Why Temperatures Matter vs Anomalies

Plot of Belize Temperatures (flat) and thermometer count

Belize Temperature Average and Thermometer Count

(You can click on the graphs to get a larger version to see the details)

Belize – on the Caribbean Coast of Central America

Belize is an interesting little country. About the size of Vermont, but without all the mountains. It is on the Caribbean coast of Central America. The Caribbean being a place of extraordinary stability in some ways. The shallow water there warms under fairly consistent sun, and the shallowness prevents significant waves. Most of the Caribbean has wonderful beaches with very little in the way of waves and tides. (Unless a hurricane is passing through ;-)

I’ve been on a beach in Jamaica when the air and water were both about 87 F all the time. Just Heaven.

So look at that chart. Darned stable temperature. At the far left edge are a couple of years of data from the 1800s. Then, for unknown reasons, there is a long gap to the 1930’s and from there the series is full for each year. About 26.x C all the time.

About the only thing you can really say about that chart is that the “lower peaks” that make excursions down to the 25 C line get ‘clipped’ in The Great Dying of Thermometers. But I doubt if any of us would look at that chart and claim Belize was burning up with a rising temperature curve.

Costa Rica

Costa Rica is a bit different from Belize, but not too much for comparisons. A bit further south. Coasts on both sides of the Central American Peninsula with both Pacific and Caribbean beaches. Stable temperatures and another little bit of Heaven on Earth for vacationers. Yes, a bit larger at about the size of Vermont, New Hampshire, and Connecticut combined. Yes, some bigger mountains down the middle. But still not much very far from water and not a lot of change.

Costa Rica Average Temperature and Thermometer Count

Costa Rica Average Temperature and Thermometer Count

Here “time begins” in 1941 with one relatively cool thermometer (one presumes in the mountains somewhere, but worth checking). In the late ’50s and early ’60s some more thermometers are added (probably near the beach, given the 25 C average temperatures). From that point onward, we see a darned stable 25 C average temperature. There is a bit of a step up in 1981 as a thermometer is deleted, but it steps back down again in 2001 with another deletion.

Again I think most folks would look at that and say “No warming here”. During each segment of stable thermometer counts, the ‘trend’ is basically flat.

Putting Belize and Costa Rica Together

Now let’s put these two on one graph together:

Belize vs Costa Rica Temperatures

Belize vs Costa Rica Temperatures

OK, I’ve added some trend lines.

Now first off, here is why I look AT THE DATA and look at tables more than graphs with trend lines. That trend line in Costa Rica looks like a real warming trend, but it hides the fact that the thermometer count jumped up right when the temps made a step function higher. It is the stable part to the right of that thermometer count change that matters. And you can see that both Belize and Costa Rica have more in common than not and that ‘temperatures change’ more in line with instrument changes than anything else.

OK, at that point the “anomaly advocates” like to toss rocks about how this is a silly approach and if you don’t use anomalies why of course you find that instrument change matters. (Often with various epithets included, since they seem to lack manners.)

Well, I look at those charts and see things that matter. I see stable segments when the instruments are stable (not what one would expect from a warming globe being steadily and relentlessly warmed via CO2). I see a ‘clipping of the down peaks’ in Belize coincident with a BIG thermometer count change. And I see a Caribbean that runs right around 25 C all the time from the 1800’s to now.

I think these things matter.

Lets look at it as Anomalies

Now lets look at the same place using the anomalies for these two countries. Realize this is based on ALL the data, unadorned. There are no interpolations, adjustments, in-fills, UHI corrections, whatever. Just the temperature data for each thermometer compared only to itself. This is done “by month” (that will not matter nearly so much for places bordering on the Caribbean as it IS so stable year round) and missing data gaps are bridged by waiting until a new value shows up in a future year so dropouts and gaps don’t have much impact.

The functions graphed here are:

dT/yr the average of the 12 monthly changes of temperature for the thermometers in the country when compared to the last valid value in that month for that thermometer. A very pure form of anomaly.

dT the running total of those anomalies. What is supposed to be the change of temperature to date.

Count is the number of thermometers in each country in a given year

And the trend lines for those dT values.

Belize vs Costa Rica Anomalies

Belize vs Costa Rica Anomalies

Here we see a strong trend line of rising anomalies for Belize while Costa Rica is shown as “warming” but much more slowly. Quite different from the actual temperatures in those countries. Anomaly processing has suppressed the “step function” in the early part of the Coast Rica series (what it’s supposed to do). And now Belize has a strong warming trend line.

