Remember the Watts et al. manuscript in 2012? Anthony Watts putting his blog on hold to urgently finish his draft? This study is now a poster at the AGU conference and Watts promises to submit it soon to an undisclosed journal.
On first sight, the study now has a higher technical quality and some problems have been solved. The two key weakness are, however, not discussed in the press release to the poster. This is strange. I have had long discussions with second author Evan Jones about this. Scientists (real sceptics) have to be critical about their own work. You would expect a scientist to focus a large part of a study on any weaknesses, if possible try to show they probably do not matter or else at least honestly confront the weaknesses, rather than simply ignore them.
Watts et al. is about the immediate surrounding, also called micro-siting, of weather stations that measure the surface air temperature. The American weather stations have been assessed for their quality in five categories by volunteers of the blog WUWT. Watts and colleagues call the two best categories "compliant" and the three worst ones "non-compliant". For these two categories they then compare the average temperature signal for the 30-year period 1979 – 2008.
An important problem of the 2012 version of this study was that historical records typically also contain temperature changes because the method of observation has changed. An important change in the USA is the time of observation bias. In the past observations were more often made in the afternoon than in the morning. Morning measurements results in somewhat lower temperatures. This change in the time of observation creates a bias of about 0.2°C per century and was ignored in the 2012 study. Also the auditor, Steve McIntyre, who was then a co-author admitted this was an error. This problem is now fixed; stations with a change in the time of observation have been removed from the study.
A Stevenson screen.
A much used type of AWS in the USA is the MMTS. America was one of the first countries to automatize its network, with then analogue equipment that did not allow for long cables between the sensor and the display, which is installed inside a building. Furthermore, the technicians only had one day per station and as a consequence many of the MMTS systems were badly sited. Although they are badly sited, these MMTS system typically measure
Two weaknesses
Weakness 1 is that the authors only know the siting quality at the end of the period. Stations in the compliant categories may have been in less well sited earlier on, while stations in the non-compliant categories may have been better sited before.pete:
Someone has a weather station in a parking lot. Noticing their error, they move the station to a field, creating a great big cooling-bias inhomogeneity. Watts comes along, and seeing the station correctly set up says: this station is sited correctly, and therefore the raw data will provide a reliable trend estimate.The study tries to reduce this problem by creating a subset of stations that is unperturbed by Time of Observation changes, station moves, or rating changes. At least according to the station history (metadata). The problem is that metadata is never perfect.
The scientists working on homogenization thus advise to always also detect changes in the observational methods (inhomogeneities) by comparing a station to its neighbours. I have told Evan Jones how important this is, but they refuse to use homogenization methods because they feel homogenization does not work. In a scientific paper, they will have to provide evidence to explain why they reject an established method that could ameliorate a serious problem with their study. The irony is that the MMTS adjustments, which the Watts et al. study does use, depend on the same principle.
Weakness 2 is that the result is purely statistical and that no physical explanation is provided for the result. It is clear that bad micro-siting will lead to a temperature bias, but this does not affect the trend, while the study shows a difference in trend. I would not know how bad or good constant siting quality would change a trend. The press release also does not offer an explanation.
What makes this trend difference even more mysterious, if it were real, is that it mainly happens in the 1980s and 1990s, but has stopped in the last decade. See the graph below showing the trend for compliant (blue) and non-compliant stations (orange).
[UPDATE. The beginning period in which the difference builds up and that since 1996 the trends for "compliant" and "non-compliant" stations is the same is better seen in the graph below computed from the data in the above figure digitized by George Bailley. (No idea what the unit of the y-axis is on either of these graphs. Maybe 0.001°C.)
]
That the Watts phenomenon has stopped is also suggested by a comparison of the standard USA climate network (USHCN) and a new climate-quality network with perfect siting (USCRN) shown below. The pristine network even warms a little more. (Too little to be interpreted.)
While I am unable to see a natural explanation for the trend difference, that the difference is mainly seen in the first two decades fits to the hypothesis that the siting quality of the compliant stations was worse in the past: that in the past these stations were less compliant and a little too warm. The further you go back in time, the more likely it becomes that some change has happened. And the further you go back in time, the more likely it is that this change is no longer known.
six key findings
Below I have quoted the six key findings of Watts et al. (2015) according to the press release.1. Comprehensive and detailed evaluation of station metadata, on-site station photography, satellite and aerial imaging, street level Google Earth imagery, and curator interviews have yielded a well-distributed 410 station subset of the 1218 station USHCN network that is unperturbed by Time of Observation changes, station moves, or rating changes, and a complete or mostly complete 30-year dataset. It must be emphasized that the perturbed stations dropped from the USHCN set show significantly lower trends than those retained in the sample, both for well and poorly sited station sets.
The temperature network in the USA has on average one detectable break every 15 years (and a few more breaks that are too small to be detected, but can still influence the result). The 30-year period studied should thus contain on average 2 breaks and likely only 12.6% of the stations do not have a break (154 stations). According to Watts et al. 410 of 1218 stations have no break. 256 stations (more than half their "unperturbed" dataset) thus likely have a break that Watts et al. did not find.
That the "perturbed" stations have a smaller trend than the "unperturbed" stations confirms what we know: that in the USA the inhomogeneities have a cooling bias. In the "raw" data the "unperturbed" subset has a trend in the mean temperature of 0.204°C per decade; see table below. In the "perturbed" subset the trend is only 0.126°C per decade. That is a whooping cooling difference of 0.2°C over this period.
Table 1 of Watts et al. (2015)
2. Bias at the microsite level (the immediate environment of the sensor) in the unperturbed subset of USHCN stations has a significant effect on the mean temperature (Tmean) trend. Well sited stations show significantly less warming from 1979 – 2008. These differences are significant in Tmean, and most pronounced in the minimum temperature data (Tmin). (Figure 3 and Table 1 [shown above])
The stronger trend difference for the minimum temperature would also need an explanation.
3. Equipment bias (CRS [Cotton Region Shelter] v. MMTS [Automatic Weather station] stations) in the unperturbed subset of USHCN stations has a significant effect on the mean temperature (Tmean) trend when CRS stations are compared with MMTS stations. MMTS stations show significantly less warming than CRS stations from 1979 – 2008. (Table 1 [shown above]) These differences are significant in Tmean (even after upward adjustment for MMTS conversion) and most pronounced in the maximum temperature data (Tmax).
The trend for the stations that use a Cotton Region Shelter is 0.3°C per decade. That is large and should be studied. This was the typical shelter in the past. Thus we can be quite sure that in these cases the shelter did not change, but there could naturally have been other changes.
4. The 30-year Tmean temperature trend of unperturbed, well sited stations is significantly lower than the Tmean temperature trend of NOAA/NCDC official adjusted homogenized surface temperature record for all 1218 USHCN stations.
It is natural that the trend in the raw data is smaller than the trend in the adjusted data. Mainly for the above mentioned reasons (TOBS and MMTS) the biases in the USA are large compared to the rest of the world and the trend in the USA is adjusted 0.4°C per century upwards.
5. We believe the NOAA/NCDC homogenization adjustment causes well sited stations to be adjusted upwards to match the trends of poorly sited stations.
