Timothy N. Stockdale

Global seasonal rainfall forecasts using a coupled ocean–atmosphere model

One conceptual model of weather is that of a series of events which are unconnected. That is, that the weather next week is essentially independent of the weather this week. However, although individual weather systems might be chaotic and unpredictable beyond a week or so, the statistics describing them may be perturbed in a deterministic and predictable way, particularly by the ocean. In the past, seasonal forceasts of atmospheric variables have largely been based on empirical relationships, which are weak in most areas of the world. More recently, atmosphere models forced by assumed or predicted ocean conditions have been used,. Here a fully coupled global ocean–atmosphere general circulation model is used to make seasonal forecasts of the climate system with a lead time of up to 6 months. Such a model should be able to simulate the predictable perturbations of seasonal climate, but to extract these from the chaotic weather requires an ensemble of model integrations, and hence considerable computer resources. Reliable verification of probabilistic forecasts is difficult, but the results obtained so far, when compared to observations, are encouraging for the prospects for seasonal forecasting. Rainfall predictions for 1997 and the first half of 1998 show a marked increase in the spatial extent of statistically significant anomalies during the present El Niño, and include strong signals over Europe.

A verification framework for interannual-to-decadal predictions experiments

Decadal predictions have a high profile in the climate science community and beyond, yet very little is known about their skill. Nor is there any agreed protocol for estimating their skill. This paper proposes a sound and coordinated framework for verification of decadal hindcast experiments. The framework is illustrated for decadal hindcasts tailored to meet the requirements and specifications of CMIP5 (Coupled Model Intercomparison Project phase 5). The chosen metrics address key questions about the information content in initialized decadal hindcasts. These questions are: (1) Do the initial conditions in the hindcasts lead to more accurate predictions of the climate, compared to un-initialized climate change projections? and (2) Is the prediction model’s ensemble spread an appropriate representation of forecast uncertainty on average? The first question is addressed through deterministic metrics that compare the initialized and uninitialized hindcasts. The second question is addressed through a probabilistic metric applied to the initialized hindcasts and comparing different ways to ascribe forecast uncertainty. Verification is advocated at smoothed regional scales that can illuminate broad areas of predictability, as well as at the grid scale, since many users of the decadal prediction experiments who feed the climate data into applications or decision models will use the data at grid scale, or downscale it to even higher resolution. An overall statement on skill of CMIP5 decadal hindcasts is not the aim of this paper. The results presented are only illustrative of the framework, which would enable such studies. However, broad conclusions that are beginning to emerge from the CMIP5 results include (1) Most predictability at the interannual-to-decadal scale, relative to climatological averages, comes from external forcing, particularly for temperature; (2) though moderate, additional skill is added by the initial conditions over what is imparted by external forcing alone; however, the impact of initialization may result in overall worse predictions in some regions than provided by uninitialized climate change projections; (3) limited hindcast records and the dearth of climate-quality observational data impede our ability to quantify expected skill as well as model biases; and (4) as is common to seasonal-to-interannual model predictions, the spread of the ensemble members is not necessarily a good representation of forecast uncertainty. The authors recommend that this framework be adopted to serve as a starting point to compare prediction quality across prediction systems. The framework can provide a baseline against which future improvements can be quantified. The framework also provides guidance on the use of these model predictions, which differ in fundamental ways from the climate change projections that much of the community has become familiar with, including adjustment of mean and conditional biases, and consideration of how to best approach forecast uncertainty.

Distinguishing the Roles of Natural and Anthropogenically Forced Decadal Climate Variability: Implications for Prediction

Abstract Given that over the course of the next 10–30 years the magnitude of natural decadal variations may rival that of anthropogenically forced climate change on regional scales, it is envisioned that initialized decadal predictions will provide important information for climate-related management and adaptation decisions. Such predictions are presently one of the grand challenges for the climate community. This requires identifying those physical phenomena—and their model equivalents—that may provide additional predictability on decadal time scales, including an assessment of the physical processes through which anthropogenic forcing may interact with or project upon natural variability. Such a physical framework is necessary to provide a consistent assessment (and insight into potential improvement) of the decadal prediction experiments planned to be assessed as part of the IPCCs Fifth Assessment Report.

Coupled Ocean–Atmosphere Forecasts in the Presence of Climate Drift

Abstract Two different coupled atmosphere–ocean GCMs are used to forecast SST anomalies with lead times of up to one year. The initialization procedure does not balance the ocean and atmosphere components, nor is the coupled model flux corrected to maintain the correct mean state. Rather, the coupled model is allowed to evolve freely during the forecast. The inevitable climate drift is estimated across an ensemble of forecasts and subtracted to give the true forecast. Although the climate drift is often bigger than the interannual signal, the method works. This is true for a drift toward both warmer and colder SSTs, as exemplified by the two models. The best way of establishing the mean bias correction from a small sample of prior forecasts is discussed. In some circumstances the sample median may be a more robust estimator than the sample mean. For the limited set of forecasts here, use of the median bias in the cross-correlated forecasts reduces forecast error, when compared to use of the mean bias.

