U.S. Climate Normals 1971-2000, Products
CLIM84 CLIMATOGRAPHY OF THE U.S. NO. 84: Daily Station Normals
Monthly Station NormalsDescription
|CODE||Data Description||CODE||Data Description|
|01||No Data||11||1990-2000 Standard Deviation|
|02||No Data||12||1990-2000 Median|
|03||Number of Estimated Values in Normals Period||13||Maximum Monthly Value in Normal Period|
|04||1971-2000 Normal||14||Year of Occurrence of Maximum Value|
|05||1971-2000 Standard Deviation||15||Minimum Monthly Values in Normal Period|
|06||1971-2000 Median||16||Year of Occurrence of Minimum Value|
|07||1980-2000 Mean||17||Precipitation 10th Percentile|
|08||1980-2000 Standard Deviation||18||Precipitation 90th Percentile|
|09||1980-2000 Median||19||Time of Observation Adjustment Factor|
The daily normals are derived by statistically fitting smooth curves through monthly values; daily data were not used to compute daily normals. As a result, the published values reflect smooth transitions between seasons. The typical daily random patterns usually associated with precipitation are not exhibited; however, the precipitation normals may be used to compute average amounts accumulated over time intervals.
A. Spline-Fit Daily Normals
Daily normals of maximum, minimum, and mean temperatures, heating and cooling degree days, and precipitation were prepared for selected stations by interpolating between the monthly normal values. The interpolation scheme was a cubic spline fit through the monthly values. Each element was interpolated independently from the other elements. The procedure is described by Greville (1967).
The series of daily values of an element resulting from the cubic spline yields a smooth curve throughout the year without requiring the use of daily data. Another property of this technique is that the average of the daily temperatures in a month equals the monthly normal and that the total of the daily precipitation or degree days in a month equals the monthly normal. In order to eliminate discontinuities between December 31 and January 1, the spline interpolation was performed on a series of 24 monthly values. This extended series was created by appending July-December normals before January and January-June normals after December. This process is applied independently to all six climatological elements. February 29 is assigned the same value as February 28.
Since each element was interpolated independently, the daily series of temperatures and degree days were adjusted using software to remove spurious inflection points caused by rounding and to ensure adherence to functional relationships among the elements. The software interrogated the data for climatologically reasonable inflection points, daily consistency between elements, monthly consistency between daily and monthly values by element, and adherence of temperature and degree day values to the formula T - 65 + H - C = 0, where T = mean temperature, H = heating degree days, and C = cooling degree days. Collectively, the processing steps for CLIM84 are shown in Figure 3.
Daily precipitation normals were published as generated by the cubic spline interpolation. The smooth curve through a month does not represent a climatologically reasonable distribution. The spreading of the monthly precipitation by the spline over all the days in a month is useful for accumulating amounts over specified time intervals. A climatologically reasonable normal precipitation, based on daily data, for any one date would be much different from the published normals.
For some dates at most locations the published degree days are shown by an asterisk. The symbol represents a value of less than one degree day, but more than zero degree days. It is used to smooth through aperiodic oscillations of zeroes and ones that are climatologically unreasonable. For example, if a station has 17, 15, and 18 normal heating degree days in June, July, and August, respectively, it is not possible to distribute the 15 July degree days evenly throughout the month using integer values (zeroes and ones) without creating unrealistic oscillations through the 3-month period. The use of fractional degree days (asterisks) does allow for a smooth transition from June through July to August.
There are several reasons for using a cubic spline fit of the monthly normals instead of averaging the daily data. First, simply averaging the observed daily values would result in a daily normal curve that has considerable variability from day to day (Guttman and Plantico, 1987), yielding an annual temperature cycle that would be considerably jagged or ragged. This climatological raggedness could result in daily normals that trend in the opposite direction from what is expected. For example, an autumn daily normal temperature could be considerably warmer than one from several days earlier, or a spring daily normal temperature could be considerably cooler than one from several days earlier. Using a cubic spline fit of the monthly normals eliminates this raggedness from the daily normals curve. Furthermore, a complete and homogeneous (i.e., no change in location, instrumentation, exposure, or observation practices) data set is necessary for the analysis to be accurate. There are very few stations that have complete and homogeneous daily records. Any change of the types indicated above would introduce a nonclimatic effect which would make the data inhomogeneous. The techniques for estimating missing daily data and adjusting daily data for inhomogeneities are complex and, for some stations, are difficult to apply. However, the estimation and adjustment techniques for monthly data are not as complex or troublesome. Hence, the official daily normals are based on monthly normals, which incorporate CLIM81 inhomogeneity adjustments.
B. Precipitation Probabilities and Quintiles:
A secondary part of the CLIM84 product is the monthly precipitation totals that correspond to the indicated probability levels. The probability levels are based on the 1971-2000 sequential monthly precipitation and are explained below. The historical precipitation data are the serially complete values (including estimated values) that were also used to compute the monthly normals (i.e., CLIM81).