Well, they already wrote "we believe". There is no evidence for this claim.
6. The data suggests that the divergence between well and poorly sited stations is gradual, not a result of spurious step change due to poor metadata.
The year to year variations in a single station series is about 1°C. I am not sure whether one would see whether the inhomogeneity is one or more step changes or a gradual change.
Review
If I were reviewer of this manuscript, I would ask about some choices that seem arbitrary and I would like to know whether they matter. For example using the period 1979 – 2008 and not continuing the data to 2015. It is fine to also show data until 2008 for better comparisons with earlier papers, but stopping 7 years earlier is suspicious. Also the choice to drop stations with TOBS changes, but to correct stations with MMTS changes sounds strange. It would be of interest whether any of the other 3 options show different results. Anomalies should be computed over a period, not relative to the starting year.I hope that Anthony Watts and Evan M. Jones find the above comments useful. Jones wrote earlier this year:
Oh, a shout-out to Dr. Venema, one of the earlier critics of Watts et al. (2012) who pointed out to us things that needed to be accounted for, such as TOBS, a stricter hand on station moves, and MMTS equipment conversion.Watts wrote in the side notes to his press release:
Note to Anthony: In terms of reasonable discussion, VV is way up there. He actually has helped to point us in a better direction. I think both Victor Venema and William Connolley should get a hat-tip in the paper (if they would accept it!) because their well considered criticism was of such great help to us over the months since the 2012 release. It was just the way science is supposed to be, like you read about in books.
Even input from openly hostile professional people, such as Victor Venema, have been highly useful, and I thank him for it.Glad to have been of help. I do not recall having been "openly hostile" to this study. It would be hard to come to a positive judgement of the quality of the blog posts at WUWT, whether they are from the pathological misquoter Monckton or greenhouse effect denier Tim Ball.
However, it is always great when people contribute to the scientific literature. When the quality of their work meets the scientific standard, it does not matter what their motivation is, then science can learn something. The surface stations project is useful to learn more about the quality of the measurements; also for trend studies if continued over the coming decades.
Comparison of Temperature Trends Using an Unperturbed Subset of The U.S. Historical Climatology Network
Anthony Watts, Evan Jones, John Nielsen-Gammon and John ChristyAbstract. Climate observations are affected by variations in land use and land cover at all scales, including the microscale. A 410-station subset of U.S. Historical Climatology Network (version 2.5) stations is identified that experienced no changes in time of observation or station moves during the 1979-2008 period. These stations are classified based on proximity to artificial surfaces, buildings, and other such objects with unnatural thermal mass using guidelines established by Leroy (2010). The relatively few stations in the classes with minimal artificial impact are found to have raw temperature trends that are collectively about 2/3 as large as stations in the classes with greater expected artificial impact. The trend differences are largest for minimum temperatures and are statistically significant even at the regional scale and across different types of instrumentation and degrees of urbanization. The homogeneity adjustments applied by the National Centers for Environmental Information (formerly the National Climatic Data Center) greatly reduce those differences but produce trends that are more consistent with the stations with greater expected artificial impact. Trend differences between the Cooperative Observer Network and the Climate Reference Network are not found during the 2005-2014 sub-period of relatively stable temperatures, suggesting that the observed differences are caused by a physical mechanism that is directly or indirectly caused by changing temperatures.
[UPDATE. I forgot to mention the obvious: After homogenization the trend Watts et al. (2015) computed are nearly the same for all five siting categories, just like it was for Watts et al. (2012) and the published study Fall et al. Thus for the data used by climatologists, the homogenized data, the siting quality does not matter. Just like before, they did not study homogenization algorithms and thus cannot draw any conclusions about them, but unfortunately they do.]
Tweet
Related reading
Anthony Watts' #AGU15 poster on US temperature trendsBlog review of the Watts et al. (2012) manuscript on surface temperature trends
A short introduction to the time of observation bias and its correction
Comparing the United States COOP stations with the US Climate Reference Network
WUWT not interested in my slanted opinion
Some history from 2010
On Weather Stations and Climate TrendsThe conservative family values of Christian man Anthony Watts
Watts not to love: New study finds the poor weather stations tend to have a slight COOL bias, not a warm one
Poorly sited U.S. temperature instruments not responsible for artificial warming
If you mean with your second and third point your claim that homogenization does not work purely based on some trends being of similar size then I think I did respond to it. It is not even wrong.
ReplyDeleteOh, noooo. After homogenizations, the well and poorly sited station trends should be roughly equivalent.
The problem arises that when a systematic binning (for microsite) is applied, what we see is a large higher group and a much smaller lower group, and that:
a.) The smaller group is better sited, and,
b.) The smaller group is adjusted upward to match the larger group.
If the larger group is showing the correct signal, then homogenization has worked and the result is improved.
But if the smaller group is showing the correct signal, then we have a problem, and homogenization-nation has become the homogenization bomb. And, yes, it's just as you described -- a gradual inhomogeneity.
As homogenization addresses the month-to-month data, it corrects, adjusts, bringing the lesser line into conformity with the greater.
You are so close, Vitor, so close. All you need to do now is to apply the same method you are using, except you must add the final capstone: that of microsite. Get it where you can. Infer it where you must (again, using homogenization methodic, inter alia).
And when/if you have done, I confidently predict you will find that all siting class trends are still roughly equivalent to each other. #B^)
Evan Jones: "That is what is happening. The poorly sited majority is having a gradual inhomogeneity over time. That is what is going wrong. That is what needs to be fixed. Then homogenization will perform as intended."
ReplyDeleteYour study does not provide any evidence for this whatsoever and you do not have proposed any physics that could produce such a gradual inhomogeneity mayhem.
It sure ain't your heat sinks. The heat sink would absorb heat in the summer and release it winter. It would dampen the seasonal cycle just as it would dampen long term temperature increases, while you claim there are increases.
What you call a "hypothesis" I would call a list of assertions. Before it is an hypothesis it would need theory to make the assertions into a coherent physically possible claim.
To claim that the "The smaller group is adjusted upward to match the larger group" does not become true by asserting it again and again. I have explained above why that is not how it works. In politics it is important to repeat your talking points again and again and apparently you can make some people believe the most obvious lies as long as you repeat it over and over and over again. In science you need evidence and arguments.
ReplyDeleteWe are now at over 200 comments. I suggest you get out your pencil and write down some equations because this discussion is not going anywhere and only you can convince you and to see you are wrong you need to clearly formulate what you are proposing with equations. Without the clarity of physical equations you seem to be able to trick yourself into believing whatever you would like to believe.
First, the other posts:
ReplyDeleteI'm talking about changes which you haven't accounted for in the stations you've kept. Obviously for stations which are genuinely unperturbed this isn't an issue.
Using either the unperturbed dataset, the pertturbed dataset, or the combination of the two, the digression of the well and poorly sited stations persists. That verifies that perturbation does not materially affect the trend digression between the poorly and well sited stations. USHCN metadata has improved greatly since Menne (2009).
Expected amplification over mid-latitude continental areas is less than 1, so CONUS satellite TLT trends should be lower than surface.