But notice that the Belize values for dT tend to make “step functions” when instrument change happens. You can see the same thing for the Costa Rica graph as well, though after that first thermometer change, the thermometers don’t change as much as the ones in Belize.

So, IMHO, using “anomalies” is a process that has some sensitivity to instrument change. I suspect it is related to the ‘edge effects’ when a series of changes is left ‘unterminated’ (that is, you take a ‘rise’ in a year, then drop the thermometer and next year you do not get the offsetting drop, so perpetually preserve that ‘rise’). They clearly do NOT present the same information you get from direct inspection of the actual temperature histories for a location.

Think about it.

OK, for the inevitable folks about to launch into a tirade about how anomalies are great and using temperatures is dumb, and I “don’t get it” et. al.: Please don’t bother. I DO “get it” and I DO know why you would use anomalies. It lets you compare widely divergent instruments with the offsets removed and it lets you average a bunch of highly divergent things with less individual instrument bias whacking your average. (It suppresses that first ‘step up’ in Costa Rica when the lone cool thermometer gets a few friends). What I’m looking at is how anomalies CAN go bad and CAN mislead. Please realize that BOTH tools have value. Real temperatures keep you anchored in an absolute reference frame. That matters. Anomalies are subject to runaway accumulation of errors as series are left unterminated and as instruments change.

Now consider that ALL the global temperature series are based on much more complicated and much more “mucked with” anomalies based on temperatures that have had all sorts of adjustments, in-fill, “correction” etc. done to them. Do you really trust that to be valid to the 1/100 C place? Even 1/10 C?

For comparison, here is the GISS view of the world for 2009 ( near the last year of the graphs above):

The GISS view of 2009 via their Anomaly Map for 2009

The GISS view of 2009 via their Anomaly Map for 2009

Belize and Costa Rica are under that +1 C and near the +2 C area of Central America.

Now when I look at the temperature graphs up top, I just don’t see it. I see peaks that are consistently the same, and bottoms that have been clipped. (And with a bit of volatility compression as a result).

And THAT is why I think it is important to look AT the temperature DATA. Sure, use anomalies. But don’t ever forget to anchor yourself back in the real temperatures.

Advertisements

About E.M.Smith

A technical managerial sort interested in things from Stonehenge to computer science. My present "hot buttons' are the mythology of Climate Change and ancient metrology; but things change...
This entry was posted in AGW Science and Background, dT/dt, Favorites and tagged , . Bookmark the permalink.

15 Responses to Why Temperatures Matter vs Anomalies

  1. j ferguson says:

    E.M.
    I wonder when you are going to come upon the country temperature records with the rapid temperature climbs which will provide the lift that your flat records lack.

    All this warming must be happening somewhere?

  2. rob r says:

    E.M.

    I haven’t got time to go find the appropriate thread but have a comment with regard to the Campbell Island dropout issue and the NZ record.

    The climate station on Campbell Island was active at least as recently as 2009. They changed the instumentation (now automated) back in about 1992. When this happened NIWA changed the reference number for the site. NIWA report both the old data (Campbell Island) and the new data (Campbell Island Aws) via the cliflo database.

    So part of the latitudinal bias in GHCN/GISTemp could be fixed relatively easily. It is worth noting that the instrumentation change did not affect the rather flat post-1950 temp trend for Campbell Island.

  3. e.m.smith says:

    @j furguson: Well, that is one of the most interesting questions… I’ve found places with dead flat trends (often ending in The Great Dying of Thermometers … along with the sporadic few showing cooling trends). I’ve found a LOT of places with anomaly “Hockey Sticks” like Belize above. Some of them have actual (very small) warming trends, some have no warming. Most of them look to have the “negative peak clip” effect (though I’ve not graphed all of them yet, so a bit more work to do).

    And yes, I’ve looked at 100% of the countries of the world now. (Most of them via looking at tables of numbers, some as finished graphs). There may be some room for revision, but not much, due to a table of numbers being a bit less clear of trend when looked at as compared to a graph with a fitted line. With that caveat: The one thing I’ve NOT seen at all is the smooth upward sloping trend line over the length of history that would be indicative of a CO2 driven warming.

