6 has somewhat faster warming than in Version 5.6, but then the latter part of the record has reduced (or even eliminated) warming, producing results closer to the behavior of the RSS satellite dataset. This is partly due to our new diurnal drift adjustment, especially for the NOAA-15 satellite. Even though our approach to that adjustment (described later) is empirical, it is interesting to see that it gives similar results to the RSS approach, which is based upon climate model calculations of the diurnal cycle in temperature.

The next plot we will examine (Fig. 4) is the gridpoint LT trends during 1979-2015. Version 6 has inherently higher spatial resolution than the Version 5 product, which had strong spatial smoothing as part of the data processing and through the nature of how LT was calculated:

The gridpoint trend map above shows how the land areas, in general, have warmed faster than the ocean areas. We obtain land and ocean trends of +0.19 and +0.08 C/decade, respectively. These are weaker than thermometer-based warming trends, e.g. +0.26 for land (from CRUTem4, 1979-2014) and +0.12 C/decade for ocean (from HadSST3, 1979-2014).

The gridpoint trends for LT in Fig. 4 are very difficult to measure accurately over land, primarily due to (1) the diurnal drift effect, which can be at least as large as any real temperature trends, and (2) how LT is computed, which in the old LT methodology required data from different view angles, and thus different geographic locations which can be from different air masses and over different surfaces (land and ocean).

As a result, users can expect that there will be differences between old and new LT trends on a regional basis. Differences are also attributable to our use of a new, more accurate land mask in Version 6. For example, going from Version 5.6 to 6.0 the Australia trend increased from +0.17 to +0.24 C/decade, but the USA48 trend decreased from +0.23 to +0.17 C/decade. The Arctic region changed from +0.43 to +0.23 C/decade. Note that trends are noisy over Greenland, Antarctica, and the Tibetan Plateau, likely due to greater sensitivity of the satellite measurements to surface emission and thus to emissivity changes over high altitude terrain; trends in these high-altitude areas are much less reliable than in other areas. Future changes, probably minor, can be expected as we refine the gridpoint diurnal drift adjustments and other aspects of our new processing strategy.

Fig. 5 illustrates the changes from v5.6 to v6.0 for a variety of regions of interest:

Notice the trends decreased the most over the Northern Hemisphere extratropics, especially the Arctic, while tropical warming trends increased somewhat, especially over land. Near-zero trends exist in the region around Antarctica.

We want to emphasize that the land vs. ocean trends are very sensitive to how the difference in atmospheric weighting function height is handled between MSU channel 2 early in the record, and AMSU channel 5 later in the record (starting August, 1998). In brief, the lower in altitude the weighting function senses, the greater the brightness temperature difference between land and ocean, mostly because land microwave emissivity is approximately 0.90-0.95, while the ocean emissivity is only about 0.50. As a result, if the AMSU channel 5 view angle chosen to match MSU channel 2 is too low in altitude, the net effect after satellite intercalibration will be a spurious warming of land areas and spurious cooling of ocean areas (at least when intercalibration is performed with land and ocean data combined). We were careful to match the MSU and AMSU weighting function altitudes based upon radiative transfer theory, and are reasonably confident that the remaining land-vs-ocean effects in the above map are real, that is, the land areas have warmed faster than the ocean regions. This is consistent with thermometer datasets of surface temperature, although our warming trends are weaker. Given the importance of the microwave oxygen absorption theory to the land-versus-ocean trends, we hope to update that portion of our processing for a future version update.

2. Major Changes in Processing Procedures with Version 6

The following is meant to provide a general introduction to the new processing steps in Version 6, emphasizing departures from past practices, and not to provide exhaustive detail. It will likely be close to two years before a peer reviewed paper with greater detail gets published in a scientific journal.

2.1 LT Calculation
We have fundamentally changed the calculation of the lower tropospheric temperature product, LT, from a multi-angle method to a multi-channel method. The main reason we changed methods for LT calculation is the old view angle method had unacceptably large errors at the gridpoint level. While the errors cancel for global averages on a monthly time scale, on a regional or gridpoint basis they can be large. The errors arise because the different view angles necessary to calculate a single LT “retrieval” sample different geographic locations, for instance radiometrically colder ocean and warmer land (see Fig. 2, above).

This would not present as big a problem if the data from the different regions were simply averaged together, but instead they are differenced. The problem is further magnified (literally) because the old LT required a weighted difference between view angles (and thus regions) with large weights (+4, -3 for the MSUs), which amplified any regional Tb differences. Compounded with the need to do diurnal drift adjustments, which can vary substantially from land to ocean, the problems with the old LT were deemed to be too large to continue the old LT calculation methodology.

So, instead of the past method of calculating LT as a weighted difference between different view angles of MSU2 (or AMSU5), we are now calculating it as a weighted difference between MSU channels 2, 3, and 4 (or AMSU channels 5, 7, and 9) at a constant Earth incidence angle. This has the very important advantage that all satellite data necessary for the LT retrieval come from the same location.