Not according to Klotzbach, et al., which says the opposite. LT trends over land are expected to by 10% higher than land surface (and a lot more over seas, but we address only land temps, so for our purposes that that is moot, levaing us with a 10% digression.)
This effect is observable in inter-annual variations, which are notably larger for surface records compared to TLT.
The annual anomalized lows are higher and the highs are lower. But as for trend, the surface observations are not the solution, they are the problem: The LST temps do not account for microsite trend effect. When not accounting for microsite effect, the USHCN trends are indeed higher than those of satellites. But when microsite is taken into account (by dropping the poorly sited stations), the trends fall right into line with Klotzbach, et al.
Uh, what? I can't find any such statement in Menne et al. 2010, and it flatly contradicts Menne et al. 2009, which gives values of -0.44 °C and -0.45 °C for the effect of transition in max and min temperatures respectively. If you have uncorrected ASOS stations in your dataset, you've got a huge systematic cooling bias lurking in there somewhere.
If that were the case, ASOS raw data (including equipment shift) would be cooling -- rapidly. It's not. So (again) the discrepancy is not mostly a result of bad equipment, it is a result of the far better siting of ASOS staions. And they do not take (unchanged) siting into effect when they calculate the discrepancy.
Besides, only ~5% of our unperturbed stations are ASOS, and the effect is further diminished by the poor distribution: 8 of these are in the Northeast and 4 in the Southeast, so that washes out much of the remaining overall effect.
To claim that the "The smaller group is adjusted upward to match the larger group" does not become true by asserting it again and again. I have explained above why that is not how it works.
ReplyDeleteActually, you explained it quite well: The only scenario where this would go wrong, as I explained on WUWT before, would be if the majority of stations would have a gradual inhomogeneity, rather than jumps.
Exactly. You have correctly identified the problem: the poorly sited stations have a gradual inhomogeneity with the poorly sited stations.
We are not talking about microsite that changes as a result of moves or an encroachment such as an added building or paved surface. That sort of change would show a jump. Yet this would affect the poorly sited stations, not the well sited set (by definition). And the effect will be minor owing to the much larger size of the unperturbed poorly sited station set -- and they receive almost no adjustment.
Instead, we are talking about microsite (good or bad) which has remained (subject to less-than-perfect, but greatly improved, HOMR metadata) constant, throughout the study periods in question (1979 to 2008, 1979 to 1998, and 1999 to 2009). And even using TOBS-adjusted (non-homogenized) data for our unperturbed set, the discrepancy between well and poorly sited stations remains constant. (JN-G originally addressed this issue was satisfied with the result.)
The poorly sited unperturbed set on average receives very little net adjustment via homogenization. It is the well sited unperturbed set that on average is greatly increased as a result of being pairwised, not with other well sited stations, but with ~80% poorly adjusted stations. That is how homogenization spuriously drags their trends up.
This is inevitable, as unchanged microsite rating is not recognized as a problem relating to trend.
And yes, as you correctly say, this paper is not a be-all, end-all study: Further testing and investigation is obviously required.
In fact, I have some further testing of the heat sink / microsite hypothesis to report regarding USCRN v. USHCN, which results I will post here separately, as I said I would earlier.
Evan Jones: "Exactly. You have correctly identified the problem: the poorly sited stations have a gradual inhomogeneity with the poorly sited stations.
ReplyDeleteOr so you have asserted, but your heat sink mechanism does not work and you have not provided any analysis that shows there is a gradual inhomogeneity in the data. Just that you find different trends is not evidence, also a jump in the data changes a trend.
Evan Jones: "We are not talking about microsite that changes as a result of moves or an encroachment such as an added building or paved surface. That sort of change would show a jump."
Exactly, when something changes close to the stations (micro-site) it produces a jump. Urbanization can produce a gradual inhomogeneity because it add up many small changes in an area of a km around the station. Changes within 20-30 m around the station are nearly by definition rare and larger, that is: a jump.
Evan Jones: "Yet this would affect the poorly sited stations, not the well sited set (by definition)."
No, the well-sited stations may have been worse sited before and also changes that do not change the stations category in your classification can produce jumps.
But really, make a hypothesis, write down equations. Only then can you make a solid case and show that all the evidence supports your hypothesis. Now all you are doing is science by assertion, which is not science.
So I was poking around on the NOAA ftp site, and I found something rather interesting:
ReplyDeletethe intent was to use only non-USHCN sites as neighbors in the PHA to homogenize USHCN stations. However, 46 USHCN stations were inadvertently used as neighbors of other USHCN stations. In version 3.3.0 no stations used in producing the USHCN data records are used as neighbors of the USHCN.
(emphasis mine)
Seems to me that conclusively kills the argument that the PHA is bringing well-sited USHCN stations into line with poorly-sited ones. Now, perhaps PHA is bringing well-sited USHCN stations into line with poorly-sited non-USHCN stations, but obviously a USHCN-only dataset can't possibly establish that.
Seems to me that conclusively kills the argument that the PHA is bringing well-sited USHCN stations into line with poorly-sited ones. Now, perhaps PHA is bringing well-sited USHCN stations into line with poorly-sited non-USHCN stations, but obviously a USHCN-only dataset can't possibly establish that.
ReplyDeleteAgreed. And there you have it.
I would use any comparison between HCN and CRN only to see how a basically badly sited net (HCN)compares with a well sited net (CRN).
It took me a little longer than I thought it would, anomalizing all that data, but I will mow post the results ...
Or so you have asserted, but your heat sink mechanism does not work and you have not provided any analysis that shows there is a gradual inhomogeneity in the data.
We have done the latter. As for the former, I will make a separate post.
Just that you find different trends is not evidence, also a jump in the data changes a trend.
That's the point: the unperturbed Class 1\2 data shows no incongruent jumps. Just a divergence over time until 1998 and a reconvergence from 1999-2008 as cooling occurs.
But really, make a hypothesis, write down equations. Only then can you make a solid case and show that all the evidence supports your hypothesis. Now all you are doing is science by assertion, which is not science.
Sure it is. Many scientific advancements start with an observation and an assertion -- a hypothesis. This is the assertion phase. #B^) We have a result. We infer the cause. I think perhaps that inference will now be better addressed, by others, as well as by our team. The equations will come.
I can provide a (very crude) analogy for heat sink effect. Please don't take this as a direct comparison with surface stations, just look at it as an example where insolation gets locally "trapped" and causes a temperature increase compared with "Class 2" conditions.
ReplyDeleteCRUDE EXAMPLE:
Me: Parked cars are deathtraps for dogs: On a 78-degree day, the temperature inside a parked car can soar to between 100 and 120 degrees in just minutes, and on a 90-degree day, the interior temperature can reach as high as 160 degrees in less than 10 minutes.
An example with “real numbers” (allegedly).
So outside the car over time (under a day in this case) of from 78 to 90 is compared with an inside-the-car increase from 120F to 160F. That is a 12F increase vs. a 40F increase inside the car.
The heat sink (the car) is warming at over twice the rate, the trend inside the car, as it is outside the car.
