Search

Producing probabilistic subseasonal forecasts of extreme events up to six weeks in advance is crucial for many economic sectors. In agribusiness, this time scale is particularly critical because it allows for mitigation strategies to be adopted for counteracting weather hazards and taking advantage of opportunities. For example, spring frosts are detrimental for many nut trees, resulting in dramatic losses at harvest time. To explore subseasonal forecast quality in boreal spring, identified as one of the most sensitive times of the year by agribusiness end users, we build a multisystem ensemble using four models involved in the Subseasonal to Seasonal Prediction project (S2S). Two-meter temperature forecasts are used to analyze cold spell predictions in the coastal Black Sea region, an area that is a global leader in the production of hazelnuts. When analyzed at the global scale, the multisystem ensemble probabilistic forecasts for near-surface temperature are better than climatological values for several regions, especially the tropics, even many weeks in advance; however, in the coastal Black Sea, skill is low after the second forecast week. When cold spells are predicted instead of near-surface temperatures, skill improves for the region, and the forecasts prove to contain potentially useful information to stakeholders willing to put mitigation plans into effect. Using a cost–loss model approach for the first time in this context, we show that there is added value of having such a forecast system instead of a business-as-usual strategy, not only for predictions released 1–2 weeks ahead of the extreme event, but also at longer lead times.

East Africa experiences chronic food insecurity, with levels varying from year-to-year across the region. Given that much can be done to prevent this level of suffering before it happens, humanitarian agencies monitor early indicators of food insecurity to trigger early action. Forecasts of total seasonal rainfall are one tool used to monitor and anticipate food security outcomes. Factors beyond rainfall, such as conflict, are key determinants of whether lack of rainfall can become a problem. In this paper, we present a quantitative analysis that isolates the value of rainfall information in anticipating food security outcomes across livelihood groups in East Africa. Comparing observed rainfall and temperature with food security classifications, we quantify how much the chance of food insecurity increases when rainfall is low. Results differed dramatically among livelihood groups; pastoralists in East Africa more frequently experience food insecurity than do non-pastoralists, and 12 months of low rainfall greatly increases the chances of "crisis" and "emergency" food security in pastoralist regions. In non-pastoralist regions, the relationship with total rainfall is not as strong. Similar results were obtained for livelihood groups in Kenya and Ethiopia, with slightly differing results in Somalia. Given this, we evaluated the relevance of monitoring and forecasting seasonal total rainfall. Our quantitative results demonstrate that six months of rainfall observations can already indicate a heightened risk of food insecurity, a full six months before conditions deteriorate. Combining rainfall observations with seasonal forecasts can further change the range of possible outcomes to indicate higher or lower risk of food insecurity, but the added value of seasonal forecasts is noticeable only when they show a strong probability of below-normal rainfall.

In this chapter, we examine how weather forecasts are made and how they may be used by the health sector. We explore why, where and when we might obtain good forecasts and why sometimes the forecasts go wrong. We consider the theoretical and practical constraints on weather forecasts and why their accuracy declines rapidly after only a few days. Knowing the limitations of weather forecasts helps us to learn how to make best use of such information.

This study explores the predictive skill of seasonal rainfall characteristics for the first rainy (and planting) season, May–June, in Central America. Statistical predictive models were built using a Model Output Statistics (MOS) technique based on canonical correlation analysis, in which variables that forecast with the Climate Forecast System version 2 (CFSv2) were used as candidate predictors for the observed total precipitation, frequency of rainy days and mean number of extremely dry and wet events in the season. CFSv2 initializations from February to April were explored. The CFSv2 variables used in the study consist of rainfall, as in a typical MOS technique, and a combination of low-level winds and convective available potential energy (CAPE), a blend that has been previously shown to be a good predictor for convective activity. The highest predictive skill was found for the seasonal frequency of rainy days, followed by the mean frequency of dry events. In terms of candidate predictors, the zonal transport of CAPE (uCAPE) at 925 hPa offers higher skill across Central America than rainfall, which is attributed in part to the high model uncertainties associated with precipitation in the region. As expected, dynamical model predictors initialized in February provide lower skill than those initialized later. Nonetheless, the skill is comparable for March and April initializations. These results suggest that the National Meteorological and Hydrological Services in Central America, and the Central American Regional Climate Outlook Forum, can produce earlier more skilful forecasts for May–June rainfall characteristics than previously stated.

In this chapter, we explore the potential value of seasonal climate forecasts to the health community. We begin by considering why the general weather conditions over a season might be predictable, before examining how seasonal forecasts are made, and why they are presented in a different way to weather forecasts. We then examine where and when seasonal forecasts work best, and emphasize that it is possible to make useful seasonal forecasts for health outcomes only in some parts of the world and for certain times of the year (and possibly only for some years). We also review some of the main sources of seasonal forecasts.

