Science blog

Example extended-range forecast

A new book on the challenges of extended-range prediction

29 January 2019
Share

Frédéric Vitart

Frédéric Vitart, Principal Scientist, Extended-Range Forecasting Group

I’ve recently edited a book with Andrew Robertson - from the International Research Institute for Climate and Society (IRI) - which highlights the latest research on extended-range forecasting, also known as sub-seasonal to seasonal forecasting. It’s a challenging timescale because it falls between medium-range (weather) and long-range (climate) prediction. However, it is of great societal interest and substantial progress has been made in recent years, which is summarised in this book.  

Introduction

Weather forecasts have greatly improved over past decades. Forecasts at day 5 are now more accurate than day 3 forecasts issued in the 1980s. Are current weather forecasting models now good enough to show some skill at weeks 3 or 4? It is of course not expected that we could predict the exact timing and location of a mesoscale event a few weeks in advance, but rather to be able to predict the probability of large-scale, long-lasting patterns. When successful, these large-scale pattern forecasts can provide useful information on the statistics of small-scale events in a probabilistic way.

When I started working on this topic in 2002, very little research had been conducted on this time range, and only one operational centre (the Japan Meteorological Agency) was issuing extended-range dynamical forecasts. Now more than 10 operational centres are issuing week 3 and 4 forecasts, and workshops on this topic attract hundreds of scientists.

The sub-seasonal to seasonal prediction project (S2S)

Forecasting at timescales beyond two weeks but less than a season, often referred to as sub-seasonal to seasonal (S2S), forms an important bridge between weather and climate prediction (Figure 1). Because of the increasing interest in forecasting at the S2S timescale the World Weather Research Programme (WWRP) and the World Climate Research Programme (WCRP) launched the Sub-Seasonal to Seasonal Prediction project (S2S project). Improving forecast skill, understanding the sub-seasonal to seasonal timescale, as well as promoting uptake by the operational centres and exploitation by the application community are the S2S project’s goals.

A schematic illustrating the S2S weather-climate prediction gap (Mariotti, Ruti and Rixen, 2018)

Figure 1: A schematic illustrating the S2S weather-climate prediction gap (Mariotti, Ruti and Rixen, 2018). Image reproduced under ​Creative Commons Attribution 4.0 International License.

Providing the basis for a book on sub-seasonal to seasonal prediction

An important outcome of this project, which Andrew Robertson and I are co-chairing, has been the creation of a database, hosted at ECMWF and the China Meteorological Administration, containing S2S real-time (with a 3-week delay) and re-forecasts from 11 operational centres. This database, accessible to all, encourages interaction between the research and operational communities.

So far more than 60 studies based on this database have been published in the peer-reviewed literature. These publications have helped to provide the basis for a book entitled “Sub-seasonal to Seasonal prediction: the gap between weather and climate forecasting”, published in October 2018 by Elsevier. The main goal of this publication is to provide an introduction, background and overview of the current status of research, operational prediction and applications in sub-seasonal to seasonal forecasting.

An example of an ECMWF extended-range forecast. The image shows 2 m temperature anomaly for 26 Feb to 4 Mar 2018 (days 15 to 21), based on a forecast from 12 Feb 2018. Derived from an ensemble of 51 members. Shaded areas are significant at the 10% level.

Figure 2: An example of an ECMWF extended-range forecast. The image shows 2m temperature anomaly for 26 Feb to 4 Mar 2018 (days 15 to 21), based on a forecast from 12 Feb 2018. Derived from an ensemble of 51 members. Shaded areas are significant at the 10% level.

Understanding sub-seasonal to seasonal predictability

The first section of the book discusses the sources of sub-seasonal to seasonal predictability, highlighting the important progress that has been made in recent years to better understand and predict these various sources of S2S predictability. One of the reasons for the recent increased interest in S2S forecasts is the significant skill displayed by the statistical and dynamical models in representing these sources of predictability, most especially the Madden Julian Oscillation (MJO, skill up to 5 weeks!) and stratospheric sudden warmings. However, there are still issues in predicting these sources of predictability (for example the propagation of the MJO over the Maritime Continent), and most importantly their impact on the extratropical weather. There is increasing evidence that these various sources of predictability are not always independent and interact with longer or shorter timescales, which makes S2S predictability a particularly complex problem.

Because the development of S2S prediction is much more recent than for medium-range and seasonal prediction, the methods to produce S2S ensemble forecasts are largely borrowed from weather prediction and seasonal prediction. S2S verification is also in its infancy. Some centres use their seasonal forecasting systems while others extend their medium-range forecasts to produce S2S forecasts. Since medium-range and seasonal forecasting have often been developed separately with the specificity of each timescale in mind, the current configuration of S2S forecasting systems is unlikely to be optimal.  More research is needed in this area. 

The benefits of sub-seasonal to seasonal prediction

The last section of the book documents several areas where S2S prediction could benefit society through new weather/climate services, early warnings and actions in response to extreme events: hydrology, agriculture, health applications, energy and more. Research on the potential use of S2S forecasts for applications has only started recently and S2S-based early warning systems with decision support have yet to emerge. Many challenges remain including the need to develop better skill estimates and user-oriented forecast products.

The future of the WWRP/WCRP S2S project

The first 5 years of the project (S2S Phase 1) (2014-2018) focused on the development and analysis of the database, the second phase (2019-2023) will focus more on S2S forecasting improvements. Phase 2 will be divided into 6 science sub-projects:

  • land initialisation and processes
  • ocean initialisation and processes
  • impact of aerosols
  • stratosphere
  • ensemble generation for S2S
  • MJO prediction and teleconnections

Phase 2 will also see more emphasis on the uptake of forecasts by the application community through the launch of a 2-year pilot project which will provide real-time S2S forecasts to several application projects.

Further reading

ECMWF extended-range forecasts 

ECMWF extended-range forecasts and the European heatwave of summer 2018