Pick a Spot

I’ve been discussing the Berkeley Earth temperature record (BEST) for the last few days.  My comments have been quite critical with me going so far as to say the BEST data has no accuracy at regional scales.  I’ve now set up an easy way to test this view.  All you have to do is pick a spot:

7-13-Coordinate-Selection

Pick any spot on that map that isn’t blue, and I will show you how BEST’s temperature estimates for that area compare to NASA GISS’s.  As an example, here’s what you’d get if you picked my house (a five year smooth is applied to each graph):

7-13-GISS-39-89

There are four BEST graphs to one GISS graph because GISS uses a 2º x 2º grid while BEST uses a 1º x 1º grid.  As you can see, there’s little point in BEST using that 1º x 1º grid as all four of its grids are nearly identical.  As you can also see, all four of those graphs are dramatically different than GISS’s.  They all have a strong warming trend not present in the GISS data.

The same is true for many other areas.  A particularly troubling one is in Atlanta, Georgia:

7-13-GISS-34-84

GISS shows that area has cooled slightly.  BEST says that area has warmed noticeably.  There is no legitimate reason for that.  One of these data sets must be wrong.

So pick a spot.  I’ll post the temperature for it, and you can decide for yourself which results you think are more believable.  Bonus points to anyone who can pick a spot BEST says is cooling.

Advertisements

32 comments

  1. Brandon,

    Am I doing something wrong? Any spot I click, blue or otherwise, links to the full size image.

    DGH

  2. Madison, AL. More specifically, the Huntsville International Airport. I live here, plus both Dr. Roy Spencer and Anthony Watts have blogged about temperature record issues here in the past.

  3. DGH, you have a bit too much faith in me. That map isn’t interactive. it’s just there to help people see the coordinate system used in the NetCDF file/what areas they can pick. They still have to say what spot they pick so I can plug it into the code and run it.

    dellwillson, interestingly, you picked a spot where BEST and GISS match nearly perfectly. You can check for yourself here. I guess BEST doesn’t need to adjust data if there’s already a notable warming trend in it.

  4. Carrick, I had to cut out the smoothing for the GISS data for your request because it was missing so many points the smoothing broke. I left it in for the BEST data though. I figure you can account for the differences caused by the difference in smoothing when looking at the graphs. One note though. I started the graph at the point both data sets had data, but I didn’t limit the endpoint the same way. That means the GISS graph has white space for part of it.

    It’s available here.

  5. So as I see it, Y=45 is 0 deg Latitude. So take Y = (Latitude+90)/2 or Y = 45+(LatNorth)/2
    X = 0 at -180 Long, 90 at 0 so X = (Long+180)/2 or X = 90 – LongWest/2
    so these are 2×2 deg cells.
    Should you pick the center of the cell?

    I am interested in a traverse east to west from
    Denver across the Continental Divide to Eagle, CO (107W) (or Grand Junction 108.5W to go the extra mile)
    from 39 or 40 N, 104 to 105 W
    to 39 or 40 N, to 107 to 109 W
    So I guess that would be y = 65 (or 64.5), x = 35.5 to 38

    I would be very interest to see what kind of smearing you see in a regional trend between the Colorado plains, the Front Range, The Continental Divide, and the Western Slope down to the agricultural basin in Grand Junction. An I-70 transect across the grid.

    (Trivia Point: There are two tunnel bores at the “Eisenhower Tunnel”, One named for President Eisenhower completed in 1973, The other is call Johnson opened in 1979. Who is Johnson? He was Governor of Colorado when the inter-state highway system was first laid out with I-70 ending west bound in Denver. It was Johnson to persuaded the system to extend I-70 westward on the US-40/US-6 routes. I cannot imagine Colorado with out I-70)

  6. Thanks Brandon, GISTEMP is pretty noisy, but it doesn’t look like the recent trend agrees with BEST. And as usual BEST result for the four sectors look like near carbon-copies of each other.

    I think they may be going a bit too far in the way their assumptions about the amount of correlation between nearby cells is driving the reconstructed temperature field.

  7. Carrick,
    RE: 1×1 deg cells look like carbon copies.

    How much change between neighboring cells should we expect to see, anyway? What is the grid supposed to represent? What is it used for?

    At first blush, there looks to be way too much smoothing, which means UHI is likely spread over the continent. Indications are that grid cells are influenced by thermometers over 1000 km away. (Support Note 1)

    In once sense, I can see a very gradual change across the globe, a noticeable gradiant in the north-south direction, very little east to west. Very gradual field that explains the zones of latitude and little else.

