Proposed Reference Solar Spectral Power Distributions for Miami and Phoenix From Three Years of Measurements

Henry K. Hardcastle III
Atlas Weathering Services Group


The solar spectral power distribution (SPD) represents a critical variable in natural and artificial weathering processes. Weathering research often requires reference SPDs representative of end-use environments or worst-case in-service conditions. Uses of SPDs include comparison of different environments, comparison of artificial light sources to natural sources, estimates for dose-damage models and characterization of materials wavelength sensitivity in natural environments. The impetus for this study was to develop more realistic reference SPDs than currently available for Miami, FL and Phoenix, AZ.

To date, the weathering industry has widely used ASTM E 891 and E 892 as reference SPDs. These models, however, are theoretical distributions based on the extraterrestrial solar spectrum (Air Mass 0). Very significant discrepancies are observed between these reference SPDs and actual measurements of SPDs at exposure laboratories. Atlas Weathering Services Group (AWSG) desired a set of reference SPDs representative of and derived from actual measurements performed in the reference exposure environments of Miami, FL (subtropical) and Phoenix, AZ (desert). Two criteria were important for this project; first, that the new SPDs be derived from actual repeated measurements, and second, that large and statistically robust numbers of measurements be utilized.

Scanning spectral radiometers were installed at exposure laboratories in Miami and Phoenix at a 45° angle from horizontal facing south throughout 1997, 1998 and 1999. Daily solar spectral data near solar noon was obtained and reduced to histograms and frequency distributions. The grouped modal values from these distributions represent the most frequently observed class of irradiance at specific wavelengths over the three-year period. By using the grouped modal values to construct reference SPDs for 45° south near solar noon, the project criteria was achieved. AWSG proposes the reference SPDs developed in this study be utilized in addition to ASTM E 891 and E 892 where applicable in weathering technology.


Instrumentation. The primary instrument used for this study was an SR18 scanning radiometer developed by The Smithsonian Environmental Research Center. The SR18 is an 18 channel scanning radiometer with a filter wheel of 18 interference filters providing 2nm-bandwidth resolution in the UV B range. The instrument measured a 180° field of view. The instrument was designed for continuous monitoring of solar radiation and collected solar UV B data in the 290 - 324 nm range over the three-year period of this study. The instrument transformed raw one-minute voltages to 12-minute average intensities. Data processing as well as scheduled instrument calibrations were performed at The Smithsonian Environmental Research Center.

Measurements. One SR18 scanning radiometer was located at AWSG's South Florida Test Service exposure laboratory near Miami, Florida (Latitude 25° 52' N, Longitude 80° 27' W) in the subtropical reference environment. Another SR18 was located at AWSG's DSET Laboratories near Phoenix, Arizona (Latitude 33° 54' N, Longitude 112° 8' W) in the desert reference environment. Both instruments were oriented at 45° from horizontal facing south throughout 1997, 1998 and 1999. An example of the daily data is shown in Figure 1.

Even though the instruments were calibrated and recorded data from 290 nm to 324 nm, only the data above 295 nm is considered in this study (295 - 324 nm). Several features of the data at wavelengths shorter than 295 nm indicated that signal-to-noise ratios may have reached unacceptable levels. Data at wavelengths shorter than 295 nm will be left for consideration in other treatments. Additionally, although the instruments recorded solar spectral intensities at 12-minute intervals, only three data points closest to solar noon were considered in this analysis. Due to spectral scans being taken on a local standard time schedule rather than a solar time schedule, the data are properly denoted as "near solar noon". Figure 2 identifies the three data points near solar noon for a clear and cloudy day in 1997 at the desert site.

Figure 1. Example of Daily Solar Spectral Power Distribution Measurements made in June 1997 in Phoenix at 45 degrees South.
Figure 2. Solar noon intensities at 315.6 nm on a clear and cloudy day.
Figure 3. Effect of wavelength calibration errors on measured solar intensities.

At frequent intervals (3-4 times each year) the instruments were returned to the manufacturer for calibration. From calibration to calibration, minor adjustments in the wave band centers were noted. In all cases these wavelength adjustments were less than ± 0.5 nm. It was important to understand the magnitude effect on spectral intensities measured due to these errors. As illustrated in Figure 3, the maximum effect on reported spectral solar intensity is estimated to be considerably less than the effects of local weather. This inaccuracy, however, is noted. The wavelength identifiers used in this study represent nominal values throughout the three-year study.

Results and Discussion

Scatter Plots

Initial data analysis involved plotting the entire population of near solar noon measurements for each wavelength on x-y scatter plots. These complex plots indicated effects of season (sun angle) and atmospheric trends (clouds) on solar intensities. Examples of these scatter plots are shown in Fiugure 4 for Miami and Figure 5 for Phoenix. The population of measured solar intensities for the three-year period at 315.6 nm near solar noon is plotted as a function of day of year. A collection of these scatter plots was generated for all 15 wavelengths considered for both Miami and Phoenix.

Figure 4. Scatter plot of solar intensities at 315.6 nm for 3 years in Miami.
Figure 5. Scatter plot of solar intensities at 315.6 nm for 3 years in Phoenix.

