|John Christy on High Temperature Records in the US13-11-2013 17:35|
|Dr. John Christy has in his presentation to Climate Change Roundtable from 30th May 2013 shown a graph with the number of high temperature records measurements from 974 USHCN stations with more than 80 years of measurement history.|
This graph is part of his internet presentation with a mixed panel (pro and con AGW). The panel were quizzed on climate change by Rep David McKinley, a Republican congressman from the coal state of West Virginia consist of:
Scott Denning..............Colorado State University
Jim Hurrell....................National Center for Atmospheric Research
Joe Casola.....................Center for Climate and Energy Solutions
Annie Petsonk..............Environmental Defense Fund
Richard Thomas............West Virginia University
Marc Morano................Climate Depot
Dennis Avery.................Hudson Institute
Myron Ebell...................Competitive Enterprise Institute
For each station the daily maximum temperatures since 1895 have been recorded. A total of 365 days of the year, with 974 stations, there are 365 * 974 = 355,510 temperature readings every year.
Christy's graph on temperature records is shown below:
The visual impression of this graph is that you have a maximum of records around year 1935, approx. 8,000 per year. Then there is a decrease in the number of records to aprox. 1,000 in 1970. Since then the number of records has increased to approx. 3,500 records per year. But in a helicopter perspective, it seems that the number of records is fairly constant.
The graph could lead to various perceptions about temperature changes since there are no guidelines for interpretation of this statistics. Before going into this subject, I would like to discuss the number of records broken.
The number of records broken
A record is defined as an observation which is larger than all previous observations. This means that the first observation is a record, and the next observation is a record if it is greater than the first observation, and so on. A new record must beat all previous records, so the bar for setting a record is always moving higher. For this reason, in a time series of high temperatures records, the total number of high temperature records will grow with the years and you will have fewer new records per year.
Assuming that there has been no change in the temperature distribution for the individual stations during the years 1895–2012, one can show:
The number of records in the n'th year will be equal to 1/n.
The expected number of records after n years equals 1+1/2+ 1/3+...+1/n.
Using this we can a calculate benchmark for the number of records at a given year.
This means, for example, that if you started counting records in 1895 with 974 stations and 365 measurements per year, then we should see:
in 1920 expect 365 * 974 / (1920-1895 +1) = 13,673 records
in 1945 expect 365 * 974 / (1945-1895 +1) = 6,971 records
In 1970 expect 365 * 974 / (1970-1895 +1) = 4,678 records
in 2012 expect 365 * 974 / (2012-1895 +1) = 3,013 records.
One would expect fewer and fewer records per year as more years are observed.
The following graphs show the cumulative number of records and number of records per year at 974 stations with 365 measurements per year using the assumption that the temperature distribution is stable for all stations:
The key point being that in the absence of a long-term global warming trend, we would expect the number of record high temperatures to decline, not remain steady. The number of records broken is discussed by Gerald A.Meehl et al.
John Christy's graph is based on standing records. The definition standing record is given in page 3, John R.Christy where he is describing how to calculate the standing records: They are calculated as the
year-by-year numbers of daily all-time record high temperatures from 970 weather stations with at least 80 years of record.
In other words, this means that for example a high temperature record from 1895, which is beaten in 1995 is not included in the number of records for 1895, but only for 1995. Old records are deleted and transferred to last year. This changes the outlook of record statistics.
The expected number of records for the current year is calculated by 1/n, but there is a small detail, namely that the expected number of records is the same for all years. Old records are moved forward in time in such a way that the expected number of these records is constant. Statisticians will talk about a uniform distribution of the number of standing records.
For statistics with standing records, it will be easy for the human eye to detect deviations from a uniform distribution. That is good.
A caveat using this kind of statistics is that two consecutive standing record statistics are not comparable, since over time you will see a decrease in the number of records with a growing balance year.
The following graph shows a simulation of 974 daily series of measurements on the assumption that all things are equal:
The horizontal lines indicate benchmark for the expected number of records.
Christy's graph differs from what might be expected in relation to the above records as follows:
During the years 1895–1925, there are fewer records per year than expected.
During the years 1926–1935, the number of heat records are largely what one should expect.
During the years 1936–2012, the number of records varies around 3,500 per year.
