COVID-19: data availability and forecasting. The winds of change.

The impact of the COVID-19 pandemic is far reaching. Our MetOps team in the US have observed increased variations in predictions in the 5-10 day range. On 7th May the World Meteorological Organization (WMO) stated that it is concerned about the quantity and quality of weather observations and forecasts, as well as atmospheric and climate monitoring. After the European Centre for Medium-Range Weather Forecasts (ECWMF) stated that aircraft reports are second only to satellite data in their impact on forecasts it became clear the dramatic reduction in air traffic was likely to be a contributing factor.

Weather models are very data hungry and the more observational data available then the better the quality of the forecast. The observations are used to help estimate the state of the Earth system at the start of model runs, so when the quantity and/or quality reduces then the uncertainty in forecasts can start to increase. In support of this WMO’s Global Observing System (GOS) serves as a backbone for all weather and climate services and products provided by the 193 WMO Member states and territories to their citizens. The GOS provides high-quality standardised observations on the state of the atmosphere, land and ocean surface for the preparation of weather analyses, forecasts and warnings.  

GOS data comes from a wide variety of land-, marine- and space based sources, with radio sondes and weather balloons remaining the most important source of “truth” in numerical models. Since 70% of the earth’s surface is ocean ships have also provided one of the few sources of data in open ‘remote’ oceans. Sadly, the WMO’s Voluntary Observing Ships (VOS) program has been in decline and there are currently about 4000 ships in the programme down from 7700 in 1984/85, with WMO reporting a reduction in data availability of about 20% compared to normal levels due to Covid-19 from this source.  

Commercial airlines usually contribute some 800,000 high quality observations per day to the WMO via the Aircraft Meteorological Data Relay (AMDAR) programme, which uses onboard sensors, computers and communications systems to automatically collect, process, format and transmit meteorological observations to ground stations via satellite or radio links. With nations in lockdown these observations have dropped by 75-80% and close to 90% in more isolated areas where other surface-based observations are also scarce, i.e. the tropics and remote areas in the oceans. Importantly, these data are heavily weighted in the models, so have a significant impact on maritime forecasts as they often provide the most significant observations of the upper atmosphere over broad expanses of open ocean.   

No alt text provided for this image

Satellite data can help mitigate this loss but while they do provide a lot of information on temperature and humidity there is less data on winds and so this does not fill in all the gaps from the current large number of missing aircraft. Meteorologists have relied upon aircraft observations to fill gaps in remote areas, so their loss can have a disproportionate impact. Similarly, other observational areas such as radiosondes, balloons and METARS can be boosted to help compensate but these do not cover remote regions. Sensitivity studies at ECMWF have shown that removing all aircraft data degrades the short-range wind and temperature forecasts at those levels by up to 15%, with significant degradations at all forecast ranges up to seven days. Overall, currently ECWMF assess a 3% impact at the surface for forecasts up to 7 days ahead. 

Missing data has impacts far beyond the location as it is rippled out from the location to verify the first guess that the model creates to fill in the gaps between observations in data assimilation schemes and to ‘perturb’ the initial conditions to create Ensemble forecasts. Our forecasters are monitoring the operational impact of this and have noticed that there is greater uncertainty in the 5 to 10 day period. Some of these differences can be put down to the model physics and how each one is handling the current scenario, but unusually there is now the loss of significant data, especially from remote locations around the globe, that is hard to replicate or fill by other means.  

According to the WMO, none of the global Numerical Weather Prediction (NWP) centres have reported catastrophic losses in accuracy due to lack of observations, but there is certainly a detrimental impact on automated onboard ship routeing systems, where there is no marine meteorologist in the loop to resolve data irregularities.

In looking ahead, if the pandemic becomes drawn-out then further impact is likely, mostly linked to reduced human activities, whether it be observing, repairing or maintaining equipment. The WMO observes that many developing nations in Africa and Central and South America have not yet transitioned to automated observation systems, so these countries still rely on observations being taken manually by weather observers. If this does transpire then there will be a consequential impact on the quality and quantity of observations in the Southern hemisphere.

Covid-19 also poses a challenge to the forthcoming Atlantic hurricane season for obvious reasons. This will be especially so for marine forecasters when optimising routes over the medium to longer term. It is certainly a time to be extra cautious and if there was a moment in time when ship observations could become more numerous and reliable then now is the time to step up! 

Scroll to Top
Share via
Copy link
Powered by Social Snap