This site was selected because Kevin Cowtan produced the
NOAA Paraguay data
video (which was also used to advertise the Climate Change Denial course).
Here, Kevin uses an older version of the software (but with a higher quality map) to examine the data.
I suggest that you follow along using my program and try to recreate the same plots.
BTW - all the images on this page can be zoomed by simply using the mouse wheel or double clicking.
The video features a prototype tool for investigating the global temperature record. This tool will be made available with the upcoming MOOC,
Making Sense of Climate Science Denial,
where we will interactively debunk myths regarding surface temperature records.
Before starting a complex analysis, here is just the station data
for Puerto Casado (the green dot on the map).
Puerto Casado location
Green dot is the location of Puerto Casado.
The 2 images below are of the same temperature record -
the red is raw, the blue is "adjusted".
The only difference between the images are the end point
of the trend lines - 2000 on the left and 1966 on the right.
The video makes a real big deal about a descending spike in 1970 -
you can see it in the red trace below.
Puerto Casado raw and adjusted temperatures
Actual temperature - red is raw, blue is adjusted, purple means they are identical
You can also see, of course, that discussing just the 1970 spike that was removed is
just misdirection - it is the change in the trend line (cooling to warming!) that is important!
Actually, the second plot shows that the slope of the 1951 to 1966 data is identical -
it is the 2°C drop (adjustment) in the early data that is important.
Sometimes this sort of thing is because the station moves to another location.
However, in this case, perhaps not - notice that the raw data is continuous
from 1966 to 1980.
With a station move, the change occurs in a single year (or two).
In addition,
they have removed 4 years of data and adjusted an additional 4 (8 total)!
Very suspicious.
It is curious that in the video Kevin compared the raw and adjusted anomalies (computed for a base period of 1961 to 1990)
rather than looking at the actual temperatures.
As you can see in the plot above, the actual temperatures from 2003 to 20014 are identical before and after the "adjustments".
As far as trendlines go, this has no effect.
However, when looking at adjustments, this makes a major 2°C "adjustment" look like only half that amount.
In other words - it looks like no big deal.
In fact, it misleadingly implies that part of the record was adjusted up .. and that part was adjusted down.
(Looks fair - doesn't it!)
Puerto Casado raw and adjusted anomalies
Notice that the raw data is continuous from 1951 to 2001
Temperature anomalies - red is raw, blue is adjusted, purple means they are identical
So, the bottom line is - when looking at adjustments to a single station, use the temperatures, not the anomalies.
As I said above, a large temperature change due to a change in station location is usually pretty obvious,
and it typically shows a transition within single month (which can become one or two years in annualized data like this).
For instance, Salto, Uruguay has a pretty obvious station move - a one- to two-year fast change in temperature.
In this case, moving the station average so that the two half's agree makes a lot of sense.
However, changing the slope of the line - from -1.179°C/decade to -0.423 -
does not make obvious sense. If you had the raw daily data, this might be completely obvious ..
but from this vantage point ???
Salto
Actual temperature - red is raw, blue is adjusted, purple means they are identical
At the bottom of this page is an image (with links) showing that
the Salto data needed to be adjusted 7 times - not just one.
Like in the video, I zoomed out and selected a region.
I tried to select similar regions for both the adjusted and raw datasets.
When doing this (and switching between datasets), I saw that there are more sites
in the raw dataset, and the extras are probably a lot less reliable.
As a result, I adjusted the selection so that both collections contained exactly the same stations.
(It was while doing this that I spotted Salto - above.)
I also inspected a number of stations and
didn't see any difference between Paraguay and the surrounding areas.
It appears that northern Argentina also had the same weird data patterns.
Closer analysis shows that a lot of stations in that area have missing data.
As a result, the surrounding stations can not be used to remove those readings.
In fact, the data indicates that more research is appropriate.
Almost next door to Salto, Uruguay is Concordia AER, Argentina
- they are both at an altitude of about 48 meters and located on opposite sides
of the Uruguay River.
Locations of Puerto Casado, Salto, and Concordia
The 3 green dots indicate Puerto Casado, Salto, and Concordia.
The Concordia data was adjusted to make it colder in the past - suspicious, but not necessarily a problem.
Concordia - trend line from 1951 to 1966
Actual temperature - red is raw, blue is adjusted, purple means they are identical
However, it becomes more interesting when it is compared to Salto,
Concordia and Salto
Raw
Adjusted
In the raw data,
notice the lack of temperature change for Concordia - the trend is 0.014°C/decade and never changes. Very steady.
