Nature today publishes a review looking at how the science of numerical weather prediction got where it is now, and where the future challenges lie. The article, co-authored by Peter Bauer (ECMWF), Alan Thorpe (ECMWF) and Gilbert Brunet (Environment Canada), aims to familiarise the wider scientific community with the progress achieved in numerical weather prediction. This progress has been gradual and has therefore not often hit the scientific headlines. The paper is also timely in that it comes exactly one year after the landmark WWOSC2014 Conference, and as the weather community is moving forward in addressing the challenges identified in Montreal.
Asked about the title chosen for the article, Peter Bauer commented:
“NWP did not get where it is today as the result of a single, or a few big discoveries as has been the case in other scientific disciplines, but rather through steady improvements in areas such as physical process representation, model initialisation, and the characterization of analysis/forecast uncertainties through ensembles. NWP has dramatically changed and improved, through fundamental, yet discrete scientific advances. This is what makes it a quiet revolution. NWP faces the additional challenge of efficiently using next generation supercomputing. “
Asked about what readers could expect to find in the article, co-author Alan Thorpe said:
“We discuss what can basically be described as the four pillars of NWP that have been fundamental aspects to this quiet revolution:
- the use of observations to provide the necessary description of the current state of the weather so that numerical predictions can start from the best estimate of the situation now,
- the description of the physics of the atmosphere (and oceans) in the prediction model, that determines for example how clouds form or how heat is transferred by convection,
- ensemble forecasting which describes the level of confidence of the predictions using a set of scenarios to define likelihoods and probabilities,
- finding the most efficient ways to use massively parallel supercomputers of today and the future to solve the mathematical equations described by the prediction model - so-called scalability.”
The article is available in full to Nature subscribers and the abstract is available freely at http://nature.com/articles/doi:10.1038/nature14956