We are, by analogy, measuring our temps largely from inside the car. And, yeah, the car cools faster at the same rate during the subsequent “cooling phase”.
So if there is no overall trend, there will be no divergence in trend as a result of spurious heat sink effect. But if there is a trend, either cooling or warming, that trend will be exaggerated.
JN-G's reply: I agree with you. Stating your observation in physical terms, the rate of heat loss depends on the excess temperature. If there's more energy going in, there needs to be more heat loss and so the excess temperature will be greater. If the energy going in is rising, the excess temperature will rise, and if the energy going in is declining, the excess temperature will fall.
1.) Regardless of whether or not there is any trend, the differences between summer and warmer temperatures should be greater for HCN than for CRN (2015-2014). This should not only occur a an overall result, but also in a large majority of years. A positive result would suggest heat sink effect in operation on an annual/seasonal basis as well as for longterm trends. (It should work for both.)
ReplyDeleteRESYLT: Yes, in all years, and increase in trend of difference.
CRN, S-W
Summer - Winter 2005: 17.30
Summer - Winter 2006: 16.03
Summer - Winter 2007: 17.35
Summer - Winter 2008: 17.38
Summer - Winter 2009: 16.75
Summer - Winter 2010; 18.35
Summer - Winter 2011: 18.55
Summer - Winter 2012: 15.40
Summer - Winter 2013: 17.60
Summer - Winter 2014: 17.95
AVERAGE: 17.3C
TREND: 0.07C/d
HCN, S-W
Summer - Winter 2005: 18.95
Summer - Winter 2006: 17.72
Summer - Winter 2007: 19.32
Summer - Winter 2008: 17.38
Summer - Winter 2009: 16.75
Summer - Winter 2010: 20.22
Summer - Winter 2011: 20.47
Summer - Winter 2012: 16.93
Summer - Winter 2013: 19.79
Summer - Winter 2014: 20.20
AVERAGE: 19.2C
TREND: 0.11C/d
2.) If there is any overall trend from 2005-2014, the TOBS-adjusted HCN trend should be larger than the CRN trend.
RESULT: Yes. By ~0.14C/decade.
USHCN Anomalized
TOBS Tmean TREND
(100/th Deg. Celsius)
2005: 20
2006: 56
2007: 22
2008: -58
2009: -52
2010: -10
2011: -7
2012: 113
2013: -51
2014: -40
Trend: -0.39C/d
USCRN (Non-Anomalized)
Obs. Tmean TREND
Deg. Celsius)
10.53
2006: 10.87
2007: 10.58
2008: 9.91
2009: 9.88
2010: 10.27
2011: 10.39
2012: 11.46
2013: 10.00
2014: 10.1
Trend: -0.25C/d
3.) If there is any overall trend from 2005-2014, HCN homogenized data should magnify that trend over HCN TOBS-adjusted data, same as it does for the unperturbed 1979-2008 set.
RESULT
Yes, but only by a little.
USHCN Anomalized
TOBS Tmean Trend
(100/th Deg. Celsius)
2005 20
2006 56
2007 22
2008 -58
2009 -52
2010 -10
2011 -7
2012 113
2013 -51
2014 -40
Trend -0.39
USHCN Anomalized
Homogenized Tmean Trend
(100/th Deg. Celsius
2005: 20
2006: 58
2007: 23
2008: -57
2009: -53
2010: -11
2011: -5
2012: 115
2013: -51
2014: -41
Trend: -0.40
All in all, very good results. Stronger for the first two than for the last.
Actually, at next glance, that point 3 looks pretty good: There was a slight adjustment increase (in a warming trend) for unperturbed Class 3\4\5s in the 1979-2008 series, and we see a slight downward adjustment (in a cooling trend) for the USHCN entire (mostly poorly sited).
ReplyDeleteSo all test results work. They do not prove the hypothesis, but they do offer support.
The heat sink was a hypothesis. It failed.
ReplyDeleteYour 3 unconnected ideas are not a hypothesis.
(A car is not a heat sink. I know the request is considered impolite at WUWT, but could you please read a physics book. If only because it is fascinating.)
The heat sink was a hypothesis. It failed.
ReplyDelete1.) How did it fail?
2.) If it is not the HS hypothesis that is causing the divergence of Class 1\2 and Class 2\4\5 (either perturbed or unperturbed, do you have an explanation for it that does not involve the quality of the siting?
JN-G seems to think we have a ground truth, here. And, bottom line, the HS hypothesis confirms the reality of at least some degree of global warming.
Point 3, the the above post demonstrates (again) that Homogenization does not correct for poor siting, but instead slightly exaggerates the effects of even badly sited stations. I'd say that was quite to the point.
(A car is not a heat sink. I know the request is considered impolite at WUWT, but could you please read a physics book. If only because it is fascinating.)
ReplyDeleteThe car absorbs, accumulates and re-radiates energy. Enough to makes its interior warmer than the air outside. For that matter, Leroy (2010) singles out parking lots as a heat source, regardless whether paved or not. He was thinking of a bunch of cars, too.
Unlike me, JN-G has read a book, and he agrees with the basic mechanism: An object absorbs more heat than the ground otherise would without it. that heat is radiated toward the sensor. It is an amplification of the normal ground effect. The offset effect on the sensor becomes greater as warming occurs (and the reverse during cooling). Therefore the trends are magnified, during either a warmig or a cooling phase.
Heat sinks reduce variability and trends. For one.
ReplyDeleteOne hypothesis would be the one in the post: you did not remove all inhomogeneities due to changes in siting. Do you have photos for all stations in 1979 to be able to assess that siting quality?
Write down some equations:
ReplyDeleteTsensor = Tregional + X
Theatsink(t+1) = Theatsink(t) + Y
Yes, NG knows physics, thus it is not credible that he would agree with your heat sink. It could be that he is more polite than I am.
Thanks yet again, Veev, Now were talkin'.
ReplyDeleteOne hypothesis would be the one in the post: you did not remove all inhomogeneities due to changes in siting.
If that were the case, the difference between well and poorly sited stations would disappear. It does not, as you can see from the perturbed set on the chart in the article you posted. The reason it does not disappear is that TOBS affects all classes roughly equally.
So the trends of both well and poorly sited perturbed stations are lower than their unperturbed counterparts (which have been largely cleansed of TOBS bias), but they show the same trend disparity between good and poor siting.
Also, if metadata were a severe problem, our unperturbed TOBS-adjusted dataset would wash out the divergence. But it does not.
A strong support for our findings is:
-- NOAA TOBS-only adjusted umperturbed data also shows the divergence between well and poorly sited stations. So it's not just the raw data that says this, and itcan't be a failure of the TOBS metadata record.
-- The divergence between well sited stations and the homogenized set remains using TOBS-only adjusted data.
-- When homogenization is applied, the divergence disappears from the TOBS-adjusted data: The well sited station trends inceased to the same level of the poorly sited stations, which are hardly adjusted at all.
If microsite were not a factor, there would be no divergence at any level. Unmomogenized results (neither raw nor TOBS-adjusted) would show it. But there is a divergence. And we can clearly thee the before-and-after effect of each level of adjustment.