No sign of any forcing. No indication of a station problem.
However, the adjacent Salto loses 3°C in twenty years.
Clearly, something is not right.
(Remember, you have to zoom the map way in to see both stations - they are very close - what ever that means on a map at this scale.)
In the adjusted data, at the beginning Salto is now about 1.5°C warmer than Concordia which is just on the other side of the river.
I preformed the same type of analysis as shown in the video
and because I was doing an actual temperature (not an anomaly) the related purple stations
must also be deselected (turned red).
I tried to select the same stations for both datasets,
got 39 and 41, and could not find the 2 extra stations.
They are probably directly under another.
At any rate, I used the following stations (green means selected).
Selected Central South American stations
Green selected - Red ignored
These are the plots
Central South America with and without Paraguay
With Paraguay
Without Paraguay
Actual temperature - red is raw, blue is adjusted, purple means they are identical
As you can see, Paraguay makes almost no difference after 1985.
Since the number of stations drops off quickly before 1951, I am not displaying the older data.
You have access to the software, so you can generate those plots yourself.
That's interesting, there appears to be extreme cooling after around 1995.
Including Paraguay does not seem to make any effect.
Central South America minus Paraguay - 1994 to 2014
Same as the previous graph, just zoomed in on the recent years with a 20 year trend line - 1994 to 2014.
Actual temperature - red is raw, blue is adjusted, purple means they are identical
If anyone ever wanted evidence for a pause - here it is.
And this time, the adjustments made it cool faster.
Of course, your results will vary because you will probably pick different "near"
stations than anyone else.
But this type of analysis is fun and you will learn to use the tool.
Note: don't pick any sites in Chile - it's on the other side of the Andes.
Well, above I followed the general procedure in the video and got different results.
Since he doesn't specifically list the selected sites, I tried to match the dots.
His map appears to match the 7 stations selected below.
Selected Central South American stations
LAS LOMITAS
FORMOSA AERO
CAMPO GRANDE
RIVADAVIA
ROBORE
PONTA PORA
CORUMBA
Green selected - Red ignored
Note that ROBORE only appears in the raw data.
Therefore, some comparisions using it are invalid.
I went through and compared all 6 other sites used in the video.
Based on what I saw,
For 3 stations, only the data from 1951 to 1990 should be used
CAMPO GRANDE should not be used
CORUMBA should only be used from 1980 to present
FORMOSA AERO has data from 1965 to present
That's enough analysis to know that the video is nonsense and that my analysis, using 27 (raw) and 30 (adj) stations outside Paraguay,
is better than his using only 7.
Specifically, he claims that the 7 justify all the adjustments to the Paraguay data -
but the larger number indicates that the adjustments are bogus.
But you have access to the software - make up your own mind.
What I found most interesting is that the GHCN adjusted data is completely missing 4 years of data
(there is a clear gap),
and that the Berkeley Earth data does not have a gap.
Above, I used Salto as an example of an obvious station move.
According to Berkeley Earth,
the Salto data needed to be adjusted 7 times!
Also notice that the GHCN and Berkeley Earth final results are quite different.
In the Berkeley Earth plots, the vertical lines with red markers are supposed to indicate documented station moves,
the black markers represent undocumented anomalies identified because the data just looks wrong.
By the way - the
Berkeley Earth adjustment
changed Concordia from -0.06°C/Century to +1.16°C/Century
The main claim in the video is that the spike in 1970 is not found in the surrounding stations.
So .. here is a map and the results .. with Paraguay removed.
I have used the raw, unadjusted data.
Raw temperatures for the region around Puerto Casado, with and without Paraguay
Temperature anomaly plots - Puerto Casado (red)
Upper right - Surrounding region with(blue)/without(green) Paraguay
Lower right - Puerto Casado (red) with 9 additional individual stations (blue)
The green dots on the map indicate the selected sites - with and without Paraguay
I think the 1970 decrease is there .. and is perhaps real!
The take away from this page is not that I have proven anything one way or the other.
Instead, I have shown one way to use the tool and external data to evaluate
discussions you might see in the press.
There is clearly no one correct answer.
Monthly data would certainly help - my tool uses yearly data
The evidence used to justify each correction is not available
Some surrounding sites might indicate that a different adjustment is more logical