When historical climate data are accumulated and examined, they generally follow a certain pattern called a statistical distribution. For example, if 30 years of June temperature data were assembled and examined, the data would display a pattern that consisted of most of the Junes having temperatures close to the normal or average value, a few Junes having very warm temperatures, and a few Junes having very cold temperatures. This kind of statistical pattern is called a Gaussian distribution and theoretically takes the form of a bell-shaped curve. Temperature data are more likely to follow a Gaussian distribution than precipitation data. This is because precipitation is zero bounded.
When historical precipitation data are examined, most of the values will be close to the middle of the distribution, but some values will be considerably higher than the middle range. On the low end of the scale, however, the smallest values will never be less than zero. In particularly dry (e.g., desert) regions, the pattern can be drastically skewed to the left-hand side of the scale, with most of the values being near zero and a few very wet values spread far to the right. This kind of pattern can be fit by a Gamma distribution. Once the statistical distribution is identified, the statistical properties of the distribution can be used to estimate the probabilities that certain values will occur, and which values can be expected at certain probability levels. For summarization purposes, the probability levels desired can be preselected at certain individual levels or at regular intervals.
The Gamma distribution is used to estimate the precipitation probability and quintile values. The probability table shows the amount of precipitation expected at 15 probability (PROB) levels (0.005, 0.01, 0.05, 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.70, 0.80, 0.90, 0.95, 0.99, and 0.995) for each month of the year and for the annual total. For example, if 1.77 inches corresponds to the 0.20 probability level, that means that, on average, 2 out of 10 years will have 1.77 inches or less of precipitation in that month. It also means that, on average, 8 out of 10 years will have more than 1.77 inches of precipitation in that month.
The second table shows the expected precipitation values at the five quintile levels (LVL): 1 (0-20%); 2 (20-40%); 3 (40-60%); 4 (60-80%); 5 (80-100%) for each of the twelve months and for the year. For example, if 2.91 and 4.07 inches are the bounds for the second quintile, then a monthly total precipitation amount for that month falling in the range 2.91 to 4.07 would be classified as a second quintile precipitation amount and the month would be considered relatively dry. The first line (LVL 0 <) in this table shows the minimum precipitation value derived from the historical record. Quintile level 0 would be used if a future precipitation observation is less than the 1971-2000 minimum. The last line (LVL 6>) shows the maximum precipitation value. Level 6 would be used if the observed precipitation value is more than the 1971-2000 maximum. The quintile table is used primarily in National Weather Service operations for composition of information that is transmitted in CLIMAT messages and published in the Monthly Climatic Data for the World publication.
This product includes normals and standard deviations for the five 30-year periods and the 70-year period between 1931-2000 for each division in a state. A division represents a region within a state that is, as nearly as possible, climatically homogeneous. Some areas, however, may experience rather extreme variations within a division (e.g., the Rocky Mountain states). The divisions have been established to satisfy researchers in hydrology, agriculture, energy supply, etc., who require data averaged over an area of a state rather than for a point (station).
The normals and standard deviations include values for each of the 12 calendar months and an annual value. The divisional data are displayed by name and number for a state or island. The states and islands include the contiguous United States, Alaska, Puerto Rico, and the Virgin Islands, and are arranged alphabetically. Hawaii is not included because the varied topography and locations of the observing stations do not allow for the establishment of homogeneous divisions. The data elements include mean temperature (degrees F), precipitation (inches), and heating and cooling degree days (base 65 degrees F).
Climatic divisions are regions within each state that have been determined to be reasonably climatically homogeneous. The maximum number of divisions in each state is 10. Monthly divisional average temperature and total precipitation data are derived using data from all stations reporting both temperature and precipitation within a climatological division. The number of reporting stations within a division varies from month to month and year to year.
Monthly temperature normals and 70-year averages for a division are computed by adding the yearly values for a given month and then dividing by the number of years in the period. The annual normal and 70-year average are computed by adding all of the monthly normal or long-term average values and then dividing by 12. Consequently, if an annual normal were computed by averaging annual values obtained for each year in the period (by adding the corresponding 12 monthly values and then dividing by 12), it may be slightly different from the average of the 12 monthly normals because of rounding differences. Precipitation normals and 60-year averages are computed in a similar manner, except that the annual values are the totals of the 12 monthly values.
Sequential monthly degree days are derived using procedures developed by Thom (1954, 1966). This technique utilizes the historical monthly average temperature and its corresponding standard deviation (over some "standardizing period") to compute degree days. The procedure for the computation of the divisional degree day normals involves the following three steps:
1. Calculate the standard deviations of the temperatures for each of the 12 calendar months over the standardizing period;
2. Use the Thom technique to compute the heating and cooling degree days for every month for every year in the period 1931-2000; and
3. Calculate the 30-year normals and 70-year (1931-2000) averages of the degree days using the procedure discussed above.
This product provides climate data from selected sites included in CLIM81, as well as statistics that have not been published elsewhere. The climatological data included in the CLIM20 make this publication the most appropriate summary for agricultural applications.