To correct this error, homogenization needs to be keyed to microsite, which is shown to have as big an effect on the USHCN at large as TOBS bias, but in the opposite direction. The two counter each other fairly neatly. Both biases accounted for roughly equals neither bias accounted for.
Do you have photos for all stations in 1979 to be able to assess that siting quality?
No, we do not. We do have (increasingly poor) satellite images of the microsite back to the mid 1990s on most (amazing how little microsite change there has been). But usually the resolution is too poor to make out the station if you go back too far. So we can usually show something of the microsite history, but usually not a distinct image of the sensor, itself. We have some interviews (but it was not helpful when NOAA removed all the curator's names from the metadata when they first took notice of the project).
Like GHCN, we do got limitations. The more recent the time period in question, the mode resources available.
But in any case, if perturbation does not remove the divergence between well and poorly sited stations, the the quality of the metadata becomes a side issue.
No, we are not talking, at least not in the sense of communicating new information. You are just repeating the same old talking points over and over again.
ReplyDeleteThen what parts do you disagree with, either in terms of the data itself, the logic, or the attribution?
ReplyDeleteAll you seem to be saying is that incomplete metadata will account for the digression. But if it did, then why does unperturbed TOBS-adjusted data show the same digression between the unperturbed Class 1\2s and Class 3\4\5s?
It is only after homogenization is applied that the digression vanishes. And it is the lower-trend class 1\2s that get the adjusting.
So how would bad metadata be the problem if TOBS-adjusted and raw show much the same divergence? And wouldn't that create jumps?
I really have no idea what I could still explain you what I did not explain again and again in my post and my dozens of comments.
ReplyDeleteThat the "unperturbed" Class 3\4\5s trend is about the same as the homogenized trend is pure coincidence and also no longer the case once you make the right, much larger MMTS corrections to your perturbed "unperturbed" data.
That the "unperturbed" Class 3\4\5s trend is about the same as the homogenized trend is pure coincidence
ReplyDeleteIn a cosmic sense I suppose this is possible. But considering the very strong statistical significance of our findings, the odds are astronomical that micrositing is at least in some way connected with whatever is causing this divergence.
That I think is what helped win JN-G over. He runs his own set of numbers independently as a check on my own (and occasionally vice-versa).
So even in the unlikelihood it is not heat sink, that would mean whatever turned out to be causing the divergence would still probably be inextricably related to microsite.
Our chance of the heat sink hypothesis being wrong is real but small. But the chance of the divergence being coincidence, unrelated to siting, is minuscule.
and also no longer the case once you make the right, much larger MMTS corrections to your perturbed "unperturbed" data.
A higher percentage of Class 3/4/5s have converted to MMTS than Class 1\2, as heavy as the latter is in CRS units. So applying MMTS adjustment slightly increases the disparity. The more you adjust MMTS up, the greater the disparity becomes between Class 1\2 and Class 3\4\5.
And "much larger"? Much larger than what? Maybe you'll squeeze 0.02C/decade more than we have already added out of MMTS alone (perhaps 0.040C/decade per MMTS anomaly each year after conversion) -- minus whatever CRS takes back. And when it comes to microsite, we are talking a 0.11C/decade -- that is much larger.
As for ramped up MMTS, it'll bring down those error bars, too. So you won't be beating our statistical significance between the unperturbed Class 1\2s vs. Homogenized down that road. Much less the unperturbed 1\2 - 3\4\5 class divide; you'll improve that.
And "unperturbed" only means that there are no recorded moves or TOBS flips. It includes all stations with MMTS conversion (and sometimes back again).
So at this point, I think I'll be applying a much larger MMTS correction. And a CRS correction, too. And show the results with both corrections and without either for comparison.
Does a.) Pure coincidence, b.) MMTS adjustments, and, C.) potentially shonky metadata adequately sum up your objections here?
The statistical significance of your trends only means that the trend is most likely not zero. It does not change that two values can be accidentally close to each other (as long as you do not implement the right MMTS corrections).
ReplyDeleteAs quoted above the MMTS corrections should be in the order of (given that you claim there were no relocations: "the average effect of the statistically significant changes (−0.52°C for maximum temperatures and +0.37°C for minimum temperatures) is close to Hubbard and Lin’s (2006) results for sites with no coincident station move."
Then the coincidental matching of the trends is gone. I could not care much at this stage that the difference between the "compliant" and "non-compliant" siting classes becomes larger. Once your work becomes more numerical it may help you too show that the effect is larger than can be explained by the other errors and thus maybe a siting effect, but in the current exploratory phase of your study it does not change the argument.
Your previous list with problems was more comprehensive.
The behaviour of Donald Trump and hubris in the comments at WUWT seem to suggest that there is a demographic that thinks pounding on your chest is convincing in itself. In most of the world, and certainly in science, over-confidence leads to loss of credibility. It irritates people and makes communication harder.
"the average effect of the statistically significant changes (−0.52°C for maximum temperatures and +0.37°C for minimum temperatures) is close to Hubbard and Lin’s (2006) results for sites with no coincident station move."
ReplyDeleteThat is not a trend adjustment. That is an offset. If it were a trend adjustment. MMTS Tmin would be adjusted to a freezing -0.15C/decade and Tmax to a broiling +0.60C/decade. Doesn't add up.
And since L&H is a 2 station to 2 station comparison over a 1 year period, it hardly seems to bear any weight, anyway. And it does not say thing one about trend. Would you accept my summer/winter HCN/CRN disparity argument if I used two stations each over one measly year? I think not. I am getting the impression that those who hype L&H don't actually read the part about exactly how the experiment is actually conducted or what the results even apply to.
Menne (2009), for all its flaws, has a much better approach -- at least he actually measures the disparity of MMTS and CRS within the USHCN as a whole. At least he has a decent-sized sample, not to mention actual coverage of the relevant time period in question. The problem with Menne is that he is adjusting MMTS to near-CRS levels, and CRS is too high.
Naturally it is an offset. That is what is happening.
ReplyDeleteYes L&H is a small sample. That is why I carefully wrote "in the order of". The values of Menne et al. (2009) are not applicable to your case. They are for the MMTS transition plus a relocation. The relocation has caused much of the bad siting in the USA, it is likely warming and partially offsets the cooling of the MMTS systems.
To compare seasonal cycles you would have to accurately match USHCN and USCRN because the seasonal cycle depends strongly on location. Furthermore, the homogenization of USHCN applies equal size adjustments to all months. The seasonal cycle itself is thus not homogenized. If you would like to use this as an argument you will have to homogenize the USHCN network with a homogenization method that adjusts the monthly values, for example, ACMANT or MASH.
If your "heat sink", or whatever else changes the variability on decadal scales, you should be able to see it in the year to year variability as well. Especially if you claim you can see it in the seasonal cycle as well. But you do not. USCRN and USHCN match very well in their year to year variability.
See also this NOAA publication.
The statistical significance of your trends only means that the trend is most likely not zero. It does not change that two values can be accidentally close to each other (as long as you do not implement the right MMTS corrections).
