TITLE OF THE DATA SET Historical Sunshine and Cloud Data in the United States DATA CONTRIBUTORS Peter M. Steurer and Thomas R. Karl National Oceanic and Atmospheric Administration National Climatic Data Center Asheville, North Carolina 28801 DOI: 10.3334/CDIAC/cli.ndp021 SOURCE AND SCOPE OF THE DATA a) Sunshine Sunshine data for 240 U.S. stations (including Puerto Rico and nine islands in the Pacific Ocean) are presented for the years 1891-1987. The periods of record of monthly and annual total hours of sunshine vary widely from station to station. Many of the stations have discontinuous sampling periods and only 49 (20%) have periods of record equal to or greater than 90 years. However, 192 stations (80%) have periods of record equal to or greater than 30 years. In addition to the total monthly and annual measurements of sunshine duration, total monthly and annual maximum possible sunshine duration and monthly and annual percentages of possible sunshine duration are also provided. Maximum possible sunshine values for the 240-station recording network range from a minimum of 4418 h at the southernmost station (Tofuna, Pacific Ocean) to a maximum of 4591 h at the northernmost station (Fairbanks, Alaska). The highest annual sunshine amount ever recorded in the United States was at Yuma, Arizona, with 4272 h (96% of possible) in 1958, and the lowest ever recorded was at Mt. Washington, New Hampshire, with 1181 h (26% of possible) in 1943. In terms of annual percentage of possible sunshine, the highest amount ever recorded was 96% at Yuma, Arizona, in 1958 and the lowest amount ever recorded was 18% at Juneau, Alaska, in 1939. In the continental United States, the lowest annual percentage of possible sunshine ever recorded was 23% at Quillayute, Washington, in 1986. b) Clouds Cloud amount data for 197 U.S. stations are presented for the years 1871-1987. Most of the stations in the cloud observing network are also present in the sunshine recorder network, but the earlier opening of several stations for cloud observing purposes, coupled with a higher percentage of stations remaining open through 1987 to measure cloud amount, has resulted in the cloud amount data set being longer and more complete than the sunshine data set. For example, 100 stations (51%) have periods of record at least 90 years in length and 194 stations (98%) have records for at least 30 years. In compiling the cloud amount data set, only monthly sunrise to sunset cloud amount averages (percentages) were used. This eliminated problems associated with nighttime measurements and also maintained consistency in any comparisons with the monthly sunshine data. c) Estimates for Missing Data The high correlation between clouds and sunshine has led to the development of a procedure that produces an estimate when data are missing in either of the two data sets. Usually, data are fitted to a least-squares regression line where a slope and y-intercept are computed. A given value in one series can then be used to predict a missing value in the other series. Similarly, suspect values can be flagged for inspection if the value falls outside of some arbitrary range. Unfortunately, these suspect values can seize control of the best-fit line and give totally misleading regression line statistics. Preprocessing of the data by eliminating all extreme values is one method of avoiding this difficulty. Another method is the resistant regression (Emerson and Hoaglin 1983). Resistant regression essentially provides resistance against unusual data points so that these points will have only a small effect on the fitted line. This is done by sorting the predictor variable (values of x) in ascending order and then dividing the data points into three groups. Within each group, the median of the x and y values are determined. A slope and y-intercept