Ronan Connolly On Urban Bias In GHCN
By Paul Homewood
I touched on the question of UHI the other day, and questioned whether NOAA/GISS were adequately allowing for it.
It is therefore perhaps worth reposting this article from 2014, which introduced a detailed study by Ronan Connolly, Urbanization bias III. Estimating the extent of bias in the Historical Climatology Network datasets
The extent to which two widely-used monthly temperature datasets are affected by urbanization bias was considered. These were the Global Historical Climatology Network (GHCN) and the United States Historical Climatology Network (USHCN). These datasets are currently the main data sources used to construct the various weather station-based global temperature trend estimates.
Although the global network nominally contains temperature records for a large number of rural stations, most of these records are quite short, or are missing large periods of data. Only eight of the records with data for at least 95 of the last 100 years are for completely rural stations.
In contrast, the U.S. network is a relatively rural dataset, and less than 10% of the stations are highly urbanized. However, urbanization bias is still a significant problem, which seems to have introduced an artificial warming trend into current estimates of U.S. temperature trends.
The homogenization adjustments developed by the National Climatic Data Center to reduce the extent of non-climatic biases in the networks were found to be inadequate, inappropriate and problematic for urbanization bias. As a result, the current estimates of the amount of “global warming” since the Industrial Revolution have probably been overestimated.
The GHCN/USHCN network is the only source of land temperature data used by GISS and NCDC, except for a handful of Antarctic stations supplied by SCAR, the Scientific Committee on Antarctic Research. GHCN is also heavily relied on by HADCRUT.
The main points raised by Ronan’s paper are:
1) Although a third of the GHCN dataset is rural, almost all have very short records. Also, note the large step in the percentage of urban sites around 1990.
2) The GHCN dataset is homogenised, with the objective of ironing out “inhomogeneities”, i.e non-climatic biases introduced into temperature records, as a result, say, of station moves or equipment change.
This homogenisation is processed via a pairwise algorithm. NCDC explain:
In version 3 of the GHCN-Monthly temperature data, the apparent impacts of documented and undocumented inhomogeneities are detected and removed through automated pairwise comparisons of mean monthly temperature series as detailed in Menne and Williams . In this approach, comparisons are made between numerous combinations of temperature series in a region to identify cases in which there is an abrupt shift in one station series relative to many others. The algorithm starts by forming a large number of pairwise difference series between serial monthly temperature values from a region. Each difference series is then statistically evaluated for abrupt shifts, and the station series responsible for a particular break is identified in an automated and reproducible way. After all of the shifts that are detectable by the algorithm are attributed to the appropriate station within the network, an adjustment is made for each target shift. Adjustments are determined by estimating the magnitude of change in pairwise difference series form between the target series and highly correlated neighboring series that have no apparent shifts at the same time as the target.
Ronan contends that, because rural stations are heavily outnumbered by urban ones, the trends of the rural sites tend to be adjusted up to match those of urban. This, obviously, is the opposite of what should be happening.
It is, of course, the homogenised, adjusted temperatures which are then fed into the GISS and NCDC global datasets.
The paper gives two examples in detail of how badly the homogenisation process is going wrong.
Valentia Observatory, Ireland
Valentia is high quality, long running station in the SW of Ireland, with continuous records dating back to 1880. Situated in genuinely rural conditions, it should boast one of the best records for long term climatological purposes.
It will come as a surprise, then, to discover that the NCDC algorithm introduces a warming trend of roughly 0.4°C/century!
Did the program figure this out by comparing it to rural stations?
No, because there are no nearby rural stations that would have a long enough record to compare it to.
Instead, the program introduces the warming so that the Valentia Observatory record better matches the records of its urban neighbours!
Figure 36 shows the effect of these adjustments. Before homogenization (top panel), the Valentia Observatory record varies between periods of cooling and periods of warming.
However, after homogenization (bottom panel), most of the cooling periods have been eradicated, and the record shows an almost continuous warming trend since the end of the 19th century.
As a result, it makes the last decade or so seem “unusually warm”. In other words, it looks pretty much like the “global temperature trends”.
With Buenos Aires, the opposite occurs. Despite being situated in a highly urbanised environment, the GHCN homogenisation process makes no adjustment at all to temperatures there.
And why? Because nearly all of the neighbouring stations, that are used to “homogenise” against, are urban.
The full paper is here.