    But then WHY should we see such high resolution in the time dimentions across broad swaths of the grid?

    Conversely, TERRAIN matters! We have to remove a terrain signature to tease out a climate signature. If that is the case, the I-70 transect across Colorado should show quite a bit of differences from one grid cell to the next.

    What is the grid used for? Does it have the rugosity we would expect for the purpose?

    Note 1: Check out the Denver page.
    http://berkeleyearth.lbl.gov/locations/39.38N-104.05W
    It list 39 active thermometers within 100 km.
    For most of the period from 1920 to 2013 there are between 2 and 4 thermometers within 50 km, (with high frequency changes, a red flag!!!)
    But look on the right side:
    “Long Term Weather Stations:”
    GLASGOW MUNICIPAL ARPT
    VILLA AHUMADA
    OZONA 1 SSW
    FAIRFIELD RANGER STN
    BLOOMFIELD 1 WNW
    ELDON

    GLASGOW MUNICIPAL ARPT – NE Montana on the Canadian Border. 8.5 deg North. 4000 ft lower in elevation.
    It has been subjected to 9 moves and 9 other breaks.

    VILLA AHUMADA – In Mexico, near El Paso, TX, 10 degrees South of Denver.
    It’s record is a MESS. Eight recording gaps. It is junk!

    OZONA 1 SSW – SW Texas, 10 deg South, 3 deg East of Denver,
    4 station moves and 5 breakpoints from 1950 – 2014, not exactly long.

    FAIRFIELD RANGER STN – SE Idaho, 3.5 deg N and 10 deg W of Denver. (about equal elevation)
    from 1948-2012 (not long), 5 station moves, but only the 2003 moves looks significant.
    peculiar chart of deviation from regional mean, skewed lower departures.
    http://berkeleyearth.lbl.gov/stations/39124

    BLOOMFIELD 1 WNW – Broomfield, Colorado, just near by…. NOPE. Broomfield IOWA!, 1 deg North, 12 deg East of Dever
    4 station moves, 7 other breaks, 1913 to 2013, but 10 year gap in the 1930s

    ELDON – Eastern Missouri, 1 deg S, 12 deg E of Denver. (1 mile lower in elevation)
    1897-2012, but 10 year gap in 1910’s, 1 move, 7 other breaks. Big Offsets, and many QC failures.

    These are the stations the BEST page highlights as the “Long Term Record Stations” to compare against Denver. Judging by their lat-long differences they are on average over 1000 km away from Denver.

    I occurs to me that the reason they include stations from so far away is to drive up the station count into the thousands to drive down the statistical uncertainty of the mean standard error — even though it is now a regional mean that is meaningless.

  8. Stephen Rasey, I actually used a couple of the formulas you describe. Because of how grid points are centered, that made me off by one row and column. You’d need to add +1 to get the right coordinates.

    Carrick, agreed. One thing I’ve noted is it seems every BEST grid cell shows warming in the recent period. No matter how different they are in the past, they all seem to agree about that. Assuming I’m not mistaken, I think that may be due in part to the way BEST regresses out variables prior to kriging.

    They regress out altitude, latitude and seasonality (I believe with splines). They do so over one set period. As I understand it, that means the temperature data for that period will have its variance decreased and trends forced further into alignment.

    Not only would that help explain heightened similiarities in the recent portion of the grid cells, it would also help explain the step change in uncertainty levels I first pointed out two years ago. But I’m still running that idea down.

  9. 86 degrees west, 37 degrees north. Middle of Kentucky.

    I may be cheating. I seem to recall Mosher has previously disclosed that the broader region of what I call the Ohio Valley was warming in the 19th, but generally cooling in the 20th century, according to BEST. When I engaged to suggest the trend arose from de-forestation suggested by re-forestation, he asked (challenged) me how I would go about testing for the hypothesis using the data.

    I dunno. I thought maybe agricultural and industrial records might have something to contribute, but the field is an overall challenge and I certainly don’t feel qualified to tackle it all.

    ANYhow, as you pick the spot, see if BEST shows which way the the Kentucky “point” moves.

  10. Pouncer, I don’t know why Steven Mosher would have told you that. It doesn’t seem to match what the data shows. See for yourself.

    So far, I haven’t seen a single gridcell where BEST showed cooling for the 20th century. A number may show cooling if you stop ~1980, but by 2000, it seems they’re all warming.

