Products Generated at the BALTEX Radar Data Centre

CONTENTS

Product definitions
    DBZ
    DBZC
    RR
    WP
    References
Product projections and areas
    Area defintions
Sample data
 

Product definitions

There are currently four BALTRAD products being generated at the BRDC. Their methodologies have been determined during the BRIDGE Pilot Phase. As of October 1, 1999 the BRDC has been creating a set of  homogenous products.

What follows is a brief description of each method. These will be complimented with more complete documentation in due course.  A report containing all algorithms has been prepared (Michelson et al. 2000), and the evaluation of some of them is reported on in Koistinen & Michelson (2002). A review on achievements made to date within the WGR, BALTRAD, and BRDC can be found in Michelson et al. (2002). (see references below)

All products are stored in the HDF5 file format.

NOTE that minor modifications to the algorithms may be introduced during the BRIDGE Base-Line Period and thereafter.
 

DBZ

This is the basic product on which the BRDC's higher level precipitation products are based. It is defined as a 500 m Pseudo-CAPPI image with 2x2 km horizontal resolution and 15 minute temporal resolution, or the closest thing to it. The stored quantity is radar reflectivity factor in dBZ. The pixel depth is 8-bits. Areas void of measurements are flagged.

The Pseudo-CAPPI is provided by all radars in the Nordic countries. The early data from the DWD and Poland can contain echoes from arbitrary scans, either as a result of quality control methods in the generation of the DWD's PL product, or from the MAXCAPPI algorithm used at Radar Katowice which selects the strongest echo in the volume above each cartesian pixel. Starting in early 2005, with the introduction of modern data from the complete Polish network, the IMGW is contributing harmonized CAPPI datasets.

The BRDC's function in managing these products is to collect all of them from the contributing institutes, their transformation to a common projection, and saving them to a common format with common characteristics. DBZ products will not be distributed unless by special agreement.
 

DBZC

This is a composite image generated every 15 minutes with a horizontal resolution of 2x2 km. Input to each DBZC are the DBZ products available for that time. The stored quantity is radar reflectivity factor in dBZ. The pixel depth is 8-bits. Areas void of measurements are flagged. Quality control routines remove pixel values, ie. they are set to zero, as opposed to flagging them.

The algorithm used in compositing the DBZ images uses a high resolution digital elevation model which is used to determine the height above the Earth's surface of each pixel in each input DBZ image. Each pixel in the composite receives the value from that radar whose pixel above that point is the shortest distance to the Earth's surface. This is used as a means of minimizing border effects in composite imagery.

A simple algorithm which makes use of Meteosat brightness temperatures and analyzed 2-m temperature fields is used to try to determine whether radar echoes can originate from precipitating clouds (Michelson and Sunhede, 2004). If the difference between the brightness temperature and 2-m temperature is greater than 20 degrees C, then any radar echoes in these regions are assumed to originate from potentially precipitating clouds and are kept; otherwise radar echoes are rejected. The use of this method is only effective during summer conditions but this is when radar data in the region is most frequently contaminated by anomalous propagation echoes. Starting in early 2006, this method has been replaced with a more modern alternative, outlined in Michelson (2006), which uses cloud-type products from the NWCSAF using data from the Meteosat Second Generation platforms. The MSG-based data are higher resolution in both time and space. The new quality control is also very simple: radar echoes are removed in areas classified as cloud-free.
 

RR

This is an accumulated precipitation analysis product based on DBZ products and synoptical observations. It is based on the method presented by Koistinen and Puhakka (1981) with a number of modifications (Michelson et al. 2000).

SYNOP for 6 and 18 UTC are accessed each day. For these times, a 12-hourly accumulated precipitation sum is derived for each radar. These sums are composited using the same algorithm as is used to derive the DBZC product. Gauge accumulations are corrected using a slightly modified version of the Dynamic Correction Model presented by (Førland et al, 1996) before they are used for the gauge adjustment (Michelson, 2004). A general relationship between log(gauge/radar) and range from the radar is derived using radar and gauge pairs from a 7-day moving time window. Depending on the number and quality of SYNOP available, this relationship can either be based on a parabolic regression or just the average log(gauge/radar) value. In areas with a relatively high gauge density, these are weighted higher in the analysis with the radar sum. In areas with low gauge densities, the general relationship dominates. (Only points from the current 12-hour term are used when determining this.) The result is a field in which the quantitative accuracy is largely determined by the gauge values and in which the spatial distribution is determined by the radar data. The regression coefficients can be saved and reused given situations with too few and/or too poor quality SYNOP data. For areas outside radar coverage, a convensional optimal interpolation algorithm is used together with the analytically derived correlation length and a fixed structure function shape. The results from the interpolated points are merged with the gauge-adjusted radar accumulation.

NOTE that, as of early 2006, the full spatial adjustment has been phased out, since its impact on the final result was only marginal, and the optimal interpolation of surface observations has also been removed.

NOTE also that a quality indicator field has been introduced during 2006 which accompanies each RR product. The information contained in this field is outlined in Michelson (2006).

At each following three hour interval up to the next 12-hour analysis, a 3-hourly composite accumulation analysis is generated based on the previously determined relationships between log(gauge/radar) and range.

Twelve-hour analyses are generated to the BALTEX area and all data outside the Baltic Sea's drainage basin is masked out. Three-hour analyses are generated to the BALTRAD area. See area definitions below.