Standing records are discussed in more details middle 2012 by John R.Christy
There are several issues in respect of data quality Christy does not inform us about. I am not claiming that these issues will invalidate the statistics completely. I am only raising questions:
The total number of years in the statistics is 128 years, but only stations with more than 81 are included. This means that for some years is that the probability of a new record is of 1/128, and from others 1/81. The latter probability is about 62.5% higher than the former.
If a measuring station with more than 81 years of observations included in the statistics starts no later than 1931 or else it ends no earlier than 1975. This means that before 1931 or after 1975 are fewer stations than for the period 1931-1975. Therefore is an inhomogeneous portfolio of stations. However this can be done by calculating benchmark for the number of standing records
USHCN monitoring stations are not spread equally over the U.S. There are many more stations in the northern and eastern parts of the United States than in the west. This could bias the statistics.
Christy does not tell us which version of the USHCN he is using.
Without that data, it will be difficult to reproduce his graph, and ensure that the 974 stations include all stations with greater than 80 years of duration.
There is a lot to do in order to use the quality of data to count the number of standing records correctly.
Reconciliation of John Christy's graph with other sources
I do not have access to John Christy's data so that I can verify the graph directly. However, I managed to find data from USCHN to qualitatively confirm the aforementioned graph. It is about 200 stations where temperature measurements are quality assured over the period 1911-2010, ie. 100 years of measurements. So there are 200 stations each with 365 measurements. This gives 200 * 365 measurements in total, to be distributed to 100 years, so that one can expect 730 standing records per. years at the steady state temperature development.
These data can be found here USCHN 200 stations with daily all-time record high.
The following graph shows the total number of standing records for the years 1911-2010, as they appear at the end of 2010 for the 200 stations.
The two graphs are obviously not identical, but they show approximately the same trend. There are plotted three horizontal lines. They represent 2.5% percentiles, mean and 97.5% fractile is, assuming that there is no development in number above records. The negative binomial distribution is used in the calculations. From these lines, it appears that there are significantly more standing records in 1930 than expected.
A reference framework
The challenge of using standing records in a climate debate is that it is difficult to set a general frame of reference , knowing the existence several heat waves. It is my assessment that global warming is a slow creeping and not a sharp sudden event that can be compared with the heat waves that can occur in the United States.
Based on One of On the Cause of the 1930s Dust Bowl that one can expect one to two heat waves like heat wave in the 1930-1940 in every century and there are examples of such a warm wave can last 38 years.
Using climate models, NASA has examined what the reason is for Dust Bowl. NASA believes that this is due to a combination of the sea water in the tropical Pacific Ocean west of the United States was colder than normal and that the sea water in the tropical Atlantic Ocean east of the United States was warmer than normal. This changed the jet streams and precipitation over the United States and has led to an elevated temperature in Midwest .Please refer to NASA Explains "Dust Bowl" drought . This pattern might be repeated at regular intervals.
It is easy to form the benchmark under the assumption that there has been no systematic temperature development and it is difficult to define a benchmark for standing records and broken records when taking heat waves into account.
It is likely that sometime in the future we will observe significant heat waves in the U.S. with temperature extremes as in 1930th partly because of global warming and partly because of meteorological conditions as above mentioned.
The is a challenge to distinguish a heat wave that is generated due to global warming and the classic naturally occurring heat waves.
The heat wave in 1930-1940 has had a significant impact on the standing records. In 1930-1940 there appeared a number of dust bowls that destroyed the lives of many people in the Midwest Wikipedia Dust Bowl. Below an icon of the consequences of the Dust Bowl is shown.
What is John Christy concluding from the graph?
John Christy has used the above graph to conclude:
"The clear evidence is that extreme high temperatures are not increasing in frequency, but actually appear to be decreasing.".
Please refer to John Christy: Senate Environment and Public Works Committee , pageno 3.
John Christy makes a comparison of extreme temperatures today with the extreme temperatures during the heat wave in the 1930th.It is not valid to do so. Today you have a natural variability plus an impact form global warming. In the 1930 you have a classical heat wave. Although John Christy mentions heat waves in 1930th, it keeps him back to draw the above conclusion.
How can valid information be taken out from the standing records?
To eliminate the effect of the heat wave in 1930th in a statistic with standing records ,one you could choose to start the statistics in 1940 or 1950. It might of course be argued that not all data is used, but with data from the 1930s, it is difficult to conclude anything.
When there is a heat wave of the future, it requires a detailed analysis using climate models to find out what the driving force behind the heat wave.