  11. @Brandon, 5:11pm
    seems every BEST grid cell shows warming in the recent period.

    Let’s return to the “zombie” station problem.

    Suppose as a function of time, we make a plot of the number of LIVE stations, categorized by (urban,rural), and (Class 1,2,(3-5)) In the “recent past” we have lost many stations. Which stations have we lost? Most likely the rural stations. (Am I wrong about this conjecture?)

    So is it reasonable to suspect that in the population of LIVE stations the percentage of the (rural and Class 1 or 2) have dropped dramatically? If we are seeing the live station population gradually become more urban and/or worse in class, then the infill of the zombie stations would also suffer from a warming bias caused by the warming bias of the live stations.

    So zombie stations are becoming more plentiful and the infill of the zombie stations is coming from a more urban mix of live stations. It should be an easy hypothesis to test.

  12. Okay, so WITH cheating (Google) instead of by memory, there’s this by Mosher on BEST regarding cooling regions.

    http://stevemosher.wordpress.com/2012/03/17/cooling-stations/

    He is of course defining “cool” by comparison to other regions identified in BEST, not by comparison to the same region in GISS. Still, I think this is the timeframe at which I got started. (And got stalled.)

    I like Mosher’s take on the so called temperature instrumental record, establishing it
    as an “index” like the Dow Jones Index of 30 Industrials. Obviously as Sears falters and Amazon surges the indexed company for “retailing” must be re-assigned. Comparably an index of some fewer, well chosen, stations representing a region, an altitude, a prevailing wind system, a ground cover type … etc would be useful, and as those factors change over a century a new station better matching the intended climate sector could be substituted.

    BEST is not there yet, even by Mosher’s own standards and goals.

  13. Stephen Rasey, thanks for the comments. I don’t think smearing from UHI is going to turn out to be that critical (because the effect can have either sign).

    But smearing from regions that have higher trends (e.g., Canada) into regions that don’t is a real issue when you make the radius so large.

    It’s easy to see this by comparing GISTEMP 250km to GISTEMP 1250 km for example.

    I’m not sure it’s necessarily rural stations that lose data. Sometimes it’s just “B-stations” that have equipment issues and are underfunded. I’ve heard of cases like rodents chewing through cables (Lord knows I’ve seen plenty of examples of that in my own field work).

  14. @Carrick
    I don’t think smearing from UHI is going to turn out to be that critical (because the effect can have either sign).

    I could not disagree more. I reject it entirely.

    UHI is no myth. Our urban areas have been growing, using more and more energy which must be release as waste heat sooner or later. The number of live temperature sensors have been becoming more urban over time as rural stations and closed and urban areas expand.

    My criticisms of BEST are many and interrelated, primarily on low-frequency signal loss, measurement and processing uncertainty that has gone missing, ridiculous and unjustified huge radius of influence for gridding…. these are supported by theory.

    But by gut feel, admittedly, BEST’s estimation of UHI influence is preposterously low. It is not believable! They are conflating UHI with climate signal. Because of confirmation bias, they don’t care.

    Carrick, you don’t have to believe me. If you think UHI doesn’t matter because “they can have either sign” [and how is that possible?] then it shouldn’t matter to you if we treat the data AS IF UHI did matter. Either UHI matters greatly and we should treat the data as if it does, or UHI is a non factor and if we treat the data as if it did matter, it won’t change a thing.

    But we know from Japanese Cheery Tree Blossom records that UHI is a big factor in urban areas, at least in the spring. Trees in the center of cities bloom a week earlier than in the rural areas and the difference has grown over the century.

    To think that UHI balances out because it could have either sign….. well I’m flabbergasted. I don’t understand that thinking at all.

  15. Stephen Rasey, where I ever say UHI is a myth?

    I have looked at it and concluded and come to the conclusion that ” smearing from UHI is going to turn out to be that critical (because the effect can have either sign).”

    What I would say is UHI is a bit of a misleading term. It incorrectly implies that the only effect of human activity on local environment is to warm it near cities.

    It’s well documented that human activity in suburban areas and in irrigated fields (an extension of the UHI concept) actually lead to a net cooling. Using the rural station to infill a more urban site can lead to negative bias to temperature relative to the actual trend, because of the bias present in the rural station.

    (So I like to extend the term to refer to any regional scale anthropogenic change in temperature, be it from irrigation, deforestation, adding reservoirs, etc.)