This required a correction for calibration drifts in MSU channel 3, especially during 1980-81, which was the original stated reason why a multi-channel retrieval method was not implemented over 20 years ago. That correction is made based upon regression of global monthly anomalies of MSU3/AMSU7 data against MSU2/AMSU5 and MSU4/AMSU9 during 1982 through 1993 (a 12 year period exhibiting two large volcanic eruptions with differential responses in the different altitude channels). We then apply the resulting regression relationship to the entire 1979-2015 period to estimate MSU3 (AMSU7) from MSU2,4 (AMSU5,9), and compare it to the raw intercalibrated global MSU3/AMSU7 time series. A difference time series of the regression estimated and the observed MSU3/AMSU7 time series is fitted with a piecewise linear estimator to give a time series of adjustment which are then applied to the MSU3/AMSU7 monthly anomaly fields. The resulting corrections cause a few hundredths of a degree per decade increase in the MSU3/AMSU7 trend (1979-2014), which ends up being very close to zero.

The following graph (Fig. 6) shows the resulting time series of LT, MT (mid-troposphere, from MSU2/AMSU5), TP (our new “tropopause level” product, from MSU3/AMSU7) and LS (lower stratosphere, from MSU4/AMSU9):

The LT computation is a linear combination of MSU2,3,4 or AMSU5,7,9 (aka MT,TP, LS):

LT = 1.538*MT -0.548*TP +0.01*LS

As seen in Fig. 7, the new multi-channel LT weighting function is located somewhat higher in altitude than the old LT weighting function. But if global radiosonde trend profile shapes (dashed line in Fig. 7) are to be believed, the net difference between old and new LT trends should be small, less than 0.01 C/decade. This is because slightly greater sensitivity of the new LT to stratospheric cooling is cancelled by even greater sensitivity to enhanced upper tropospheric warming.

Specifically, we see from Fig. 7 that application of the old and new LT weighting functions to the radiosonde trend profiles (average of the RAOBCORE and RATPAC trend profiles, 1979-2014) leads to almost identical trends (+0.11 C/decade) between the new and old LT. These trends are a good match to our new satellite-based LT trend, +0.114 C/decade.

The new LT weighting function is less sensitive to direct thermal emission by the land surface (17% for the new LT versus 27% for the old LT), and we calculate that a portion (0.01 C/decade) of the reduction in the global LT trend is due to less sensitivity to the enhanced warming of global average land areas. The same effect does not occur over the ocean because all of these channels’ microwave frequencies are not directly sensitive to changes in SST since ocean microwave emissivity decreases with increasing SST in such a way that the two effects cancel. This effect likely also causes a slight enhancement of the land-vs-ocean trend differences. Thus, over ocean the satellite measures a true atmosphere-only temperature trend, but over land it is mostly atmospheric with a small (17%, on average) direct influence from the surface. One might argue that a resulting advantage of the new LT is lesser sensitivity to long-term changes in land surface microwave emissivity, which are largely unknown.

The rest of the reduction in the LT trend between Versions 6.0 and 5.6 (-0.016 C/decade) is believed to be partly due to a more robust method of LT calculation, and the new diurnal drift adjustment procedure, described later. It is well within our previously stated estimated error bars on the global temperature trend (+/- 0.040 C/decade).

2.2 Monthly Averaging Methodology
In order to compute gridpoint values of LT, we must first compute gridpoint averages of the three channels used to compute LT. We have a new methodology for computing monthly gridpoint averages from MSU channels 2, 3, 4 (AMSU ch. 5, 7, 9) which is based upon initially computing monthly gridpoint averages from all channels’ view angles separately: 6 view angles from 11 footprints of MSU, or 15 view angles from 30 footprints of AMSU, which are separately averaged in 2.5 deg. lat/lon bins during the month.

The resulting monthly Tb gridpoint averages for each of the three channels are then fitted as a function of Earth incidence angle with a second order polynomial. The Tb for any desired Earth incidence angle is then estimated from the fitted curve, rather than from the raw view-angle averages.

An example of this fit is shown in Fig. 8, for AMSU channel 5 for a single gridpoint for a single month from a single satellite (NOAA-15):

This new averaging procedure has the following advantages:

1) All of the different view angle Tb measurements are included in the optimum estimation of the Tb at the desired Earth incidence angle, reducing sampling noise.

2) The resulting average calculation for a gridpoint location is based only upon data from that location, a new feature that avoids sampling noise inherent in the old calculation of LT from geographically different areas.

3) The orbit altitude decay effect (which has been large only for calculation of the old LT), as well as different satellites’ altitudes, is automatically handled since we use routine satellite ephemeris updates to calculate Earth incidence angles, which are the new basis for Tb estimation, not footprint positions per se.

4) Working from monthly grids of separate view angle averages allows rapid reprocessing of all of the data from 1979 forward, allowing us to efficiently test the use of different nominal view angles for the products, matching of the MSU and AMSU view angles, changes in diurnal drift estimation, etc.