    FWIW, Canada looks to be one of THE most extremely “warming” places in the anomaly and absolute temperature averages. I’m still trying to work out exactly why. Some of it shows an up to 8 C average “warming” (that I think we’d have noticed if it was real…) My working thesis at this point is that high latitude places have more volatility so as you move thermometers to ever less volatile places the “haircut” is more extreme than at places like dead flat Belize.

    One other thing that’s pretty clear. You get the “warming signal” (that in the anomaly graphs is a hockey stick) when you have instrument change. You do not get it when there is no change of Mod Flags.

    @Rob R:

    Thanks. I’ll go looking in the data and see if I can dig out a current Campbell Island. It might be enlightening ;-)

  4. Rob R says:

    Hi again,

    Email me here and I will send the Campbell Island data to you as an excel spreadsheet. That will give you an idea what to look for if you want to navigate into the NIWA cliflo database yourself.

    I am currently sorting through the database to get a complete mean air temp dataset for the South Island. From this search the max number of climate stations reporting monthly temperature in any single year in the South Island now stands at over 90 in the late 1970’s. Looks like it might get up to 100. The max is likely to be greater around this time in the North Island.

    The South Island hit the 20 station mark in 1921.

    The South Island had at least 43 climate stations actively reporting monthly air temperature during 2009. A good number also report soil temperature.

    I have started some station to station first difference calculations for several long-lived stations that are not in the GHCN database. This is to check for unannounced site changes. On balance it appears that a significant warming in air temperature occurred in the South Island during the early1950’s, but there has been little change since then.

    Shallow soil temperatures follow the monthly and annual mean air temp closely, also near trendless since about 1954. Deeper soil temperatures tend to show a slightly slower response to the early 1950’s air temp increase. Since about 1960 the 1 m deep soil temps have pretty much flatlined as well.

    9 am air temps have flatlined since 1954.

    Some sites appear to be slightly less frosty (grass minimum temp) in recent times, but that might not be climatic (undecided and haven’t followed it up in detail). Max temps have also flatlined since 1954. Hotest day of the year has flatlined since 1954.

    A number of the oldest records from NZ are from urban areas. I have been looking at these one by one to see if much useful information can be extracted. My initial impression is that the deep past (pre 1900) does not seem to have been as cold as some others would like it to have been.

    REPLY:[ I’ll see about putting up a N.Z. thread Real Soon Now ;-) but until then, you can email me at pub4all ATSIGN aol.com -E.M.Smith ]

  5. Malaga View says:

    Sure, use anomalies. But don’t ever forget to anchor yourself back in the real temperatures.

    Can’t agree with you more…. when only the temperature anomalies are published then I know this is NOT science just obscuration.

    Thank you for your wonderful work.

  6. j ferguson says:

    It occurs to me that when the various participants in these discussions say that they have tested some of the observations we read here and found them “spurious”, what they have done is to run a computer analyses on one of the datasets with filtering for one thing or another.

    They have not examined the dataset in detail as E.M. is doing here, and likely they have run anomaly analyses in much the same way they do on the total data.

    one might imagine that the results they get could be a product of their analytical method and not dependent entirely on the specific components of the data.

    REPLY: [ For many cases, yes. They do some “analysis” using anomalies of some sort and say it is divergent, so I must be wrong. This, of course, can never find any systematic bias or error in the anomaly method itself (as it is assumed to be valid and unbiased on the face of it.) Though there are some other cases.

    Generally the complaints fall into 2 broad categories:

    1) Using an actual temperature average is bogus compared to anomalies because you average a hot place and a cold place, then add another one of either and the results are not comparable. This misses WHY I do the actual temperature average (to SEE that bias and estimate how much there is) and it misses how you can VALIDLY read that AveT chart – look at the trends during times of stable thermometer counts. THAT is the real change of temperature. Basically, it tells you what is really happening in the data.

    2) The results which I get don’t match some Toy World or Hypothetical Cow argument. For those, I really just don’t care. What I’m trying to do is not to admire Hypothetical Cows nor live in Toy Worlds. I want to know exactly 2 things. What is REALLY happening in the REAL world. And just how close is the real world temperature code, like GIStemp, to that real world actual result. None of the Toy Worlds nor Hypothetical Cow arguments can ever get you there.

