By Steven Mosher,
AC Osborn made an interesting comment about airports that will give me an opportunity to do two things: Pay tribute to Willis for inspiring me and give you all a few more details about airports and GHCN v4 stations. Think of this as a brief but necessary sideline before returning to the investigation of how many stations in GHCNv4 are “ruralish” or “urbanish”. In his comments AC was most interested in how placement at airports would bias the records and my response was that he was talking about microsite and I would get to that eventually. Also a few other folks had some questions about microsite versus LCZ, so let’s start with a super simple diagram.
We can define microsite bias as any disturbance/encroachment at the site location which biases the measurement up or down within the “footprint” of the sensor. For a thermometer at 1.5meters, this range varies from a few meters in unstable conditions to hundreds of meters in stable conditions . In the recent NOAA study, they found bias up to 50 meters away from a disturbance. I’ve drawn this as the red circle, but in practice, depending on prevailing wind, it is an ellipse. The NOAA experiment (more on that in a future post) put sensors at 4m, 50m, and 124m from a building and found
The mean urban bias for these conditions quickly dropped from 0.84 °C at tower-A (4 m) to 0.55 and 0.01 °C at towers-B` and -C located 50 and 124 meters from the small-scale built environment. Despite a mean urban signal near 0.9 °C at tower-A, the mean urban biases were not statistically significant given the magnitude of the towers standard 2 deviations; 0.44, 0.40, 0.37, and 0.31 °C for tower-A, -B, -B’, and -C respectively.
While not statistically significant, however, they still recommend precaution and suggest that the first 100m of a site be free of encroachments. In field experiments of the effect of roads on air temperature measured at 1.5m, a bias of .1C was found as far as 10m away from roads. At airports this distance should probably be increased. At an airport where the runway is 50m+ wide, the effect the asphalt has on the air temperature is roughly 1.2C at the edge of the runway and diminishes to ~.1c by 150m away from the runway. (Kinoshita, N. (2014). An Evaluation Method of the Effect of Observation Environment on Air Temperature Measurement. Boundary-Layer Meteorology) Exercising even more caution, I’ve extended this out to 500m, although it should be noted that this could classify good sites as “bad” sites and reduce differences in a good/bad comparison. Obviously, this range can be tested by sensitivity analysis.
Outside the red circle I’ve depicted the “Local Climate Zone”. Per Oke/Stewart this region can extend for kilometers. In simple terms you can think of two kinds of biases: Those biases that arise from the immediate vicinity within the view of the sensor and have a direct impact of the sensor, and those that are outside the view of the sensor and act indirectly– say that tall set of buildings 800m away that disturb the natural airflow to the site. In the previous post, we were discussing the local scale; this is the scale at which we would term the bias “UHI.”
There is another source of bias, from far away areas, and I will cover that in another post. For now, we will use airports to understand the difference between these two scales. Let’s do that by merely picturing some extremes in our mind: An airport in Hong Kong, and an airport on a small island in the middle of the ocean. Both airports might have microsite bias, but the Hong Kong temperature would be influenced by the urban local climate zone with its artificial ground cover. The airport on the island is surrounded by nonurban ocean, with no UHI from the ocean. Simplistically, the total bias a site might be seen as a combination of a micro bias, local bias, and distant bias.
There are, logically, six conditions we can outline:
||No Micro Bias
||Warm Micro Bias
||Cool Micro Bias
||No Micro Bias
||Warm Micro Bias
||Cool Micro Bias
It is important to remember that micro disturbances can bias in both directions, cooling by shading for example. And note that logically you could find a well sited site in an urban location. This was hypothesized by Peterson long ago:
“In a recent talk at the World Meteorological Organization, T. Oke (2001, personal communication) stated that there has been considerable advancement in the understanding of urban climatology in the last 15 years. He went on to say that urban heat islands should be considered on three different scales. First, there is the mesoscale of the whole city. The second is the local scale on the order of the size of a park. And the third scale is the microscale of the garden and buildings near the meteorological observing site. Of the three scales, the microscale and local-scale effects generally are larger than mesoscale effects….
Gallo et al. (1996) examined of the effect of land use/ land cover on observed diurnal temperature range and the results support the notion that microscale influences of land use/land cover are stronger than mesoscale. A metadata survey provided land use information in three radii: 100 m, 1 km, and 10 km. The analysis found that the strongest effect of differences in land use/land cover was for the 100-m radius. While the land use/land cover effect ‘‘remains present even at 10,000 m….
Recent research by Spronken-Smith and Oke (1998) also concluded that there was a marked park cool island effect within the UHI. They report that under ideal conditions the park cool island can be greater than 5 C, though in midlatitude cities they are typically 1 –2C. In the cities studied, the nocturnal cooling in parks is often similar to that of rural areas. They reported that the thermal influence of parks on air temperatures appears to be restricted to a distance of about one park width….
Park cool islands are not the only potential mitigating factor for in situ urban temperature observations. Oceans and large lakes can have a significant influence on the temperature of nearby land stations whether the station is rural or urban. The stations used in this analysis that were within 2 km of the shore of a large body of water disproportionally tended to be urban (5.8% of urban were coastal versus 2.4% of rural).