Seasonal Forecasting of Tropical Storms Using Coupled GCM Integrations

The ECMWF Seasonal Forecasting System, based on ensembles of 200-day coupled GCM integrations, contains tropical disturbances that are referred to as model tropical storms in the present paper. Model tropical storms display a genesis location and a seasonal cycle generally consistent with observations, though the frequency of model tropical storms is significantly lower than observed, particularly over the North Atlantic and the eastern North Pacific. Several possible causes for the low number of model tropical storms are discussed. The ECMWF Seasonal Forecasting System produces realistic forecasts of the interannual variability of tropical storm frequency over the North Atlantic and the western North Pacific, with strong linear correlations and low rms error obtained when comparing the forecasts to observations. The skill of the seasonal forecasting system in predicting the frequency of tropical storms is likely to be related to its skill in predicting sea surface tem

An Ensemble Generation Method for Seasonal Forecasting with an Ocean–Atmosphere Coupled Model

Seasonal forecasts are subject to various types of errors: amplification of errors in oceanic initial conditions, errors due to the unpredictable nature of the synoptic atmospheric variability, and coupled model error. Ensemble forecasting is usually used in an attempt to sample some or all of these various sources of error. How to build an ensemble forecasting system in the seasonal range remains a largely unexplored area. In this paper, various ensemble generation methodologies for the European Centre for Medium-Range Weather Forecasts (ECMWF) seasonal forecasting system are compared. A series of experiments using wind perturbations (applied when generating the oceanic initial conditions), sea surface temperature (SST) perturbations to those initial conditions, and random perturbation to the atmosphere during the forecast, individually and collectively, is presented and compared with the more usual lagged-average approach. SST perturbations are important during the first 2 months of the forecast to ensure a spread at least equal to the uncertainty level on the SST measure. From month 3 onward, all methods give a similar spread. This spread is significantly smaller than the rms error of the forecasts. There is also no clear link between the spread of the ensemble and the ensemble mean forecast error. These two facts suggest that factors not presently sampled in the ensemble, such as model error, act to limit the forecast skill. Methods that allow sampling of model error, such as multimodel ensembles, should be beneficial to seasonal forecasting.

Did the ECMWF seasonal forecast model outperform statistical ENSO forecast models over the last 15 years

Abstract The European Centre for Medium-Range Weather Forecasts (ECMWF) has made seasonal forecasts since 1997 with ensembles of a coupled ocean–atmosphere model, System-1 (S1). In January 2002, a new version, System-2 (S2), was introduced. For the calibration of these models, hindcasts have been performed starting in 1987, so that 15 yr of hindcasts and forecasts are now available for verification. Seasonal predictability is to a large extent due to the El Nino–Southern Oscillation (ENSO) climate oscillations. ENSO predictions of the ECMWF models are compared with those of statistical models, some of which are used operationally. The relative skill depends strongly on the season. The dynamical models are better at forecasting the onset of El Nino or La Nina in boreal spring to summer. The statistical models are comparable at predicting the evolution of an event in boreal fall and winter.

Tropical Atlantic SST prediction with coupled ocean-atmosphere GCMs

Abstract Variations in tropical Atlantic SST are an important factor in seasonal forecasts in the region and beyond. An analysis is given of the capabilities of the latest generation of coupled GCM seasonal forecast systems to predict tropical Atlantic SST anomalies. Skill above that of persistence is demonstrated in both the northern tropical and equatorial Atlantic, but not farther south. The inability of the coupled models to correctly represent the mean seasonal cycle is a major problem in attempts to forecast equatorial SST anomalies in the boreal summer. Even when forced with observed SST, atmosphere models have significant failings in this area. The quality of ocean initial conditions for coupled model forecasts is also a cause for concern, and the adequacy of the near-equatorial ocean observing system is in doubt. A multimodel approach improves forecast skill only modestly, and large errors remain in the southern tropical Atlantic. There is still much scope for improving forecasts of tropical Atla...

Evaluation of atmospheric fields from the ECMWF seasonal forecasts over a 15 year period

Abstract Since 1997, the European Centre for Medium-Range Weather Forecasts (ECMWF) has made seasonal forecasts with ensembles of a coupled ocean–atmosphere model, System-1 (S1). In January 2002, a new version, System-2 (S2), was introduced. For the calibration of these models, hindcasts have been performed starting in 1987, so that 15 yr of hindcasts and forecasts are now available for verification. The main cause of seasonal predictability is El Nino and La Nina perturbing the average weather in many regions and seasons throughout the world. As a baseline to compare the dynamical models with, a set of simple statistical models (STAT) is constructed. These are based on persistence and a lagged regression with the first few EOFs of SST from 1901 to 1986 wherever the correlations are significant. The first EOF corresponds to ENSO, and the second corresponds to decadal ENSO. The temperature model uses one EOF, the sea level pressure (SLP) model uses five EOFs, and the precipitation model uses two EOFs but e...

Atmospheric initial conditions and the predictability of the Arctic Oscillation

The Arctic Oscillation (AO) is the leading mode of variation in the northern hemisphere winter circulation. Despite its importance for winter temperatures, seasonal forecast models typically suggest that its predictability is low. Nonetheless, we show that an operational forecast model has high skill in predicting the AO, with a correlation of 0.61 for the period of 1981–2010. Experimentation covering a recent 8 year high-skill period demonstrates the predictability of the model AO to be dominated by atmospheric initial conditions, although surface forcing does have increasing influence later in the winter. Results suggest that the stratosphere is an important carrier of model predictability during the early winter. These results challenge the conventional paradigm of surface forcing being the dominant source of predictability on seasonal time scales but are compatible with the results showing stratospheric influence on winter circulation. They also suggest that model representation of stratospheric to tropospheric coupling needs urgent improvement.