Existing Climate Data Sources and Their Use in Heat Island Research

The heat island over Phoenix, Arizona, has been the focus of intense research, both because the location of the city and the characteristics of its environs are ideal for heat island analysis and because the city has experienced tremendous growth in recent years.³

Balling and Brazel (1987b) analyzed long-term temperature records from the cooperative network in Phoenix. Twelve stations were selected, all of which experienced location changes of less than a kilometer and elevation changes of less than 30 m over their period of operation. The locations of these stations in the Phoenix area are shown in Figure 2. Averages of daily temperature maxima and minima for the summer months of June, July, and August were examined for trends over the period 1949-1985. The significance of observed trends was tested using the Mann-Kendall Rank Statistic (Mann 1945, Mitchell et al. 1966). The findings of Balling and Brazel are presented in Table 8. Significant increases in summer averages of daily maximum and minimum temperatures are observed at eight and six of the twelve stations, respectively. Climate records at the three stations located in the center of greater Phoenix (Mesa Experiment Farm, Phoenix WSFO, and Tempe) all showed significant increases in both maximum and minimum temperatures, indicative of heat island effects. However, significant increases were also observed at Buckeye, Casa Grande Ruins, Florence, Gila Bend, and Sacaton, which are all distant from the urbanized areas of Phoenix.

In the same work, the authors examine daily maximum and minimum temperatures recorded between 1980 and 1985 at 35 sites. Amateur weather observers supervised by ASU performed measurements at 29 of these sites. These sites are predominantly located in residential areas and supplement data from the other six sites, all operated by the National Weather Service. Contour maps using this data reveal average heat island profiles over the Phoenix areas with high values 1 - 2K (2 - 4°F) for daily maxima and 3 - 4.5K (6 - 8°F) for minima.⁴ Distances as large as 3 km in the urban area and 15 - 30 km in the surrounding areas separate the measurement locations. This makes an interpretation of the linear contour maps presented by the authors problematic since the city environs, which are used to characterize synoptic weather characteristics and thus define the magnitude of the heat island, are insufficiently characterized.

The creation of the PRISMS network made possible a much more sensitive temporal and spatial characterization of the heat island of Phoenix. This network has been used for a small number of research works. Blumberg and Brazel (1992) demonstrated some potential uses of the network with two case studies. In one, they studied the synoptic weather patterns over the Phoenix areas during October 27 - 29, 1991, which helped interpret data from an AVHRR satellite overpass on October 28.⁵ Another work (Verville et al. 1992) compared temperature measurements made at 6 and 20 feet at both the Alameda PRISMS site and a residential site one mile away. They concluded that PRISMS site temperature measurements made at 6 feet above ground are approximately equal to temperature measurements made 6 feet above canopy height. In other words, the PRISMS measurements are relatively free of microclimate effects.

We performed an exploratory analysis of PRISMS data to study the heat island of Phoenix. Our analysis initially focused on a sample of data for the months of May and June in the years 1992 and 1993. We relied on the data quality control techniques used by researchers at ASU. All data that were flagged as questionable were removed from the data set and replaced with interpolated values between valid data before and after the questionable ones. We supplemented this data with insolation, wind speed and direction, and air temperature and humidity measurements at the General Motors Testing Ground north of Phoenix and with wind speed and direction, air temperature and humidity, and sky cover measured at Phoenix Sky Harbor Airport.

Analysis and Results

Synoptic weather conditions were characterized by averaging data collected at Corbell, Falcon, and Kay, three rural stations. Temperature data were averaged over the 31 days of May 1992 to create average diurnal temperature profiles both for each individual station and for the rural stations above. In the comparison of some urban and rural average diurnal temperature profiles, clear heat island signals were discerned. For example, night-time site temperatures were 1 - 1.5K (2 - 2.5°F) higher at Alameda, Arcadia, Pera, Falcon, Pringle, and Stapley, all located near the urban center of Phoenix, than at the three rural stations. The predominance of high average temperatures at stations near the city center and low average temperatures at rural stations agrees with expected heat island effects. This comparison is presented in Figure 3.

Figure 3: Urban verus rural temperatures at PRISMS sites. The thick solid line is the hourly running average temperature of six weather stations located near the urban center of Phoenix, Arizona. The dashed line is the hourly running average of three rural weather stations. The light solid line represents the difference in temperatures between the two averages.

While the average diurnal profiles discussed above exhibit clear heat island signals, the urban-rural temperature difference was found to vary from day to day. We began our analysis with Arcadia station, which is located within the Phoenix city center and exhibited clear signs of heat island effects. In Figure 4, the solid line is the mean urban-rural temperature difference at Arcadia for May 1992, averaged by hour. The variation in the urban-rural temperature difference, as measured by the standard deviation of the 372 measurements that are averaged in each hour (twelve readings per hour for 31 days), is 0.5 - 1K (0.9 - 2°F). The dashed lines in Figure 4 are one standard deviation from the mean. However, given the large number of measurements, the estimated standard deviation error in the mean hourly difference is only approximately 0.5K (0.9°F), shown by the dotted lines.

A clear and significant positive correlation was found between rural temperatures, which were assumed to be influenced by synoptic conditions, and the magnitude of afternoon urban-rural temperature differences. However, no correlations were found between urban-rural temperature differences and various measurements of wind speed, precipitation, or insolation. Continue to: Chapter IV: Conclusions and Suggested Directions for Future Work

Figure 4: Characteristics of population of differences between Arcadia and rural average temperatures for May 1992, separated by hour. The solid line is the population means. Dashed lines are one standard-deviation from mean. The dotted lines are the estimated errors of mean.

---

³The population of Phoenix has risen from 150,000 in 1945 to 1.8 million in 1984 (Balling and Brazel 1986).

⁴PRISM data neither indicate nor quantify the impact of the microclimate on the recorded local air temperatures. The microclimate can have an impact of as much as 2 or 3K (5 or 6°F) on local air temperatures (Smith et al. 1996).

⁵In another study, they examined the night-time heat island of Phoenix using contour maps made from interpolation of temperatures from all 16 sites.