    Somehow this hard and fast focus on reality baffles some folks…

    I come from an accounting and forensics background. I just don’t let go of what the numbers really say. You could never walk in and say “The books are fine, the accountant said they follow GAAP (Generally Accepted Accounting Principles) and he has a degree from a good school so it’s ok.” and you would certainly never say “The programmer wrote some code and it says there is $1,000,000 in the vault, and even though I looked and couldn’t find it, we ought to trust the program because it has a very elegant design.” That is the stuff of which Madoff, Enron, and the Fanny Mae / Freddy Mac / Lehman home loans were made…

    The “Right Way” is a very literal: “Show Me The Money!”… then you count every penny and compare it to what the printed books say THEN compare it to a surprise re-run of the computerized accounting system THEN you compare it to the bank statements for the past 10 years THEN you ask to see the vault books for who has been in there the last year dragging bags in and out and THEN you inspect all the bag receipts for what was supposed to be in them and THEN you look at …. until you’ve looked at every single thing you can get… And if the vault has 4,000 dollar coins and was supposed to have 10,000 quarters, that IS a red flag even though it’s $4,000 in both cases. Somebody either swapped some bags or screwed up the accounting.

    And that is why I start with looking at the temperatures. Because while it doesn’t make a lot of sense to look at the average number of coins in your pocket without looking at the denominations, if someone says they put $1,000,000 of coins in the vault, the average number of coins better not be smaller than 1,000,000 as we don’t have any $2 coins any more… (Or, bluntly, forensics isn’t “climate science”, and I’m damn proud of that.) -E.M.Smith ]

  7. harrywr2 says:

    “FWIW, Canada looks to be one of THE most extremely “warming” places in the anomaly and absolute temperature averages. I’m still trying to work out exactly why. Some of it shows an up to 8 C average “warming” (that I think we’d have noticed if it was real…) My working thesis at this point is that high latitude”

    I’d propose another hypothesis. Places with significant snow cover are more susceptible to Albedo changes.

    I.E. Deadhorse and Point Barrow Alaska show significant warming. Both had what one would consider ‘bush’ landing strips prior to the 1980’s. They both now have Class I Commercial runways. Both have 9 months a year with temperatures below 0C.

    We don’t see the same trend in the Antarctic. There are no paved runways in the antarctic.

    REPLY: [ Gee… we plant the thermometers in grassy fields that are covered in snow much of the time. Then change the place to be black tarmac, jet wash, and snow ploughs … and “find warming”. Yeah, I could see that pretty easily… but we’ll see what the data say. -E.M.Smith ]

  8. E.M.Smith says:

    That initial cold reading was nagging at me, so I went to take a look:

    First, find the country code for Costa Rica in the 28 December 2009 data download from NCDC:

    Snow-Book:~/Desktop/GHCN/GHCN28Dec09 chiefio$ grep COSTA v2.country.codes
    405 COSTA RICA

    OK, it’s country code 405. Now go looking in the v2.mean temperature file for the earliest records:

    Snow-Book:~/Desktop/GHCN/GHCN28Dec09 chiefio$ grep ^405 v2.mean | sort -n -k 1.13,1.16 | head
    4057876200101941-9999-9999-9999-9999-9999-9999-9999 217 220 216 202 201
    4057876200101942 199 202 219 222 227 218 211 214 212 207 207 201
    4057876200101943 196 190 206 216 213 212 205 210 210 201 186 178
    4057876200101944 171 174 190 196 196 202 198 197 192 183-9999 174
    4057876200101945 188 189 201 208 209 217 203 208 208 202 198 190
    4057876200101946 192 196 199 206 216 210 206 204 208 204 202 207
    4057876200101947 192 198 206 204 212 212 208 209 206 208 206 192
    4057876200101948 190 191 200-9999 214 221 210 206 210 207 202 191
    4057876200101949 185 185 192 204 202 195 196 196 196 192 187 183
    4057876200101950 178 178 201 204 208 208 203 201 208 202 192 189

    so it’s station ID 78762 with sub-station of 001 and Modification History Flag of 0.

