More Information on Minnesota Temperature Trends, Ada Mn.

Dave Dixon is doing his usual sterling work in revisiting the early charts he constructed for certain weather stations in Minnesota.  There will be a series of these but the first one is Ada, Minnesota, up in the northern part of the state.  You may recall from the earlier work that Minnesota, which is a long state, showed an interesting pattern, with warming in the north, no real change in the middle and cooling in the southern part of the state.  We wanted to explore this pattern further.  The charts are constructed as “anomalies”.  You select a baseline, in this case 1980 to 2010, and see how the temperature in any particular year varied from that baseline.

Here is the big picture in Ada.  The only change in temperature is in the minimums, which typically occur during the night and may reflect greater cloud cover leading to less cooling.  More clouds may in turn just be part of a natural cycle.  I have seen no good explanation for why CO2 increases would lead to higher minimum temperatures but not higher max temps.  In Minnesota, higher minimum temps would be welcomed by most of the population.

What is also apparent for Ada is that for the last 50 years there is no trend in the minimum raw temperature data.  Even with the adjustments there is little trend.  There was an uptick in the years before that.  For maximum temperatures, over the entire period there is cooling and in recent decades it appears to be flat.  Adjustments miraculously turn that into slight warming.  The average unadjusted temp is also flat in recent decades and has slight warming over the entire period.

There are a lot of unanswered questions in the data and data gathering methods, including such basic things as how the thermometers work and at what point did they begin to automatically record high and low temps for a day, regardless of the time of observation.  That would appear to be the soundest approach.  For any given station there can be lots of unique issues about the reliability of a thermometer, siting of the thermometer, who records the data, how it is transmitted, etc.  In addition, a lot of concerns have been raised about the adjustments to raw data, particularly the time of observation adjustment, which made some potentially wrong assumptions about how the adjustment should be done, and the homogenization adjustment.  Skepticism is warranted because somehow these adjustments always seem to produce an increasing temperature trend.

Dave’s notes:

• Recently we had reexamined temperature trends for several MN weather reporting stations, here: https://healthy-skeptic.com/2024/08/20/real-temperature-trends-in-minnesota/. This past effort used adjusted daily data as published by NOAA. However, our prior data source was missing data from some Minnesota stations of interest. This lead us to explore of data sources, resulting in the analysis presented here today.
• The weather stations of interest are part of the United States Historical Climatology Network (USHCN). These stations are claimed to have been “selected according to their spatial coverage, record length, data completeness, and historical stability.” Note that these stations, while tending to be rural, are not claimed to be free of environmental influences such as building, pavement, etc. See here for more information: https://www.ncei.noaa.gov/products/land-based-station/us-historical-climatology-network
• NOAA publishes data for each USHCN site here: https://www.ncei.noaa.gov/pub/data/ushcn/v2.5/. For each site there are data files for monthly averages of the daily minimum (tmin), average (tavg), and maximum (tmax) temperatures. In addition, there are three versions of data for each of these measurements; raw, TOBS adjusted, and homogenized. The ‘raw’ data is presumably unadjusted, although we have not found any documentation stating this.
• The TOBS adjustment process is likewise not documented. However, TOBS (time of observation) adjustment is intended to correct for bias introduced by changing the time of observation from 2 PM to 11 AM during the 1950’s. An explanation we have seen claims that the older temperature measurement equipment stored the highest and lowest temperature experienced in the last 24 hours, which are reported as the prior days temperatures. Reporting temperatures at 2 PM, if the temperature were higher than the actual prior day’s maximum temperature, would cause maximum temperatures prior to the 1950’s to be higher than actually occurring. Higher older temperatures would cause the long term temperature rise to appear to be lower than actual. The TOBS adjustment counteracts this by effectively increasing the long term temperature trend. This adjustment is somewhat controversial, as seen by a simple internet search.
• The homogenized data is adjusted by comparing the temperature record of each station with nearby stations. It is said that this homogenization process can detect and correct for anomalies in temperature readings caused by changes in local environments. Homogenization also infills missing data. For the data for Ada, MN, presented here there are 12 years of missing data from  1904 to 1916. The homogenization process is described here: https://www.ncei.noaa.gov/pub/data/ushcn/papers/williams-etal2012.pdf and here: https://www.ncei.noaa.gov/pub/data/ghcn/v3/techreports/Technical%20Report%20NCDC%20No12-02-3.2.0-29Aug12.pdf. These references describe the adjustment process, and show that homogenization was performed on both TOBS adjusted and TOBS unadjusted data as part of the investigation. However, we have been unable to find any documentation stating whether or not the NOAA=published homogenized data has been TOBS adjusted or not.
• For the charts shown here, we downloaded the NOAA monthly data here: https://www.ncei.noaa.gov/pub/data/ushcn/v2.5/. For each data type we computed a monthly temperature anomaly by comparing the reported temperature to a baseline temperature. The baseline temperature for each data was computed from the 30 year period from Jan. 1981 through Dec. 2010. For example, the baseline minimum temperature for the month of Jan. was calculated by taking the average of the Jan. temperatures from 1981 through 2010. The anomaly then is then calculated by subtracting the baseline for each month from the actual monthly temperature reported by NOAA.  In order to make the charts more readable we then calculated an annual anomaly by calculating the average of the monthly anomalies for each month. A simple linear analysis was applied to the monthly anomalies to compute the long term trends for each data type.
• Fig. 1, Comparison of Annual Anomalies of Minimum Temperatures, Ada, MN USH00210018: This chart shows the trends in daily minimum temperatures, for raw (green curve), TOBS adjusted (blue curve), and homogenized data (orange curve). Note that the raw and TOBS adjusted data have missing data from 1904 to 1916, but this data has been infilled for the homogenized data curve. Note that the TOBS adjustment adds approximately 1 Deg. F of warming per century, increasing the warming rate from 3.00 Deg. F per century to 3.98 Deg. F per century. However, the final homogenized adjusted data has a warming rate of 2.50 Deg. F per century. It is interesting to note the relatively cold temperatures that can be seen early in the record, from 1893 to 1904, which drives a large portion of the overall warming trend. Meanwhile, there does not appear to be much warming, if any, since the 1980’s. It is curious that the TOBS adjusted data has a higher warming trend than the raw data, since published justification for the TOBS adjustment claims it corrects for warming bias in maximum temperatures.
• Fig. 2, Comparison of Annual Anomalies of Average Temperatures, Ada, MN USH00210018: This chart shows the trends in the daily average temperature, as reported by NOAA as monthly averages. Similar to the minimum temperature chart in Fig. 1, the TOBS adjusted data has 0.8 Deg. F higher warming trend than the raw data trend. In this case, the homogenized trend of 2.03 Deg. F per century warming is 0.69 Deg. F. per century higher than the raw data trend.
• Fig. 3, Comparison of Annual Anomalies of Maximum Temperatures, Ada, MN USH00211018: Finally, this chart displays the trends in daily maximum temperatures for the Ada, MN site. The raw data actually show a cooling trend of 0.26 Deg. F per century, adjusted to 0.36 Deg. F per century warming for TOBS adjusted data, and 1.57 Deg. F warming per century for homogenized data. The homogenized data is adjusted colder than raw for older dates, and warmer than raw for more recent dates, generating a warming trend. Infilling the missing 12 years of data also appears to contribute to the adjusted warming trend, since the older data is colder than more recent adjusted data.
• Analysis of this first site raises many questions about the data adjustment process. The TOBS adjustment increases the minimum temperature trend, which shouldn’t be affected by time of observation bias the way we understand it. The homogenization process greatly increases the warming trend in maximum temperatures.