ReplyDeleteIt means (specifically) that the divergence between unperturbed well and poorly sited is well under 1% likely to be a result of random chance. That's what I think convinced JN-G.
Then the coincidental matching of the trends is gone.
The trend disparity would not disappear. It would increase. There is a higher percentage of Class 2\4\5 that converted to MMTS than there are Class 1\2s. I will no doubt up the ante on MMTS in any case, come what will -- but watch for the CRS side of life.
I could not care much at this stage that the difference between the "compliant" and "non-compliant" siting classes becomes larger. Once your work becomes more numerical it may help you too show that the effect is larger than can be explained by the other errors and thus maybe a siting effect, but in the current exploratory phase of your study it does not change the argument.
We are still in exploratory phase. But given that the higher percentage of badly sited stations are MMTS-converted, then it is inevitable that increasing the MMts adjustment would likewise increase the difference between the well and poorly sited stations will increase. Obviously. Not science, just basic arithmetic.
Your previous list with problems was more comprehensive.
Much to be examined, much to be improved, much to be resolved.
The behaviour of Donald Trump and hubris in the comments at WUWT seem to suggest that there is a demographic that thinks pounding on your chest is convincing in itself.
Both sides do that all the time. It is a sword that cuts both ways. Chests will be pounded. But we are men of the world, you and I, and are interested in what is actually going on.
In most of the world, and certainly in science, over-confidence leads to loss of credibility. It irritates people and makes communication harder.
In the climate argument, it has gone way beyond that. Irritation as a result of overconfidence is the least of it. It's down to name-calling, calculated misrepresentation, and personal attacks, even litigation. All bets are off.
And maybe a little genuine confidence is not entirely misplaced, at that. In most of the world, and certainly in science. I made three predictions regarding HCN vs. CRN without knowing what the result would be. That took a lot of confidence not to mention putting me out on a limb. But I did it anyway, and it cost me at least 20 hours of work. And you did note the results, I take it? So I am pleased. But I would have lived with the results if I had not been pleased with them, too. There will be more such risks in future. I only hope I maintain the confidence to live up to them.
Hubbard and Lin 2004 is a 1-year comparison of two stations. Hubbard and Lin 2006 is a 33-year comparison of 279 stations. Their inclusion criteria were: a) USHCN stations, b) no station moves, c) no instrument height changes, d) no other type instruments used in the period from 1970 to 2003, and e) for MMTS stations, at least 171 months of data on either side of the transition. That left them with 163 CRS stations, and 116 MMTS. They found, on average, that the CRS-MMTS transition resulted in a .57 °C cooling bias in max temperatures, a .35 °C warming bias in min temps, and no difference in trend over the following 171 months. They also cautioned that, due to the large variance in their results, global adjustments were not appropriate, but rather that CRS-MMTS transition bias should be evaluated on a case-by-case basis.
ReplyDeletetl;dr: correcting for MMTS bias is a) non-trivial, and b) necessary.
Very good, MM. And thanks. Veev, make note. Sorry about length. These comments may be quoted and are "part of the record". So I have to be pretty specific, and that takes words.
ReplyDeleteOne thing you can say for H&L is that they are succinct. I notice they use USHCN metadata for MMTS (always very good), and TOBS-adjusted for pairwise. H&L-06 makes it very clear 5that we are talking straight offset, here. NOT trend. If Menne uses homog for MMTS (keyed to CRS, of course), he will get a nice trend adjustment. That would explain what I aw earlier.
Writ large, I kinda like his method. I think maybe I'll replicate it using regional gridding and individual station adjustment using our unperturbed set. Nothing like doing it oneself -- especially with all the metadata improvements since 2006.
Evan Jones: These comments may be quoted and are "part of the record".
ReplyDeleteMaybe people will quote it, but at least for any paper that may come out of this, your comments do not matter at all. If your manuscript is of scientific quality in its description of the work, all that matters is that write-up and how it came together is irrelevant, with the exception of fraud.
Maybe people will quote it, but at least for any paper that may come out of this, your comments do not matter at all. If your manuscript is of scientific quality in its description of the work, all that matters is that write-up and how it came together is irrelevant, with the exception of fraud.
ReplyDeleteYes.
But I may have to deal with any quotes.
BTW, I'm not talking about merely replicating L&H-06. I want to go all the way down the road he suggested and get individuals for all the stations. Not a promise, but I think I can dope it out. His numbers may (or may not) be off, but his basic approach appears sound.
Evan Jones writes: "The trend disparity would not disappear. It would increase. There is a higher percentage of Class 2\4\5 that converted to MMTS than there are Class 1\2s. I will no doubt up the ante on MMTS in any case, come what will -- but watch for the CRS side of life."
ReplyDeleteAs Hubbard & Lin wrote, you cannot apply the MMTS Bias corrections globally - they require individual site adjustment based on specific temperatures, solar radiance, and wind. Your supposition is due to thinking you can apply a *global* offset - you can't. Unless you believe that your two subsets are homogeneous already, per the aforementioned criteria, any suppositions about the effect of the MMTS Bias adjustment are just a WAG. And your whole point has been the two sets are *not* homogeneous.
You also forget that the Hubbard and Lin MMTS Bias adjustments are *not* linear.
As Hubbard & Lin wrote, you cannot apply the MMTS Bias corrections globally.
ReplyDeleteI am looking at using his own methods, pairwising with stations in the same region. Pretty much the same as H&L (but with some differences I'll explain later).
Interesting to note that PRT shows an even greater absolute-value trend sensitivity than MMTS (and especially) CRS using non-converted MMTS & CRS stations from 2005 - 2014.
Well, for our MMTS issue, here's how the procedure needs to occur.
ReplyDeleteWe do need to address CRS. L&H-06 gives us his 2006-metadata version of jumps for CRS vs. MMTS. But he also says CRS looks wonky and needs to be adjusted. (I'll provide data with both CRS adjusted and CRS not adjusted, for discussion).
So first, CRS must be adjusted for trend. How to do this? We don't want to simply match to sats. That reduces the argument to surface vs. sats. We want to connect it to CRN, which doesn't go back that far or this whole issue would be moot.
But what we can do is:
First compare the magnitude of the trend of UAH6(latest) with CRN (2005 - 2014). See how well it tracks is it closer to CRN than USHCN. There is 15% less trend for UAH (less cooling, in the 2005-1014 case). That suggests a baseline for CRS adjustment at 15% higher than UAH. So that's what we shoot for. (This is applied to the anomalies, not the raw data, because we want to reduce the cooling as well as the warming.) To support this, CRS cooled at a rate just about as great (as CRN percentage) during the cooling as it warmed from 79-08, another support for heat sing hypothesis.
So anyway, then we pairwise (5 years forward and back) to calculate our jumps. An MMTS can be used for comparison, but any station included in the pairwise must not have equipment conversion within 5 years of its inclusion in either direction. All such eligible stations in a region will be used. I'd like to do it Class 1\2 to Class 1\2, but there are simply not enough of them. So I will do it all-to-all, even though this will work against the class 1\2s.
Short version:
1.) Adjust CRS trend.
2.) Do pairwise.