Histograms and Frequency Distributions

Although interesting, the scatter plots were difficult to apply and therefore were reduced to histograms. For each wavelength, the range of solar intensities was determined. The intensity range was then divided into ten classes. For ease of comparison, the same set of classes was used for both the Miami and the Phoenix data. The number of occurrences of solar intensity within each class was counted.

Each class count was then ratioed to the total count of measurements to obtain a frequency for each class of intensities within each wavelength. An example of this histogram frequency distribution is shown at 315.6 nm for Miami and Phoenix in Figures 6 and 7 respectively. A collection of these histograms was generated for the 15 wavelengths considered for both Miami and Phoenix.

Figure 6. Histogram frequency distribution of solar intensities at 315.6 nm for 3 years in Miami (n = approx. 2780).
Figure 7. Histogram frequency distribution of solar intensities at 315.6 nm for 3 years in Phoenix (n = approx. 2970).

Calculation of Grouped Modes

The distributions obtained displayed non-normal behavior. In all cases, the distributions were skewed. Researchers desired a statistic that would represent the most frequently observed intensity at each of the 15 wavelengths. For each wavelength, a grouped mode was calculated from the frequency distribution using the formula:

Mode = LMo + [d1 / (d1 + d2)] w


LMo = lower limit of the modal class d1 = frequency of the modal class minus the frequency of the class directly below it d2 = frequency of the modal class minus the frequency of the class directly above it w = width of the modal class interval

Table 1 shows the grouped modes calculated at the nominal wavelengths for Miami and Phoenix, the proposed reference solar spectral power distributions for Miami, FL and Phoenix, AZ - 45° South Near Solar Noon.


Miami Subtropical Environment Phoenix Desert Environment
Nominal Wavelength-nm Modal Value in milliwatts/m2-nm Nominal Wavelength-nm Modal Value in milliwatts/m2-nm
296.2 0.35 296.2 0.16
297.6 1.21 297.6 0.38
299.7 6.83 299.7 1.04
301.4 13.64 301.4 16.63
303.7 35.72 303.8 41.75
305.7 55.04 305.8 64.95
307.5 78.32 307.5 96.3
309.7 99.71 309.71 123.76
312.3 164.82 312.3 193.74
314.0 185.77 314.0 217.92
315.7 201.96 315.7 235.72
317.9 237.82 317.9 277.68
320.4 281.32 320.5 323.17
322.2 385.31 322.2 325.86
323.3 301.92 323.3 345.27

Proposed Reference SPDs

The data in Table 1 were then plotted in x-y format to obtain the proposed reference SPDs as shown in Figure 8 and Figure 9. These SPDs are properly denoted as hemispherical solar spectral power distributions for a 45° south facing surface near solar noon in Miami and Phoenix.

It is unclear why these distributions indicate that Miami has a higher grouped mode than Phoenix at wavelengths shorter than 300 nm. Errors due to wavelength calibration do not appear to account for this observation. One possibility may be that this is an artifact of signal-to-noise ratios in either of the two instruments. Another may be that these values are truly representative of differences between the environments. No matter what the root cause of this observation, users of this data are strongly cautioned about the values at wavelengths shorter than 300 nm until these data are successfully explained.


One application envisioned for the proposed SPDs is comparison of the most frequently observed values near solar noon in actual real world environments to values measured from artificial light sources and values measured at different conditions in the real world. Figure 10 shows a graphical comparison of several different types of data to the reference SPDs developed in this study.


This study enabled exposure laboratories in Miami and Phoenix to obtain reference spectral power distributions based on empirical measurements in the subtropical and desert reference environments. Sufficient measurements were considered (daily measurements for approximately 3 years) to constitute statistically robust data sets. The reference SPDs were derived from grouped modes and represent the most frequently observed class of intensities for the period of measurements. Only intensities near solar noon for wavelengths between 295 and 324 nm were considered for the 45° tilted surfaces facing south.

Sources of instrumental error appeared less than variation due to environmental factors. It is unclear why the Miami data displayed higher intensities than Phoenix data at wavelengths shorter than 300 nm. The proposed reference SPDs may be useful for comparing artificial light sources to real world measured solar intensities.


The UV B data reported in this paper were obtained by AWSG in partnership with The National Institute of Standards and Technology consortium on Service Life Prediction. AWSG would like to acknowledge Dr. Jonathan Martin, Consortium Director, for providing processed data, Larry Kaetzel of K Systems for facilitation in processing the SR18 measurements and The Smithsonian Environmental Research Center for instrument development and calibration services.

Figure 8. Proposed reference solar spectral power distribution for Miami, FL - 45
Figure 9. Proposed reference solar spectral power distribution for Phoenix, AZ - 45
Figure 10. Comparison of the proposed reference SPDs to other light sources in weathering technology.


  1. Storey, R.F.; Shoemake, K.A.; Chisholm, B.J. J. Polym. Sci.: Part A: Polym. Chem. 1996, 34, 2003.
  2. Storey, R.F.; Chisholm, B.J.; Brister, L.B. Macromolecules 1995, 28, 4055.
Related Links