Impacting the entire globe, the COVID-19 pandemic has had profound disruptions to our definition of "normal." Schools, businesses, bars, restaurants, companies, travel, and essentially all non-essential entities have been closed. Thousands around the world have unfortunately died from the virus, with many more actively impacted by the disease. Businesses large and small are on the brink of going belly up, and the stock market has tumbled into a "bear market" after one of the best runs in the history of Wall Street. Aside from all of that, this crisis is possibly going to have an impact on the quality and accuracy of weather forecasts, believe it or not!


The Coronavirus has caused a worldwide reduction of air traffic, with thousands of planes grounded due to over a 60% reduction in passengers. This major drop is likely going to have an impact on the data assimilation for weather modeling. Before we get into the planes themselves, let's go over a brief recap of how weather "models" actually work. There are a lot of different weather models that meteorologists use to forecast the weather, some that run only out to 12 hours forward in time from now, others, like the "GFS" model that run out to 384 hours (16 days). There are even sub-seasonal computers that try to simulate upper air weather patterns out several weeks in advance. Below is an example of one model we use to look at the flow of the Jet Stream (30,000 feet) - valid for tomorrow at 5PM local time.


While every model has calculus and physics geared towards a certain type of weather or time-frame, one thing they all share is that all models require data of the atmosphere from right now. Meaning, temperature, dew point, wind speed, precipitation, etc. Most obviously, these variables are collected from official National Weather Service reporting stations, either at airports or actual forecast offices of the NWS across the United States. This data is then continually "fed into" the models so they can start with as accurate of a representation of the atmosphere right now that they can. The accuracy and amount of data given to models at "time 0" (right now) increases the forecast accuracy the further out in time they get. This is the perfect reference of "garbage in = garbage out" -- "quality data in = quality out."


Surface weather observations are only part of the weather model recipe. Upper air data estimated from satellites and actually observed from airplanes also get ingested into models. Even though we don't live up where planes fly at 30,000 feet, knowing the composition of the Jet Stream at that level is VITAL to being able to figure out what will subsequently happen down at the ground. Basically every commercial aircraft has weather instruments on them that record temperature, pressure, wind speed, and dew point on every flight taken.


The "European model" (ECMWF) is unequivocally our most accurate model that meteorologists use to predict the weather. The folks who run that model in the United Kingdom estimate that the number of weather observations from planes has been reduced by nearly 80% from the pandemic. On average, AMDAR (Aircraft Meteorological Data Relays) on planes collect nearly 750,000 weather observations every day. That data plays a crucial part in the accuracy of the model. The folks over the pond at the ECMWF HQ estimate that near a 15% decline in accuracy may occur in the model because of this lack of data entry.

Many NWS offices and other groups are trying to counter act this reduction by launching more weather balloons to make up for the missing upper atmosphere data from aircraft. The COVID-19 impacts are very far reaching, not just on the ground!