    Snow-Book:~/Desktop/GHCN/GHCN28Dec09 chiefio$ grep 40578762001 v2.temperature.inv
    40578762001 SAN JOSE/CENTRAL OFFICE 10.00 -84.10 1141 1327U 391MVxxno-9x-9TROP. SEASONAL C

    So it’s San Jose, Costa Rica, at about 10 LAT 84.1 LONG and an altitude of 1141 m recorded or 1327 m from a terrain map. So as expected we start with one thermometer at some modestly cool altitude, then add them to the beach resorts over time…

    But there are more interesting things to be seen here… The 5 digit StationID is supposed to be one physical place (roughly) with the substation meaning a “different station, but very nearby”. So you might have “At the Airport” and “At the Tower” or even “At the beach NEAR the airport”, but would not have “On the Mountain” vs “On the Beach”. So I went looking for what thermometers were “near each other” in stationID 78762. There is the one we have seen above. Cool and on the mountain somewhere. Supposedly right next to it ought to be:

    Snow-Book:~/Desktop/GHCN/GHCN28Dec09 chiefio$ grep ^40578762000 v2.temperature.inv
    40578762000 JUAN SANTAMAR 10.00 -84.22 939 1060S 33MVxxno-9A 2TROP. SEASONAL C

    Both are at LAT 10.00 and the LONG is a little different at 84.22 W but with the same 78762 StationID “in the same place”. But look at the Altitude! 939 Meters reported or 1060 m via a terrain map. We’re a couple of hundred meters further ‘down slope’. So, any impact? First “upslope”

    4057876200101963 189 195 201 209 213 214 207 210 214 212 202 197
    4057876200101964 194 212 209 211 212 210 209 210 208 203 200 192
    4057876200101965 185 199 199 203 223 214 208 208 214 207 205 200
    4057876200101966 202 200 202 214 218 216 216 215 215 212 194 197
    4057876200101967 190 194 189 202 203 202 198 200 198 200 196 195

    Then “downslope”.

    4057876200011963 212 219 229 221 221 220 222 223 212 216 208 222
    4057876200011964 217 226 230 224 224 208 208 209 205 204 216 215
    4057876200011965 214 219 214 227 228 226 228 213 213 219 221 216
    4057876200011966 227 236 231 234 218 219 220 224 216 205 217 219
    4057876200011967 224 224 225 231 232 215 218 218 208 218 221 224

    So January is running about 18.x to 19.x, at altitude, while a bit further down the hill it’s 21.x to 22.x C. Nice couple of degrees C higher.

    and when does the first thermometer end? 1980

    4057876200101979 184 185 199 198 204 197 197 195 196 202 196 192
    4057876200101980 190 186 199 209 213 207 203 201 198 197 188 180

    just about when we start our temperature flattening and rising some more on deletions. Though someone was doing some experimenting then. There was another Mod Flags at that location, though curiously, only for 10 years between 1961 and 1970:

    4057876200111968 188 192 190 197 197 202 204 199 199 198 195 192
    4057876200111969 191 193 208 210 220 207 200 204 204 198 196 196
    4057876200111970 198 184 202 204 200 205 218 196 194 193 185 187

    The readings are a little different from the original:

    4057876200101968 195 196 195 206 215 216 209 206 209 209 204 198
    4057876200101969 198 200 213 217 226 222 200 204 204 198 196 196
    4057876200101970 198 184 202 204 200 205 197 196 193 193 185 186

    They look like a bit cooler, then catch up. Like some experimental apparatus being tuned, then that entire location is abandoned.

    During the same period of time, the lower station has a similar “experimental” decade (though it gets “mod flag 2”:

    4057876200021961 215 217 222 224 219 213 208 216 198 205 205 212
    4057876200021962 213 214 220 220 212 206 201 210 208 198 206 212
    4057876200021963 212 219 229 221 221 220 222 223 212 216 208 222
    4057876200021964 217 226 230 224 224 208 208 209 205 204 216 215
    4057876200021965 214 219 214 227 228 226 228 213 213 219 221 216
    4057876200021966 227 236 231 234 218 219 220 224 216 205 217 219
    4057876200021967 224 224 225 231 232 215 218 218 208 218 221 224
    4057876200021968 225 218 227 230 214 213 221 221 213 214 214 220
    4057876200021969 211 224 238 235 230 217 225 212 216 205 211 222

    it is followed by “mod flag 3” going forward to date, as though someone had made a decision that this one was a ‘keeper’ and went production. The “mod flag 1” from that location was identical for that time period as was the mod flag 0 for the same time period.

    Mod flag 1 runs from 1956 to 1981 while mod flag 0 runs from 1956-1991 (when it leaves in The Great Dying of Thermometers).

    Curious. Very curious. As though there were two original records (perhaps one person reading in the morning and one at night? Or…). Then some experimental apparatus is put in both places for a decade (or perhaps just a different process?).