is then found using these median values. This initial line is then adjusted by an iterative process using the actual data points. The iterations continue until the adjustments cause less than a 1% change in the slope. The resulting line provides resistance to those unusual data points that depart substantially from the primary pattern. It is important to note that the fewer the number of extraneous points, the closer the regression and resistant lines become with regard to slope and y-intercept. Seasonal and annual resistant line statistics were computed for all stations that recorded at least 10 years of simultaneous monthly cloud and sunshine data during the 1953-1987 period. This reference period was chosen because of the impression that from 1953 onward, all stations in the sunshine recording network were using the photoelectric type of sunshine recorder. Because the performance characteristics of the three types of sunshine recorders used through the years are known to vary, only the modern-day instrument period was desired for deriving the resistant line statistics. Subsequent to the calculation of these statistics, it was realized through detailed inspection of station histories that most stations had not implemented the photoelectric recorders by 1953. Many did so through the latter years of that decade, and some stations did not switch over until the mid-1960s (see Table 1 of the accompanying documentation). Two sets of seasonal and annual statistics were computed: one using clouds and the other using sunshine as the predictor variable. The seasons were based on the standard climatological periods (i.e., winter being December, January, and February; spring being March, April, and May; etc). An estimate for a missing value in one series was produced by using the corresponding monthly value and seasonal predictor statistics of the other series when both were available for a given month. The cloud amount data set has far fewer missing observations than does the sunshine data set, which in the case of some stations, has blocks of several years in which estimates were substituted. These estimates should prove very useful, as they often create continuous monthly and annual records for many stations that would otherwise suffer from numerous gaps. However, some caution should be used with these data, since, in the case of some stations, they were derived from regression statistics that may have been adversely affected by using sunshine data from a thermometric recorder together with those from a photoelectric recorder. DATA FORMAT Thirteen files are provided in this subdirectory including this documentation file (NDP021.TXT): four FORTRAN input/output routines (NDP021R1.F2,3,4, and 5); four SAS input/output routines; one file containing monthly and annual values of measured hours of sunshine (NDP021R1.F10); one file containing monthly and annual values of the maximum possible hours of sunshine (NDP021R1.F11); one file containing monthly and annual values of measured sunshine expressed as a percentage of that possible (NDP021R1.F12); and one file (NDP021R1.F13) containing monthly and annual cloud amounts (percentage). MEASURED HOURS OF SUNSHINE FILE The measured hours of sunshine file (NDP021R1.F10) contains monthly and annual hours of sunshine for each year from each of the 240 sunshine stations. The data are sorted by WBAN (Weather Bureau Army Navy) station number, with each record containing the WBAN station number, information on any previous WBAN station number, state name and station name, monthly values of hours of sunshine, and an annual value of hours of sunshine. Missing sunshine amounts are set to -99. The file may be read using the following FORTRAN format: INTEGER WBAN, PWBAN, MON, YEAR, HRSSUN(12), TOTAL CHARACTER*18 STNAME READ (5,100,END=99)WBAN,PWBAN,MON,STNAME,YEAR,(HRSSUN(I),I=1,12), + TOTAL 