Gravette Measuring Station in Arkansas
As an example of a station where both the number of standing records and the mean is calculated, I have simply chosen Gravette Measuring Station in Arkansas . There is no conscious choice of this station other than it should be a station in Midwest. I have just clicked on the map and selected state AR and a measuring station: USCHN WEB interface.By looking at one station, it becomes easy to understand and therefore easier to discuss.
The following graph shows the number of standing records for the years 1991-2010. There is 365 measurements in total to be distributed over 100 years. This means that the expected number of these records is 3.65. On the graph, the expected number of standing records and 97.5 % percentile are plotted. 2.5 % percentile coincides with the x-axis. Poisson distribution used in these calculations.
It is clear from this graph that there is a significantly larger number of standing records in the years around 1935. In the follow years great the number of standing records is close to the expected ones.
From this graph, one could be led to believe that there has been no increase in temperature. However, there is a significant increasing mean temperature in the period 1898-2012. This is illustrated by the following graph.
GRAVETTE measuring station with this post involved in the climate debate. It will be interesting to see whether this station "goes from the ashes" into the "fire" in the climate debate.
What is expected in the number of standing records in a warming world?
My assumption is that the number of high temperature records will grow by a rising temperature and number of standing records will be skewed to the left. To investigate this further, I have used the average temperature for US 48 contiguous states for the years 1895 - 2013, as shown in the following graph:
The temperature curve for US 48 contiguous states originates from NOAA.
The following figure shows a simulation of the number of high temperature records for each station using the temperature curve as above. There is a wide range of options for making a simulation, but I have chosen a simulated normal distribution plus the temperature curve as above, but my guess is that others will qualitatively achieve the same results.
The horizontal lines indicate benchmark for the expected number of records. This figure shows how the number of standing records will develop as at the end of the years 1920, 1945, 1970 and 2012.
The figure as at the end of 1920 shows a constant number of standing records corresponding to a constant temperature.
The figure as at end of 1945 shows a growing number of standing records corresponding to an increasing temperature over the years 1920–1945.
The figure as at end of 1970 shows that there is a maximum around 1935 and then a decreasing trend similar to the decreasing temperature until 1965.
The figure as at the end of 2012 shows qualitatively the same as Christy's graph corresponding to a small peak around 1930 to 1945 and then a growing trend.
The number of standing records does not follow the temperature curve directly; however, if you look at the graph as at end of the years 1920, 1945, 1970 and 2012 and combine them, there is a connection.
I have used the graph of the number of standing records as Christy has shown. However, it would be desirable to be able to verify Christy's counting of the number of standing records. This means that one needs to know which data series used and the start and end dates for each temperature series. This could change his statistics of standing records.
It is believed that John Christy's graph on standing records is more or less consistent with data from USCHN.
There is a significant effect on the standing records from heat wave in the 1930th.
You cannot compare the extreme temperatures in 1930th with what is observed today, since different mechanisms are behind the heat wave in the 1930s and today.
It is not possible to immediately identify whether a heat wave is driven from the classic meteorological mechanisms or of the global warming. One must resort to the methods of analysis were used to explain the heat wave in the 1930th
It is my opinion that John Christy is misusing and misinterpreting the standing record statistics.
The graph used by Christy is not accurate and it could be misleading depending on the context it is used.
There is no clear message to the public in this graph and this is maybe the reason for using it.
In the climate skeptic circles, Christy's graph will certainly be used as a proof that there is no global warming. However, reality is that the temperature is increasing.
The occurrence of temperature records in the U.S. is also treated in Does cold weather disprove global warming?
Records statistics are of great interest to the public since any record attracts attentions from the news media. In many cases they will produce nice pictures on TV and there might be large consequential losses. However, one will lose a lot of information by only looking on the record statistics in the discussion about global warming. The whole picture of data must be used.
However, I would like to thank Christy for bringing an important subject up for discussion. Hopefully, I have clarified the use of standing records in the discussion on global warming.
|global temperature||4||15-11-2020 04:08|
|Darwin Airport homogenizing of temperature data||36||20-10-2020 20:28|
|New High Temperature Records Again Outpace Lows||31||02-09-2020 23:24|
|New Temperature High Recorded in Death Valley||19||19-08-2020 02:27|
|Another week, more new high temperature records||11||17-08-2020 01:52|
|John McCain: Remarks on Climate Change Policy|