Archive for April, 2019

UAH, RSS, NOAA, UW: Which Satellite Dataset Should We Believe?

Tuesday, April 23rd, 2019

This post has two related parts. The first has to do with the recently published study of AIRS satellite-based surface skin temperature trends. The second is our response to a rather nasty Twitter comment maligning our UAH global temperature dataset that was a response to that study.

The AIRS Study

NASA’s Atmospheric InfraRed Sounder (AIRS) has thousands of infrared channels and has provided a large quantity of new remote sensing information since the launch of the Aqua satellite in early 2002. AIRS has even demonstrated how increasing CO2 in the last 15+ years has reduced the infrared cooling to outer space at the wavelengths impacted by CO2 emission and absorption, the first observational evidence I am aware of that increasing CO2 can alter — however minimally — the global energy budget.

The challenge for AIRS as a global warming monitoring instrument is that it is cloud-limited, a problem that worsens as one gets closer to the surface of the Earth. It can only measure surface skin temperatures when there are essentially no clouds present. The skin temperature is still “retrieved” in partly- (and even mostly-) cloudy conditions from other channels higher up in the atmosphere, and with “cloud clearing” algorithms, but these exotic numerical exercises can never get around the fact that the surface skin temperature can only be observed with satellite infrared measurements when no clouds are present.

Then there is the additional problem of comparing surface skin temperatures to traditional 2 meter air temperatures, especially over land. There will be large biases at the 1:30 a.m./p.m. observation times of AIRS. But I would think that climate trends in skin temperature should be reasonably close to trends in air temperature, so this is not a serious concern with me (although Roger Pielke, Sr. disagrees with me on this).

The new paper by Susskind et al. describes a 15-year dataset of global surface skin temperatures from the AIRS instrument on NASA’s Aqua satellite. ScienceDaily proclaimed that the study “verified global warming trends“, even though the period addressed (15 years) is too short to say much of anything much of value about global warming trends, especially since there was a record-setting warm El Nino near the end of that period.

Furthermore, that period (January 2003 through December 2017) shows significant warming even in our UAH lower tropospheric temperature (LT) data, with a trend 0.01 warmer than the “gold standard” HadCRUT4 surface temperature dataset (all deg. C/decade):

AIRS: +0.24
GISTEMP: +0.22
ECMWF: +0.20
Cowtan & Way: +0.19
UAH LT: +0.18
HadCRUT4: +0.17

I’m pretty sure the Susskind et al. paper was meant to prop up Gavin Schmidt’s GISTEMP dataset, which generally shows greater warming trends than the HadCRUT4 dataset that the IPCC tends to favor more. It remains to be seen whether the AIRS skin temperature dataset, with its “clear sky bias”, will be accepted as a way to monitor global temperature trends into the future.

What Satellite Dataset Should We Believe?

Of course, the short period of record of the AIRS dataset means that it really can’t address the pre-2003 adjustments made to the various global temperature datasets which significantly impact temperature trends computed with 40+ years of data.

What I want to specifically address here is a public comment made by Dr. Scott Denning on Twitter, maligning our (UAH) satellite dataset. He was responding to someone who objected to the new study, claiming our UAH satellite data shows minimal warming. While the person posting this objection didn’t have his numbers right (and as seen above, our trend even agrees with HadCRUT4 over the 2003-2017 period), Denning took it upon himself to take a swipe at us (see his large-font response, below):


First of all, I have no idea what Scott is talking about when he lists “towers” and “aircraft”…there has been no comprehensive comparisons of such data sources to global satellite data, mainly because there isn’t nearly enough geographic coverage by towers and aircraft.

Secondly, in the 25+ years that John Christy and I have pioneered the methods that others now use, we made only one “error” (found by RSS, and which we promptly fixed, having to do with an early diurnal drift adjustment). The additional finding by RSS of the orbit decay effect was not an “error” on our part any more than our finding of the “instrument body temperature effect” was an error on their part. All satellite datasets now include adjustments for both of these effects.

Nevertheless, as many of you know, our UAH dataset is now considered the “outlier” among the satellite datasets (which also include RSS, NOAA, and U. of Washington), with the least amount of global-average warming since 1979 (although we agree better in the tropics, where little warming has occurred). So let’s address the remaining claim of Scott Denning’s: that we disagree with independent data.

The only direct comparisons to satellite-based deep-layer temperatures are from radiosondes and global reanalysis datasets (which include all meteorological observations in a physically consistent fashion). What we will find is that RSS, NOAA, and UW have remaining errors in their datasets which they refuse to make adjustments for.