100 FORMAT(2(1X,I5),1X,I2,2X,A18,2X,I4,12(2X,I3),3X,I4) or by using the SAS format: INPUT WBAN 2-6 PWBAN 8-12 MON 14-15 STNAME $ 18-35 YEAR 38-41 JAN 44-46 FEB 49-51 MAR 54-56 APR 59-61 MAY 64-66 JUN 69-71 JUL 74-76 AUG 79-81 SEP 84-86 OCT 89-91 NOV 94-96 DEC 99-101 TOTAL 105-108; Stated in tabular form, the contents include the following. Variable Variable Starting Ending Variable type width column column WBAN Numeric 5 2 6 PWBAN Numeric 5 8 12 MON Numeric 2 14 15 STNAME Character 18 18 35 YEAR Numeric 4 38 41 JAN Numeric 3 44 46 FEB Numeric 3 49 51 MAR Numeric 3 54 56 APR Numeric 3 59 61 MAY Numeric 3 64 66 JUN Numeric 3 69 71 JUL Numeric 3 74 76 AUG Numeric 3 79 81 SEP Numeric 3 84 86 OCT Numeric 3 89 91 NOV Numeric 3 94 96 DEC Numeric 3 99 101 TOTAL Numeric 4 105 108 where WBAN is the 1987 WBAN station number or the number during the final year of a station's record; PWBAN is the actual WBAN number that was assigned during that year in the record (set to 0 after the first year in which the latest WBAN number was assigned, set to 99999 for years in which the WBAN number was not known or not assigned); MON is the month when a new WBAN number went into effect (set to 0 for years when no change took place); STNAME is the state abbreviation and station name; YEAR is the year of the data; JAN-DEC are the monthly measured hours of sunshine values (nearest whole hour); and TOTAL is the total hours of sunshine recorded in that year (nearest whole hour). MAXIMUM POSSIBLE HOURS OF SUNSHINE FILE The maximum possible hours of sunshine file (NDP021R1.F11) contains monthly and annual values for the maximum hours of sunshine that could be received at each station. The file is sorted by WBAN station number, with each record containing WBAN station number, state and station name, monthly and annual maximum possible sunshine amounts (hours) for non-leap years, and an adjustment factor (hours) to be added to the February and annual values in leap years. The file may be read using the following FORTRAN format: INTEGER WBAN, MAXSUN(12) REAL TOTAL, ADJ CHARACTER*18 STNAME READ (5,100,END=99) WBAN, STNAME, (MAXSUN(I),I=1,12), TOTAL, ADJUST 100 FORMAT(1X,I5,1X,A18,12(1X,I3),1X,F6.1,1X,F4.1) or by using the SAS format: INPUT WBAN 2-6 STNAME $ 8-25 JAN 27-29 FEB 31-33 MAR 35-37 APR 39-41 MAY 43-45 JUN 47-49 JUL 51-53 AUG 55-57 SEP 59-61 OCT 63-65 NOV 67-69 DEC 71-73 TOTAL 75-80 ADJUST 82-85; Stated in tabular form, the contents include the following. Variable Variable Starting Ending Variable type width column column WBAN Numeric 5 2 6 STNAME Character 18 8 25 JAN Numeric 3 27 29 FEB Numeric 3 31 33 MAR Numeric 3 35 37 APR Numeric 3 39 41 MAY Numeric 3 43 45 JUN Numeric 3 47 49 JUL Numeric 3 51 53 AUG Numeric 3 55 57 SEP Numeric 3 59 61 OCT Numeric 3 63 65 NOV Numeric 3 67 69 DEC Numeric 3 71 73 TOTAL Numeric 6 75 80 ADJUST Numeric 4 82 85 where WBAN is the WBAN station number; STNAME is the state abbreviation and station name; JAN-DEC are the monthly values of the maximum hours of sunshine possible at that station (nearest whole hour); TOTAL is the maximum hours of sunshine possible at that station in a non-leap year (given to the nearest tenth of an hour, e.g., 4447.3); and ADJUST is the number of hours to add to FEB and TOTAL for leap years (also given to the nearest tenth of an hour, e.g., 11.4). PERCENTAGE OF POSSIBLE SUNSHINE FILE The percentage of possible sunshine file (NDP021R1.F12) contains monthly and annual values of percentage of possible sunshine for each year from each of the 240 sunshine stations. The file is sorted by WBAN station number, with each record containing WBAN station number, state and station name, monthly sunshine percentages (missing values are set to -9999) with each month's four respective data flags, and an annual sunshine percentage (the mean