Weather forecasts play essential parts in economic activity. Assimilation of meteorological observations from aircraft improves forecasts greatly. However, global lockdown during the COVID-19 pandemic (March-May 2020) has eliminated 50-75% aircraft observations and imperils weather forecasting. Here, we verify global forecasts against reanalysis to quantify the impact of the pandemic. We find a large deterioration in forecasts of surface meteorology over regions with busy air flights, such as North America, southeast China and Australia. Forecasts over remote regions are also substantially worse during March-May 2020 than 2017-2019, and the deterioration increases for longer-term forecasts. This could handicap early warning of extreme weather and cause additional economic damage on the top of that from the pandemic. The impact over Western Europe is buffered by the high density of conventional observations, suggesting that introduction of new observations in data-sparse regions would be needed to minimize the impact of global emergencies on weather forecasts.