From late 1998 through 2004, there were two satellites operating: NOAA-14 with the last of the old MSU series of instruments on it, and NOAA-15 with the first new AMSU instrument on it. In the latter half of this overlap period there was considerable disagreement that developed between the two satellites. Since the older MSU was known to have a substantial measurement dependence on the physical temperature of the instrument (a problem fixed on the AMSU), and the NOAA-14 satellite carrying that MSU had drifted much farther in local observation time than any of the previous satellites, we chose to cut off the NOAA-14 processing when it started disagreeing substantially with AMSU. (Engineer James Shiue at NASA/Goddard once described the new AMSU as the “Cadillac” of well-calibrated microwave temperature sounders).

Despite the most obvious explanation that the NOAA-14 MSU was no longer usable, RSS, NOAA, and UW continue to use all of the NOAA-14 data through its entire lifetime and treat it as just as accurate as NOAA-15 AMSU data. Since NOAA-14 was warming significantly relative to NOAA-15, this puts a stronger warming trend into their satellite datasets, raising the temperature of all subsequent satellites’ measurements after about 2000.

But rather than just asserting the new AMSU should be believed over the old (drifting) MSU, let’s look at some data. Since Scott Denning mentions weather balloon (radiosonde) data, let’s look at our published comparisons between the 4 satellite datasets and radiosondes (as well as global reanalysis datasets) and see who agrees with independent data the best:

Trend differences 1979-2005 between 4 satellite datasets and either radiosondes (blue) or reanalyses (red) for the MSU2/AMSU5 tropospheric channel in the tropics. The balloon trends are calculated from the subset of gripoints where the radiosonde stations are located, whereas the reanalyses contain complete coverage of the tropics. For direct comparisons of full versus station-only grids see the paper.

Clearly, the RSS, NOAA, and UW satellite datasets are the outliers when it comes to comparisons to radiosondes and reanalyses, having too much warming compared to independent data.

But you might ask, why do those 3 satellite datasets agree so well with each other? Mainly because UW and NOAA have largely followed the RSS lead… using NOAA-14 data even when its calibration was drifting, and using similar strategies for diurnal drift adjustments. Thus, NOAA and UW are, to a first approximation, slightly altered versions of the RSS dataset.

Maybe Scott Denning was just having a bad day. In the past, he has been reasonable, being the only climate “alarmist” willing to speak at a Heartland climate conference. Or maybe he has since been pressured into toeing the alarmist line, and not being allowed to wander off the reservation.

In any event, I felt compelled to defend our work in response to what I consider (and the evidence shows) to be an unfair and inaccurate attack in social media of our UAH dataset.

A Simple Model of the Atmospheric CO2 Budget

Thursday, April 11th, 2019

SUMMARY: A simple model of the CO2 concentration of the atmosphere is presented which fairly accurately reproduces the Mauna Loa observations 1959 through 2018. The model assumes the surface removes CO2 at a rate proportional to the excess of atmospheric CO2 above some equilibrium value. It is forced with estimates of yearly CO2 emissions since 1750, as well as El Nino and La Nina effects. The residual effects of major volcanic eruptions (not included in the model) are clearly seen. Two interesting finding are that (1) the natural equilibrium level of CO2 in the atmosphere inplied by the model is about 295 ppm, rather than 265 or 270 ppm as is often assumed, and (2) if CO2 emissions were stabilized and kept constant at 2018 levels, the atmospheric CO2 concentration would eventually stabilize at close to 500 ppm, even with continued emissions.

A recent e-mail discussion regarding sources of CO2 other than anthropogenic led me to revisit a simple model to explain the history of CO2 observations at Mauna Loa since 1959. My intent here isn’t to try to prove there is some natural source of CO2 causing the recent rise, as I think it is mostly anthropogenic. Instead, I’m trying to see how well a simple model can explain the rise in CO2, and what useful insight can be deduced from such a model.

The model uses the Boden et al. (2017) estimates of yearly anthropogenic CO2 production rates since 1750, updated through 2018. The model assumes that the rate at which CO2 is removed from the atmosphere is proportional to the atmospheric excess above some natural “equilibrium level” of CO2 concentration. A spreadsheet with the model is here.

Here’s the assumed yearly CO2 inputs into the model:

Fig. 1. Assumed yearly anthropogenic CO2 input into the model atmosphere.

I also added in the effects of El Nino and La Nina, which I calculate cause a 0.47 ppm yearly change in CO2 per unit Multivariate ENSO Index (MEI) value (May to April average). This helps to capture some of the wiggles in the Mauna Loa CO2 observations.

