|National Solar Radiation Data Base User's Manual (1961-1990)|
2.0 Data Base Product Options
A sample printout of data in the NSRDB synoptic format is shown in Figure 3-1. When data for multiple years are requested in a single file, the years will follow each other hour by hour with no end-of-file until the end of the last year. The first record in each file contains the WBAN number, city, state, time zone, latitude (degrees and minutes), longitude (degrees and minutes), and elevation (meters). Time zones are indicated in terms of the number of hours by which the local standard time lags or leads Universal Time (UT). For example, Mountain Standard Time at Albuquerque is designated as -7, to indicate a lag of 7 hours from UT (i.e., a UT of 1700 corresponds to a time of 1000 [10 AM] in Albuquerque). The field positions and definitions of these header elements are given in Table 3-1 along with the FORTRAN format required to read the header.
Following the header, there are 8,760 or 8,784 (for leap years) hourly data records in each station-year. Each of the elements found in each data record are defined in Table 3-2. The FORTRAN format for reading the data records is given at the bottom of Table 3-2.
RReDC Note: The solar energy portion of the synoptic format (Hourly Data Files) is available on the RReDC. The full synoptic data set can be ordered from the National Climatic Data Center (NCDC).
TD-3282 is a special version of the TD-3280 format designed for the NSRDB. TD-3280 is the archival format used by NCDC for Surface Airways Hourly data. These weather observations, made in support of aircraft operations, have been taken since the earliest days of commercial aviation. They include data from NWS and military stations.
In contrast to the synoptic formats, which contain all elements within each hourly data record, TD-3280 and TD-3282 could be described as element interleaved formats. Each record contains hourly data for one day and one element; all elements for each day are grouped together or interleaved. A TD-3282 station-year file contains all of the data for one station and one year. You also have the option of requesting the data for multiple years in a single file.
The record for each element begins with an identification (ID) portion, 30 characters in length, as shown in Figure 3-2. This is followed by the data portion of the record, which consists of the time, the sign of the data, the data value, and two flags, repeated as many times as necessary to contain one day of record. For TD-3282, all records for all elements contain 24 hourly values and are 322 characters in length (including a four character control word and the 30-character ID portion). The control word is used by the computer to determine record length. The data are blocked in lengths of 6,000 characters, each block containing data for all elements for one day.
The fields in the ID portion of the record are defined in Table 3-3. The data element type codes, data positions in the fields, and data definitions are given in Table 3-4. Flag 1 (FL1 in the data portion) is the data source flag and Flag 2 (FL2) is the data uncertainty flag, both of which are defined later in this section. Definitions of the TD-3282 units codes are given in Table 3-5.
One of the advantages of the TD-3282 format is the flexibility it offers in ordering data. You can order only those elements of value to your application. For example, if global horizontal solar radiation data for July 1988 at Denver, Colorado are all that is needed, then your order can be specifically limited to that data.
The information needed to view and use the statistical products from the NSRDB is provided in this section.
Figure 3-3 is an example of the daily statistics as they will appear on a "wide-screen" display2 or wide hardcopy print from the file. The header identifies the station by a five-digit WBAN number, city, and state. The header identifies the time zone (TZ) for the station by indicating the number of hours by which the local standard time lags (-) or leads (+) Universal Time (UT) (e.g., Eastern Standard Time is designated as -5). Latitude and longitude for the station are given in degrees and minutes, station elevation is in meters, and the mean atmospheric pressure is given in millibars.
The next line in each file identifies the year(s) for which the next section of data applies. In each file, the first group of data provides daily statistics for the 30 years from 1961 to 1990. Statistics for each year for the entire period of record follow. As an example, Figure 3-3 gives 30-year statistics for Albuquerque, New Mexico. January and annual statistics are also shown for 1961 and 1990. The FORTRAN formats for reading the header, year(s), and data records are given at the bottom of Figure 3-3.
The standard deviations of solar radiation elements (e.g., SDGLO) for individual years and for the period of record (61-90) are not the same. For individual years, the standard deviations provide a measure of daily variability. For the period of record, the standard deviations provide a measure of the interannual variability of monthly and annual averages.
The NSRDB mean values for meteorological elements may not be identical with the means published by NCDC in the Annual Summaries of Local Climatological Data. The small differences expected are the result of different computational methods and differences in methods used to replace missing data.
The hourly statistics are presented in the form of means and standard deviations of the hourly values for each hour (from which diurnal profiles can be formed) and distributions generated by binning hourly values to determine the number of hours for which the radiation fell within twenty four 50-Wh/m2 ranges (e.g., 0-50, 50-100, 100-150,..., 1100-1150, 1150-1200). The bin data have been normalized to indicate the percentage (in tenths) of all daytime hours for which the radiation fell within each bin. Figure 3-4 provides an example of the hourly statistics as they will appear on a "wide-screen" display or a wide printout of the file. The header information is the same as that used for daily statistics except that the year(s) represented by each file has been added as the last field of the header.
Following the header, the next record identifies the solar radiation element (e.g., global horizontal radiation in Wh/m2) and the statistic (e.g., means) for which the following data records apply. Each file contains data for all three solar radiation elements and each of the three statistics, as indicated in Figure 3-4.
The first two fields in each data record designate the month (13 indicates annual statistics) and the source and uncertainty flags that apply to each monthly profile and distribution.