Note both jumps for both the CRS adjusted and non-adjusted.
We will see if that wipes out the difference between the Class 1\2s and the adjusted record as you say it will. I don't know what the results will be, so it is exciting. This is taking L&H06 to its conclusion (individuals for each station and CRS accounted for).
(Anyone who thinks these discussions are without value needs to re-evaluate.)
Results pending.
Over at Nick Stokes Olof commented:"The latest Ratpac A data is out now. The troposphere temperature 850-300 mbar is skyrocketing, winter (2 months of 3) is up by 0.25 C from autumn.
ReplyDeleteAt the peak of the 1998 el Nino (spring season), Ratpac and UAH v6 were quite similar in 1981-2010 anomalies. Now,in the present winter season Ratpac leads by 0.4 C..."
Someone will need to analyze microsite on those radiosondes - I'm sure that must be the problem.
Empirical demonstration that adjusted historic temperature station measurements are correct, because they match the pristine reference network: Evaluating the Impact of Historical Climate Network Homogenization Using the Climate Reference Network, 2016, Hausfather, Cowtan, Menne, and Williams. http://www-users.york.ac.uk/~kdc3/papers/crn2016/background.html
ReplyDeleteEvan, you got some splainin to do.
How strange that Evan has not 'splained in response to my most recent comment. Maybe he does not understand the issue, so I'll be explicit:
ReplyDeleteWatts's and Evan's entire endeavor's sole nominal purpose is to create a semblance of a "pure" surface station temperature record--a record that is unsullied by microsite influences. They are attempting to approximate the record that would have resulted from stations lacking microsite "problems." Their attempt involves sampling and adjustments that they admit are destined to result in a record less pure than what would be gotten from physical stations that always lacked microsite "problems."
But there is in fact such a "pure" set of physical stations--the Reference Network. Watts's and Evan's constructed record can only be hoped to be nearly as pure as that one.
The new paper by Hausfather, Cowtan, Menne, and Williams shows unequivocally and more thoroughly than any efforts before, that the current homogenization scheme already succeeds in making the record indistinguishable from the reference network. So the existing homogenization scheme already accomplishes what Watts and Evan nominally are struggling to accomplish.
So Watts and Evan are wasting their time. Their nominal goal already has been met by other people.
But of course their real goal is to show that the current homogenization scheme does not produce a record nearly identical to the reference network. So they have failed to meet their real goal, and there is no hope of them meeting that goal--ever. Ever. Really.
It would be genteel for them to admit they were wrong.
Obviously Evan Jones hasn't shown up to tell us the results of latest adjustments, but it's quite instructive to read his most recent comments in his safe haven at WUWT on Feb. 17-19.
ReplyDelete"All surface station data (arguably other than CRN) is bad data."
Yes, so let's discuss Hausfather et al 2016.
"One cannot achieve perfection. But sometimes one can identify issues and improve otherwise downright incorrect data to the level of usefulness.
Or, as in the case of NOAA, make it worse."
Disregarding every piece of research ever done on the topic.
"When VeeV wants me to feed my stuff into his black box, my impulse is to infer (from the results) what is going into the black box and create on my own a cruder, but entirely transparent and understandable box that can be understood and discussed — positively or negatively — by anyone."
So, homogenization is a blackbox?
"The best way to ensure the best result is complete transparency of method. No black boxes need apply."
I really don't think Evan understands the concept of 'blackbox.'
"The adjustments made to sat data are large but fairly simple."
LOL. You can't make this up. Has any layman ever actually taken the raw satellite data and duplicated any satellite temperature series? Anyone?
There's no indication Evan and company have really changed at all. They have a hypothesis and *every* new piece of data confirms it. The 2012 "pre-release" paper had an uncounted number of errors, but none of them changed the conclusion. Every piece of research shows that hommogenization works, but it doesn't change their conclusion. USCRN matches USHCN, but that doesn't change their conclusion.
Basically, there is *zero* evidence that can ever change their conclusion.
Hilarious. Homogenization a black box??? Just because they refuse to look into it?
ReplyDeleteThe homogenization algorithm of NOAA, they keep on fretting about, is open source software.
If that is too complicated for them, they can have a look at the method of Kevin Cowtan, who only needed 150 lines of code for a basic homogenization method.
If something is a black box, it is all the computations WUWT makes for their press releases on their manuscripts that are always about to be submitted. Since 2012.
I ticked off the results of homogenization. You made your own list. It's quite clear what is going on, writ large.
ReplyDeleteAnd, as I have said, it will work -- provided always there is not a systematic bias in the data. There is. And the results are just as expected and rather easily explained. But it can be fixed, if you are willing to do it. Account for Microsite bias. Account for CRS Tmax bias. Then you're golden. Not until.
Empirical demonstration that adjusted historic temperature station measurements are correct, because they match the pristine reference network
ReplyDeleteI ran 'em myself (2005 - 2014). They don't. UAH6.0 is closer.
Obviously Evan Jones hasn't shown up to tell us the results of latest adjustments
ReplyDeleteJust finished. Results are not far from Quayle (1993).
All: Tmax -0.29, Tmin:+0.26 (Quayle is -0.4, +0.3)
Class 1\2: Tmax -0.20, Tmin: +0.09
Class 3\4\5: Tmax +0.30, Tmin +0.30
This varies by region, obviously, but that's the average for all unperturbed stations.
As you can see, the Class 1\2 trends are increased by the jump and Class 3\4\5s are largely unaffected. So the full brunt of the jump adjustments fall upon the well sited stations.
And, to maximize the jumps I applied the results before adjusting for CRS, not after.
As with both Quayle and H&L, results can be large for individual stations and in either direction. They got that right.
The results will be applied immediately, as indicated in my posts above.
Any questions?
And, as I have said, it will work -- provided always there is not a systematic bias in the data.
ReplyDeleteThe validation with simulated data by Claude Williams and colleagues (2012) shows that the homogenization algorithm of NOAA can remove temperature trend biases. The Time of Observation bias is also a cause of a trend bias in the USA. If you do not explicitly correct this bias, the homogenization method of NOAA does it. Clearly it can remove biases.
Claude Williams Jr, Matthew J. Menne, and Peter W. Thorne, 2012: Benchmarking the performance of pairwise homogenization of surface temperatures in the United States. Journal Geophysical Research, 26, no. 3, pp. 345–381, doi: 10.1029/2011JD016761.
UAH6.0 is closer.
ReplyDeleteHard to believe UAHv6.0 has a better fit with USCRN than the fit of USHCN with the USCRN. The fit of the USHCN is very close.
I guess it really has to be version 6.0, right? Given the large differences between the versions. Do you have computation that actually shows this closer fit?
Any questions?
ReplyDeleteYes, how did you compute these adjustments? The methods section of a paper is normally the most important one.
Yes, quite. A sine qua non. I am currently checking my results (I'll report any changes) and fill y'all in on the method.
ReplyDeleteEvan - I'll ask again regarding your statement: ""The adjustments made to sat data are large but fairly simple."
ReplyDeleteMy response was: "LOL. You can't make this up. Has any layman ever actually taken the raw satellite data and duplicated any satellite temperature series? Anyone?"
Well, has any layman ever reproduced a satellite temperature series from raw data? Anyone? Ever?
The satellite series are just one 'blackbox' after another per your interpretation.
As for sats, compare UAH6.0 and USHCN trends with CRN trends from 2005 - 2014.
ReplyDeleteYou will find that HCN exaggerates the cooling trend by ~50%.
As USHCN was discontinued as the official ongoing record in 2014, it seems that NOAA is happy to ride HCN trend exaggeration up -- but not ride it down. Instead, it changes the bus to CRN. You will also find the CRS bias demonstrated: MMTS cools over 20% faster than MMTS -- and this is for all Classes of stations (1 to 5).
UAH is much closer in terms of trend to CRN than is USHCN.
UAH6.0 shows ~13% less cooling from 2005 - 2014 than CRN, which suggests that during a warming trend, UAH will show ~13% less warming than is actually occurring on the ground. So surface should be a somewhat higher trend than sats during our study period.
So when adjusting for CRS bias we do not baseline the CRS record to UAH, but to 13% higher than UAH.
Evan, could you express these percentages in °C? That would make it clear to the reader how small the numbers are that you are interpreting as real. For example, the differences in trend between USHCN and USCRN. I did not dare to interpret them.
ReplyDeleteUSCRN is not "official ongoing record". It is much too short to be climatologically useful yet.
Didn't the mitigation skeptical movement learn anything from their "hiatus" debacle? I know Anthony Watts like to cover up his mistake by calling it "Karlisation", what it actually is is another case of excessive (proclaimed) convinced in minute deviations. If there is anything Karl et al. (2015) showed it was how enormously fragile that deviation WUWT & Co. likes so much was. And how stupid it is not to consider the arguments of experts trying to help you and avoid making a fool of your movement.
Why don't you tell the people you real reasons to be against mitigation. (Not here. This is a science blog.) These reasons convinced you, why wouldn't these reasons convince others as well?
Okay, here is the method for determining equipment jumps:
ReplyDeleteCompare the jump in each station within each of the 9 Climate regions with the average of the jump of the average of all eligible stations within the region.
The interval for determining the jumps is five years in either direction, and applied starting the year after conversion. This is necessary as the jumps are not instantaneous and to calculate the jumps from the month of conversion would reduce the jump. A jump takes ~2 years to manifest itself fully, so applying the full results a year later balances the equation very nicely.
Tmax and Tmin are calculated separately. Mean is problematic: For a station with little trend in either max or min will be
Eligibility is determined as follows: Unperturbed stations are used. A station can be any sort of equipment (CRS, MMTS, or ASOS), however, there must be no equipment change within 5 years prior or within five years subsequent, or station being used for pairwise will be contaminated by equipment change, itself.
As with both Q-91 and H&L-06, USHCN metadata is used. (However, to NOAA's credit, there has been a vast improvement in metadata since 2006.)
This allows us a much larger station base than H&L-06 or Q-91, with at least 10 stations used for each pairwise, and in many cases, more than 30.
The mechanism is as follows:
Jp = 5Yb - 5ya (regional)
Js = 5Yb - 5ya (subject station)
Jf = Js - Jp
Where:
Js = Jump (subject station)
Jp = Jump (regional)
Jf = Final jump (to be applied to each month starting 1 year after conversion).
5Yb = Average of 5 years prior to jump
5Ya = Average of five years subsequent to the jump
Jumps are applied before CRS-bias adjustment or else the jumps would be a lot smaller. I round this against our hypothesis for the nonce.
Also, raw data shows greater jumps than anomalized. So we use raw. Despite its drawbacks, raw data must be used, as we do not want the 10-year comparison interval to be contaminated by events outside the 10-year comparison span. They must be strictly separated.
We use all classes station to determine pairwise owing to the relative dearth of Class 1\2 stations. This skews the results somewhat against the Class 1\2s.
And indeed, the resulting jumps net out at near-zero for Class 3\4\5s, and falls exclusively on the Class 1\2s, raising Tmean trend by ~0.04 (prior to CRS trend adjustment), which is close to the findings in Menne (2010).
My results suggest is indeed possible that there is an overlap effect of microsite effect and jump effect in the Class 3\4\5 stations. However, we are concerned more with the results of the Class 1\2s, because that more closely resembles the "real-life" trend.
Jumps for individual stations of all classes have wide variation, consistent with the findings of both Q-91 and H&L-06. In any case, one cannot reliably apply the same jump to all stations within a region. It must be worked out individually, as above, and even so, there is some risk of conflation with unrelated events concerning individual stations. But that is a vital first step.
In followup, I plan to expand the pairwise sample to include stations with "partial" unperturbed records. This would not only make the pairwise yet more robust, but also our basic trend findings, as well, in regions with lower coverage. But as that involves varying baselines for varying start-points, it will be dealt with in followup.
If anyone has any questions as to the above methods, please feel free to ask (shout, tear out hair, whatever), and I will be happy to answer.
Yes, of course.
ReplyDeleteTrend CRN 2005 - 2014 (C/decade): Tmean -0.270, Tmax -0.221, Tmin:-0.318
Trend USHCN (TOBS data):
TMEAN Class 1\2: -0.360, 3\4\5: -0.408
TMAX Class 1\2: -0.448, Class 3\4\5: -0.492
TMIN Class 1\2: -0.248, Class 3\4\5: -0.336
TMEAN CRS-only, all classes: -0.551
Trend UAH6.0 CONUS: Tmean -2.40
That CRS is a headbanger. Both on the way up from 1979 and on the way down from 2005.
In order to improve accuracy on the individual station reliability level, one might well correct station change by running the above process through multiple iterations, a sort of crude homogenization process. Won't shake the regionals much, and it would rope the mavericks. Sans systematic error, it is to be hoped. #B^)
ReplyDeleteMaybe I'll do it for followup. (Via the Excel hand crank.)
Evan Jones writes: "Trend UAH6.0 CONUS: Tmean -2.40"
ReplyDeleteAnd vs RSS V4?
Or STAR?
Sorry, slipped a decial point. -0.240 UAH6 (compared with -0.270 USHCN).
ReplyDeleteThe mechanism is as follows:
ReplyDeleteJp = 5Yb - 5ya (regional)
Js = 5Yb - 5ya (subject station)
Aack! Typo. make that:
Jp = 5Ya - 5yb (regional)
Js = 5Ya - 5yb (subject station)
Finally finished my end of the draft. Those equipment adjustments take time.
ReplyDeleteEventually, I'll do up a set of individual station graphs with MMTS/ASOS/CRS-adjusted data to show the effects -- yet more followup.
This paper is a first step. "Further work is necessary."
One thing's fer sher: I would never have been prodded into tackling equipment head-on were it not for the discussion that occurred above. Working out the methods and then applying them was both exiting (and a little hair-raising).
So I want to personally thank you all for your kind assistance and stimulating criticisms, and especially my dear VeeV who has provided this forum. Homogenization is a potentially very powerful tool, but, as you must know, it is tricky, particularly prone to systematic artifacts.