Changes to the AMV BUFR sequence

by James Cotton — last modified May 13, 2014 08:46 AM

This page aims to coordinate and summarise discusions on proposed changes to the AMV BUFR sequence.

IWW11.1. All AMV producing centers are encouraged to investigate how to provide enhanced situation-dependent error estimates of wind and pressure with new derivation techniques. NWP centers are encouraged to work with producers on the evaluation.

Page maintained by James Cotton

Email discussion: iwwg_errors(at)ssec.wisc.edu distribution list

\\

A proposal has been put forward by NOAA for a new BUFR sequence for use with future GOES-R data and includes some new entries which have been proposed for testing and validation puposes. The original proposal document can be viewed at http://www.wmo.int/pages/prog/www/ISS/Meetings/IPET-DRMM_Tokyo2013/Documents/IPETDRMM-I_Doc3-2_4_SatelliteWind.docx.

This is a topic that would benefit from discussion within the whole winds community to make sure that proposed changes fit with the future plans of the main AMV producers and users. Therefore, feedback was requested from members of the IWWG on the proposed BUFR sequence, as well as suggestions of what other parameters arising from the derivation which could be provided/useful (e.g. for input to obs error schemes or QC). The idea was to have an email discussion via the smaller "iwwg_errors" email list and then for the main points to be summarised on this wiki page.

Document - Winds_BUFR_Sequence_20140508.docx

The proposed sequence has been updated largely following inputs received from the Met Office and ECMWF. See the document for full details of the changes in this and previous iterations.

Document - NOAA_revised_20131112.docx

The height assignment method code table has been added back via a new descriptor (0-02-162) called "Extended height assignment method". This is 6 bits in length (plenty of room for growth) and the first 14 entries contain what is currently in 0-02-163 ("Height assignment method"). More can be added e.g. entry 15 is "optimal estimation". Still need to decide the best way to define the method used to analyse the pixel level cloud heights in the target scene and arrive at a final AMV height assignment. Another descriptor (with an associated code table) may need to be added to do this.

Document - EUMETSAT_inputs_to_change_the_BUFR_wind_sequence.docx

More detailed and technical comments have been received from EUMETSAT in response to the original proposal. This also includes a new revised wind sequence.

NWP centres welcome the idea of changing the template to include additional information.
The template should fit with plans of other producers and be suitable for both geostationary and polar wind products.

Any deviations from the former sequence should be fully explained in the proposal.

Do we really know which parameters are the most useful to add? Studies such as Nebuda et al. (conference paper) from IWW11 and the work on expected error give some indications.
The coordinate system should be coherent for the whole BUFR message: U/V wind components or speed/direction. For NWP, U/V components are more practical from a statistical viewpoint (better Gaussian distribution). However there is also some support for using speed/direction instead: more conventionally, a wind direction may be measured independently from the speed (vane/anemometer) and each measurement has its own error.

Is it useful to include model parameters? NWP centres have access to more up-to-date fields, probably at higher vertical and spatial resolution than was used in the AMV processing. This information could be useful for non-NWP users but are they using BUFR data?
A layer depth of +/-200 hPa is too large for calculating the model wind shear and temperature gradient. +/- 100 hPa or even +/- 50 hPa is more appropriate. How to handle for winds near the tropopause and surface?

The "Height assignment method" indicator has been deleted. This entry should be maintained, at least for those centres for which it remains meaningful. It is also required for the derivation of winds in the WV channels.

For derivation schemes that use pixel CTH from a L2 cloud product a single HA method no longer makes sense. In this case there may still be ways to indicate a 'fundamental' retrieval method used to retrieve the cloud top heights of the largest cluster (nested tracking) or the pixels identified as being used in tracking (CCC). The current table is full so may have to create an "Extended height assignment method" descriptor to include the current entries plus any new ones.

May also include the method by which the CTP in the target scene are used to assign AMV heights (e.g. median CTPs of selected pixels).

It would be preferable, as at present, to still have both the QI with and without the FG check.
The 2 QI values should be followed by an estimate of the expected error in the wind components and pressure.

Does the low level inversion flag indicate that an inversion-height assignment has been applied, or simply that an inversion has been detected in the model profile?
A new flag or bit table could be created, merging the proposed low level inversion flag, a proposed orographic flag and possibly a inversion at the surface flag.
Reintroduce surface type (land/sea etc)

Instead of 'size of largest cluster' use a more generic wording to store the number of pixels used in the tracking/HA
The SZA in constant for GEO but for LEO winds we need to store one SZA for each image used

Model best-fit pressure

Is tracking correlation coefficient included?

Number of clusters from nested tracking
A measure that characterises the shape of the correlation surface around the maximum. E.g. is the maximum broad or well-defined? Are there secondary maxima?
Cloud depth e.g. from OCA