Automated Surface Observation Systems (ASOS) installation is rapidly nearing completion. This is changing the way in which weather information is gathered. These changes will not only be felt by aviation interests and forecasters, but by anyone doing climatological studies. In order to do objective climatological studies one must understand the differences of ASOS temperature and dew point measurements as compared to SAO measurements taken previously at that location. Since temperature and to a lesser extent dew point are the most important meteorological measurements in climatological studies, this study has restricted itself to these two meteorological variables. Once differences between the two systems are discovered and understood, then climatologists and meteorologists can better interpret ASOS data.
This study examines 32 days of hourly data generated from both ASOS and SAO data gathering systems at the Evansville Regional Airport. Several statistical calculations were made to find the differences between the two and possibly discern the cause of the differences. Previous studies of ASOS-SAO comparisons, Lamberty and Noel (1994), Lashley (1994), and Woodworth (1994), indicated that ASOS temperature and dew point measurements were, on average, cooler/dryer than SAO temperature and dew point measurements. Lashley indicated that sensor location may be a factor explaining this difference. Griffith (1995) also investigated location as a factor in the explanation of the difference between high and low temperatures recorded by the two systems. Griffith's study restricted itself to examination of these daily extremes. This study examines hourly data and uses a more critical approach in an attempt to discover possible explanations for the difference.
Official weather records began to be kept in Evansville on January 1, 1877 at an unknown site in the city. This continued until November 30, 1897. The point of observation was then moved to the old Federal Building in Evansville at Second and Vine Streets where it remained through the end of 1915. The site was then moved to the Citizens National Bank, located at Forth and Main,
January 1, 1917 until June 30, 1926. On July 1, 1926 it returned to the Old Federal Building where the observations were taken until August 30, 1940. It was at this time that the office was moved to an airport site, the Aerological Building at the Municipal Airport.
On February 21, 1951, the official site was moved to Dress Regional Airport on the far north side of Evansville where it stayed until the weather service office closed January 31, 1996. The site of the temperature/dew point observation was, however, moved on March 3, 1992, from center field of the airport to its final location 10 yards east of the former weather office.
The National Weather Service ASOS site at Evansville, Indiana is located approximately 1.1 miles northeast of the weather office, about 20 yards to the right of the approach of Runway 22. The surrounding area is grassland. A large hill is clearly visible about 500 yards southeast of the ASOS site. The entire airport slopes up at a rate of 3 feet per 22 yards as one walks north (Figure 1).
Figure 1. Map of Evansville Regional Airport with NWS and ASOS Sites.
The HO83 hygrothermometer used for the SAO observations was located about 10 yards to the east of the former weather office, 5 feet away from the office parking lot, in a small grass plot. The parking lot is asphalt and used by cars and small trucks. The surrounding area, adjacent to the old hygrothermometer, is flanked by buildings to the west and the airport field to the east.
ASOS and SAO temperature and dew point data sets were gathered over a 32 day period extending from October 3 to November 4, 1995. All data sets were divided in four ways: 1) whole data sets, 2) daytime observed data (6 a.m. to 6 p.m.), 3) nighttime observed data (6 p.m. to 6 a.m.), and 4) each data set was divided into 32 individual sets consisting of the hourly data for each day (midnight to midnight) from which a daily average was calculated. Daily means one day's hourly data.
Once these divisions were made, an average was calculated for each data set. Then for each pairing, the SAO average was compared with ASOS. The SAO average was subtracted from the ASOS to get the difference, if any. This order was chosen because SAO data was the official reported data used at the time of this study. All comparisons in this study have been done in this way. The difference between each individual ASOS and SAO data point was calculated for all data sets and their respective divisions.
From this information, an average difference was calculated along with the standard deviation of these differences. The standard deviation is a statistical measure used to quantify the variability of the individual data points from the mean (average). The correlation coefficient (CC) was then calculated for ASOS vs. SAO data for each of the four data divisions. The correlation coefficient measures pattern similarity/dissimilarity between ASOS and SAO data. For a more detailed discussion of these methods consult Neter (1988), Wasserman (1988), Whitmore, (1988). It is understood that small sample sizes will have some effect on the results of this study. Since the data do not meet criteria required for a random sample, statistical significance tests could not be performed. The above statistics were used because a random sample is not required.
Two tables were prepared to display the results of the investigation. Each was divided into four sections, one for each data division. Part A concerns statistics calculated for the whole ASOS and SAO data sets. Part B displayed statistics calculated for daytime data. Part C displays the statistics calculated for nighttime data and Part D displays statistics calculated for the daily average data. Table 1 is devoted to temperature data, Table 2 to dew point data.