    Regarding your cherry tree example—yes it’s true that urban environments have a higher temperature than the surrounding country side.

    But it’s not true that this results in a long-term difference in measured trend. It’s more of a step-function change (compared to the length of the long-duration records) as the area begins to urbanize. That makes it relatively easier to detect as long as you have rural stations to compare against.

    I have looked at UHI pretty carefully in the past, and it’s my personal opinion that it might amount to a few percent at most.

    This comment is based on noting that it certainly does not explain the zonal (latitudinal) variation in temperature trend:

    figure.

    The dominant variance comes from polar amplification, an understood natural phenomenon.

    Where UHI to be a dominant effect, you’d expect a hump around 45-60° N. There’s no evidence for even a kink in the trend, so my take home is the net effect must be small compared to the measurement uncertainty, at least if you make some effort to correct for it using your full network of stations.

  16. I should extend my comment, to prevent misunderstanding to say:

    “But it’s not true that this results in a long-term difference in measured trend after the initial urbanization period

    You have a certain trend before the urbanization, you have a period of rapidly increasing temperature, so large trend. Then you get nearly the same trend after urbanization, which was the point I was trying to make.

    Of course if you have a step increase in your measured temperature, that will affect your trend, but the effect gets progressively smaller as you increase the time window of your fit.

    Secondly this:

    “it might amount to a few percent at most relative to the measured long-term (e.g. 60 year) trend“.

  17. There was an exchange over at Judith Curry’s blog I want to address. Carrick had said:

    I think they should detrend before computing the correlation, otherwise the fact that the series has trends dominates the computation of the correlation, and this results in biased-high estimates of the trends. They are also assuming axial symmetry which I believe is a mistake.

    To which mwgrant responded:

    BEST does detrend in some manner with respect to latitude and elevation though I must admit find available discussion extraordinarily obtuse. See

    http://www.scitechnol.com/2327-4581/2327-4581-1-103.pdf [pp.2-4]
    http://www.scitechnol.com/2327-4581/2327-4581-1-103a.pdf ) p.2, p.4, p.8]

    I posted a comment there addressing this issue, but it was stuck in moderation the last time I checked. The issue is worth bringing up on its own anyway. As I said there:

    Saying they detrend the data is misleading. They did not detrend it. They estimated climatological parameters for latitude, altitude and season. They then removed those. That’s not detrending in the sense most people would interpret it as it has no time component.

    It’s really just a way of anomalizing the data. It can remove absolute differences in temperatures, but it cannot, by definition, remove differences in trends of temperatures.

    To expand upon this point, the BEST NetCDF file I used has a “climatology” variable in it. There are twelve different sets of values in it, one for each month. Those are what BEST subtracted from each grid cell prior to its kriging step. I’ve uploaded a simple map showing the data for January so you can get an idea (available here).

    Each of the 12 maps available via that variable are subtracted from the data for their entire record. That means data from 1780 has the same values subtracted as data from 1980. It won’t do anything to remove the variance in trends caused by areas warming at different rates. That means it will not mitigate the problems of spatial smearing I’ve raised in my posts.

  18. Thanks Brandon. I was never able to find a discussion of detrending and was just guessing that I wasn’t reading it carefully enough.

    So this wouldn’t fix the problem with the low-freqency portion (“the trend”) dominating the estimate of the correlation coefficient, when what you really want is just the the high-frequency portion unadorned by the trend from another region.

  19. Indeed.

    Even worse, they perform their regression (or whatever they call it) over a single period. As we know, regressing over a single period artificially deflates the variance in that period relative to the rest of the series unless steps are taken to account for that. The BEST methodology does not. This means the variance in the BEST record is fundamentally biased. This bias in variance translates into bias in their empirical breakpoint calculations. It may also translate into bias in the iterative weighting process they use for combining stations into a temperature field.

    I can’t estimate the magnitude of that effect without doing a lot of work, but I think it is remarkable BEST would allow it to exist.

  20. Steven Mosher commented in the thread at Judith Curry’s where this topic has been discussed. One of his comments is remarkable for saying (in part):

    not likely mw.
    Way and I were working last night on some satellite data.
    He mentioned somebody suggesting that GISS was suppressing the trends with their hokey method. I laughed recalling the work of JeffId and RomanM. I scanned the list of names for a sane one, and yours popped up.

    Long term goal will be to add something similar to Tomas’ work and push the res down to 1km

    I’m not sure which part is most screwy. It’s very confusing he believes people claimed GISS suppressed trends with a hokey method when that’s pretty much the exact opposite of what people were saying, but… one kilometer spatial resolution!?