The following link contains a detailed description of the summaries
Monthly Station Climate Summaries Documentation
The Supplement No. 1 publication presents the monthly and annual precipitation values (in inches) corresponding to three probability levels: 0.10, 0.50, and 0.90. The stations are listed alphabetically. There is a separate volume of this publication for each state.
Monthly and annual precipitation probabilities are also available on microfiche and in digital format. The values are summarized in two tables. The first table shows the amount of precipitation expected at 15 probability levels (0.005, 0.01, 0.05, 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.70, 0.80, 0.90, 0.95, 0.99, and 0.995) for each month of the year and for the annual total. The second table shows the expected precipitation values at the five quintile levels:
First Quintile: 0-20%
Second Quintile: 20-40%
Third Quintile: 40-60%
Fourth Quintile: 60-80%
Fifth Quintile: 80-100%
The probability tables in this product are determined by fitting the 1971-2000 historical monthly precipitation to a Gamma distribution (Crutcher et al., 1977; Crutcher and Joiner, 1978). The process was performed with the historical data for each of the twelve months and separately with the annual values to produce 13 sets of probability values for each station.
Monthly and annual degree day normals are available on microfiche and in digital format. The heating degree day normals are to the following bases: 70, 65, 60, 57, 55, 50, 45, 43, 40, 35, 32, and 30. The cooling degree day normals are to the following bases: 80, 75, 70, 65, 60, 57, 55, 50, 45, 43, 40, and 32.
The main contents of this publication are freeze/frost probability tables for each station, listed by state. The tables contain the dates of probable first and last occurrence, during the year beginning August 1 and ending July 31, of freeze-related temperatures; probable durations (in days) where the temperature exceeds certain freeze-related values; and the probability of experiencing a given temperature, or less, during the year period August 1 through July 31. For the fall and spring dates of occurrence, and freeze-free period, probabilities are given for three temperatures (36, 32, and 28 degrees F) at three probability levels (10, 50, and 90 percent). A series of maps present calendar data related to the probability of occurrence of freeze at two temperature thresholds.
Extended tables of freeze/frost data, which contain the dates for probabilities of 0.1 through 0.9 in increments of 0.1 versus temperature thresholds of 36, 32, 28, 24, 20, and 16 degrees F, are available on magnetic tape or on microfiche (by state) for all of the sites given in the publication.
Each month, averages of temperature and precipitation are calculated for U.S. Climate Divisions by simple averaging of data from all stations within the division that record both temperature and precipitation. A division represents a region within a state that is climatically quasi-homogeneous or, in some cases, a semi-homogeneous dranage basin (as described by CLIM85).
The average monthly temperature and precipitation for a state are derived from the divisional values by weighting each division by its percentage of the total state area, including the 48 contiguous states, Alaska, Hawaii, Puerto Rico, and the Virgin Islands. The District of Columbia is treated as part of Maryland.
The nation was divided into nine census divisions as defined and used by the Census Bureau. The divisions and states they comprise are as follows:
The areal weights used to produce monthly and regional temperatures are also shown. These weights were obtained by dividing the area of each state by the total regional area. A particular regional monthly temperature value was obtained by multiplying the corresponding state temperature within a region by the approriate wieght and adding all of the products. Annual values were obtained by taking the average of the monthly values. Monthly and annual temperatures for the nine census divisions are presented in tables following the weights.NORTHEAST REGION
New England Division: Maine, New Hampshire, Vermont, Massachusettes, Rhode Island, Connecticut
Middle Atlantic Division: New York, New Jersey, Pennsylvania
East North Central Division: Ohio, Indiana, Illinois, Michigan, Wisconsin
West North Central Division: Minnesota, Iowa, Missouri, North Dakota, South Dakota, Nebraska, Kansas
South Atlantic Division: Delaware, Maryland, District of Columbia, Virginia, West Virginia, North Carolina,
South Carolina, Georgia, Florida
East South Central Division: Kentucky, Tennessee, Alabama, Mississippi
West South Central Division: Arkansas, Louisiana, Oklahoma, Texas
Mountain Division: Montana, Idaho, Wyoming, Colorado, New Mexico, Arizona, Utah, Nevada
Pacific Division: Washington, Oregon, California, Alaska, Hawaii
The national temperatures were devied by areally weighting the temperature values for the nine census divisions. The national value, therefore, covers only the contiguous United States.
The population weights for U.S. Climate Divisions are computed from the 2000 Census county and metropolitan populations in that division. Divisional population totals are summed from 2000 county totals for counties residing completely within a given division. For counties residing in more than one division, 2000 county populations are divided proportionally by overlaying the climate divisions on a one-kilometer squared population database based on the 1990 census and provided by the Socioeconomic Data Application Center (SEDAC). Approximately 25%, or about 800 out of 3200 counties, require division in this manner. Once divisional totals are determined, their proportion in the context of the state, division, region, and nation are determined.