    Finally, at the end of the time, everything is scrapped except that one lone “Mod Flag 3” station at the lower location. And it has more holes in the data and looks to me like it runs hotter:

    4057876200031987 229 239 246 246-9999 239 232 235 232 228 228 223
    4057876200031988 236 240 239 244 237-9999 227 220 220 219 226 225
    4057876200031989 231-9999 232 246 236 228 230 239 222 227 227 226
    4057876200031990-9999 230 241 249 241 238 230 233 230 224 231 230
    4057876200031991 232 235 244 242 237-9999 239 235 234 228 231 230
    4057876200031992 231 235 245-9999 244 236 232 221 225 228 227 232
    4057876200031993 229 232 236 242 234 234 238 236 225 230 232-9999
    4057876200031994 227-9999 241 244 237-9999 238 233 235 229 225-9999
    4057876200031995 232 237 240 242 232-9999 233 225 229 222 233 231

    In particular notice those 23.x in January and the 24.x in February 1988.

    Frankly, this little example leaves me with the hypothesis:

    It’s all about The Splice.

    We splice on a Mod Flag 3 that is hotter, then GIStemp tries to do an “in-fill” back to the cooler up slope (based on comparisons between earlier cooler mod flags for the adjustments) and ends up finding more warmth all over.

    And once again we have the point that:

    In calorimetry it’s a bad idea to play with the instrumentation.

  9. dougie says:

    Hi E.M.

    just for the brain fried lurkers at here, Lucia ,etc… like me 8-)
    can you give your interpretation (or point to previous post) of the explaining your interpretation of the anomaly method & it’s advantage/disadvantages & why this method is used to show global warming/cooling/standstill by the climate scientists?
    yet it is never/seldom mentioned in the MSM (always higher temps/doom/gloom).
    what does an anomlily mean in layman terms? give me an everyday example i can relate too (if possible).

    thanks for your effort to inform/educated joe public on all the posts on your very informative site.
    keep on digging (like a female bunny).

    ps
    do bunnies hibernate, cos our wild bunny has been missing for 2 mth (UK)
    R.I.P. Jack (sad post at the time, but now i understand).

    REPLY:[ Bunnies are crepuscular so you will only see them at sunrise and sunset most of the time. As those times shift, you may miss the “usual clock time”. There are a very large number of wild bunny species, I don’t know about all of them. Domestic bunnies do not hibernate, nor do ‘snowshoe’ bunnies of several sorts, so I think not. They will stay out of bad weather and under cover from rain and wind.

    Anomalies:

    If you have widely different and divergent series, a simple average gives very misleading answers (which is why you keep an eye on the count of objects being averaged and watch out for ‘ringers’ as they change).

    Take three thermometer series.

    1 1 1 1 1 1
    2 2 2 2 2 2
    3 3 3 3 3 3

    The average over time would be:

    2 2 2 2 2 2

    But if, instead, you had fragmented records (where 0 means ‘missing’)

    1 1 1 0 0 0
    2 2 2 2 2 2
    0 0 0 3 3 3

    The simple average (remember where there is 0 you divide by 2 as 0 is ‘missing’)

    1.5 1.5 1.5 2.5 2.5 2.5

    So ‘missing segments’ gave you what looks like a 1.5 to 2.5 rising temperature average.

    To ‘fix this’, you use anomalies. Each value is compared to the prior value in each series then those are averaged. You can compare things to the starting value (called First Differences) or to the average of the whole series or to an average of a sample (what GISS does in GIStemp and what CRU does in CRUT with the ‘baseline’). Well do an average.

    Average of all ones is “1”. Average of all 2s is “2”. Average of all 3s is “3”. Difference between ones and “1” is nil. Difference between twos and “2” is nil. Threes and “3”, nil. So in all cases our set of difference would be nil

    0 0 0 0 0 0
    0 0 0 0 0 0
    0 0 0 0 0 0

    And we would find “no trend” even in the case with missing data. (gaps.)

    If we suddenly had our last values increase by one, or table of ‘anomalies’ would become:

    0 0 0 0 0 1
    0 0 0 0 0 1
    0 0 0 0 0 1

    and we would get a valid indication that we went up by 1 degree in the last period. For a simple average, we would get 2+3+4 = 9 that we’d divide by 3 to get 3, and compare to the prior 1+2+3 = 6 /3 = 2 and we’re at a 1 C increase as well.

    So the only real difference comes in when you have change of the things you are averaging together. (Those “gaps” where we had a bogus 1 C rise due to data drop outs in segments).

    OK, but anomalies are not perfect either. If you had a baseline in the first 2 values and
    had:

    1 1 1 1 1 1
    2 2 2 2 2 2
    0 0 0 3 3 3

    your baseline would be 1.5 (as a simple average) and you would now compare 1 to 1.5 in time zero vs 1 to 1.5 in time 6 and get ‘same as time zero’. Similarly 2 vs 1.5 and 2vs 1.5, no change. But what do you do for that “3”? Compare it to 1.5 and find 1.5 of warming? Compare it to ??? (So GIStemp makes up a basket of things it averages together and compares that to the baseline… 3+2+1 = 6 /3 = 2 … gee, that’s 0.5 C of warming… In reality, there’s an attempt to adjust that 3 to ‘what it ought to have been in time period 1’ but just how good is that ‘homogenized’ and ‘in-filled’ value?)

    If you run it the other way:

    1 1 1 1 1 1
    2 2 2 2 2 2
    3 3 3 3 3 0

    We have a baseline of 6/3 = 2 (as a simple example) and we have 2+1 =3 / 2 = 1.5 at the end… so we ‘cooled’ by 0.5 C compared to the ‘baseline basket’ due only to station change. Again, you could try to ‘make up’ what that last 3 ought to have been, but then you are not measuring ‘anomalies’ so much as you are measuring ‘imaginings’.

    First differences has a zero as the first “comparison” as a thing matches itself in the first period. After that, the “difference” is all that is computed and kept. For data:

    1 1 1 1 2 1
    2 2 3 2 1 2
    3 3 4 3 4 0

    We would get:

    0 0 0 0 1 -1
    0 0 1 -1 -1 1
    0 0 1 -1 1 –

    You can see that the first line adds to zero (even though we took one hot year), no net change, same for the second (though it is a neutral year with one hot and one cold). The third line probably ought to have been 3 3 4 3 4 3 given what the others were (and that’s what the GIS Reference Station Method would probably plug into it – though even that would depend on if you compared to 1, 2, or the average..) but we don’t really know. It could have been a 2 or a 4 or… but as it stands, we have a net +1 in the last line. So which is right? That first line that found “no change” but did have a single warm year? The last line that found one warm year, but we don’t really know if it ought to have been neutralized in the next year (that was missed)? FWIW, I think that last line effect is how station drops warm the planet. Leaving a warmed series ‘unclosed’ by dropping at a non-random time.

    So both ‘have issues’ which is why I look at both. Best would be a selected set of long lived thermometers without a lot of dropouts. If only we had them…

    Hope this helps more than hurts understanding ;-0 -E.M.Smith ]

  10. A C Osborn says:

    The comments are even more informative than the Graphs, thanks for that.

  11. dougie says:

    thanks for the reply E.M. clear as ever.
    helps me get a handle on the issues (i hope).

  12. E.M.Smith says:

    Yeah, I run things a little differently than other places. I’m not trying to generate hit counts or post little “sound bites” of tease to get a bunch of wolves howling. Just posting what I’m doing / finding (as much to document and to share) and helping folks to find some truth where it can be found. Then explaining it when some part was ‘less than clear’ (which helps me, too, as it lets me know what I was skipping over too fast ;-)

    The end result is that the comments often end up with more interesting stuff in them from the expansion and interactions than did the original postings.

    Kind of like watching a football match on TV with friends. Yeah, a nice running play can be interesting, but it’s the discussion of the set up and strategy by chums in “the peanut gallery” about what to do going forward on the field that’s the real fun…

  13. boballab says:

    EM

    There is also a neaby station 78760 Puntarenas that sits down by the beach at a whopping 3M and the record ran from 1961 to 2000. GISS Coords are 10N by 84.8W, basically just down the mountain from the other two.

