Professor Deliang Chen, Department of Earth Sciences at the University of Gothenburg, local host for the research programme Advanced Prediction of Earth and Climate through Big Science, Big Data and Big Computing, introduced the distinguished chair of the programme, Professor Fuqing Zhang, who is appointed Professor both in meteorology and in statistics at the Pennsylvania State University.
Fuqing Zhang explained that the programme will have five public lectures, held by different invited scientists who are experts in the field, and will also host the workshop Third Pole Environment. His lecture was called How predictable are our daily weather and long-term climate change? The weather seems to be a bit weird in many places of the earth at moment. The modern modelling of weather prediction began with Carl-Gustav Rossby, a Swede who moved to the U.S. and started the first American department of meteorology at the MIT in 1928. In the late 1940s he shared his time between the University in Chicago and SMHI in Sweden, and 1954 he gave the first operational weather prediction.
Stretching the limits of weather predictability
So, how to predict future weather and climate? Mathematically, it is handled by partial differential equations, which work for any dynamical system. The atmosphere and the time is finely chopped into small pieces, and the conservation of mass, momentum and energy are considered. Both practical initial condition uncertainties and model errors remain significantly big in present-day forecast system. Fuqing Zhang told about extremely violent tornados, where the models can predict their behaviour a few hours ahead, and of surprise snowstorms, where there are almost uncontrollable errors in the initial conditions. Generally, storms would be much easier to predict without water – the moisture causes much instability.
The ultimate limits of daily weather predictability with state of the art global NWP (numerical weather predictability) models are today 9–10 days. Likely, 3–5 more days can be gained through better information coefficients. There has been a quiet revolution of numerical weather prediction, where the forecast skills in the range from 3 to 10 days ahead has been increasing by about one day per decade. For the last year, Fuqing Zhang has been working with hurricane forecasts. The track forecasts have improved drastically over the past 25 years, so that a 3-day forecast today is as accurate as a 1-day forecast in 1989. The intensity forecasts have however remained generally stagnant, except for the last few years.
Hurricane Forecast Improvement Program
NOAA (National Oceanic and Atmospheric Administration) has launched the Hurricane Forecast Improvement Program to reduce the average track and intensity forecast errors by half for the 1-to-5-days range. Some means to make better input to the hurricane models are through airborne inflight measurements, dropsondes, and Doppler radar winds. The tools are not only to serve the U.S., but the whole world, to support economy and save lives. The damages of the hurricane Harvey amounted to the equivalent to 1/3 of the Swedish GDP only in insurance costs.
In conclusion, predictability of our daily weather including hurricanes and severe storms is very limited at all scales due to the chaotic systems of moisture. Advances might come from the areas of observations, models, data assimilation and computing. Future public lectures in the programme will be more about climate, and the limits of climate predictability at the regional scales.
Text and photos: Setta Aspström