Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Clim Change
Impact Factor: 4.571

Seasonal climate predictability and forecasting: status and prospects

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Seasonal climate forecasts occupy an intermediate zone between weather forecasting and climate projections. They share with the numerical weather prediction the difficulty of initializing the simulations with a realistic state of the atmosphere and the need to periodically verify different aspects of their quality, while additionally are burdened by uncertainties in feedback processes that also play a central role in constraining climate projections. Seasonal predictions have to deal also with the challenge of initializing all the components of the climate system (ocean, sea ice, and land surface). The value of skilful seasonal forecasts is obvious for many societal sectors and is currently being included in the framework of developing climate services. Seasonal forecasts will in addition be valuable by increasing the acceptance of climate projections among the general public. This advanced‐review article presents an overview of the state‐of‐the‐art in global seasonal predictability and forecasting for climate researchers and discusses fundamental advances to increase forecast quality in the near future. The article concludes with a list of challenges where seasonal forecasting is expected to focus on in the near future. WIREs Clim Change 2013, 4:245–268. doi: 10.1002/wcc.217

Conflict of interest: The authors have declared no conflicts of interest for this article.

Figure 1.

Correlation between the ERSST SST Niño 3.4 index (average temperature over 5°N–5°S, 170°–120°W) and the GPCCv5 gridded precipitation over the period 1960–2009. (a) March to May, (b) June to August, (c) September to November, and (d) December to February.

[ Normal View | Magnified View ]
Figure 2.

Correlation of the ensemble mean for boreal autumn (September to November) surface air temperature of 3‐month lead hindcasts (start date first of May) performed with the EC‐Earth2 prediction system over the period 1991–2005 where the sea‐ice initial conditions are taken from the restart files of a sea‐ice simulation forced by a specified forcing from reanalyses (a) with interannual variability and (b) taking the climatological mean. The reference data are taken from ERAInterim. Dots are used for the 95% confidence significant correlations, where a two‐sided test is applied using a bootstrap method with a sample of 1000.

[ Normal View | Magnified View ]
Figure 3.

Monthly predicted anomalies of the average western tropical Indian Ocean SST for (a) a statistical model based on linear regression, (b) an 11‐member ensemble from ECMWF System 3, (c) a 24‐member ensemble from NCEP CFSv2, and (d) the forecast assimilation combination of the dynamical systems using the statistical system as a prior. The predictions are for the target month of October with a lead time of 5 months. The plots show the reference values (HadSST1.1, black solid line), the mean predicted values (red solid line), the 95% predicted interval (gray area) and the climatological value of November (black dashed line). The correlation with the reference of the mean prediction, the Brier skill score and its reliability and resolution components for dichotomous events of SST anomalies exceeding the climatological median and the upper quartile are shown in the upper left corner. The statistical model was trained in forecast mode, where the first set of parameters were estimated for the period 1951 to 1981 (the first training period), and the predictions were performed for the target years 1982–2010, extending the training period by 1 year at a time mimicking an operational context.

[ Normal View | Magnified View ]
Figure 4.

Hovmöller latitude‐time diagrams of West African precipitation climatology (mm/day) for GPCP (top left), where the precipitation has been longitudinally averaged over 10°W–10°E. The rest of the left column shows the ECMWF‐System4 for three different start dates: May (zero lead time), February (3‐month lead time), and November of the previous year (6‐month lead time). The corresponding systematic error, estimated as the model minus GPCP climatology, appears immediately in the right column. The annual cycle of the equatorial Atlantic SST averaged over the region 4°S–4°N/ 15°W–10°E is shown (top right) for ERSST (gray bars) and System4 for the three start dates: May (solid black), February (dashed black), and November (dotted black). The period of study is 1982–2008.

[ Normal View | Magnified View ]
Figure 5.

Correlation of the ensemble mean of one‐month lead surface air temperature (top row) and precipitation (bottom row) from the ENSEMBLES multimodel seasonal predictions in boreal summer (June to August, left column) and winter (December to February, right column). The predictions have been performed over the period 1980–2005 with five different forecast systems, each one running nine‐member ensembles. The reference data are taken from ERAInterim for temperature and GPCP for precipitation.

[ Normal View | Magnified View ]
Figure 6.

Correlation of the ensemble mean of the predictions of total monthly precipitation, the 90th percentile of the daily precipitation and the number of days in a month where the precipitation is larger than the 90th percentile of the climatological distribution of the month of August for the 3‐month (May start date, forecast time 4 months, left column) and zero‐month (August start date, forecast time 1 month, right column) lead time seasonal predictions of the perturbed‐parameter DePreSys system. The predictions were performed over the period 1960–2005 and verified against the E‐Obs data set interpolated bilinearly on the DePreSys grid.

[ Normal View | Magnified View ]
Figure 7.

Surface downward solar radiation and 10‐m wind seasonal 1‐month lead boreal summer (June to August) 15‐member ensemble predictions from ECMWF's System 4. Panels (a) and (b) show the correlation between the ensemble‐mean prediction and ERAInterim over 1981–2010. Panels (c) and (d) display probabilistic forecasts for the most likely tercile event (where below normal, normal and above normal are considered) in summer 2011, where blue (yellow‐red) colors correspond to probabilities for the event below (above) the normal summer values. Panels (e) and (f) show the ERAInterim mean values for summer 2011. Panels (g) and (h) depict examples of the ensemble predictions of anomalies (with respect to the period 1981–2010) for specific sites for every summer, where the central red box corresponds to the interquartile range of the ensemble, the thick horizontal bar to the median, the whiskers the 5th and 95th percentiles and the small dots some outliers, while the blue line is for the ERAInterim value. These panels show the correlation of the ensemble‐mean prediction with ERAInterim in the top left corner.

[ Normal View | Magnified View ]

Browse by Topic

Climate Models and Modeling > Earth System Models
Climate Models and Modeling > Knowledge Generation with Models
The Social Status of Climate Change Knowledge > Climate Science and Decision Making

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts