Customer reported a potential issue with GFS tropopause height

[HuiyaChuang-NOAA]

In Aug 2024, an user reported GFS tropopause height was higher than the obs tropopause height derived by NCAR using radio sound data. See attached email.

National Oceanic and Atmospheric Administration Mail - Re_ Significant bias in NOAA-derived tropopause height.pdf

[HuiyaChuang-NOAA]

@ChristopherHill-NOAA was assigned to work on this issue. Adding @WenMeng-NOAA and @GeorgeGayno-NOAA for awareness. Huiya corresponded with NCAR scientist and found the algorithm which NCAR used to derive tropopause height to be significantly different from the WMO tropopause height definition used by EMC. She sent the attached email reply to the customer who reported this issue. NOAA Mail - Re_ Significant bias in NOAA-derived tropopause height.pdf

Chris is preparing a slide to document his findings.

[ChristopherHill-NOAA]

Findings to date:

• The user indicated that the tropopause height represented by the GFS was anomalously high over some regions with some frequency, specifically citing a difference with the NCAR calculated value from the radiosonde observation over southern Texas on 1200 UTC 23 August 2024.
• A study from Homeyer et al 2010 highlighted a subset of anomalously high tropopause height values represented by the GFS compared against available radiosonde observations during the period of April-June 2008. The authors determined that the GFS had insufficient vertical resolution - at the time - to capture the primary tropopause level where the local temperature lapse rate was observed to be nearly equal to the critical value (-2 K / km) identified in the WMO (1957) algorithm for defining the tropopause.
• Compared to the current operational version GFS, the current version of the ECMWF has a similar model top (~ 80 km) and has more vertical levels (137 vs 127), from which a finer vertical resolution could be inferred. The ECMWF post-processing system calculates a tropopause pressure value based on a similar application of a temperature lapse rate algorithm, while a calculated potential vorticity value equal to 2 PV units is considered a more appropriate indication of the tropopause height for latitude positions > 30º N and S.

[HuiyaChuang-NOAA]

Chris will POC or website for ECMWF products and inform Huiya. Huiya will ask around on ECMWF data availability on wcoss2

[HuiyaChuang-NOAA]

Meeting notes from 10/7/2024 tag-up

1. Chris will plot comparisons of GFS vs ECMWF tropopause height and share with Huiya and Wen
2. If ECMW tropopause height does not have high anomaly, Huiya will reach out to ECMWF to inquire about their algorithm

[ChristopherHill-NOAA]

Bo Cui and Jiayi Peng provided insight of ECMWF datasets that are ported daily to WCOSS. A sample of ECMWF data files from 2024102712 was selected for review for potential inclusion of tropopause data.

Specific focus was made with GRIB file U1D1027120010280001, which contained deterministic model output for forecast hour 12 at 0.25º resolution from the ECMWF. A total of 110 variables were identified from the datafile through 'wgrib' and 'wgrib2' analysis. Of the 110 variables, none was found to explicitly contain tropopause information, and 6 variables were left 'undefined' by wgrib/wgrib2 analysis. The GRIB file was converted to a GEMPAK grid file for data analysis, and none of the 6 undefined variables was found to feature values resembling height or pressure data - let alone to resemble tropopause information.

The tropopause pressure variable is featured among many other variables in the complete deterministic dataset available directly from ECMWF. This complete dataset is stated to be available to ECMWF-member nations only, and otherwise available via license (commercial or research) to other entities. Specific POC information for potentially acquiring such data will need to be uncovered.

An October 2021 implementation document from the ECMWF website provides information pertaining to the introduction of their 'tropopause pressure' variable and its computation, which could potentially yield a method for reconstructing tropopause pressure values from freely available ECMWF model data - and thereby allow a direct comparison of GFS vs ECMWF tropopause data.