  21. “Steven Mosher commented in the thread at Judith Curry’s where this topic has been discussed. One of his comments is remarkable for saying (in part):…”

    As I indicated at CE I responded to a Mosher ‘duh’ and that short comment apparently showed and then went away. FYI my quip was to the effect that is since he had shown up that ‘duh’ would likely become a ‘Doh!’. No big deal, except it stresses Monfort. No missing technical exchange.

  22. Brandon Shollenberger
    July 14, 2014 at 2:40 pm wrote (responding to mwgrant)

    “…
    To which mwgrant responded:

    BEST does detrend in some manner with respect to latitude and elevation though I must admit find available discussion extraordinarily obtuse. See

    http://www.scitechnol.com/2327-4581/2327-4581-1-103.pdf [pp.2-4]
    http://www.scitechnol.com/2327-4581/2327-4581-1-103a.pdf ) p.2, p.4, p.8]

    …”

    Funny that above you fail to indicate that in that same comment noted another comment of mine on the matter:

    Just to be clear (and I think you understood) I have no knowledge whether “they are properly detrending the data before computing correlations.” :o) I try to keep my suffering to a minimum these days.

    That is unfortunate as the omitted part is a relevant caveat that I indicated and its omission changes the tenor of of my response.

  23. mwgrant, it may change the tenor, but I wasn’t interested in the tenor of your comment. I was just interested in pointing out the difference between detrending in spatial dimensions and detrending in temporal dimensions. BEST only does the former, but many people would take your comment as indicating the latter (as Carrick did).

    I try not to make tone be an issue unless it gets in the way of a point.

  24. @Carrick at 2:05 pm

    It is a fair point that “UHI” includes the anthropomorphic cooling influences of agriculture and inrrigation.

    “It’s well documented that human activity in suburban areas and in irrigated fields (an extension of the UHI concept) actually lead to a net cooling.

    But is it really balanced to a near negligible effect? “Park Cooling Islands” have restricted influence. Irrigation is highly seasonal and should be seen in differences in trends by month and season.

    Where are the live theremometers? Most are in places where the neighborhood is, 1) increasing in population, 2) increasing in structures, 3) increasing in pavement, 4) increasing in traffic, and 5) increasing in energy use over a broad length of time.

    True there are thermometers in places of declining population. These are the stations more likely to be shut down and become zombie stations in the database (now 45% of the database). The growth in the zombies is a bias toward the Urban readings.

    bias present in the rural station.
    Wow. So the rural stations are biased — negatively? I don’t believe you can separate out human influences from natural influences.

    You seem to put great store in the concept that as an area urbanizes, the temperatures rise quickly, the return to the overall climate trend. Why? Where is the justification and the data? When you look at the BEST station data for Tokyo, they had 4.5 deg C/century. After local regional changes they reduce it down to over 2 deg/Century, But that is still higher than their Global by 1 deg/Century. An urban heating of 3.5 deg/Century is likely. And Tokyo is moderated by the ocean. Even if the population density doesn’t change much over the past 50 years, its energy use, it’s traffic most certainly have. UHI effects are gradual and longterm because our civilization’s development and land use changes have been long term. It is certainly possible there are abrupt microsite changes that might be detected and confused for UHI. But UHI changes are big and long term.

    Where UHI to be a dominant effect, you’d expect a hump around 45-60° N.
    Why? Look at these maps of population by latitude.
    http://www.themarysue.com/world-population-latitude-longitude/

    BTW, do we have any charts of thermometers by latitude?
    And do we have Latitude of live stations by decade? So as to see a bias in the measurements by latitude over time.

  25. Stephen Rasey, not a lot of time today, so this has to be brief:

    But is it really balanced to a near negligible effect?

    That’s an operational question… I would put the answer as “it’s not negligible, but corrected-for UHI is not an important source of systematic error for global temperature trend estimates”.

    However, I think it is could still be big player (as all systematics are) when you look at regional scale variability.

    Wow. So the rural stations are biased — negatively?I don’t believe you can separate out human influences from natural influences.

    Natural influences shouldn’t correlate with the spatial and temporal patterns associated with human activity. So I think we can.

    When you look at the BEST station data for Tokyo, they had 4.5 deg C/century. After local regional changes they reduce it down to over 2 deg/Century, But that is still higher than their Global by 1 deg/Century.

