Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

AI weather tool cuts moisture errors where balloon data is scarce, study finds

Study title: Satellite and Atmospheric Data Fusion for Improved Characterization of Thermodynamic Profiles

Evidence type
Controlled or validated comparison
Sample
n = 68,887
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2539cite or check it

Study title: Satellite and Atmospheric Data Fusion for Improved Characterization of Thermodynamic Profiles

Researchers built TAC-Thermo, an AI system that fuses satellite retrievals with numerical weather prediction output to sharpen temperature and moisture profiles of the atmosphere up to about 12 km, with a focus on regions that have few weather balloon launches. It was trained on 68,887 global radiosonde and collocated satellite profiles from a two-year period.

Read the study abstract

Abstract This paper presents Total Atmospheric Characterization–Thermodynamics (TAC-Thermo), a global artificial intelligence (AI) data fusion system to improve satellite-retrieved thermodynamic profiles with a focus on data-sparse regions. TAC-Thermo uses available satellite retrievals and numerical weather prediction model output to upscale satellite-retrieved temperature and moisture profiles that can capture thermodynamic features (up to ∼12 km) on sigma vertical coordinate levels (i.e., 1.0–0.2). TAC-Thermo…

The gain

In a validation against radiosonde soundings, the TAC-Thermo AI system had smaller dewpoint temperature errors than NUCAPS and the GFS at every sigma level, and filled in sounding data where radiosondes are absent.

In brief

Checked against radiosonde soundings in all climate regimes, its dewpoint errors were smaller than those of NOAA's NUCAPS and the Global Forecast System at every sigma level analysed. For temperature, its largest standard deviation error (1.97°C) was close to the best values from NUCAPS (1.76°C) and the GFS (2.28°C).

The profiles can feed weather-model data assimilation and, as 0.25° gridded maps every 3 hours, help forecasters spot temperature and moisture inversions and cloud layers. The authors say the soundings could also give product developers and mission planners a global record of the planetary boundary layer.

Main points

  1. TAC-Thermo was trained on 68,887 global radiosonde and collocated satellite profiles (microwave, hyperspectral infrared and radio occultation) over two years.
  2. Its largest temperature standard deviation error was 1.97°C at the 0.96 sigma level, compared with a minimum of 1.76°C for NUCAPS and 2.28°C for the GFS.
  3. Dewpoint standard deviation errors were lower than NUCAPS and the GFS at all sigma levels in the analysis.
  4. In a 24-hour test of the gridded product against radiosondes, 98% of sigma levels had an absolute median temperature error under 1.0°C.
  5. For dewpoint temperature, 85.3% of sigma levels had an absolute median error under 1.0°C in the same test.
  6. Gridded products give global feature maps at 0.25° resolution every 3 hours.

The full story

TAC-Thermo combines available satellite retrievals with numerical weather prediction output to upscale temperature and moisture profiles on sigma vertical coordinate levels from 1.0 to 0.2, capturing features up to about 12 km. It targets data-sparse regions where radiosonde launches are limited.

The system was validated against radiosonde soundings and benchmarked against NUCAPS and GFS analysis profiles in all climate regimes. Its temperature errors were comparable rather than uniformly better: its worst error sat near the best values of the other two datasets.

Gridded versions were examined over a 24-hour period against radiosonde data. The authors say the products support operational forecasting by identifying temperature and moisture inversions and cloud layers, and that the soundings support data assimilation in weather prediction models.

Sources

1 on file
  1. Artificial Intelligence for the Earth Systemsdoi:10.1175/aies-d-25-0092.1
    Peer-reviewed2026-10-01

Talk it over

Four questions to start onWho gets this benefit, and who was left out of the study or story?

Start with a question

AI weather tool cuts moisture errors where balloon data is scarce, study findshttps://truvace.com/item/satellite-and-atmospheric-data-fusion-for-improved-characterization-of-thermodyn

  1. Who gets this benefit, and who was left out of the study or story?

    The gain In a validation against radiosonde soundings, the TAC-Thermo AI system had smaller dewpoint temperature errors than NUCAPS and the GFS at every sigma level, and filled in sounding data where radiosondes are absent.

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a controlled comparison with a sample of 68,887, from Artificial Intelligence for the Earth Systems.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat The gridded product was examined for only a 24-hour period, and its temperature errors were close to, not clearly below, the best NUCAPS and GFS values.

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Climate