Best Practices for Weather Station Accuracy: Lessons from Urbana, Illinois

“How accurate is your data?” 

This is one of the big, never ending, unresolved questions around using and interpreting atmospheric data. Of course uncertainties have existed and will continue to exist in all forms of environmental data. But in order to understand our weather history, and our weather future, there is a need to define, measure, and understand these uncertainties. 

There are obvious pieces in the weather observation process that can produce variations and uncertainties. These include

1. instrument accuracy
2. changes in station location
3. changes in observers
4. changes and/or malfunctions of instruments
5. changes in observation and reporting times
6. changes in exposure and site characteristics, and
7. variations in the height of measurements and ground cover.

But there are a couple of weather station locations in the United States that offer some best practices for keeping accurate records even during changes in location, observers, and instruments. One of these is the weather station in Urbana, Illinois. Looking more closely at the records and history of the Urbana station reveals how various changes in sites, instruments, and the station’s environment affect a historical climate record. It also encourages some best practices for maintaining accurate records even when changes have to occur. And this provides a foundation for looking at other weather stations to assess the long term accuracy of their records.

The Urbana, Illinois station

Climate data has been collected at Urbana, Illinois, since 1888. All together, these records form one of the state’s longest datasets. And while the Urbana station has been identified in the past as one of the best weather stations in the nation, it has undergone some pretty big changes. Fortunately, all of the alterations made in the station’s location and its instruments have been well recorded. This is key for maintaining accurate climate records. Importantly, overlapping temperature and precipitation measurements were made for each of the shifts in the station’s location and instrumentation. This makes it one of the few weather stations in the nation that has experienced dedicated dual record keeping over its entire period of operation. 

The operation and care of the Urbana station reflect several important climate monitoring principles:

1. the need for parallel testing of old and new systems
2. full documentation of each system and operations
3. assessment of data quality and homogeneity, and
4. maintenance of a long-term commitment to the observations and data quality.

In 1961, the nationally recognized quality and value of the station led the U.S. Weather Bureau’s top climatologists of the mid-twentieth century, Helmut Landsberg and Murray Mitchell, to select the Urbana station as one of 26 benchmark stations in the nation for monitoring the nation’s climate (Mitchell 1961). Of added importance is the fact that from the beginning, atmospheric measurements included a comprehensive set of conditions including humidity, wind, sky cover, pan evaporation, evapotranspiration (ET), and soil temperatures. 

Quality Observers and Station Changes

The high quality of data collection and recording of operational events at the Urbana station were related to those operating the station. The Urbana weather station was started by the University of Illinois on its south campus, operated as part of the new agricultural experiment station started in 1887. The College of Agriculture maintained the station, provided the best instrumentation, and leading scientists supervised its operations from its beginning in June 1888 until 1947. The ET and soil temperatures were measured in special devices designed and constructed by university staff. From 1948 to the present, the Illinois State Water Survey has maintained and staffed the station as part of its research and services program. In 1902 the Urbana station was identified by the U.S. Weather Bureau as one of its cooperative substations.

The Morrow Plots at the University of Illinois at Urbana-Champaign

In early 1897 the station was moved 600 feet from its original site and installed alongside the famed “Morrow Plots,” which have been agricultural experimental plots since 1876 and are recognized as a National Historic Landmark. Temperature and precipitation records were made at both sites for a year, revealing no differences. During 1899–1902 several instruments were added (rain, temperature, humidity, and wind). The weather shelter, which contained the liquid-in-glass thermometers, had been four feet above the ground from 1888 to 1897, but in 1898 the thermometers were installed in a shelter on a platform 10 feet above the ground where it remained until 1948 (Changnon and Boyd 1963). In 1948 the weather shelter was again mounted at a standard height, 4 feet above the ground, and comparisons were made of temperatures at both heights for a year. The station’s recording rain gauge, a tipping-bucket type, was changed in 1948 to a weighing-bucket-style gauge. In 1984 the weather station was established at a rural setting 1.3 miles southwest of the 1897-1984 Morrow Plots site. All forms of weather data were collected at both the Morrow Plots site and the new location for 3 years (1984-86) to measure differences.

One of the factors that can affect the accuracy of weather data are the weather observers who collect the data. Since the inception of the Urbana station in 1888, there have been skilled scientists supervising the station and its operations. The institutional staff employed and trained the station’s observers. There is little reason to suspect that observer errors are a problem in the historical data of Urbana. Other elements that may impact data are the timing of collection, and the land cover around the instruments. At the Urbana station, observation times have been in the morning hours, a condition that is unlikely to affect the continuity of the values over time. The land cover at all station sites was grass. All changes in instrumentation were also carefully monitored, so as to uncover any variations in data. This includes taking care to account for major alteration in precipitation and temperature instrumentation, as detailed below. 

Precipitation

The official daily precipitation was measured in standard 8-inch diameter non-recording rain gauges using measuring sticks from 1888 to the present. All gauges were at the same height above ground, and all gauge exposures were relatively open and very similar throughout the station’s 118-year history. There are no instrumental or siting reasons for temporal shifts in the official precipitation values. The 1897 station relocation was checked by data collected at both sites and the difference was less than 1%. The 1984 relocation was also checked and the 3-year difference in amounts was 3%, which was within the precipitation variability found across a local dense rain gauge network.