of the monthly values, if none are set to missing; otherwise set to the missing indicator -9999). The file may be read using the following FORTRAN format: INTEGER WBAN, YEAR, SUNPCT(12), MEANC CHARACTER FLAG1(12), FLAG2(12), FLAG3(12), FLAG4(12) CHARACTER*18 STNAME READ (5,100,END=99) WBAN, STNAME, YEAR, + (SUNPCT(I),FLAG1(I),FLAG2(I),FLAG3(I),FLAG4(I),I=1,12), MEAN 100 FORMAT(1X,I5,1X,A18,1X,I4,12(I5,4(A1)),1X,I5) or by using the SAS format: INPUT WBAN 2-6 STNAME $ 8-25 YEAR 27-30 JAN 31-35 JANFL1 $ 36 JANFL2 $ 37 JANFL3 $ 38 JANFL4 $ 39 FEB 40-44 FEBFL1 $ 45 FEBFL2 $ 46 FEBFL3 $ 47 FEBFL4 $ 48 MAR 49-53 MARFL1 $ 54 MARFL2 $ 55 MARFL3 $ 56 MARFL4 $ 57 APR 58-62 APRFL1 $ 63 APRFL2 $ 64 APRFL3 $ 65 APRFL4 $ 66 MAY 67-71 MAYFL1 $ 72 MAYFL2 $ 73 MAYFL3 $ 74 MAYFL4 $ 75 JUN 76-80 JUNFL1 $ 81 JUNFL2 $ 82 JUNFL3 $ 83 JUNFL4 $ 84 JUL 85-89 JULFL1 $ 90 JULFL2 $ 91 JULFL3 $ 92 JULFL4 $ 93 AUG 94-98 AUGFL1 $ 99 AUGFL2 $ 100 AUGFL3 $ 101 FEBFL4 $ 102 SEP 103-107 SEPFL1 $ 108 SEPFL2 $ 109 SEPFL3 $ 110 MARFL4 $ 111 OCT 112-111 OCTFL1 $ 117 OCTFL2 $ 118 OCTFL3 $ 119 OCTFL4 $ 120 NOV 121-125 NOVFL1 $ 126 NOVFL2 $ 127 NOVFL3 $ 128 NOVFL4 $ 129 DEC 130-134 DECFL1 $ 135 DECFL2 $ 136 DECFL3 $ 137 DECFL4 $ 138 MEAN 140-144; Stated in tabular form, the contents include the following. Variable Variable Starting Ending Variable type width column column WBAN Numeric 5 2 6 STNAME Character 18 8 25 YEAR Numeric 4 27 30 JAN Numeric 5 31 35 JANFL1 Alphanumeric 1 36 36 JANFL2 Alphanumeric 1 37 37 JANFL3 Alphanumeric 1 38 38 JANFL4 Alphanumeric 1 39 39 FEB Numeric 5 40 44 FEBFL1 Alphanumeric 1 45 45 FEBFL2 Alphanumeric 1 46 46 FEBFL3 Alphanumeric 1 47 47 FEBFL4 Alphanumeric 1 48 48 MAR Numeric 5 49 53 MARFL1 Alphanumeric 1 54 54 MARFL2 Alphanumeric 1 55 55 MARFL3 Alphanumeric 1 56 56 MARFL4 Alphanumeric 1 57 57 APR Numeric 5 58 62 APRFL1 Alphanumeric 1 63 63 APRFL2 Alphanumeric 1 64 64 APRFL3 Alphanumeric 1 65 65 APRFL4 Alphanumeric 1 66 66 MAY Numeric 5 67 71 MAYFL1 Alphanumeric 1 72 72 MAYFL2 Alphanumeric 1 73 73 MAYFL3 Alphanumeric 1 74 74 MAYFL4 Alphanumeric 1 75 75 JUN Numeric 5 76 80 JUNFL1 Alphanumeric 1 81 81 JUNFL2 Alphanumeric 1 82 82 JUNFL3 Alphanumeric 1 83 83 JUNFL4 Alphanumeric 1 84 84 JUL Numeric 5 85 89 JULFL1 Alphanumeric 1 90 90 JULFL2 Alphanumeric 1 91 91 JULFL3 Alphanumeric 1 92 92 JULFL4 Alphanumeric 1 93 93 AUG Numeric 5 94 98 AUGFL1 Alphanumeric 1 99 99 AUGFL2 Alphanumeric 1 100 100 AUGFL3 Alphanumeric 1 101 101 AUGFL4 Alphanumeric 1 102 102 SEP Numeric 5 103 107 SEPFL1 Alphanumeric 1 108 108 SEPFL2 Alphanumeric 1 109 109 SEPFL3 Alphanumeric 1 110 110 SEPFL4 Alphanumeric 1 111 111 OCT Numeric 5 112 116 OCTFL1 Alphanumeric 1 117 117 OCTFL2 Alphanumeric 1 118 118 OCTFL3 Alphanumeric 1 119 119 OCTFL4 Alphanumeric 1 120 120 NOV Numeric 5 121 125 NOVFL1 Alphanumeric 1 126 126 NOVFL2 Alphanumeric 1 127 127 NOVFL3 Alphanumeric 1 128 128 NOVFL4 Alphanumeric 1 129 129 DEC Numeric 5 130 134 DECFL1 Alphanumeric 1 135 135 DECFL2 Alphanumeric 1 136 136 DECFL3 Alphanumeric 1 137 137 DECFL4 Alphanumeric 1 138 138 MEAN Numeric 5 140 144 where WBAN is the WBAN station number; STNAME is the state abbreviation and station name; YEAR is the year of the data; JAN-DEC are the monthly percentages of possible sunshine (nearest whole percent), with missing values set to -9999; and MEAN is the mean of the monthly sunshine percentages, if all 12 are available; otherwise set to the missing indicator -9999. Flag codes for the data The use of flags in the percentage of possible sunshine file was generally modeled after the U.S. Historical Climatology Network (HCN) format, examples of which may be found in Karl et al. (1990). This meant that each monthly data value had four flag positions. For consistency, these flag positions are retained in the percentage of possible sunshine file, but their use has been simplified