The resulting fit to the Mauna Loa data required an assumed “natural equilibrium” CO2 concentration of 295 ppm, which is higher than the usually assumed 265 or 270 ppm pre-industrial value:

Fig. 2. Simple model of atmospheric CO2 concentration using Boden et al. (2017) estimates of yearly anthropogenic emissions, an El Nino/La Nina natural source/sink, after fitting of three model free parameters.

Click on the above plot and notice just how well even the little El Nino- and La Nina-induced changes are captured. I’ll address the role of volcanoes later.

The next figure shows the full model period since 1750, extended out to the year 2200:

Fig. 3. As in Fig. 2, but for the full model period, 1750-2200.

Interestingly, note that despite continued CO2 emissions, the atmospheric concentration stabilizes just short of 500 ppm. This is the direct result of the fact that the Mauna Loa observations support the assumption that the rate at which CO2 is removed from the atmosphere is directly proportional to the amount of “excess” CO2 in the atmosphere above a “natural equilibrium” level. As the CO2 content increases, the rate or removal increases until it matches the rate of anthropogenic input.

We can also examine the removal rate of CO2 as a fraction of the anthropogenic source. We have long known that only about half of what is emitted “shows up” in the atmosphere (which isn’t what’s really going on), and decades ago the IPCC assumed that the biosphere and ocean couldn’t keep removing excess CO2 at such a high rate. But, in fact, the fractional rate of removal has actually been increasing,¬†not¬†decreasing.And the simple model captures this:

Fig. 4. Rate of removal of atmospheric CO2 as a fraction of the anthropogenic source, in the model and observations.

The up-and-down variations in Fig. 4 are due to El Nino and La Nina events (and volcanoes, discussed next).

Finally, a plot of the difference between the model and Mauna Loa observations reveals the effects of volcanoes. After a major eruption, the amount of CO2 in the atmosphere is depressed, either because of a decrease in natural surface emissions or an increase in surface uptake of atmospheric CO2:

Fig. 5. Simple model of yearly CO2 concentrations minus Mauna Loa observations (ppm), revealing the effects of volcanoes which are not included in the model.

What is amazing to me is that a model with such simple but physically reasonable assumptions can so accurately reproduce the Mauna Loa record of CO2 concentrations. I’ll admit I am no expert in the global carbon cycle, but the Mauna Loa data seem to support the assumption that for global, yearly averages, the surface removes a net amount of CO2 from the atmosphere that is directly proportional to how high the CO2 concentration goes above 295 ppm. The biological and physical oceanographic reasons for this might be complex, but the net result seems to follow a simple relationship.

Mid-April Blizzard to Clobber the Upper Midwest

Monday, April 8th, 2019

A near-repeat of March’s “bomb cyclone” will bring up to 30 inches of snow this week to portions of Minnesota and South Dakota, with blizzard conditions and a threat of severe thunderstorms.

Roughly the same area that experienced flooding rains in March — and still trying to dry out enough to plant corn and soybeans — will see another round of heavy rain and heavy snow. The forecast location of the intense cyclone as of Thursday morning April 11 shows it taking a similar path to the record-setting March storm:


Forecast locations of strong low pressure and precipitation patterns Thursday morning, April 11, 2019 (Weatherstreet.com).

Forecast snowfall totals by midday Friday April 12 indicate the heaviest snowfall (up to 30 inches) over southern Minnesota, with 12-16 inches for Minneapolis:


Forecast total snowfall from the GFS model by midday Friday April 12 (graphic courtesy of Weatherbell.com).

The European ECMWF forecast model adds similarly heavy (~30 inches) snow totals in eastern South Dakota. Much of Wisconsin and northern Michigan are forecast to receive 6 to 12 inches.

The energy for such intense cyclones comes from the strong temperature contrast between two air masses. For example, by late Wednesday the temperatures in Nebraska will range from the 70s in the southeast to the 20s in the northwest, simultaneously feeding both blizzard conditions and a severe thunderstorm threat within the state.

Strange lights over Norway

Friday, April 5th, 2019

Famed Arctic and aurora photographer Ole C Salomonsen has reported in the last hour strange lights over Tromso, Norway. Ole says the sight is the “weirdest stuff I’ve seen”.

I’ve taken the liberty of increasing the brightness of two of the images he posted:

Strange lights in the sky photographed by famed aurora photographer Ole Salomonsen around midnight, Saturday April 6 2019 from Tromso, Norway.

I can’t imagine what this is, but I suspect it’s related to some sort of rocket-borne experiment. But the spatial distribution of the lights is very strange. I assume Ole will update us with time lapse photography in the near future.

UPDATE: Frank Olsen, also in Norway, posted the following photo, and said that this was indeed rocket-borne experiments containing special chemicals:

Photo by Frank Olsen, Norway.

Australia Surface Temperatures Compared to UAH Satellite Data Over the Last 40 Years

Wednesday, April 3rd, 2019

Summary: The monthly anomalies in Australia-average surface versus satellite deep-layer lower-tropospheric temperatures correlate at 0.70 (with a 0.57 deg. C standard deviation of their difference), increasing to 0.80 correlation (with a 0.48 deg. C standard deviation of their difference) after accounting for precipitation effects on the relationship. The 40-year trends (1979-2019) are similar for the raw anomalies (+0.21 C/decade for Tsfc, +0.18 deg. C for satellite), but if the satellite and rainfall data are used to estimate Tsfc through a regression relationship, the adjusted satellite data then has a reduced trend of +0.15 C/decade. Thus, those who compare the UAH monthly anomalies to the BOM surface temperature anomalies should expect routine disagreements of 0.5 deg. C or more, due to the inherently different nature of surface versus tropospheric temperature measurements.

I often receive questions from Australians about the UAH LT (lower troposphere) temperature anomalies over Australia, as they sometimes differ substantially from the surface temperature data compiled by BOM. As a result, I decided to do a quantitative comparison.

While we expect that the tropospheric and surface temperature variations should be somewhat correlated, there are reasons to expect the correlation to not be high. The surface-troposphere system is not regionally isolated over Australia, as the troposphere can be affected by distant processes. For example, subsidence warming over the continent can be caused by vigorous precipitation systems hundreds or thousands of miles away.

I use our monthly UAH LT anomalies for Australia (available here), and monthly anomalies in average (day+night) surface temperature and rainfall (available from BOM here). All monthly anomalies from BOM have been recomputed to be relative to the 1981-2010 base period to make them comparable to the UAH LT anomalies. The period analyzed here is January 1979 through March 2019.

Results Before Adjustments

A time series comparison between monthly Tsfc and LT anomalies shows warming in both, with a Tsfc warming trend of +0.21 C/decade, and and a satellite LT trend of +0.18 C/decade:

Fig. 1. Australia average surface temperature (red) and satellite lower tropospheric temperature (LT, blue) anomalies from January 1979 through March 2019.

The correlation between the two time series is 0.70, indicating considerable — but not close — agreement between the two measures of temperature. The standard deviation of their difference is 0.57 deg. C, which means that people doing a comparison of UAH and BOM anomalies each month should not be surprised to see 0.6 deg. C differences (or more).

Part of the disagreement comes from rainfall conditions, which can affect the temperature lapse rate in the troposphere. For reference, the following plot shows Australian precipitation anomalies for the same period:

Fig. 2. Australia precipitation anomalies from January 1979 through March 2019.

If we take the data in Fig. 1 and create a scatter plot, but show the months with the 25% highest precipitation anomalies in green and the lowest 25% precipitation in red, we see that drought periods tend to have higher surface temperatures compared to tropospheric temperatures, while the wettest periods tend to have lower surface temperatures compared to the troposphere:

Fig. 3. Scatterplot of the data in Fig. 1, but with color coding of those months with the 25% highest (green) and lowest (red) precipitation departures from average.

A More Apples-to-Apples Comparison

Comparing tropospheric and surface temperatures is a little like comparing apples and oranges. But one interesting thing we can do is to regress the surface temperature data against the tropospheric temperatures plus rainfall data to get equations that provide a “best estimate” of the surface temperatures from tropospheric temperatures and rainfall.

I did this for each of the 12 calendar months separately because it turned out that the precipitation relationship evident in Fig. 3 was only a warm season phenomenon. During the winter months of June, July, and August, the relationship to precipitation had the opposite sign, with excessive precipitation being associated with warmer surface temperature versus the troposphere, and drought conditions associated with cooler surface temperatures than the troposphere (on average).

So, using a different regression relationship for each calendar month (each month having either 40 or 41 years represented), I computed a satellite+rainfall estimate of surface temperature. The resulting “satellite” time series then changes somewhat, and the correlation between them increases from 0.70 to 0.80:

Fig. 4. As in Fig. 1, but now the satellite data are used along with precipitation data to provide a regression estimate of surface temperature.

Now the “satellite-based” trend is lowered to +0.15 C/decade, compared to the observed Tsfc trend of +0.21 C/decade. I will leave it to the reader to decide whether this is a significant difference or not.