The biggest changes in precipitation records took place in the rain gauge data. The tipping-bucket gauge used from 1902 to 1948 produced lesser amounts at high rainfall rates than did the weighing-bucket gauge used since 1948. This is because the tipping mechanism in such gauges was slow enough that it recorded lesser rain values when heavy rainfall rates occurred than the weighing mechanism in bucket-type gauges. Measuring the differences during heavy rainfall rates of 1 inch or more per hour showed that the tipping bucket measured 15%- 35% less than the weighing-bucket gauge (Kurtyka 1953). The heavier the rainfall, the bigger the difference was between the two gauges. This is a factor in the noted increase in heavy rainfall amounts that have occurred since 1940-50 in Illinois (Huff and Changnon 1987). The shift from the wide use of tipping-bucket gauges to weighing-bucket types by the Weather Bureau occurred during the 1940-50 period, and the 1-day, 2-year rainfall frequencies in Illinois showed an increase of 17% from 1901-40 to 1941-80. Unlike the rain, however, annual snowfall amounts during 1888-1988 at Urbana showed no sign of changes in measurement and observations due to station shifts (Changnon 1990).

Temperature

The annual mean temperatures showed no effects from changes in station sites, instrumentation, and urban environment. The curves show 1) a major temperature increase that began in the 1920s; 2) a peaking of values from 1933 to 1956; 3) a rapid descent in values from 1957 into the 1960s; and 4) fluctuations around a rather constant value for 1961–2004.

But again, there were changes in instrumentation that had a measurable impact. 

Liquid-in-glass (LIG) thermometers were used for maximum and minimum daily values from 1888 until 1988 when maximum-minimum temperature sensors (MMTS) became the official Weather Bureau method for obtaining the daily maximum and minimum temperatures at Urbana and most other cooperative stations (Quayle et al. 1991). This shift resulted in a change in the Urbana values with maximum temperatures from MMTS being 0.6°C lower than LIG values, whereas the minimum MMTS values averaged 0.06°C higher than LIG values (Wendland and Armstrong 1993). The shift in average daily mean temperature was 0.3°C lower.

Another factor was the shift in shelter heights from four feet to 10 feet from 1904 to 1948. This change resulted in a lowering of annual mean temperatures by 0.17°C. This was determined from measurements of temperatures taken during 1947-48 in the shelters at each height.

One other key piece that affected air temperatures at Urbana was the local growth of Champaign and Urbana communities. In 1888 the local total population was only 7700 and located more than half a mile from the weather station. By 1960 the population had reached 100,000 and the urban settlement had grown very close to the Morrow Plots area. The potential for effects from the urban heat island was real. Fortunately, deep soil temperature data were collected at a nearby rural site during the 1889-1952 period, allowing assessment of the heat island effect on local air temperatures. A study of the air and soil temperatures and their differences over time revealed that the urban heat island had increased the annual mean air temperature by 0.7°C from 1900 to 1952 (Changnon 1999). Ensuing growth from 1952 to 1980 had less effect and raised the annual mean temperature by another 0.2°C by 1980.

In 1984, the Urbana station was moved again, this time to a more rural area. Again, the data were monitored closely for any changes. This took place over a 3-year period, as measurements were taken at both the Morrow Plots site and the new site. This produced a reduction in values with annual mean temperatures being 0.8°C lower. This new rural site effectively eliminated the urban heat island effect.

The overlapping measurements made during each change in station location allowed observers to identify what shifts were due to sites and equipment, and what was due to climatic change. The changes in annual temperature (in degrees Celsius) for each are

• higher shelter from 1898 to 1948 lowered by 0.17°,
• urban heat island from 1900 to 1983 increased by 0.9°,
• rural setting from 1984 to the present lowered values by 0.8°, and
• shift to MMTS from 1988 to the present lowered values by 0.3°.

These are illustrated for the 118-year record using a curve constructed to reveal the decrease in 1898 (higher shelter), the continuing increases thereafter (urban heat island), the increase in 1948 (lower shelter), the sharp decrease in 1984 (station moved), and the decrease in 1988 (MMTS installed). These adjusted annual values were applied to the actual measured annual temperatures to create an “adjusted” curve shown below, which also has the unadjusted curve based on the annual mean temperatures recorded for the 4-year periods from 1889 to 2004.

From 1989 to 2004, the annual values were raised by 0.3°C to adjust for the MMTS effect. The effect of the urban heat island during 1900-80 was used to lower values but it was partially reduced by the increases due to MMTS and the lower temperature resulting from the higher shelter during 1898-1947. The adjustments to the time series produced some important changes in the interpretation of the multidecadal variability at this site. Although 1953-56 was the warmest 4-year period, several recent periods were nearly as warm and comparable to other warm periods of the 1930s. Also, warming temperatures after the cool period of 1961-80 were not evident in the unadjusted actual time series.

Urbana’s Best Practices for Long term Station Accuracy

Analysis of the extensive Urbana station climate records revealed several significant changes in temperature values from 1888-2005. This has been shown to be a result of instrument shifts, a station relocation, and urban growth. The official precipitation amounts were measured over the 118-year period by the same type of non-recording rain gauges and at similar sites, leading to no human-made shifts. A change in the type of recording rain gauges in 1948 produced a shift in the sampling of heavy rainfall, leading to increased heavier rainfall amounts.

Few other long-term U.S. stations are apt to have records that compare with those at Urbana. Particularly important to the Urbana assessment is the thoughtful keeping of time-overlapping measurements when changes were made. These have allowed quantification of the changes in values resulting from shifts in instruments and station locations. 

Most importantly, they allow one to understand the uncertainties inherent in the historical climate record everywhere. These simply cannot be measured because of the lack of detailed record keeping and the important collection of dual data for shifts in instruments or sites. Comparable studies should be made of other weather stations that possess quality records documenting station location and instrument shifts plus overlapping measurements of temperature and precipitation when shifts were made.

This article has been edited specifically for the AMS Weather Band from a longer article by Stanley Changnon and Kenneth Kunkel. Any errors or omissions may be attributed to AMS staff. Copyright remains with the AMS.