    Also Juan Santamar is probably the chopped off version of Juan Santamaria as in Juan Santamaria International Airport which sits about 20km oustide San Jose and is the second busiest airport in Central America: http://en.wikipedia.org/wiki/Juan_Santamar%C3%ADa_International_Airport

    I did to these Costa Rican Stations that I did to Alert Canada and compared them to the gridded GISS trends and found, just like you, it’s a step function around 1985. When you graph out the data from 1942 to 1980 you get a very small trend and then do 1981-2009 you get a flat trend. So I decided to lop out the baseline period and just glue the data from 1981-2009 onto the 1942-1950 data and got basically the entire trend for 1942-2009. To see it all here is the link:
    http://boballab.wordpress.com/2010/03/16/costa-rican-warming-a-step-artifact/

    REPLY: [ You just KNEW I wanted to do an analysis like that and simply had no time to get to it, didn’t you? So much guilt wiped away in one wonderfully done analysis. I think your result came out better than what I was thinking of doing, too. IMHO, your posting could serve as a gold plated Poster Child for where all the ‘warming’ is coming from and WHY all the “by altitude” and “by latitude” thermometer change studies “matter”. Just Love It. -E.M.Smith ]

  14. E.M.Smith says:

    FWIW, I have just (finally) put in the “hair graphs” for Japan on:

    https://chiefio.wordpress.com/2010/03/01/japan-poster-child-for-the-smith-effect/

    And they are very interesting. Clear “step functions” with thermometer change. There is a very interesting one just a couple of years after the end of WWII (about as long as one would expect for “thermometers” to become an issue and to be brought into conformance with non-Imperial standards…) with both the pre-war and post-war trends dead flat (though step function offset). Then, in 1990, we get a step function higher and a strong warming trend with a drop from 157 thermometers to only 50.

    I think it is becoming quite clear that “global warming” is nothing more than the typical poor science problems of “Splicing data” and “changing the instrument in the middle of a test run”. Things you ought to learn not to do in High School Chemistry class. ( I did.)

  15. dougie says:

    with ref to above
    just to keep you updated to comments at Lucia

    {~snip Lucia is all about “Toy Worlds” and “Hypothetical Cow” arguments. I’m entirely NOT about them. We are orthogonal. I see no reason for cross posting discussions that will NEVER be about the same basic area in that Toy World and Real World will never meet. -E.M.Smith }

    torn8o (Comment#38215) March 16th, 2010 at 12:51 pm

    The posters who are claiming that the station dropoff in GHCN is because those stations don’t report regularly are just plain wrong. Just as an example, 64 South American stations drop out of GHCN after 1992. More than 20 of them have continued to report regularly since then, and the data can be found in — wait for it — another NCDC product!

    Which product?

    There are multiple types of regular reports. There are SYNOP reports; there are also monthly CLIMATs. In order to get into the GHCN, you need to issue CLIMATs. These are the ones with the monthly means.

    The stations that dropped at 1990 were not issuing CLIMATs at the time GHCN was put together. Now, some of those stations have started issuing CLIMATs in the last few years; not all of those have been added to the GHCN yet. This includes Bolivia, some of Canada, and scattered others. Some of these could be added this year.

    To search the CLIMATs, go here
    http://www.ogimet.com/gclimat.phtml.en
    The JMA includes all CLIMAT data. See maps here.
    http://ds.data.jma.go.jp/gmd/tcc/climatview/

    ps. i know you are more interested/time confined in your own ‘digging’ so just tell me to leave it.

    they are watching you 8-) & your posts/comment discussions.

    if i was in MSM i would hire you in a second for your wide ranging knowledge/opinions. have any shown interest?

    Funny how bunnies see to crop up in my mind (watts up doc).

    REPLY:[ For all the watching they seem short on the understanding part ;-) Maybe because the real world is significantly more complex than the imaginary worlds and real world issues are not subject to removal by “deeming them dealt with”. So Lucia and her hangers on will never be satisfied here, and I will never play “Toy World Games” as they are just not very relevant at the point where Real World begins; and we must simply agree to go our separate ways. Pointers to things like “Missing CLIMAT reports available here” are useful, so interesting to share.

    It would be fun to be the the “Tech Guy” for a MSM outlet, but I fear that they would not appreciate being pulled back to the facts when they had a good story all lined up to spin ;-) but if they would listen politely as I explained something was unphysical, I’d be OK if they then ignored me to make the story better. Most people have no desire to know the truth and little interest in the facts; they are much more interested in “wardrobe accidents” and “Who’s Ox was gored”. (And not so much interested in the question of why oxen are being gored or if the wardrobe accident was such that it could not happen by accident…) But no, no enquiries. One TV station in N. Cal asked for contact info, though. -E.M.Smith ]

Comments are closed.