Feb 29 2008
Miami, AZ: Undocumented Station Move?
I missed this post at ClimateAudit a few weeks ago. Another UFA sighted in Arizona. It’s about the surface station in tiny Miami, Arizona. Truth be told, I didn’t even know there was a Miami in Arizona. Next I suppose they’re going to try to tell me there’s a South Beach. The crux of the ClimateAudit post is that the surface station moved, and it introduces a positive bias in the trend. In this post, I will show that the main conclusion reached by Anthony Watts is true, but that it was not the documented move that causes a bias. Instead, it is an undocumented move.
Parking Lot = Warm Bias
First, let’s have a looksee at where the station is to see if there is/are potential source/s of bias. This has already been discussed at length in the thread at ClimateAudit. The short answer: yes.
|
Image courtesy Anthony Watts, surfacestations.org, and Nicholas Meyer.
The station history [type "Miami, AZ" in the search field, no quotes] shows that the last station move occured in 1965. Since then, there are multiple entries in the station history, but the location and height of the remain the same. There is an entry for February 1996 that seems interesting. Actually, it seems pretty boring if just looking at the history. However, if we look at the NASA GISS temperature record for this station, there is a large jump that roughly coincides with this time period.
It is difficult to assess whether the apparent jump in data is due to a station move or some other factor when looking at the yearly GISS data. Therefore, I will show the monthly data directly from the USHCN. I will also be showing the maximum temperatures and not the means (because I still have a problem reading the mean files that I haven’t tried to figure out), but this should not influence the conclusion. I also plot in degrees Fahrenheit, because that’s what USHCN reports.
Miami vs. Tucson
First, I want to look at how the raw data from Miami, AZ compares to surrounding stations. This is the data directly from the USHCN with minimal processing; for extreme outliers and such.
|
What is plotted is the difference between the monthly temperature at Miami, AZ and the closest 10 surrounding surface stations. I chose the 10 nearest stations to Miami, AZ to get a representation of what is happening regionally. Some of these stations also have issues (such as Tucson), but these should get averaged out. Values greater than zero indicate that the Miami, AZ station has a temperature greater than the regional average.
The 1996 station “move” is highlighted as the red dashed line. There is clearly a jump in the data at approximately the same time as the station “move”, and it would be indistinguishable if looking at the yearly data. However, there is approximately 9-12 month gap between the station “move” and the step in the time series. I can think of three hypothesis why this could be. I welcome more in the comments.
- The station move is documented wrong in the history
- There was an undocumented station move in addition to the 1996 move
- There was an undocumented change of equipment, but no station move
|
The above plot is extremely similar to the first plot, except now I’m plotting the FILNET data. This is supposed to account for station moves and other biases in the data. Again, the red dashed line shows the documented “move” in the station history. Clearly, whatever algorithms that are supposed to adjust for station moves missed this one.
Conclusions
From the recent examples of Lampasas, TX and Miami, AZ, it seems evident that whatever algorithm that is used to adjust the USHCN data for station moves does not work as well as expected. In both cases, an actual station move to a warmer location has introduced a positive bias into the temperature trends. The difference between the two cases is the the Lampasas station move was documented in the station history.
While there was an entry in the Miami, AZ station history that roughly corresponds to the time period where the jump in temperatures is seen, there is a several month gap. There are several possible reasons for this, and I’ve outlined them above. It seems likely that the sensor was moved to a warmer location, this change was not reflected in the station history, and thus was not corrected in the FILNET data.
(No Ratings Yet)
Loading ...Related Posts:
7 Responses to “Miami, AZ: Undocumented Station Move?”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
Is there a specific reason why you can rule the “jump” out as being natural (i.e., a real temp. jump) vs. a move induced or land use induced jump?
It could also simply be that they spread a fresh patch of that black gravel all around the vicinity. A previous commenter who worked there said it was a product of the mining process.
Such an event could easily affect temperature given its proximity. And of course, NOAA/GISS/NCDC would have no record of it, and corrections would miss it wholesale without a station move reference to flag it.
See the aerial view here.
That sensor sits in the middle of an island of blackness compared to surrounding soil albedo. Sometimes simple explanations exist. I’m surprised Atmoz didn’t pick up on this. It is a darker albedo than even the Tucson U of A weather station in the parking lot.
It is also on the south (sunlit side) of a metal building with a concrete vault next to it, which makes a splendid night time heatsink.
Microsite bias. I’d say for this sensor, it is surely man-made warming
Eric:
Steps like that almost never happen naturally. I’ll “zoom in” to get the exact month of the switch, then you can ask if they repaved (or similar) at that time.
Anthony:
You’re absolutely right. Although changes of this sort should be recorded somewhere.
Another dull question from me. Will the jump continue to stand out no matter what happens over the next 20 years — say even if temps drop? I guess what I am getting at … is there any data squish or extra effect the choice of graph itself is having on the data?
For the purposes here, I am going to assume there are 3 spatial scales that matter: global, regional, and local. Of course, in reality there is a continuum of spatial scales from really small to really big.
What I am looking at in the plots is how the local temperature differs from the regional temperature. I defined the regional temperature as the average of the 10 closest stations It’s actually slightly different than this, but it’s insignificant for this discussion. If the global temperature goes up/down, both the regional average and the local temperature go up/down. If the regional temperature goes down/up then the regional average and the local temperature go down/up. In both cases, the difference between the regional temperature anomaly and the local temperature anomaly is zero.
The only situation where the difference is non-zero is if there is a difference between the regional average and the local temperature. This can occur because of deviations in the regional average or because of deviations in the local temperature. Because of the way I defined the regional average, local effects at the regional stations will have only a small effect on the average. Thus, changes seen on the plot are most likely the result of local temperature changes not seen at the regional and global scales.
The jump will always appear in the plot of the raw data. The jump will disappear in the filnet (supposedly fixed) data when/if it actually gets corrected for this type of bias.
check the daiy temps, tmax and tmin.
[...] used the same methodology as described in a previous post. Basically, I look at the monthly temperature anomaly at station X and compare it to the anomalies [...]