Coordinated response mitigates loss of aircraft-based weather data

A coordinated response involving EUMETNET (a network of 31 European national meteorological services), national meteorological services and private companies has helped to mitigate any adverse effects of the COVID-19-related loss of aircraft-based observations on weather forecasts. In March and April, due to the COVID-19 pandemic, there was a sharp drop in flights and thus in the aircraft-based observations available to weather prediction centres. The continued availability of complete sets of satellite observations from EUMETSAT, ESA and other space agencies ensured that there was no severe impact from the loss of aircraft observations, as satellite data remain the most important observations. Aircraft reports include temperature and wind and in some cases humidity and turbulence. They are used together with many other observations to help estimate the state of the Earth system at the start of forecasts.

Responses to the drop in observations include the use of previously untapped aircraft-based observations; an increase in the number of radiosonde launches from some locations; and the assimilation of additional satellite data. In an example of successful collaboration with the private sector, the companies FLYHT and Spire stepped in to provide additional aircraft-based observations and radio occultation satellite data, respectively.

The impact

Between mid-March and mid-April, the number of aircraft reports received at ECMWF went down by about 75% before levelling out and then slowly picking up again. A data denial experiment run at ECMWF in 2019 suggested that removing all aircraft data has most impact at aircraft cruise levels (250–200 hPa or 10–12 km altitude). Here 12-hour wind and temperature forecasts became about 10% worse in the northern hemisphere extratropics. At the surface, 3- or 4-day forecasts of mean sea level pressure deteriorated by about 3% on average. There was, however, no clear signal in ECMWF forecast verification in April and May that could be linked to the decrease in aircraft numbers. Possible explanations are:

• Day-to-day and seasonal variability in forecast skill make it difficult to pin down the impact of reduced data availability
• The impact per observation increases somewhat with reduced data density, so one would expect to see less than 75% of the ‘no aircraft’ impact from the current configuration
• Additional data from a range of sources, partly made available in response to the drop in aircraft data, helped to mitigate any detrimental effects. 

The response 

EUMETNET has played a key role in coordinating the response to the loss of data among its members. It assessed the impact of the COVID-19 restrictions on the European Composite Observing System (EUCOS) and put in place a coordinated mitigation plan amongst EUMETNET members. Part of the response was to optimise and expand the use of aircraft observations that continued to be available:

• EUMETNET changed its configurations to make the most of any flights that E-AMDAR-equipped aircraft make. E-AMDAR is the EUMETNET AMDAR programme.
• In collaboration with EUMETNET and the UK Met Office, the company FLYHT made its aircraft observations available for free for a limited period of time. On 12 May, ECMWF started actively using some of these data.
• The European Meteorological Aircraft Derived Data Center (EMADDC) at the Royal Netherlands Meteorological Institute (KNMI) has been processing ‘Mode-S’ air traffic control signals to derive wind and temperature information. Following a recent meeting of experts, ECMWF has been working on processing the data and can now use them in a test version of the forecasting system (see separate article in this Newsletter). 

In other developments regarding in‑situ weather observations:

• Coordinated by EUMETNET, several of its members increased the frequency of radiosonde ascents from some of their stations. However, as the radiosonde figure shows, in some other places the numbers of radiosonde ascents dropped, perhaps due to supply difficulties.
• As part of a longer-term study, ECMWF is looking at the quality and possible assimilation of radiosonde descent data after balloon burst. In June 2020, ECMWF started the operational assimilation of most descent data from German radiosondes; the impact is currently modest as the numbers are relatively small, but it is expected to increase once descent data from other countries are assimilated too.
• In some areas, the number of surface reports from airfields (METARs) went down.
• In general, SYNOP weather station reports have stayed relatively constant.

At ECMWF, in normal times aircraft reports are second only to satellite data in their impact on forecasts. There are several recent additions to the satellite data assimilated at ECMWF which helped to make up for the loss of aircraft data:

• In 2019, ECMWF started assimilating data from several important instruments onboard EUMETSAT’s Metop-C satellite.
• In January, ECMWF began to assimilate wind observations from ESA’s ground-breaking Aeolus satellite.
• In March, the Centre started to use GNSS radio occultation (GNSS-RO) measurements from the FORMOSAT-7/COSMIC-2 mission, increasing the number of occultation profiles available for operational assimilation from around 3,000 a day to around 8,000.
• In mid-May, the number of occultations rose again by another 5,000 profiles a day when ECMWF began to assimilate RO data from the data and analytics company Spire.

Spire had offered to provide its RO observations free of charge for a limited period of time to help mitigate the loss of aircraft data. Building on previous work on the data by the European Space Agency (ESA), EUMETSAT and the UK Met Office, ECMWF was able to quickly move from quality assessment and passive monitoring to active assimilation. The additional RO data improved forecasts against a range of metrics. In terms of the relative impact of different types of observations on forecasts, between the end of March and the end of May aircraft-based observations went down substantially and RO data went up in broadly equal measure.