    But that’s Tokyo, one of the largest metropolitan areas in the world. That’s as atypical of an example as you could possibly pick, so I don’t know what you think it proves.

    But UHI changes are big and long term.

    On a local scale yes. I don’t have time to dig them out for you, but there are papers where they include UHI corrections and compare that to leaving the UHI corrections out. If my memory is correct, over the longer time intervals that I’m discussing, you get maybe a 10% effect on trend.

    If you only fix 2/3s of that with the UHI correct, that still leaves you with about a 3% systematic error in global trend.

    Where UHI to be a dominant effect, you’d expect a hump around 45-60° N.
    Why? Look at these maps of population by latitude.

    Because prior to circa 2000, that’s where planet-scale industrialization was centered. As I’ve shown there is no measurable effect.

    In the last several decades, China and India have been rapidly growing. If UHI is important there, it certainly doesn’t show up in the global temperature trend.

    So that’s the problem:

    There are distinctive spatial and temporal patterns associated with the warming observed from e.g. 1880 – 2010. These spatial and temporal patterns are not well explained by UHI. So UHI must be relegated to a relatively minor role in the discussion of systematic uncertainty in global mean temperature.

  26. @Carrick 7/16 at 10:47 am
    I am unconvinced.

    there are papers where they include UHI corrections and compare that to leaving the UHI corrections out.
    Sounds like circular reasoning to me.

    But that’s Tokyo, one of the largest metropolitan areas in the world. That’s as atypical of an example as you could possibly pick, so I don’t know what you think it proves.
    For the record, Zeke picked it — for another purpose.
    http://wattsupwiththat.com/2014/06/28/problems-with-the-scalpel-method/#comment-1672999

  27. They are both wrong.
    One minimizes the error of prediction.
    Take the 4 or 5 thousand of stations that are not in
    Giss or best.
    Compare them to the predicted fields.
    Publish your results

  28. Steven Mosher, if you have something to contribute, contribute. If all you have are vague comments and hand-waving, just stay away and don’t waste our time.

    Also, don’t tell people what to do. You have no room to give anyone orders.

  29. Alright, I’m done cleaning for the day so I can take the time to make a more informative response to Steven Mosher. He starts by saying:

    They are both wrong.

    This is a tautology in that all models are wrong. Pointing it out contributes nothing to people who understand the subject.

    One minimizes the error of prediction.

    This statement is pure assertion. There is no factual basis for it. There is no logical underpinning to it. It is perfectly possible neither BEST nor GISS “minimizes the error of prediction.”

    Even worse, that “possibility” is a actually a logical necessity. Neither GISS nor BEST comes close to having perfect predictions, thus neither has the minimal amount of error possible. Neither could possibly minimize that error. The most you could say is one minimizes the error more than the other does.

    Take the 4 or 5 thousand of stations that are not in
    Giss or best.
    Compare them to the predicted fields.

    From what I’ve seen Mosher say elsewhere, his idea would be to take the extra stations, run the GISS and BEST methodology on them and see how the results compare to the results we get now. That would be stupid. A test like that could not possibly address the issue of this post – spatial resolution. That test can only measure accuracy. It cannot measure precision. Given this post has been primarily about precision, a test which cannot measure precision is stupid.

    That said, Mosher’s phrasing says to compare individual temperature stations to temperature fields. I think that’d be a useful test. We don’t need external data though. We can compare the temperature stations used by BEST and GISS to their temperature fields. I did that for BEST at the very beginning of this series of posts. It showed BEST was terrible.

    If we take what Mosher said literally, he’s proposing a test I’ve already performed. If we interpret it differently, he’s proposing a test that would be completely meaningless. Either way, his “contribution” is negligble.

    Publish your results

    Not long ago, Mosher told me if I wanted my concerns addressed, I should publish them in a peer-reviewed journal. He then hinted it was likely BEST represenatives would be asked to review the paper. Call me crazy, but I don’t think requiring people publish in a peer-reviewed journal, where you can actively attempt to block their publication is reasonable.

    Now that I’ve written all that, we can look at what Mosher said as a whole. Notice he wrote six brief sentences, and none of them contributed anything of value. Explaining what was wrong with them took far more effort because he didn’t explaing anything or provide any basis for what he said. That shows Mosher is forcing people to spend far more time and effort addressing his nonsense than he is putting into writing it.

    In other words, Steven Mosher is a troll, and he’s BEST’s mouthpiece.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s