The nominal footprint position we use for all MSU channels (see Fig. 2) is footprint position 4 and 8 (Earth incidence angle 21.59 deg), rather than nadir position 6; and nominal footprint position 6.33 and 24.66 for AMSU channel 5 (Earth incidence angle 34.99 deg.); and Earth incidence angles 13.18 deg. for AMSU7 and 36.31 deg for AMSU9. The choice of MSU prints 4,8 is because the resulting sampling at those footprint positions gives approximately 28 measurements evenly distributed in longitude around the Earth twice a day, rather than only 14 samples if the nadir position (MSU print #6, or AMSU print #15,16) was used as the reference. We find this greatly reduces sampling noise in the middle latitudes caused by coincidental phasing of moving weather systems with the satellite orbital sampling patterns.

Nevertheless, a few months in the record still exhibit mid-latitude striping patterns (especially over the southern oceans) when the precession of satellite orbits combined with warm and cold air mass movements happen to lead to non-random sampling patterns, even with as many as three satellites operating. So, we apply a +/- 2 gridpoint smoother in the east-west direction for the monthly gridded anomaly grid fields, which is applied over land and ocean separately to prevent “bleeding” of signals between land and ocean.

2.3 Diurnal Drift Calculation
As the 1:30 satellites drift to later local observation times (an indirect result of orbit decay), the MSU2 (AMSU5) Tb tend to cool, especially over land in certain seasons, due to the day-night cycle in temperature. As the 7:30 satellites drift to earlier observation times, the Tb tend to warm for the same reason. These average relationships change at very high latitudes because the ascending and descending satellite passage times converge — while they are ~12 hours apart at the equator, they approach the same local time at high latitudes.

These diurnal drift effects are empirically quantified at the gridpoint level by comparing NOAA-15 (a drifting 7:30 satellite) to Aqua (a non-drifting satellite), and by comparing NOAA-19 against NOAA-18 during 2009-2014, when NOAA-18 was drifting rapidly and NOAA-19 had no net drift. The resulting estimates of change in Tb as a function of local observation time are quite noisy at the gridpoint level, and so require some form of spatial smoothing. Since they also depend upon terrain altitude and the dryness of the region (deserts have stronger diurnal cycles in temperature than do rain forests), a regression is performed within each 2.5 deg. latitude band between the gridpoint diurnal drift coefficients and terrain altitude as well as average rainfall (1981-2010) for that calendar month, then that relationship is applied back onto the gridpoint average rainfall and terrain elevation within the latitude band. Over ocean, where diurnal drift effects are small, the gridpoint drift coefficients are replaced with the corresponding ocean zonal band averages of those gridpoint drift coefficients.

Fig. 9 shows an example of the diurnal drift coefficients (in deg. C per hour of ascending node time drift) used for MSU ch. 2 at nominal footprint 4 (and for AMSU ch. 5, a nominal footprint position between #6 and #7) for the month of June:

The reason why the drift coefficients change sign at high northern latitudes is a combination of early sunrise time in June, late sunset time, and the fact that the ascending and descending orbit satellite observations at high latitudes approach the same time, instead of being 12 hours apart as they are at the equator.

We also compute and apply diurnal drift coefficients for MSU channels 3 and 4 (AMSU channels 7 and 9), but the drifts and resulting adjustments are very small.

3. Final Comments

This should be considered a “beta” release of Version 6.0, and we await users’ comments to see whether there are any obvious remaining problems in the dataset. In any event, we are confident that the new Version 6.0 dataset as it currently stands is more accurate and useful than the Version 5.6 dataset.

The new LT trend of +0.114 C/decade (1979-2014) is 0.026 C/decade lower than the previous trend of +0.140 C/decade, but about 0.010 C/decade of that difference is due to lesser sensitivity of the new LT weighting function to direct surface emission by the land surface, which surface thermometer data suggests is warming more rapidly than the deep troposphere. The remaining 0.016 C/decade difference between the old and new LT product trends is mostly due to the new diurnal drift adjustment procedure and is well within our previously stated range of uncertainty for this product’s trend calculation (+/- 0.040 C/decade).

We have performed some calculations of the sensitivity of the final product to various assumptions in the processing, and find it to be fairly robust. Most importantly, through sensitivity experiments we find it is difficult to obtain a global LT trend substantially greater than +0.114 C/decade without making assumptions that cannot be easily justified.

The new Version 6 files are located here:

Lower Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tlt
Mid-Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tmt
Tropopause: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/ttp
Lower Stratosphere: http://vortex.nsstc.uah.edu/data/msu/v6.0beta/tls

The explosive double eruption of Calbuco Volcano in Chile late yesterday afternoon and last night, combined with mostly clear skies, allowed some spectacular photos and videos, even from a drone.

Here’s one of the prettiest photos of the first eruption, which occurred before sunset:

The two mature eruption plumes as seen by the GOES weather satellite show plume top temperatures so cold that the stratosphere was surely penetrated, a necessary (but not sufficient) condition for effects on global climate (arrows indiate volcano location, satellite imagery courtesy of CIRA at ColoState):

From low Earth orbit, this was how the area around the snow-capped volcano looked just hours before the eruption yesterday from NASA’s Terra satellite, and then again this morning. Note the large region where the snow-capped mountains and the valleys are covered in volcanic ash: