(Updated description including developments for version 6.0 (v2018), with examples)


Table of contents

1. Goal of HRW algorithm
2. Summary description of HRW algorithm 
3. Description of HRW inputs 
4. Description of HRW outputs
5. Examples of HRW output visualization


Access to “Algorithm Theoretical Basis Document for the Wind product processors of the NWC/GEO (HRW v6.0)” for a detailed description


1. Goal of HRW algorithm

NWC/GEO-HRW algorithm aims to provide detailed sets of Winds (Atmospheric Motion Vectors, AMVs) and Trajectories from:

  • Up to seven MSG/SEVIRI satellite channels (HRVIS, VIS06, VIS08, IR108, IR120, WV062 and WV073).
  • Up to three GOES-N/IMAGER satellite channels (VIS07, IR107 and WV065).
  • Up to six Himawari-8/9/AHI satellite channels (VIS06, VIS08, IR112, WV062, WV070 and WV073).

Note: The adaptation to GOES-R/S satellites is now under way and will be delivered as a patch for NWC/GEO v2018 software package throughout the year 2019.

The AMVs and Trajectories are calculated through the displacement of cloudiness features in successive images from all these channels, and of humidity features in successive images from the water vapour channels. The AMVs and Trajectories can be calculated 24 hours a day, considering:

  • 15 minute image cycles ("Nominal scan mode") or 5 minute image cycles ("Rapid scan mode") with MSG satellites.
  • 15 minute image cycles (in the "Continental United States" region) or 30 minute image cycles (in the "North America" region) with GOES-N satellites.
  • 10 minute image cycles with Himawari-8/9 satellites.

The product includes pressure level information and a quality control flagging, giving some indication of its error in probabilistic terms. It has been developed by AEMET (the Spanish National Weather Service) in the framework of the EUMETSAT's "Satellite Application Facility on support to Nowcasting and Very short range forecasting" (NWC SAF).

The AMVs and Trajectories can be calculated for up to two different scales at the same time: "Basic scale" and "Detailed scale". The corresponding dimension of the tracers to be tracked can be defined by the user, with a default value of 24 pixels for the "Basic scale". Many other processing parameters can also be configured, in particular the region and the satellite channels to be processed for the AMV calculation.

This product can be useful in near real time applications including:

  • Nowcasting: Watch and warning of dangerous strong wind situations; Monitoring of the general flow, of convergence and divergence at low and high levels, or of small scale circulation and wind singularities.
  • Forecasting applications: Assimilation of HRW outputs in NWP models, or in applications where the real displacement of atmospheric features is needed.

Several different HRW outputs (as BUFR or netCDF bulletins) can be provided. One of them is the common traditional AMV BUFR output, used by other AMV products at the WMO Global Telecommunications System. Winds, trajectories and the related parameters are calculated with a level 2 of processing. A level 3 of processing (as a grid interpolation of the winds or a meteorological analysis based on the data) is not included. 


2. Summary description of HRW algorithm

Eight different steps are considered in the processing of HRW algorithm: 

1. Initialization of data: next matrices are read or calculated for the start of the processing, by the NWC/GEO-NWCLIB library, for the running slot and region in which HRW algorithm is calculated:

  • Normalized reflectances for the MSG/GOES-N/Himawari-8/9 visible channel images, and brightness temperatures for the MSG/GOES-N/Himawari-8/9 infrared and water vapour channel images.
  • Latitude, longitude, solar and satellite angle matrices.
  • Temperature, wind and geopotential profiles from NWP data.
  • NWC/GEO Cloud type (CT), Cloud top temperature and height (CTTH) and/or Cloud microphysics (CMIC) outputs (if so configured, for a better AMV and Trajectory height assignment).

2. Tracer determination: Two methods are used consecutively:

  • Gradient method: with a fast and efficient search of well defined cloud edges.
  • Tracer characteristics method: which covers holes in the search of tracers with Gradient method, with a longer but still reasonable computing time; two different tests are run based on finding a brightness threshold separating the cloudiness in front from the background, and finding a well defined shape of the feature to be tracked to avoid too linear elements.

An option inside the HRW algorithm process permits that new tracer centres for the running cycle are defined at the final locations of AMVs calculated for the previous cycle. This way, a set of “Persistent tracers” can be successively tracked in consecutive images, and “Trajectories” are defined through the progressive positions of these tracers throughout the time.

Since NWC/GEO-HRW v2018, an option has been implemented to increase the spatial density of tracers and AMVs at low levels. This is done reducing to a half the distance betwen tracers related to low and very low clouds. 

3. Tracer tracking and Wind calculation: One of two well known methods can be used for this process:

  • Euclidean distance.
  • Cross correlation (default option).

The tracking area in which tracers from a previous image are looked for in the current image, can be defined through the extrapolation of the linearly interpolated NWP wind guess for a quicker running of the algorithm. Nevertheless, this option is not used as default one to reduce the dependence from the NWP data. The best three tracking centres are kept, from which one is taken at the Quality Control step.

A “mixed calculation method” is also available since NWC/GEO-HRW v2018, considering at the same time short and long time intervals, through which the tracking process is verified in short time intervals, but the AMVs are calculated considering displacements in long time intervals. This process is useful for the calculation of AMVs with high resolution images, and to improve the quality of the calculated AMVs.

4. Height assignment: Two different methods can be used:

  • Brightness temperature interpolation method: calculates the cloud top pressure and cloud base pressure, interpolating the brightness temperature of the corresponding satellite channel (or MSG/IR120, GOES-N/IR107, Himawari-8/9/IR112 channel in the case of visible channels) to a NWP vertical temperature profile. Then, the AMV/Trajectory pressure is defined through one of these two values, taking into account the cloud type related to the tracer.
  • Cross Correlation Contribution (CCC) method (default option): defines the AMV/Trajectory pressure, considering only the pressure of the pixels contributing most to the image correlation (which is available only for Cross correlation tracking if NWC/GEO CT and CTTH outputs are available). A correction of the pressure level for cloudy AMVs/Trajectories is also implemented, considering the AMV/Trajectory Liquid water path/Ice water path (obtained from the corresponding NWC/GEO Cloud Microphysics output, and related to the depth of the cloud which is being tracked).

5. Quality Control: the “Quality Indicator method”, developed for the AMV calculation at EUMETSAT/MPEF, has been adapted for HRW algorithm. Several consistency tests are if possible computed, considering:

  • Temporal consistency test: vector/direction/speed consistency with neighbour AMVs in the previous slot.
  • Spatial consistency test: vector consistency with neighbour AMVs in the current slot.
  • Forecast consistency test: vector consistency with the NWP forecast winds.
  • Two scale consistency test (only for the “Detailed scale”): vector consistency with simultaneous AMVs from the “Basic scale”.

Three individual “Quality Indices” are calculated for each AMV/Trajectory, considering the normalized statistical fitting functions described by the method and the weighted sum of the partial consistency tests:

  • Quality Index with forecast (QIF): Quality index including the forecast consistency test.
  • Quality Index without forecast (QINF): Quality index excluding the forecast consistency test.
  • Common Quality Index (QIC): defined through a common calculation module used by all AMV centres. The Common Quality Index has shown to provide a real skill in filtering AMVs from different AMV algorithms, for a better consistency. 

6. Orographic Flag: an “Orographic flag” is also calculated, related to the detection of land tracers, and tracers blocked by orography or related to orographic waves (for which the AMVs/Trajectories are not displacing with the corresponding atmospheric flow), with next possible values:

  • 1: AMV wrongly located below the lowest representative pressure level in the vicinity (basically due to microphysics corrections in the AMV pressure). This AMV is always skipped.
  • 2: Very important orographic influence at the current location of the AMV.
  • 3: Less important orographic influence at the current location of the AMV.
  • 4: Very important orographic influence at any previous location of the AMV.
  • 5: Less important orographic influence at any previous location of the AMV.
  • 6: No orographic influence is expected at any current or previous position of the AMV.

7. AMV Selection: a comparison is made among all AMVs calculated for the same tracer, when more than an AMV has been calculated, for the final definition of the AMV related to that tracer (only one AMV per tracer can be kept). This comparison is made considering the behaviour of each one of those AMVs for each one of the following criteria:

  • From the AMV calculation process: tracking correlation.
  • From the Quality control process: temporal, spatial and forecast consistency tests; two scale consistency test also for the “Detailed scale”.
  • From the Orographic flag process: orographic flag.

Possible test values for each case are:

  • 3: When the AMV is the best of all for the corresponding criterion.
  • 2: When the AMV is slightly worse than others for the corresponding criterion.
  • 1: When the AMV is rather worse than others for the corresponding criterion.
  • 0: When the comparison could not be defined for the given AMV.

The AMV selection is based on the AMV with the best comparison tests. In case of draw, the AMV with the best forecast consistency test is selected, or else the AMV with the best tracking correlation.

8. Autovalidation of AMVs: considering requests from the users, NWC/GEO-HRW v2018 is able to calculate validation statistics for the AMVs with the algorithm itself, using as reference NWP forecast winds (in real time or reprocessing processes) or NWP analysis winds (in reprocessing processes).

The validation statistics are written for each running slot in a specific file S_NWC_HRW-STAT_*.txt, including information per channel, per level or as a whole on the number of AMVs (NC), the mean wind speed (SPD), the normalized bias (NBIAS), the normalized mean vector difference (NMVD), and the normalized root mean square difference (NRMSVD) with respect to the reference winds. 

The NWP reference wind at AMV level and best fit pressure level, and the vector differences with both, can also be calculated and stored in the AMV output for further processing and research.


3. Description of HRW inputs

  • NWC/GEO configuration files: HRW Model Configuration file (*.cfm) and Region Configuration file (*.cfg).
  • MSG or GOES-N or Himawari-8/9 satellite data: Full Resolution MSG/HRIT or GOES-N/GVAR or Himawari-8/9/HSD brightness temperatures for the infrared and water vapour channels to be used, and reflectances for the visible channels to be used, for the working region for the current and previous slot.
  • NWP data: Temperature fields and longitudinal and latitudinal component wind fields for as many as possible of the following pressure levels: 1000, 925, 850, 700, 500, 400, 300, 250, 200, 150, 100, 70, 50, 30, 10 hPa. Also geopotential fields in case the Orographic flag is to be calculated.
  • NWC/GEO Cloud Type (CT) and Cloud Top Temperature and Height (CTTH) output for the working region and slot (optionally, for a better height assignment).
  • NWC/GEO Cloud Microphysics (CMIC) output for the working region and slot (optionally, for a microphysics correction of the AMV pressure level, if CCC method is used for the height assignment). This option is not available for GOES-N satellite series.
  • List of Tracers, Predecessor winds and Trajectories calculated at the previous slot (in case they are available).


4. Description of HRW outputs

Depending on configuration, AMVs and Trajectories are provided in the form of BUFR or netCDF bulletins, considering separately each processed region and each tracer scale used (“Basic scale” and “Detailed scale”).

Two different HRW BUFR AMV bulletin formats are available: one similar to the format used by all previous versions of HRW algorithm (named S_NWC_HRW-WIND_*.bufr), and another one equivalent to the format used by other AMV bulletins distributed by other AMV centres through the WMO Global Telecommunications System (named S_NWC_HRW-WINDEUM_*.bufr).

In the first case, which is the default option for HRW algorithm, if Trajectories are calculated, an additional BUFR bulletin (named S_NWC_HRW-TRAJ_*.bufr) is also provided with the content of the corresponding Trajectories.

HRW netCDF bulletins provide information on both AMVs and Trajectories, and are named S_NWC_HRW_*.nc.

Note: An additional BUFR format for AMVs, defined by “BUFR Sequence 310067” was approved by the World Meteorological Organization in November 2017. This new BUFR format was to be implemented as new common AMV BUFR format by all AMV centres. However, the defined sequence was to be redefined to be accepted by all AMV processing centres through a new “BUFR Sequence 310077”, which on 1st November 2018 was still not formally approved. Once the situation is clear on this, NWC/GEO-HRW v2018 code will be properly updated in the Spring 2019 (together with the option for GOES-R/S satellites).


The content of these netCDF or BUFR bulletins includes next parameters:

General processing indicators:

  • Satellite identifier.
  • Generating centre and application.
  • Wind computation method.
  • Origin of first guess.
  • Satellite instrument used.
  • Satellite channel central frequency and width.
  • Year, month, day, hour, minute of final image.
  • Time increment from initial to final image.
  • Satellite cycle of final image.
  • Quality control characteristics.
  • Horizontal and vertical tracer size in pixels.

Processing indicators for each AMV/Trajectory sector:

  • AMV/Trajectory identifier.
  • Predecessor AMV identifier and amount of Trajectory sectors (if any, with option calculating “Persistent tracers”).
  • Horizontal and vertical tracer size in metres.
  • Tracer tracking method (“Euclidean distance” or “Cross correlation”).
  • Initial latitude and longitude.
  • Latitude and longitude increments.
  • Wind direction and speed.
  • Wind temperature and pressure level.
  • Quality indices (QIF, QINF, QIC).
  • Type of tracer (Basic tracer; Detailed tracer related to a wide basic tracer; Detailed tracer related to a narrow basic tracer; Detailed tracer unrelated to a basic tracer).
  • Type of tracer (Cloudy or Clear air tracer) and Height assignment method used.
  • Number of AMVs computed for the tracer (1 to 3).
  • Correlation and Quality tests: which or the partial QI consistency tests (two scale, temporal, spatial, forecast, orographic test) have been applied, and in case several AMVs were calculated for the tracer, if the selected one is better, slightly worse or rather worse than the other AMVs for the considered test.
  • Number of NWP levels used.
  • Number of predecessor AMVs.
  • Orographic flag.
  • Cloud type.
  • MSG/GOES-N/Himawari-8/9 satellite channel used.
  • Correlation (if “Cross correlation” is used for the Tracer tracking).
  • Pressure error (if “CCC method” is used for the Height assignment).
  • Pressure correction (if “CCC method” is used for the Height assignment).
  • Validation NWP reference wind at AMV level: direction, speed, direction difference and speed difference.
  • Validation NWP reference wind at best fit pressure level: direction, speed, direction difference, speed difference and pressure level.
  • Validation flag (against NWP analysis or forecast). 


5. Examples of HRW output visualization

Three different HRW algorithm configuration files (*.cfm) have been defined as default options for the running of NWC/GEO-HRW algorithm, repectively for use with MSG satellites (safnwc_HRW_MSG.cfm), GOES-N satellites (safnwc_HRW_GOESN.cfm), and Himawari-8/9 satellites (safnwc_HRW_HIMA.cfm). 

Next figures show the output of HRW algorithm AMVs for three specific examples with these satellites, in which HRW algorithm has been respectively run every 15 minutes with MSG-2 in the “European and Mediterranean region”, every 15 minutes with GOES-13 in the “Continental United States region”, and every 10 minutes with Himawari-8 in the "China/Korea/Japan region". 

NWC/GEO-HRW v2018 Basic AMVs for “European and Mediterranean region“ (MSG-2, 14 May 2010, 12:00 UTC). Colours based on the AMV pressure

NWC/GEO-HRW v2018 Basic AMVs for “Continental United States region“ (GOES-13, 1 July 2010, 17:45 UTC). Colours based on the AMV pressure

NWC/GEO-HRW v2018 Basic AMVs for “China/Korea/Japan region“ (Himawari-8, 2 April 2018, 00:00 UTC). Colours based on the AMV pressure