Information on Surface Marine Monthly Statistics

Statistics on global surface marine meteorological and oceanographic data are compiled monthly by personnel of NOAA's NWS/NCEP/Central Operations. These data include those which are observed/measured by ships (including those vessels that are part of the U.S. Voluntary Observing Ship Project), moored and drifting buoys, Coastal Marine Automated Network (CMAN) stations, and tide gauge stations (tide gauge data are provided by NOAA's National Ocean Service/Center for Operational Oceanographic Products and Services (NOS/CO-OPS)). Separate statistics are compiled monthly for specified vessels in the Gulf of Alaska and for specified member vessels of the International SeaKeepers Society. The data for which these statistics are compiled include sea-level pressure, air temperature, and wind direction and speed (for the surface meteorological data), and sea surface temperature (for the oceanographic data).

Only 1 year's worth of statistics will be available at any given time on this web site. Please note that the monthly statistics are updated on the first business day of each month. I.e. if the 1st falls on a weekend or on a Federal Government holiday, the statistics are updated on the first business day which follows the weekend or holiday.

The global monthly statistics are sent to the UK Met Office in Bracknell, England each month (the UK Met Office has been designated by the World Meteorological Organization (WMO) as the lead center for monitoring surface marine observations), where they are consolidated with additional statistics and forwarded to platform managers. The platform managers may then investigate platforms listed in the statistics to see if instrumentation needs to be corrected. Please note that the statistical files available via this web site are not identical to those passed to Bracknell, in that the ship/platform names are included in the files on this web site (they are not included in the Bracknell files).

The monthly statistics produced for the Gulf of Alaska vessels and for the International SeaKeepers Society vessels are made available to those program managers via this web site, so that they may evaluate the performance of select vessels in their area of interest.

All of the monthly statistics are computed using differences between the observed data and first guess fields from one of 2 NWS numerical models. This file explains the method of comparison between the data and first guess fields, and also provides the criteria used in determining which platforms will be listed in the NCEP Central Operations (NCO) global marine statistics each month.

Standard of Comparison

Data from each platform (ships, moored (fixed) buoys, drifting buoys, CMAN platforms, and tide gauge stations) are compared to forecasts from one of 2 NWS numerical models for each of 4 daily synoptic periods. Each synoptic period is a 6 hour window centered on the times given in the table below. The first guess valid times are also given below. All data observed/measured during each synoptic period are compared to the first guess valid for that synoptic period. The first guess is interpolated to the platform's location and differences between observation and first guess are computed. Differences between observed data and forecasts are archived four times daily.

If the GDAS is unavailable for a particular synoptic period, a 6-hour forecast from the previous run of NWS's Global Forecast System (GFS) is used instead.

Each synoptic period includes data with observation times +/- 3 hours as follows:

No time interpolation between observed data and first guess is done (i.e. an observation from 1000 UTC or 1400 UTC will be compared to a first guess valid at 1200 UTC).

Criteria for Specific Parameters

• Global Monthly Statistics:

  For each of the 5 parameters (sea-level pressure, air temperature, wind speed, wind direction, and sea surface temperature) included in the monthly statistics, the following criteria must be met in order for a particular platform to be listed in the NCO global monthly statistics:

  Sea-level pressure (slp):

  • 20 or more slp observations for the month, and 1 or more of the following:
    • absolute value of bias (mean difference between observation and forecast) is at least 4.0 mb
    • standard deviation of difference is at least 6.0 mb
    • percentage of reports that were gross errors is at least 25% of the total number of reports received for this parameter (see Definition of Gross Errors, below).

  Air temperature:

  • 20 or more air temperature observations for the month, and 1 or more of the following:
    • absolute value of bias is at least 4.0 deg C
    • standard deviation of difference is at least 6.0 deg C
    • percentage of reports that were gross errors is at least 25% of the total number of reports received for this parameter.

  Wind direction:

  • 20 or more wind direction observations for the month, and 1 or more of the following:
    • absolute value of bias is at least 30.0 degrees
    • standard deviation of difference is at least 80.0 degrees
    • percentage of reports that were gross errors is at least 25% of the total number of reports received for this parameter.

  Wind speed:

  • 20 or more wind speed observations for the month, and 1 or more of the following:
    • absolute value of bias is at least 5.0 m/s
    • percentage of reports that were gross errors is at least 25% of the total number of reports received for this parameter.

  Sea surface temperature (sst):

  • 20 or more sst observations for the month, and 1 or more of the following:
    • absolute value of bias is at least 4.0 deg C
    • standard deviation of difference is at least 6.0 deg C
    • percentage of reports that were gross errors is at least 25% of the total number of reports received for this parameter.
  
  
• Gulf of Alaska Monthly Statistics and International SeaKeepers Society Monthly Statistics:

  Specific callsigns were given to NCO/DMQAB to indicate those vessels which are to be included in these monthly statistics. No other criteria must be met, except that the vessels must have at least one observation during the desired month. If a vessel reports only once during a specific month, then the bias for any reported parameter is not a mean bias but is merely the difference between observed and first guess values for that parameter. Standard deviation is only computed when a vessel has 2 or more reports for any particular parameter in the month. As with the global monthly statistics, gross errors are not included in computations of bias or standard deviation (see Definition of Gross Errors, below).

With the exception of wind direction, the bias for each parameter is always computed as (observed data - first guess). A mean difference (bias) for slp of -4.8 mb would mean that the observed slp for that platform is, on average, 4.8 mb lower than the first guess slp. Also, a low standard deviation (SD) of the difference would mean that the platform's bias is consistent (the closer to 0 the SD is, the more consistent that platform is for that parameter). For example, if the slp bias was -4.8 mb and standard deviation of the difference was 1.0, that would imply that the platform consistently was around 4.8 mb too low with its slp. If the SD is high, then the platform has a lot of variability for the parameter in question, with regards to its comparison with the first guess.

Calculation of Bias and Standard Deviation for Wind Direction

Beginning with the February 2000 statistics, the computation of bias and standard deviation for wind direction is being done using a new algorithm. This new algorithm is the same one used by Meteo-France's Centre for Marine Meteorology (Centre de Meteorologie Marine (CMM)) in computation of wind direction statistics for moored and drifting buoys. For platforms reporting wind direction that is consistently close to 180 degrees off the first guess wind direction, using the regular algorithm of (observed wind direction - first guess wind direction) would result in a very large SD and bias, which would make it difficult to detect the wind direction error. The ideal for this situation would be to have a small SD with a bias on the order of 180 degrees. This is possible if CMM's algorithm is used. This algorithm computes bias and SD for the four following series:

• (Obs - FG)
• (Obs + 90) - FG
• (Obs - 90) - FG
• (Obs + 180) - FG

where Obs = observed wind direction and FG = first guess wind direction.

The bias and SD are computed for each of the above 4 series. The smallest of the 4 SD values is selected and then (bias - i) (where i is 0, 90, -90, and 180, respectively) is selected as the actual bias if the series ((Obs + i) - FG) was the series associated with the smallest SD. See the following table for more details.

Using this algorithm for all wind direction bias and SD computations will ensure that the most accurate representation of the difference between observed wind direction and first guess wind direction will be used and displayed in the monthly statistics.

Sample lines from global monthly statistics (and explanation)

          NCEP CENTRAL OPERATIONS (NCO) PLATFORM STATS
                         NOVEMBER 2001
                SEA-LEVEL PRESSURE FOR ALL HOURS

PLATFORM   SHIP/PLAT. NAME           TYPE   NO. REPORTS    MEAN DIFF    SD DIFF    # GROSS ERR

VVMA       UNKNOWN                   SHP         20           11.2        2.1           1

The above lines were taken from the global monthly statistics produced for November 2001. The platform shown has the callsign VVMA. Its name is unknown. The letters SHP in the TYPE column indicate that the platform is a ship (see Platform Types Seen in Monthly Statistics, below). The number of reports used in calculation of the sea-level pressure statistics for ship VVMA is 20. The mean difference is 11.2 mb and the standard deviation of the difference is 2.1 mb. The number of gross errors for sea-level pressure from ship VVMA is 1 (see Definition of Gross Errors, below). Please note that the total number of sea-level pressure reports received from ship VVMA for the month of November is 21 (the sum of the numbers listed under NO. REPORTS and # GROSS ERR, as seen above).

Caveat

Please note that a platform's appearance in the monthly statistics does not necessarily mean that the platform's data are bad. A high bias can be indicative of a persistent problem with the first guess at the platform's location. While the model's grid resolution and topographical resolution are 1 degree x 1 degree, this is still not fine enough to depict small-scale local meteorological and oceanographic effects. In addition, the first guess in general performs more poorly in the Southern Hemisphere than in the Northen Hemisphere, due to lower data counts (much less land south of the Equator, so not nearly as many land-based surface and upper air observations, plus major shipping lanes are primarily in the Northern Hemisphere).

Definition of Gross Errors

An observation is considered to be a gross error if it differs from the first guess by the following values:

Please note that, prior to August 1998, the gross error limits for wind direction and wind speed were the following:

The wind direction and wind speed gross errors were changed in August 1998 to conform to gross error limits used by other international centers producing similar statistics.

By convention, gross errors are not used in the computation of the mean difference or standard deviation.

Platform Types Seen in Monthly Statistics

Statistics are not computed for platforms with the following callsigns: 'SHIP', 'BUOY', 'BBXX', 'PLAT', 'RIGG', or 'METEO'. These callsigns are not unique (i.e. more than one platform may have the same callsign), so statistics for these callsigns would be meaningless.

Ship and Platform Names

The ship and platform names are obtained from several sources. Since callsigns and ship names change from time to time, a particular callsign may be associated with 2 or more different names, or a ship name may be associated with several callsigns. Although we will update the names file several times a year, we may not know which callsign/name is current.

A good web site for querying ship names and/or callsigns is that of the Ship-Of-Opportunity Programme (SOOP), which is a joint program of the WMO and the Intergovernmental Oceanographic Commission (IOC). On their query page, one may enter a full or partial ship name or callsign, and an attempt will be made to find a match on either. To access the query page, go to the SOOP website, and click on "Network operations" in the left frame, then click on "Ships" in the left frame. The right frame will then update. Click on "Search for Ship Particulars" in the right frame and you'll be taken to the query page, where you may then enter the callsign or ship name.

Statistics Updates

• Aug. 1998 - gross error limits for wind direction and wind speed are modified (see Definition of Gross Errors, above).
• Feb. 2000 - new algorithm for computation of bias and standard deviation for wind direction is implemented (see Calculation of Bias and Standard Deviation for Wind Direction, above).
• Sep. 2000 - tide gauge data are now being archived (effective Sep. 5, 0600 UTC synoptic period).
• Oct. 2000 - tide gauge data are present in monthly statistics file for September 2000.
• Feb. 2001 - separate statistics are now displayed for select vessels in the Gulf of Alaska.
• Jun. 2001 - separate statistics are now displayed for International SeaKeepers Society vessels (data first received at NCEP/NCO on May 15, 2001).
• Apr. 2003 - renaming Aviation model (AVN) to Global Forecast System (GFS) per changes made by NCEP's Environmental Modeling Center.
• May 2003 - added table and links for Gulf of Alaska data counts.

NOTE: If you are seeking climatological marine data such as average temperature, wave heights, or prevailing winds for a particular oceanographic region, please contact NOAA's National Climatic Data Center (NCDC). For marine forecasts, please contact NOAA/NWS/NCEP's Ocean Prediction Center (OPC). Thank you.