Buoy Data Documentation

The DFG funded Project “Atlas of Antarctic Sea ice Drift” includes buoy data from the main sources:

·       collected at Alfred Wegener Institute for Polar and Marine Research (AWI) 1986-2000 (Kottmeier and Lüdemann, 1997, Sellmann and Hellmer, 1997,  AWI 1998/1999, Sellmann et al., 1997)

·       collected at the International Project of Antarctic Buoys (IPAB) of the World Climate Research Programme (1995-2001) (WCRP)

·       collected at the Australian Antartic Division (AAD)  (1985-2001) 


All archives contain data of buoys maintained by research institutes from various countries gathered between 1985 and 2001. Buoy equipment, sensor types, and measurement principles differ considerably depending on manufacturer and research focus on either oceanography, meteorology or sea ice glaciology. Standard meteorological buoys at least monitor buoy position, air temperature (typically about 1m over ice surface) and air pressure.

The variety of buoy manufacturers and operators involved in data acquisition results in a quite inhomogeneous data structure. Within this project the buoy data have been reformatted to one consistent raw data format as bases for calculations of daily means. No corrections have been applied during the production of the consistent raw data set, not to increase the number of different buoy data versions. Three-hourly means for position and drift, near surface air temperature and pressure are available within this reformatted raw data structure.  For the atlas we used the daily means of buoy data after some corrections, being applied to dates, buoy positions and drift components. The following section describes the corrections in calculating daily means:


1. Buoy measurements and data transmission

Buoy data logging systems collect and average the sensor data usually over ten minutes and transmit them to two polar-orbiting satellites every 60 or 90 seconds. Depending on latitude and area of coverage during the satellite overpass, about 16 to 22 independent data transmissions are available for each platform per day. Only the data set with the largest number of identical transmissions within a satellite pass is saved in preprocessing the data (Kottmeier et al., 1997, IPAB).

Buoy positions are determined from Doppler shifts of transmitter signals observed from different satellite positions. The accuracy of the position depends on transmitter stability and satellite constellation geometry. Data transmission and buoy localisation are done under responsibility of the Argos Collect Localisation Satellites Company with accuracies of about 350 m (newer buoys 150 m). AWI Buoys since 1994 (Kottmeier et al., 1997) usually carry a GPS (Global Positioning System) receivers with a positioning accuracy of better then 50 m.

Atmospheric pressure is measured with various buoy devices, including aneroid cells, beam balance quartz crystal transducers, and piezo-resistive transducers. The inlet or pressure port allows the ambient pressure to be transmitted freely into the sensing element. The error or pressure change is related to the shape and configuration of the inlet. In most cases, the pressure error is related to the square of the wind velocity. Atmospheric pressure measurements from a small drifting buoy in the open ocean are difficult to obtain due to natural atmospheric pressure variation, wind interaction with the sensor, sensor error, sensor calibration and long-term drift, telemetry bit errors, and other factors (IPAB).

Air temperature measurements are typically conducted with a thermistor contained within a radiation-shielded housing about one meter above the ocean surface. Thermistors also measure sea surface temperature. A source of error is the heating effect produced when solar and infrared radiation strike the air temperature transducer (IPAB). This error can be significant during direct insolation and low winds, but is no means for proper correction.


1.1 Buoy Data library of Alfred Wegener Institute Polar and Marine research

The AWI deploys automatic drifting sea ice buoys in the Antarctic since 1986. Sensor equipment strongly depends on research objectives and influences the format of provided data. Standard parameters are buoy position, atmospheric pressure, air temperature and drift velocity components. 

In a first step, corrections for outliers and calculations of drift velocities are performed. Regularly spaced time series are then generated using cubic splines with a time resolution of three hours. Gaps up to a few days are filled by linear interpolation, longer gaps indicated as data inexistent. Some Antarctic weather stations included in this archive, and can be identified from their constant position.

AWI data are preferred to their reformatted version in IPAB- and AAD-libraries. AWI offers the buoy data used in the “Atlas of Antarctic Sea Ice Drift” in higher temporal resolution on the   PANGAEA web site (PANGAEA - Network for Geological and Environmental Data). For use with existing programmes and because of good documentation and error adjustment the atlas anyway uses the interpolated 3-, 6- or 12 hourly data.


1.2 Buoy data archive of International Project of Antarctic Buoys

The World Climate Research Programme (WCRP) International Programme for Antarctic Buoys (IPAB), through participating research organizations in various countries, maintains a network of drifting buoys in the Antarctic sea ice zone to support a better understanding of sea ice motion, meteorology, and oceanography. The IPAB Antarctic Drifting Buoy Data archive, presently spanning the years 1995 to 1998, includes measurements of buoy position, atmospheric pressure, air temperature, and sea surface temperature.  Data were collected from buoys initially deployed in three study regions: East Antarctica, the Weddell Sea, and the Bellingshausen, Amundsen, and Ross Seas and are provided as instantaneous raw values, daily averages, (every 24 hours), and three-hourly averages. This project exclusively uses buoy positions, atmospheric pressure and air temperature from the three-hour averages.  


1.3 Buoy data from Australian Antarctic Division 1985-2001

At the end of the Project “Atlas of the Antarctic Sea Ice Drift” Ian Allison and Petra Heil (Australian Antarctic Division) delivered relevant buoy data from their archive to us, including buoys from other institutes e.g. AWI, Lamont Doherty Earth Observatory, Geophysical Institute, Univ. Alaska, Fairbanks etc., too. In cases where AAD, IPAB and AWI data overlap, the hierarchy prefers AWI to IPAB to AAD, because the AWI archive contains the most original AWI buoy data, and AWI and IPAB data are well tested at our institute, while AAD buoys are just included to our archive during the last days of the project and thus sparsely checked for individual errors, which haven’t been corrected for other buoys before. Usually the archive offers 24-hourly data, an interpolated version with  2h, 3h or 6h temporal resolution, and the original data with irregular temporal spacing.


2. Reformatting buoy data to a uniform possibly raw format

The AWI provides its buoy data depending on the sensor configuration in three different data formats separated in one file for each month. Details about the data are described by Kottmeier et al. (1996).
The IPAB archive separates the buoy data in files per years (one buoy per year in each file), using a uniform data format provided as 24hourly, 3hourly or irregular spaced data. Both archives provide three-hourly interpolated data, with some exception of only 6- or 12-hourly interpolated data from AWI.  

AAD buoys are available with the original, irregular temporal spacing, daily means and an interpolated version with time steps of 2h, 3h or 6h between the data lines, depending on the measuring year. Last mentioned ar used for producing the Atlas’ raw format data.

The sea ice atlas requires possibly long, continuous time series of drift components and buoy positions. Dates, buoy positions, air temperatures, air pressures, and drift components are reformatted to files of one year duration. File names are combinations of the year of operation and WMO buoy ID (yyyy_{buoy ID}_3.txt). For IPAB buoys no drift components are available.

Data example:

!         IDall        yearall        dateall       Datumall          Monat         lonall         latall      udriftall      vdriftall           pall        tairall      
        3293.000       1986.000      212.000          1.000          8.000        358.132    -63.746   -0.030         -0.050        972.070     -18.340 
        3293.000       1986.000      212.250          1.250          8.000        358.139    -63.784    0.170         -0.190        969.900     -11.850
        3293.000       1986.000      212.500          1.500          8.000        358.196    -63.796    0.110          0.010        973.640     -10.710  

Fortran Format: 11F10.3 (for IPAB only 9F10.3)


3. Association to roman date

IPAB uses day of year (numbers from 1 to 365, leap years 366) for temporal assignment. AWI archive primarily uses Julian days beginning with “0.0” for the beginning of January 1. Some data exceptionally start with day “1” (especially files older than 1994). Date association (day of year (start day “1”) or Julian days (start day “0”)) is reconstructable manually depending on separation into monthly files by AWI (first and/or last date of the monthly files of one buoy).

The raw format  to download from this sites contains all data referenced to Julian days starting with day “0”. Additionally a roman date (day, month, year after Roman calendar) has been included to simplify correlation to SSM/I sea ice drift data. Three-hourly interpolated IPAB data are available consequently. More than half of AWI data are three-hourly available, too. Especially older buoys from AWI archive are solely provided as 6- or 12-hourly data. Those have not been interpolated to three hours, but are marked in the file name.

The file name convention is: {year}_{PPT#}_{Interpolation intervall}.txt. Files contain PPT#, year, Julian day, roman date separated in day of month and month, longitude, latitude, meridional drift component, longitudinal drift component, air pressure, air temperature. 


4. Calculating daily means

From the uniform dataset daily means were calculated, excluding temperature and pressure being not directly used in the atlas. Drift components and positions and roman dates have been extensively been checked to emerge calculation errors resulting from data gaps, double data lines, outliers in parameter time series.  

To achieve a consistent data archive daily means are even calculated where 24hourly data exist (IPAB and AAD archives) because comparison of calculated to existing daily drift means obviously differ, and procedures for calculation of the IPAB and AAD daily means is not documented in detail.


Data example:

! IDallmean dateallmean Datumallmean   Monatallmean  yearallmean lonallmean latallmean udriftallmean vdriftallmean & Datensaetzeall
            3293.000        212.000          1.000          8.000       1986.000         -1.816        -63.780          0.105         -0.055          4.000
            3293.000        213.000          2.000          8.000       1986.000         -1.593        -63.833          0.160         -0.140          4.000
            3293.000        214.000          3.000          8.000       1986.000         -1.216        -63.993          0.245         -0.183          4.000
            3293.000        215.000          4.000          8.000       1986.000         -0.947        -64.018          0.050          0.087          4.000

Fortran Format: 10F10.3

The parameter “Datensätze” is an indicator for the statistical significance of the means. It counts the contributing measurements included in the means after eliminating outliers in drift components. After uniformly formatting all buoy data daily means have been calculated.  Dummy values in buoy positions and drift components are eliminated by linear interpolation between the measurement before and after the gap. For longer transmission gaps this simple method may lead to unrealistic values. Higher accuracy requires more complicated programmes, but visual examinations and comparisons with neighbouring buoys from the archive didn’t show obvious problems for simple linear interpolation.
Daily means of longitudes are corrected for errors at transitions between 0° and 360° and 180° and -180°. Longitudes greater than 360° are transferred to 0° to 360°.


The buoy data were processed in four  groups: AWI1, AWI2,  IPAB (including most of AAD) and special AAD.

IPAB data require the least correction effort, because of their uniform format.
AWI1 are the older AWI buoys usually deployed as buoy arrays and show the highest necessity for corrections. AWI2 are mainly later single deployed AWI buoys. “AWI buoys” therefore only means “from AWI archive”. Strictly temporal separation of AWI1 and AWI2 is not possible. The series are cleared from double data lines (indicator: date). AAD buoys in general require same corrections as IPAB buoys. Data where in absence of the interpolated data file the irregular temporal resolution data had to be used where treated like IPAB too.

All data of one buoy and one year are separated in subseries of one day from 0 UTC to 24 UTC. For AWI1 dummy values in buoy positions and drift components are eliminated by linear interpolation between the measurement before and after the gap (Maximum for longitude:360°, Max. for latitude: 0°, max: for udrift/vdrift: modulus 5).

For longer transmission gaps this simple method may lead to unrealistic values. Higher accuracy requires more complicated programmes, but visual examinations and comparisons with neighbouring buoys from the archive didn’t show obvious problems for simple linear interpolation. Modulus greater than 10 m/s in daily means of drift components indicate unrealistic interpolations of invalid drift components. Therefore daily means of drift components are calculated from buoy position. If this new drift velocities again have a modulus greater 10 m/s, the hole dataline is omitted, suggesting wrong buoy position.  

Finally longitudes and latitudes larger than 500° (used to indicate erroneous or missing data) and outlying daily mean drift components with absolute values greater than 5 m/s (ship motion before deployment) are eliminated from the daily mean series of one buoy year mean.

For AWI2 and IPAB data, the first and last day of one buoy year were treated separately by omitting dummy values before calculation of means (instead of interpolating).  

Another standard problem is that the first and last measured positions of each buoy are often embedded in a long interval no reported data (filled with dummies). Such position measurements in some cases don’t seem to be trustable compared to later/earlier transmitted buoy positions. Especially doubtful positions are left in the raw data but excluded in daily means.


4.1  Correction and Calculation of drift components

For AWI data, drift components udrift  and vdrift exist in raw data format with  three-hourly resolution calculated from buoy
positions at the University of Hannover and at AWI  (Kottmeier et al., 1997). For Sea Ice Atlas daily means have been calculated from buoy positions via

leads to:


where l(t2) – l(t1) and f(t2) - f(t1) are the changes of longitude and latitude between t1 and t2,  RE  is the radius of the Earth and  is the mean of start and end latitude of a buoy.  l(t2) and f(t2) are longitude and latitude at the end of one day, l(t1) and f(t1) are longitude and latitude at the beginning of one day.

Where daily means of a single drift component are erroneous (larger than 10 m/s), replacement values are calculated from buoy positions with the same method as for IPAB data. Regrettably such drift components are not identifiable via “Datensätze “. The critical values of 5 m/s and 10 m/s for drift components are arbitrary, but seem to correctly differentiate between daily series with many (critical value 10 m/s) or few (critical value 5 m/s) invalid data.

For three of the AAD buoys of 1987 (1987_1165a, 1987_1165b,  1987_1166) only 24 hourly means are available. Drift calculations for these buoys were not performed because drift vectors calculated from buoy positions from two different days are not easily to assign to one of both used days without time, and including a longer period not directly comparable to other drift vectors.

5. Missing data

The buoy data archive for the Atlas of Antarctic Sea ice drift aimed to be as complete as possible. Compared to the AWI Technical documentations and the IPAB Buoy lists (IPAB), some data are missing to both archives completely, however. Reasons are not known to the author. Furthermore IPAB Archive includes AWI buoys, too. In cases of overlap, the more original AWI data are preferred.

Interruptions of transmissions of one ore more sensors are marked with dummy values by the data providers (AWI, IPAB). AWI uses 888.88, IPAB replacement method are 99.99, 999.999 and so on, depending on the format fixed for a special parameter.


6. Data Errors and Problems

  • Air temperature seemed to be hull temperature or temperature of the ocean surface for AWI buoys 1995_8060, 1996_14955, 1996_14956, 1997_8059, 1997_8060, 1997_8061, 1997_8064 and 1997_8068.
  • The erroneous temperatures have been replaced by correct values after personal communication with Lutz Sellmann, being responsible at AWI for buoy data analyses and the problem has been reported to IPAB. The data on the IPAB-ftp-server (three- hourly data) haven’t yet been repaired when finishing this project. Above mentioned measurements therefore will differ from IPAB-data concerning air temperatures.
  • AWI-Buoys 1987_101113, 1987_111213 and 1987_121416 are not included in the AWI data documentation (Kottmeier and Lüdemann, 1997). Missing documentation together with special name conventions prohibit certain allocation to year, while buoy year is not part of parameters in the transmitted data.
  • Year assignment is fixed from file naming conventions for buoys 1990_5893 and 1991_5892, although they are not documented by AWI (Sellmann et al., 1997).
  • Buoys 1992_1, 1992_2, 1992_3, 1992_4, 1992_5 have inconvenient PTT#s and are not included in (Sellman et al., 1997). Reasons for both are not known to the author.

  • 1995_24774, 1995_24777 are each combinations of two IPAB files with identically PTT#.


Some of the AAD buoys combine different drift paths to one PTT#. Such cases are indicated with an a and b character after PTT# in the file name, e. g. 1987_1165a_24pT.txt.

AAD and IPAB buoys are used in the interpolated (2h, 3h or 6h depending on availability) versions. Sometimes the Interpolation seems too fail, resulting in buoy longitude differences of more than 100° for 3h intervals. Such highly doubtable longitudes are replaced by dummies by our formatting routine.