The data fields for means and standard deviations contain these hourly statistics in Wh/m2 for each of the 24 hours of the day. Values for hour 01 represent the mean or standard deviation of the total radiant energy measured from midnight (2400) to 1:00 AM (0100). The data fields for the distributions designate the percentage of hours, in tenths of one percent, for which the average radiation fell within the 50-Wh/m2 bins described above.
The FORTRAN formats for reading the header, statistic identification, and data records are given at the bottom of Figure 3-4.
The persistence of weather events and the effect this has on the availability of solar radiation energy can affect many solar energy applications. In particular, the persistence of solar energy can affect energy storage requirements and the need for backup energy sources.
A persistence statistic calculated for the NSRDB is the number of sequential days (runs) in a month during which the daily total solar energy exceeded or fell below 12 energy thresholds (see Figure 3-5). The run lengths vary from 1 to 15+ days. The total number of runs over the entire 30-year period from 1961 to 1990 were determined for each month. The decision to compute persistence on a monthly basis resulted in the truncation of runs at the end of each month. Although this procedure caused some distortion of the statistics, it provided important information on seasonal changes in persistence.
An example of the persistence statistic is shown in Figure 3-5. The thresholds for diffuse horizontal radiation (0 to 5,000 Wh/m2 in 11 steps) are one-half those used for global horizontal and direct normal radiation (0 to 10,000 Wh/m2 in 11 steps). The header information gives the WBAN number, city, state, month, and the solar radiation element.
Each number in the matrices on the left side of Figure 3-5 gives the total number of times the daily total solar energy exceeded the threshold indicated for that row for no more or no less than the number of days indicated for that column. Each number in the matrices on the right gives the total number of times the daily total solar energy was less than the threshold indicated for that row for no more or no less than the number of days indicated for that column.
The numbers in these matrices can be used in a variety of ways. For example, the sum of all the numbers in the sector enclosed by the dotted line box on Figure 3-5 indicates that the daily total global horizontal energy fell below 6,000 Wh/m2 for four or more days, 31 times from 1961 to 1990. These matrices can be sectored in any manner that produces information useful to specific applications.
The FORTRAN formats for reading the headers, number of days, thresholds, and numbers of events are given at the bottom of Figure 3-5.
Quality flags are attached to each hourly solar radiation and meteorological element. These flags provide information on the source and uncertainty of a data element, allowing each user to evaluate its usefulness. The flags are further described in the following sections.
Two flags are used to define the quality of each solar radiation element. The first flag gives the user information about the source of each hourly value for each solar radiation element, including the methods and input data used to derive model estimates. Solar radiation source flags are defined in Table 3-6. The flags are ranked roughly from highest quality to lowest quality data. However, this ranking may not hold for an individual datum. For example, if the quality assessment of data from a station measuring all three elements of solar radiation (Source Flag A) shows a large probable error in the data, then a large uncertainty will be assigned to the hourly values for each element. This might be a larger uncertainty than that assigned to a modeled value with good quality input data.
The second flag designates the uncertainty attached to each hourly value. Uncertainty as used here provides an estimate of the interval around a measured or modeled value within which the true value will lie 95% of the time. The flags for each interval are defined in Table 3-7. In Version 1.0 of the NSRDB, no flags as low as 1 or as high as 9 were assigned. A special meaning defined in Section 8.3.2 is attached to the uncertainty of modeled data.
In general, the uncertainties assigned to measured solar radiation data show considerable variability because of instrument failures and human factors. The uncertainties assigned to modeled data will be higher, on average, than measured data, but will show lower variability for a given source category because the model is applied uniformly at all times. Only changes in the uncertainty of the input data will significantly affect the uncertainty of modeled data.
The flags which originally accompanied most of the NWS meteorological data have not been incorporated into the NSRDB. Instead, the source flags defined in Table 3-8 have been assigned.
The resources available for producing Version 1.0 of the NSRDB did not allow for a quantitative evaluation of the uncertainty of the meteorological elements. Rather, the relative uncertainties defined in Table 3-9 were assigned.
The source and uncertainty flags for meteorological elements are not included in the NSRDB synoptic product. They are included in the TD-3282 product.
The dominant source and uncertainty flags of the hourly data are assigned to each of the daily and hourly solar radiation statistics described previously. These flags provide a limited measure of the quality of the hourly data for each station-month and station-year. The quality statistics described in Section 3.4.4 provide a more comprehensive assessment of the quality of data available for each station in the data base.
An example of the quality statistics that are available for each station is given in Figure 3-6. The header for the quality statistics files contains the same information as does the header for daily statistics. Each record contains the percentages (in tenths of one percent) of the hourly values for each year to which the indicated source and uncertainty flags were assigned. The first record for each element is for the 30-year period from 1961 to 1990. FORTRAN formats for the header, element ID, and data records are given at the bottom of Figure 3-6.
If you have a choice of locations or years from which to select data, these quality statistics can be used to select the best quality data. Appendix C contains the quality statistics for 1961-1990 for each of the Primary stations.
4.0 Brief History of Solar Radiation Measurements in the United States
Table of Contents
2If your monitor/software does not provide a "wide-screen" display, the statistics may wrap around on your screen or may be truncated.
Return to RReDC Homepage ( http://www.nrel.gov/rredc )
4.0 Brief History of Solar Radiation Measurements in the United States
Table of Contents