|National Solar Radiation Data Base User's Manual (1961-1990)|
4.0 Brief History of Solar Radiation Measurements in the United States
NCDC provided all of the meteorological data for the entire period of record. NCDC also provided solar radiation data that had been collected by NWS. Solar radiation data were also acquired from WEST Associates (a consortium of Southwest utilities), the University of Oregon, three DOE SEMRTS (Solar Energy & Meteorological Research & Training Sites) and the NREL Historically Black College and University (HBCU) network in the Southeast. Although more solar radiation data are known to exist, and the period of solar data collection by the NWS extends back to 1950, budgets and time did not allow for its use in Version 1.0 of the NSRDB. Periodically, solar radiation data from other sources may be acquired and added to the data base, creating updated versions.
Most of the meteorological data required for the data base were available from the TD-3280 tape deck files at NCDC. TD-3280 files contain data from surface airways hourly observations. Copies of NCDC archive tapes were sent to NREL, where the variables of interest were extracted. The units were changed to Standard International units, and gaps in the data record were filled using methods noted in Section 5.2.1.
Development of the NSRDB benefitted from a recalculation of precipitable water as part of a DOE-funded climate-change project at Columbia University. The calculations were performed at NCDC, under a subcontract from Columbia University, for the period from 1948 to 1988. These data were made available to NREL, through NCDC, for more than 70 stations in the United States and Canada. Calculations for 1989 and 1990 and for stations not included in the Columbia University project were done by NCDC under a subcontract from NREL.
The data, as received from NCDC, provided precipitable water within five atmospheric pressure bands (1013-1000 mb, 1000-850 mb, 850-700 mb, 700-500 mb, and 500-300 mb). The values in the five layers were summed to obtain total precipitable water from the surface to 300 mb. The data indicate that no more than 1 mm of precipitable water is likely to exist above the 300-mb level.
Autocorrelation of the total precipitable water data yielded values between 0.9 and 0.6 for lags of 12 and 24 hours, respectively. Based on this, linear interpolation between soundings was used to obtain hourly values of precipitable water. These hourly values were input to the model used to estimate solar radiation.
Snow depth data were used for estimating ground albedo. The high surface albedos resulting from snow cover increase diffuse radiation from the clouds and atmosphere. Under partly cloudy skies this can greatly increase diffuse radiation over the levels that exist when the ground is bare. Snow depth data were extracted by NREL from copies of TD-3210 archive tapes supplied by NCDC.
Ozone data are not generally available for most locations for the entire period from 1961 to 1990. However, ozone has a relatively small effect on the transmittance of solar radiation between 0.3 µm and 3.0 µm. Thus, monthly mean values of ozone for geographic regions defined by state boundaries and latitude (Alaska, California, and Texas) were used as input to the solar radiation model described in Section 6.0. The monthly mean data were obtained from surface and satellite data (Morris and Barras 1977 and Schneider et al. 1991). These values were assumed to be the same for each year, and were incorporated into the model as a look-up table.
In order to produce serially complete hourly values of solar radiation data, some meteorological data required as input to the solar radiation model had to be derived or created. Data were found to be missing for periods ranging from an hour to a year. The causes for missing data are many and include a NOAA cost-saving move in effect from 1965 to 1981, which called for digitizing only every third hourly observation. The NREL station selection criteria eliminated most stations with TD-3280 data gaps longer than one month, but some exceptions were found.
In addition to gaps in the data record, there were data elements that were not available at any time for some stations. For one element, aerosol optical depth, the available data were few in number and of uncertain quality. The manner in which each of these situations were handled is described in this section.
Short gaps in the data record, for those meteorological elements needed to perform model estimates of solar radiation, were filled by linear interpolation between data points on each side of the gap. Interpolated values were rounded to exhibit the same number of significant figures reported for the measured/observed data.
The definition of a short gap was a function of the known rate of change of the element. Sky cover, for example, was linearly interpolated over gaps as long as five hours. Based on an autocorrelation analysis that is described in Volume 2 of the NSRDB documentation, total precipitable water was linearly interpolated over gaps as long as 60 hours (five missed soundings). Interpolation between individual soundings was used to obtain hourly precipitable water data.