Looking at airports will also help you cement the difference between the micro and the LCZ in your thinking. With that in mind we will turn to airports and look at various pictures to understand the difference between the micro and the local- the nearby city or the nearby ocean or field.
First a few details about airports. In my metadata I have airports classified as small, medium and large
First, the small: some are paved. Pixels (30m) detected as artificial surface are colored orange:
Some are dirt
Now large airports
We will get to medium, but first a few other airports by water, a 10km look, the blue dot is the station, red squares are 30meter urban cover
The medium airport I choose was one of Willis’ favorite airports, discussed in this post. Before we get to that visual, I encourage you all to read that post, because it put me on a 6 year journey. Willis is rather rare among those who question climate science. He does his own work, and he raises interesting testable questions. He doesn’t merely speculate; he looks and reads and does actual work. He raised two points I want to highlight:
Many of the siting problems have nothing to do with proximity to an urban area.
Instead, many of them have everything to do with proximity to jet planes, or to air conditioner exhaust, or to the back of a single house in a big field, or to being located over a patch of gravel.
And sadly, even with a map averaged on a 500 metre grid, there’s no way to determine those things.
And that’s why I didn’t expect they would find any difference … because their division into categories has little to do with the actual freedom of the station from human influences on the temperature. Urban vs Rural is not the issue. The real dichotomy is Well Sited vs Poorly Sited.
It is for this reason that I think that the “Urban Heat Island” or UHI is very poorly named. I’ve been agitating for a while to call it the LHI, for the “Local Heat Island”. It’s not essentially urban in nature. It doesn’t matter what’s causing the local heat island, whether it’s shelter from the wind as the trees grow up or proximity to a barbecue pit.
Nor does the local heat island have to be large. A thermometer sitting above a small patch of gravel will show a very different temperature response from one just a short distance away in a grassy field. The local heat island only needs to be big enough to contain the thermometer, one air conditioner exhaust is plenty, as is a jet exhaust
I think we both agree that the micro, what he calls local, is important. However, the area outside of the immediate area cannot be discounted: Hong Kong airport next to a huge city is going to be influenced by that locale, whereas, a large airport ( see above) on an island next to the sea, is arguably not going to be biased as much.
The second point Willis made was about the problems with 500meter data. In particular the MODIS classification system which required multiple adjacent pixels before a pixel was classified as urban. At that time we did not have a world database at 30m; Today we can look at that station and calculate the artificial area using 30m data. The next 4 images show the site at various scales: 500m, 1000m, 5000m and lastly 10000m. At the microscale ( <500meters) it classified as greater than 10% artificial, at 1km greater than 10% artificial, and at 5km and 10km it was less than 10% artificial.
There were some concerns about the temperature at this station being used. However, there has never been enough data from this station to include in any global series, even Berkeley’s. Nevertheless, it lets us see the kind of improvements that can be made now that higher resolution data is available for the entire world. Also, even when airports are included in the data analysis, the bias can be reduced in some cases. Here a 2C bias is removed.
One last small airport to give you some kind of idea of that data that we can produce today.
AC Osborn also wanted to know just how many airports were in GHCN v4; and, I think it’s safe to say that many skeptics believe that the record is dominated by airport stations. Well, is it? We can count them and see. For this count I will use 1km as a distance cut off. There are couple ways to “determine” if a station is at an airport. The least accurate way is to look at the names of the stations. This misses a large number of airports. To answer the question I use GPS coordinates compiled for over 55000 airports world wide, including small airports, heliports, balloon ports, and seaplane ports. I then calculate the distance between all 27K stations and the 55K airports and select the closest airport. I then cross check with those stations in GHCN that have a “name” that indicates it is an airport.
For this we consider a 1km distance for being “at an airport”. While this is farther than the microsite boundary, the point of the exercise is to illustrate that not all the stations are at airports.
Using 1km as a cut off, I find there are 1,129 stations by small airports, 1830 by medium airports, and 267 by large airports. That’s from a total of ~27,000 stations.
To assess the ability of the 30m data to detect airport runways and other artificial surfaces we can look at the stations that are within 500 meters of a large airport and ask? Does our 30m data show artificial surface?. There are 131 stations within 500m of an airport. We know that no sensor data/image classification system is perfect, but we can see that in the aggregate the 30m data performs well.
We can also ask how many large airports are embedded in Local climate Zones that have less than 10% artificial cover out to 10km. As expected large airports are in local areas that are also built up at levels above 10%. You don’t get large airports where there are no people.
Conversely, you get small airports embedded in local zones that are not heavily built out, a few cases of small airports embedded in Local Climate Zones that are heavily built out.
Here are the points that I would like to emphasize.
1. We can discuss or differentiate between at least 2 types/sources of bias: the close and immediate and those sources more distant
2. Bias at the short range (micro) can be more important than bias at the long range.
3. A good site can be embedded in a “bad” area or “good” area, similarly for a bad site.
4. 30m data is better than 500m data
5. Skeptics should not argue that all the sites or a majority are at airports. They are not.
6. There are different types of airports.
7. One way to tell if there is a bias is by comparing Airports with Non airports.