The current climatology includes the surface–300-mb precipitable water. The Storm Prediction Center Sounding Climatology page now contains the real-time point sounding information that was previously included on this page.

The monthly top 50 PW values (in a PDF file below) were created using data from the (1) NCDC Radiosonde Database of North America CD-ROMs, 1948–1996/97 and (2) NOAA/ESRL on-line radiosonde archive, 1996/97–present (https://esrl.noaa.gov/raobs/). The climatology for some of these sites includes two or more stations in order to get a lengthy time series for calculating statistics. For example, the climatology for ABR (Aberdeen, SD) was derived from upper-air observations using ABR and HON (Huron, SD), because from 1953 to 1995 the upper-air site was at HON and from 1995 to present the site has been at ABR. For all but 16 (7) cases, differences in station locations do not exceed 60 (80) miles for these “combined” climatologies. In general, the combining of sites is not expected to be problematic since these plots represent a synoptic signal and not a mesoscale signal.

Updated PDFs for the prior year are expected each January (last update, 1/2/2024).

Monthly top 50 PW values (118 sites)

CONUS PW Plots: Latest, 0000 UTC, 1200 UTC

Alaska PW Plots: Latest,  0000 UTC, 1200 UTC

NAM Forecasts of Precipitable Water (1st row) and PW % of Normal (2nd row) – updated twice daily

Precipitable Water Algorithm

The data were processed as follows: (1) questionable soundings were thrown out based on gross (objective and subjective) quality control measures*; (2) precipitable water values were calculated for all available soundings for the period of record, including 00z, 03z, 12z, 15z, and special observations; (3) the data were put into monthly bins; (4) the maximum, minimum, 99th percentile, 75th percentile, 50th percentile (median), and 25th percentile values were obtained for each month; and (5) the mean and standard deviation (SD) also were calculated. Note that for normal (or Gaussian) distributions, 95% of the values lie within ± two standard deviations of the mean/average value, so when you reach +2SD, you have a fairly rare event. Since the sample size is quite large (i.e., ~1000 per month), the Gaussian assumption is reasonable. Nevertheless, there are sites as well as times when the PW distribution is less Guassian than others, and indeed is positively skewed. Hence, at times it may be just as likely to observe +3SD as it is to observe –2SD.

*Caveat:  There are some soundings with bad and/or questionable data in the sounding archive. The automated QC algorithms did not catch all of them, and in fact, may have mistakenly removed some good soundings. This does not adversely affect the climatology, but has implications for extreme values.

The equations that were used for calculating precipitable water (units of cm) are as follows:

   a)  e = 6.112 * exp [ (17.67 * Td) / (Td + 243.5) ]  -- where e is vapor pressure in mb (or hPa) and Td is dewpoint in °C [refer to Bolton (1980, MWR, pp. 1046-1053 for details]

   b)  vapor density = [ (e / (Rv * T)] * [10^5 g Pa / kg hPa] -- where e is from above, Rv = 461.5 J/K/kg is the gas constant for water vapor, T is the temperature in K, and the vapor density is in g/m^3. The conversion factor, 10^5 g Pa / kg hPa, is needed to convert from (hPa kg / J) to (g / m^3). Recall that J = Pa m^3.

   c)  pw (layer) = [mean vapor density over layer (g/m^3)] * [layer thickness (m)] * [1 / 1,000,000 g/m^3] * [100 cm / 1 m] -- where the layer thickness is simply the distance between two adjacent sounding levels, the 3rd term is the density of liquid water, and the 4th term converts the pw (layer) from m to cm

   d)  pw (sfc to 300 mb) = summation of pw (layer) from the sfc to 300 mb for any given sounding -- again, these equations return the pw in cm

If you have any questions or comments regarding the raw precipitable water data, please send them to Matthew.Bunkers@noaa.gov. Thanks!