as follows. (JAN-DEC) FL1 is a general data type flag. The codes are as follows: Z = value has been estimated by resistant regression on cloud data, and Blank = value based on observations (see Flag 2 for details). (JAN-DEC) FL2 is the data source code. The codes are as follows: 3 = Manuscript-Original National Climatic Data Center Records; 7 = LCD-Local Climatological Data, published monthly by the National Climatic Data Center (NCDC), Asheville, North Carolina; T = NCDC Tape Deck 9788; I = NCDC Tape Deck 9788-value is an estimate because of incomplete data for the month [i.e., at least one missing day (actual number not available)]; C = NCDC Tape Deck 9788-value from deck has been edited after comparison with the monthly cloud amount from the station; and Z = value has been estimated by resistant regression on cloud amount. (JAN-DEC) FL3 has the same meaning as Flag 1. (JAN-DEC) FL4 has the same meaning as Flag 1. CLOUD AMOUNT FILE The cloud amount file (NDP021R1.F13) contains monthly and annual cloud amounts (percentage of sky cover) for each year from each of 197 stations. The file is sorted by WBAN station number, with each record containing WBAN station number, state and station name, monthly cloud amounts (missing values are set to -9999) with each month's 4 respective data flags, and an annual cloud amount (the mean of the monthly values if none are set to missing; otherwise set to the missing indicator -9999). The file may be read using the following FORTRAN format: INTEGER WBAN, YEAR, CLDPCT(12), MEAN CHARACTER FLAG1(12), FLAG2(12), FLAG3(12), FLAG4(12) CHARACTER*18 STNAME READ (5,100,END=99) WBAN, STNAME, YEAR, + (CLDPCT(I),FLAG1(I),FLAG2(I),FLAG3(I),FLAG4(I),I=1,12), MEAN 100 FORMAT(1X,I5,1X,A18,1X,I4,12(I5,4(A1)),1X,I5) or by using the SAS format: INPUT WBAN 2-6 STNAME $ 8-25 YEAR 27-30 JAN 31-35 JANFL1 $ 36 JANFL2 $ 37 JANFL3 $ 38 JANFL4 $ 39 FEB 40-44 FEBFL1 $ 45 FEBFL2 $ 46 FEBFL3 $ 47 FEBFL4 $ 48 MAR 49-53 MARFL1 $ 54 MARFL2 $ 55 MARFL3 $ 56 MARFL4 $ 57 APR 58-62 APRFL1 $ 63 APRFL2 $ 64 APRFL3 $ 65 APRFL4 $ 66 MAY 67-71 MAYFL1 $ 72 MAYFL2 $ 73 MAYFL3 $ 74 MAYFL4 $ 75 JUN 76-80 JUNFL1 $ 81 JUNFL2 $ 82 JUNFL3 $ 83 JUNFL4 $ 84 JUL 85-89 JULFL1 $ 90 JULFL2 $ 91 JULFL3 $ 92 JULFL4 $ 93 AUG 94-98 AUGFL1 $ 99 AUGFL2 $ 100 AUGFL3 $ 101 FEBFL4 $ 102 SEP 103-107 SEPFL1 $ 108 SEPFL2 $ 109 SEPFL3 $ 110 MARFL4 $ 111 OCT 112-111 OCTFL1 $ 117 OCTFL2 $ 118 OCTFL3 $ 119 OCTFL4 $ 120 NOV 121-125 NOVFL1 $ 126 NOVFL2 $ 127 NOVFL3 $ 128 NOVFL4 $ 129 DEC 130-134 DECFL1 $ 135 DECFL2 $ 136 DECFL3 $ 137 DECFL4 $ 138 MEAN 140-144; Stated in tabular form, the contents include the following. Variable Variable Starting Ending Variable type width column column WBAN Numeric 5 2 6 STNAME Character 18 8 25 YEAR Numeric 4 27 30 JAN Numeric 5 31 35 JANFL1 Alphanumeric 1 36 36 JANFL2 Alphanumeric 1 37 37 JANFL3 Alphanumeric 1 38 38 JANFL4 Alphanumeric 1 39 39 FEB Numeric 5 40 44 FEBFL1 Alphanumeric 1 45 45 FEBFL2 Alphanumeric 1 46 46 FEBFL3 Alphanumeric 1 47 47 FEBFL4 Alphanumeric 1 48 48 MAR Numeric 5 49 53 MARFL1 Alphanumeric 1 54 54 MARFL2 Alphanumeric 1 55 55 MARFL3 Alphanumeric 1 56 56 MARFL4 Alphanumeric 1 57 57 APR Numeric 5 58 62 APRFL1 Alphanumeric 1 63 63 APRFL2 Alphanumeric 1 64 64 APRFL3 Alphanumeric 1 65 65 APRFL4 Alphanumeric 1 66 66 MAY Numeric 5 67 71 MAYFL1 Alphanumeric 1 72 72 MAYFL2 Alphanumeric 1 73 73 MAYFL3 Alphanumeric 1 74 74 MAYFL4 Alphanumeric 1 75 75 JUN Numeric 5 76 80 JUNFL1 Alphanumeric 1 81 81 JUNFL2 Alphanumeric 1 82 82 JUNFL3 Alphanumeric 1 83 83 JUNFL4 Alphanumeric 1 84 84 JUL Numeric 5 85 89 JULFL1 Alphanumeric 1 90 90 JULFL2 Alphanumeric 1 91 91 JULFL3 Alphanumeric 1 92 92 JULFL4 Alphanumeric 1 93 93 AUG Numeric 5 94 98 AUGFL1 Alphanumeric 1 99 99 AUGFL2 Alphanumeric 1 100 100 AUGFL3 Alphanumeric 1 101 101 AUGFL4 Alphanumeric 1 102 102 SEP Numeric 5 103 107 SEPFL1 