Table 1 shows that ASOS temperature measurements were cooler than SAO temperature measurements in all circumstances, except at night. In that case, ASOS temperatures were warmer than SAO temperatures. Correlation coefficients were excellent in all the temperature data sets examined, ranging from 0.933 (Part A) to 0.993 (Part D) showing a nearly perfect pattern similarity between ASOS and SAO temperature data.
The mean difference calculations in Part A indicated that the average ASOS temperature reading was 1.72 oF cooler than the average SAO temperature. The variability of the differences was 4.4oF (Part A), which decreased some at night (Part C). When the daily averages were examined (Part D), variability, as measured by the standard deviation, decreased significantly. Variability is important because it is a measurement of the consistency of the difference. Therefore the more variable these differences, the more unreliable the mean difference.
The dew point data (Part A of Table 2) indicated that ASOS measured dew points were, on average, 0.65oF dryer (cooler) than the SAO's. Correlation coefficients varied from 0.956 (Part B) to 0.991 (Part C) which denotes an almost perfect pattern similarity between ASOS and SAO data, better than that for temperature. Dew point measurements also showed less variability ranging from 3.45oF (Part B) during the day to 2.58oF (Part C) at night.
Daily averaged data (Part D of both Tables 1 and 2) revealed a much smaller variation around the mean difference as measured by the standard deviation than that calculated using hourly data which is due to the averaging process. Statistics calculated using hourly data will highlight the detail of the variability of each individual hourly data point.
Figure 2 and 3 depict the differences between the daily averages of both SAO and ASOS temperature (Figure 2) and dew point (Figure 3) data. Clearly the average daily difference between the two systems appears to be small.
Both temperature and dew point data reveals that ASOS (temperature/dew point) data (Part A of both Table 1 and Table 2,) is (cooler/dryer) than SAO (temperature/dew point) data. However, when the temperature data is broken up into day and nighttime sets, ASOS temperature data was cooler only during the daytime hours. Fluctuation, measured by the standard deviation of the nighttime data, for both temperature and dew point temperature data was also less.
The question is what is the cause of different nighttime results? Equipment differences are ruled out because those differences would be systematic, hence time independent. One likely cause would be the location of the sensors. The ASOS temperature/dew point sensor is located in an area surrounded by grass and a hill to the southeast. The sensor used for temperature and dew point measurements in SAO reports was located near a parking lot. The ASOS site location, dominated by vegetation, will have a temperature that is cooler by day and warmer by night than a vegetation free site. This has been determined in micro climatological studies (Geiger 1961). Therefore, the ASOS site is conducive to cooler temperatures during the day and warmer temperatures at night. The SAO site, dominated by asphalt, is conducive to warmer temperatures during the day due to rapid heating of asphalt by sunlight.
To further emphasize the influence of the two sites, the parking lot near the SAO sensor serves as a heat source which does not exist at the ASOS site. The differences between the two sites, as measured by the mean difference and standard deviation, were greater during the day than at night possibly due to the presence of the parking lot. In fact, the temperature difference reversed itself at night indicating the loss of the influence of the parking lot's heat source as well as the influence of the vegetation surrounding the ASOS site. The above results show the primary factor causing the differences between SAO and ASOS data appears to be the location of the sensors used in making the measurements.
The following conclusions can be made from the results.
The author would like to thank the staff at WSO Evansville, Indiana and NWSO Paducah, Kentucky for valuable incites need to prepare this paper. Special thanks goes to Mr. Ernie Block of WSO Evansville, Mr. Pat Spoden SOO, Mr. David Humphrey, forecaster and Mr. Dennis Sleighter of NWSO Paducah for their input in furthering the investigation and examination of this study.
Geiger, R., 1961: The Climate Near the Ground, Forth Ed., Harvard University Press, Cambridge, 611 pp.
Griffith, J.R., 1995: A Comparison of High and Low Temperatures Measured by ASOS and the HO83 Hygrothermometer at Grand Island, Nebraska. Central Region Applied Research Papers, 15-01, 1-5pp.
Lamberty G.L. and J.J. Noel, 1994: Comparison of Automated Surface Observation System (ASOS) and Standard Aviation Observation (SAO) for Temperature, Dew Point and Visibility at Illinois NWS Stations. Central Region Applied Research Papers, 13-07, 58-66pp.
Lashley, S.L., 1994: A Comparison of the Objective Observed Elements Between ASOS and Manual SAO's, Central Region Applied Research Papers, 13-08, 67-74pp.
Neter, J., W. Wasserman, and G.A. Whitmore, 1988: Applied Statistics, Third Ed., Allyn and Bacon, 1006pp.
Woodworth, K.M., 1994: A Comparison of ASOS vs. Manual Observations at Detroit Metropolitan Airport, Central Region Applied Research Papers, 13-09, 75-85pp.