NOTE that, as of early 2006, all RR products are generated to the BALTRAD area only.

Horizontal resolution is 2 x 2 km and pixel depth is 32-bits.

An evaluation of the RR product's accuracy can be found in Koistinen & Michelson (2002).
 

WP

Doppler radars are actually better suited to measuring winds than measuring precipitation and this has been exploited by many BALTRAD radars. These radars generate vertical profiles of wind speed and direction using either the Velocity Azimuth Display (VAD) or Velocity Volume Processing (VVP) techniques. In this context, the BRDC's function is to collect as many of these profiles as possible, store them in a common format, and make them available to research activities, for example data assimilation of these winds into numerical weather prediction models.

The vertical resolution can vary among radars, as can the number of stored quantities. From Swedish and Finnish radars the wind's direction, speed, standard deviation of direction, and standard deviation of speed are stored at the BRDC. These profiles are generated every 15 minutes. The values are stored in 32-bit depth.

NOTE that, with the increased availability of WRWP (Weather Radar Wind Profiles) over the GTS, as a result of efforts in Eumetnet OPERA, the BRDC is still receiving wind profiles within NORDRAD and POLRAD, but they are only being made available by special arrangement.
 

 References

Førland E.J., Allerup P., Dahlström B., Elomaa E., Jonsson J., Madsen H., Perälä J., Rissanan P., Vedin H., and Vejen F., 1996: Manual for operational correction of nordic precipitation data. DNMI Report nr. 24/96. Norwegian Meteorological Institute. P.O. Box 43, Blindern, Oslo, Norway. 66 pp.

Koistinen J. and Puhakka T., 1981: An Improved Spatial Gauge-Radar Adjustment Technique. Preprints 20th AMS Conf. on Radar Met. Nov. 30-Dec. 3, Boston, MA. pp. 179-186.

Koistinen J. and Michelson D.B., 2002: BALTEX weather radar-based products and their accuracies. Boreal Env. Res. 7. p. 253-163.

Michelson D.B., Andersson T., Koistinen J., Collier C.G., Riedl J., Szturc J., Gjertsen U., Nielsen A., and Overgaard S., 2000: BALTEX Radar Data Centre Products and their Methodologies. SMHI Reports Meteorology and Climatology RMK Nr. 90. SMHI, SE-601 76, Norrköping, Sweden. 76 pp.

Michelson D.B., Koistinen J., Bennartz R., Fortelius C., and Thoss A., 2002: BALTEX radar achievements at the end of the main experiment. Proc. ERAD (2002). Copernicus GmbH. p. 357-362.

Michelson D.B. and Sunhede D., 2004: Spurious weather radar echo identification and removal using multisource temperature information. Meteorol. Appl. Vol. 11, No. 1, p. 1-14.

Michelson D.B., 2004: Systematic correction of precipitation gauge observations using analyzed meteorological variables. J. Hydrol. Vol. 290, p. 161-177.

Michelson, D., 2006: The Swedish weather radar production chain. Proc. ERAD 2006: online PDF.
 

Product projection and areas

All spatial BRDC products are created in a Lambert Azimuthal Equal Area projection with the following characteristics:

origin: 20 degrees east, 60 degrees north

pixel spacing: 2000 x 2000 metres

spherical Earth model with a radius of 6370997 metres

Each product has its own definition specifying the centres of the south-westernmost and north-easternmost pixels in so-called surface coordinates which are related to the projection's origin. Surface coordinates are expressed in metres in both X and Y dimensions. This definition is the product's so-called area_extent attribute which is documented in the BALTRAD file format.

Each product has a related definition specifying the exact geographic coordinates of the south-western and north-eastern image corners. This definition is the product's so-called corners attribute and is also documented in the BALTRAD file format.
 

Area definitions

The BALTRAD and BALTEX product areas cover the BALTRAD network area and the Baltic Sea drainage basin area, respectively. The area characteristics are:

 
For users who wish to know the exact geographic coordinates for each pixel in products, files containing this information are available here, on-line. These files are written in BALTRAD format, the intention being that users will be able to become aquainted with this format as a preparatory step to managing BRDC data.

Each file contains two layers of data, both stored as 32-bit float values in little-endian byte order. The data is ordered from top (N) to bottom (S) and left (W) to right (E). The first layer contains longitude values and the second layer contains latitude. Two sets of files have been created:

Compressed with ZLIB:

BALTRAD (6.1 Mb)

BALTEX (5.9 Mb)

Uncompressed:

BALTRAD (7.8 Mb)

BALTEX (7.5 Mb)

Reduced size JPEGs of the same (for simple visual verification of the fields):

BALTRAD (longitude, latitude)

BALTEX (longitude, latitude)

Sample data

BALTRAD products are generated in delayed mode and can be viewed here around 2.5 days after their valid times. Not all products will be viewable on the web in this fashion. The idea is to showcase each day's highlights. This area is under construction and new products will hopefully turn up soon.

Products

Composite animation based on DBZC products
12-hour accumulated precipitation (RR)
Time-height wind profile plots based on WP products can be viewed at the Eumetnet Winprof CWINDE Data Hub.

If you keep getting the same day's content in your browser window, use the Reload button to get the latest stuff. If you still have problems then check your browser's Preferences.