Alphanumeric 1 108 108 SEPFL2 Alphanumeric 1 109 109 SEPFL3 Alphanumeric 1 110 110 SEPFL4 Alphanumeric 1 111 111 OCT Numeric 5 112 116 OCTFL1 Alphanumeric 1 117 117 OCTFL2 Alphanumeric 1 118 118 OCTFL3 Alphanumeric 1 119 119 OCTFL4 Alphanumeric 1 120 120 NOV Numeric 5 121 125 NOVFL1 Alphanumeric 1 126 126 NOVFL2 Alphanumeric 1 127 127 NOVFL3 Alphanumeric 1 128 128 NOVFL4 Alphanumeric 1 129 129 DEC Numeric 5 130 134 DECFL1 Alphanumeric 1 135 135 DECFL2 Alphanumeric 1 136 136 DECFL3 Alphanumeric 1 137 137 DECFL4 Alphanumeric 1 138 138 MEAN Numeric 5 140 144 where WBAN is the WBAN station number; STNAME is the state abbreviation and station name; YEAR is the year of the data; JAN-DEC are the monthly cloud amounts (nearest whole percent of sky cover); and MEAN is the mean of the monthly cloud amounts, if all 12 are present; otherwise set to the missing indicator -9999. Flag codes for the data The use of flags in the cloud amount file was also generally modeled after the U.S. HCN format (Karl et al. 1990). The four flag positions normally associated with each monthly value are retained in the file, but their use has been customized as follows. (JAN-DEC) FL1 is a dual purpose code pertaining to either (1) the number of daily values not available in computing the monthly mean cloud amount or (2) how an estimate of the monthly mean cloud amount has been produced. The codes and their meanings are as follows: A,B,C, . . . I corresponds to 1,2,3 . . . 9 days missing; Blank = no missing days; X = value has been estimated from manuscript entries (see Flag 2); and Z = value has been estimated by resistant regression on percentage of possible sunshine. (JAN-DEC) FL2 is the data source code. The codes and their meanings are as follows: 1 = NCDC Tape Deck 3210, Summary of the Day, First Order; 3 = Manuscript-Original National Climatic Data Center Records; 5 = Climate Record Book-for a description, see History of Climatological Record Books, U.S. Department of Commerce, Weather Bureau, U.S.G.P.O., 1960; 7 = LCD-Local Climatological Data, published monthly by NCDC; M = Monthly Weather Review, U.S. Weather Bureau, U.S.G.P.O., 1872-1966; R = Report of the Chief of the Weather Bureau, U.S. Weather Bureau, U.S.G.P.O., 1891-1934; United States Meteorological Yearbook, U.S.G.P.O., 1935-1949; or Climatological Data, National Summary, U.S. Weather Bureau, Asheville, North Carolina, 1950-1980; and Z = value has been estimated by resistant regression on percentage of possible sunshine. (JAN-DEC) FL3 is the location code. The codes and their meanings are as follows: 0 = primary station location is current; U = previous station number unknown; 1 = prior city or airport location; 2 = second city or airport location; 3 = third airport location; and Z = value has been estimated by resistant regression on percentage of possible sunshine. (JAN-DEC) FL4 is an additional source qualifier. The codes and their meanings are as follows: C = computed by technician from data available in the nondigital source specified by Flag 2; Blank = data are directly from the source specified by Flag 2; and Z = value has been estimated by resistant regression on percent of possible sunshine. REFERENCES Emerson, J. D., and D. C. Hoaglin. 1983. Resistant lines for y versus x. pp. 129-65. IN D. C. Hoaglin, F. Mosteller, and J. W. Tukey, eds., Understanding Robust and Exploratory Data Analysis. John Wiley & Sons. Karl, T. R., and P. M. Steurer. 1990. Increased cloudiness in the United States during the first half of the twentieth century: Fact or fiction? Geophysical Research Letters 17(11):1925-28. Karl, T. R., C. N. Williams, Jr., and F. T. Quinlan. 1990. United States Historical Climatology Network (HCN) serial temperature and precipitation data. ORNL/CDIAC-30, NDP- 019/R1. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge, Tennessee.