E. Linacre and B. Geerts
Very short-range forecasts for up to 12 hours have improved considerably between 1978-98 for mesoscale events such as squall lines, because of improved nowcasting tools (such as the network of next generation Doppler radars -nexrad- in the USA) and improvements in high-resolution modelling (1). Small, intense and brief events such as tornadoes, hail storms and flash floods remain difficult to predict out to 12 hours in advance, but mainly due to nexrad the alert lead times have at least doubled since 1978 in most cases, such that emergency measures can be taken. Forecasts have improved where the weather is affected by fixed features such as irregular terrain, land-use and land-sea contrasts.
Faster computers, better observations and improved assimilation of new data sources into NWP models have improved short-range forecasts of 12-72 hours. Forecasts of rain 36-60 hours ahead have become as accurate as 12-36 hour forecasts were in the 1970’s. Mesoscale features like sea breezes, pre-frontal low-level jets and mesoscale convective systems are now resolved in models that run out to 48h at a horizontal resolution of 30 km or less.
Medium range forecasts of 3-10 days have improved significantly, providing useful warnings of major disturbances, for instance. Temperature forecasts show considerable skill on day 3, becoming marginal by day 7. Predictability beyond 7 days is generally small, because of the inherently chaotic nature of the atmosphere. ‘.. no verifiable skill exists or is likely to exist for forecasting day-to-day weather changes beyond two weeks. Claims to the contrary should be viewed with skepticism.’ (1)
(1) AMS Council 1998. Policy statement on weather analysis and forecasting. Bull. Amer. Meteor. Soc. 79, 2161-3.