For longer gaps in radiosonde data, calculations of precipitable water were made using surface vapor pressure derived from surface temperature, relative humidity, and atmospheric pressure. When the surface data were also missing, long gap methods were applied to obtain the required data.
Long gaps in the TD-3280 meteorological data (sky cover, temperature, and relative humidity) were subdivided into two categories: 6 to 47-hour gaps and 48-hour to one-year gaps. For gaps 6 to 47 hours in length, data from adjacent time periods, for identical periods (e.g., beginning at 0600 and ending at 2300) were selected to fill the gap. These segments of data were adjusted to match the end-point values of the gap.
For gaps of 48 hours to one year, data from other years for the same time periods were selected to fill the gap. The selection was based on finding a year for which the data before and after the period of the gap had the best match with data before and after the actual gap. Best match was determined by characterizing three time slices for several days adjacent to the actual gap and comparing them to a corresponding period of time in candidate years. The larger the gap, the greater the number of days included in the characterization, up to four weeks.
No effort was made to fill gaps in the snow depth or present weather data. These discontinuous weather events did not lend themselves to any kind of interpolation or substitution methods. When missing, snow depth was set to zero.
There were times and many locations for which no precipitable water data were available from radiosonde soundings. From the work of Garrison and Adler (1990) and others, it was known that long-term monthly means of surface vapor pressure are well correlated with monthly means of total precipitable water.
Research conducted under this project showed similar correlations between hourly surface vapor pressure measurements and precipitable water calculated from individual radiosonde soundings. Therefore, surface temperature, relative humidity, and pressure data were used to derive hourly values of precipitable water for times and locations for which radiosonde data were not available.
Aerosol optical depth is, and will probably continue to be, the most difficult input parameter to obtain for models that estimate solar radiation at the earth's surface. Although aerosol optical depth measurements have been made at selected wavelengths using sunphotometers, these data are considered to have large uncertainties that become larger when extrapolated to broadband values for the entire solar spectrum (Hallaron 1982; Cachorro, de Frutos, and Casanova 1987; and Frohlich 1980). Furthermore, such data are only available for a limited number of locations for limited periods of time.
Given this situation, a decision was made to link broadband aerosol optical depth to the METSTAT model used to estimate solar radiation (METSTAT is described in Section 6.0). METSTAT algorithms were used to calculate direct normal transmittances for ozone absorption (TO), Rayleigh scattering (TR), absorption by uniformly mixed gases (TUM), and water vapor absorption (Tw). These were combined to obtain a value for molecular transmittance, TM,
IO = the extraterrestrial direct normal radiation.
where m is relative air mass.
This method produced broadband aerosol optical depths for locations and times for which measured direct normal data were already available. The calculated aerosol optical depths were then used to derive seasonal functions for estimating aerosol optical depths for any day of the year. A sine function was found to provide the best fit to the calculated values, as illustrated in Figure 5-1.
Coefficients for the sine functions were then mapped to establish climatic/geographic relationships. These relationships were used to define seasonal functions for all Primary and Secondary stations. Volume 2 contains more information on this derivation of aerosol optical depths, including the national maps of the coefficients for the seasonal functions.
As shown in Figure 5-2, the calculated aerosol optical depths clearly revealed the atmospheric loading that resulted from the volcanic eruption of E1 Chichon between February 28 and March 4, 1982. Results from Lidar measurements of the effects of other volcanic eruptions (Mendonca, Hanson, and DeLuisi 1978) were used to determine the effects of other volcanic eruptions from 1961 through 1990. Although the effects of Mt. St. Helens were undoubtedly very large at locations in the path of its plume, they were short lived. No clear signature of Mt. St. Helens could be derived from the available direct normal data so the effects of this eruption could not be included.
The effects of volcanic eruptions were used to form a look-up table of daily optical depth increases. The combination of values derived from seasonal functions and values from the volcanic look-up table were used to estimate aerosol optical depth for each day from 1961 through 1990 for each of the stations. These values were input to the model used to estimate solar radiation when measured data were not available.
4 Beer's law in its simplest form expresses the transmittance (T) of solar radiation passing through a medium as
T = exp(-M), where is the optical depth of the medium, and M is the optical path length through the medium. The law applies rigorously only for monochromatic radiation.
6.0 Model Estimates of Solar Radiation
Table of Contents