To make the differences in Fig. 4 a little easier to see, we can plot the difference time series between the two temperature measures:

Fig. 5. Difference between the two time series shown in Fig. 4.

Now we can see evidence of an enhanced warming trend in the Tsfc data versus the satellite over the most recent 20 years, which amounts to 0.40 deg. C during April 1999 – March 2019. I have no opinion on whether this is some natural fluctuation in the relationship between surface and tropospheric temperatures, problems in the surface data, problems in the satellite data, or some combination of all three.

Conclusions

The UAH tropospheric temperatures and BOM surface temperatures in Australia are correlated, with similar variability (0.70 correlation).
Accounting for anomalous rainfall conditions increases the correlation to 0.80. The Tsfc trends have a slightly greater warming trend than the tropospheric temperatures, but the reasons for this are unclear. Users of the UAH data should expect monthly differences between the UAH and BOM data of 0.6 deg. C or so on a rather routine basis (after correcting for their different 30-year baselines used for anomalies: BOM uses 1961-1990 and UAH uses 1981-2010).

UAH Global Temperature Update for March, 2019: +0.34 deg. C.

Monday, April 1st, 2019

The Version 6.0 global average lower tropospheric temperature (LT) anomaly for March, 2019 was +0.34 deg. C, down slightly from the February, 2019 value of +0.37 deg. C:

We have made two changes in satellite processing starting with the March 2019 update. First, we have decided to stop processing of NOAA-18 data starting in 2017 because that satellite has drifted in local observation time beyond the ability of our Version 6 diurnal drift correction routine to handle it acccurately, as evidenced by spurious warming (not shown) in that satellite relative to the Metop-B satellite (which does not drift). By itself, this change reduces the trends very slightly. Secondly, we have applied a diurnal drift correction to NOAA-19, which previously did not need one because it had not drifted very far in local observation time. By itself, this increases the trends slightly.

The net effect of these two changes is virtually no change in trends (the global trend for 1979-2019 remains at +0.13 C/decade). However, individual monthly anomalies since January 2017 have changed somewhat, by amounts that are regionally dependent. For example, the standard deviation of the difference between the old and new monthly anomalies since January 2017 is 0.03 deg. C for the global averages, and 0.07 deg. C for the USA48 averages.

Various regional LT departures from the 30-year (1981-2010) average for the last 15 months are:

YEAR MO GLOBE NHEM. SHEM. TROPIC USA48 ARCTIC AUST
2018 01 +0.29 +0.51 +0.06 -0.10 +0.70 +1.39 +0.52
2018 02 +0.25 +0.28 +0.21 +0.05 +0.99 +1.21 +0.35
2018 03 +0.28 +0.43 +0.12 +0.08 -0.19 -0.32 +0.76
2018 04 +0.21 +0.32 +0.10 -0.14 +0.06 +1.01 +0.84
2018 05 +0.16 +0.38 -0.05 +0.02 +1.90 +0.13 -0.24
2018 06 +0.20 +0.33 +0.06 +0.12 +1.11 +0.76 -0.41
2018 07 +0.30 +0.38 +0.22 +0.28 +0.41 +0.24 +1.49
2018 08 +0.18 +0.21 +0.16 +0.11 +0.02 +0.10 +0.37
2018 09 +0.13 +0.14 +0.13 +0.22 +0.89 +0.22 +0.28
2018 10 +0.20 +0.27 +0.12 +0.30 +0.20 +1.08 +0.43
2018 11 +0.26 +0.24 +0.28 +0.45 -1.16 +0.67 +0.55
2018 12 +0.25 +0.35 +0.15 +0.30 +0.25 +0.69 +1.21
2019 01 +0.38 +0.35 +0.41 +0.36 +0.53 -0.15 +1.15
2019 02 +0.37 +0.47 +0.28 +0.43 -0.02 +1.04 +0.06
2019 03 +0.34 +0.44 +0.25 +0.41 -0.55 +0.96 +0.59

The UAH LT global anomaly image for March, 2019 should be available in the next few days here.

The new Version 6 files should also be updated at that time, and are located here:

Lower Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0/tlt/uahncdc_lt_6.0.txt
Mid-Troposphere: http://vortex.nsstc.uah.edu/data/msu/v6.0/tmt/uahncdc_mt_6.0.txt
Tropopause: http://vortex.nsstc.uah.edu/data/msu/v6.0/ttp/uahncdc_tp_6.0.txt
Lower Stratosphere: http://vortex.nsstc.uah.edu/data/msu/v6.0/tls/uahncdc_ls_6.0.txt