In writing this book, we have prioritized shorter timescales of weather and climate variability because these have most direct relevance to the types of operational and planning decisions made in the health sector. However, in this chapter, we focus on longer-term health decisions that require climate knowledge and information at timescales beyond a year to multiple decades (up to 50 years). This time span incorporates both natural decadal variability and the influences of anthropogenic climate change. It is an extremely challenging timeframe at which to work because of the very limited operational predictive skill at multiannual timescales, and limited confidence about how rainfall may change on longer timescales. Despite these challenges, the timescales under consideration, especially the next five to 20 years, are particularly important when considering how to adapt to climate change. Here we discuss what we do and do not know about the climate in the coming years. We then focus on the practical use of climate information in understanding observed climate impacts on health as well as the prediction of multi-annual to multi-decadal risks. The multi-annual timescale is often referred to as ‘decadal’, even though the forecasts are for periods less than a decade.

World Meteorological Organization (WMO) guidance on verification of operational seasonal forecasts.

Potential and real predictive skill of the frequency of extreme rainfall in southeastern South America for the December–February season are evaluated in this paper, finding evidence indicating that mechanisms of climate variability at one time scale contribute to the predictability at another scale; that is, taking into account the interference of different potential sources of predictability at different time scales increases the predictive skill. Part I of this study suggested that a set of daily atmospheric circulation regimes, or weather types, was sensitive to these cross–time scale interferences, conducive to the occurrence of extreme rainfall events in the region, and could be used as a potential predictor. At seasonal scale, a combination of those weather types indeed tends to outperform all the other candidate predictors explored (i.e., sea surface temperature patterns, phases of the Madden–Julian oscillation, and combinations of both). Spatially averaged Kendall’s t improvements of 43% for the potential predictability and 23% for real-time predictions are attained with respect to standard models considering sea surface temperature fields alone. A new subseasonal-to-seasonal predictive methodology for extreme rainfall events is proposed based on probability forecasts of seasonal sequences of these weather types. The cross-validated real-time skill of the new probabilistic approach, as measured by the hit score and the Heidke skill score, is on the order of twice that associated with climatological values. The approach is designed to offer useful subseasonal-to-seasonal climate information to decision-makers interested not only in how many extreme events will happen in the season but also in how, when, and where those events will probably occur.

In 2008, the International Federation of Red Cross and Red Crescent Societies (IFRC) used a seasonal forecast for West Africa for the first time to implement an Early Warning, Early Action strategy for enhanced flood preparedness and response. Interviews with disaster managers suggest that this approach improved their capacity and response. Relief supplies reached flood victims within days, as opposed to weeks in previous years, thereby preventing further loss of life, illness, and setbacks to livelihoods, as well as augmenting the efficiency of resource use. This case demonstrates the potential benefits to be realised from the use of medium-to-long-range forecasts in disaster management, especially in the context of potential increases in extreme weather and climate-related events due to climate variability and change. However, harnessing the full potential of these forecasts will require continued effort and collaboration among disaster managers, climate service providers, and major humanitarian donors.

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.

Procedures for verification of seasonal and longer-range forecasts.

This paper evaluates the quality of real-time seasonal probabilistic forecasts of the extreme 15% tails of the climatological distribution of temperature and precipitation issued by the International Research Institute for Climate and Society (IRI) from 1998 through 2009. IRI’s forecasts have been based largely on a two-tiered multimodel dynamical prediction system. Forecasts of the 15% extremes have been consistent with the corresponding probabilistic forecasts for the standard tercile-based categories; however, nonclimatological forecasts for the extremes have been issued sparingly. Results indicate positive skill in terms of resolution and discrimination for the extremes forecasts, particularly in the tropics. Additionally, with the exception of some overconfidence for extreme above-normal precipitation and a strong cool bias for temperature, reliability analyses suggest generally good calibration. Skills for temperature are generally higher than those for precipitation, due both to correct forecasts of increased probabilities of extremely high (above the upper 15th percentile) temperatures associated with warming trends, and to better discrimination of interannual variability. However, above-normal temperature extremes were substantially underforecast, as noted also for the IRI’s tercile forecasts.

Humanity recognised millennia ago the importance of climate variability to the sustenance of life, whether that variability was expressed in the form of droughts, floods, heat, cold, or wind. Coping strategies, developed to handle the consequences of climate variability, helped ensure mankind’s survival, although the historic record indicates that not all societies successfully overcame past challenges imposed by long-term droughts, extensive flooding, and the like. Early coping strategies included migration, invasion, appropriation and storage. In addition many, probably most, perhaps all, societies developed indigenous knowledge or belief systems that they felt enabled them to foresee or control those elements of the climate that are so critical for maintaining water and food supplies.

An overview of Regional Climate Outlook Forums.

Although it is impossible to forecast the weather more than a few days in advance, the science of seasonal climate forecasting is premised upon an ability to predict the general weather conditions over a prolonged period of time, without trying to predict the precise weather at any specific time during that period. The forecasting is possible only because sometimes, and primarily within tropical latitudes, the atmosphere is sensitive to unusual conditions at the earth’s surface, and especially at the sea surface. El Niño, and its counterpart La Niña, are the primary examples of such forcing conditions: during El Niño events, much of the equatorial Pacific Ocean is unusually hot (cold during La Niña), and the consequent changes to the heat and moisture supplied to the atmosphere can disrupt weather conditions in many parts of the globe. However, all seasonal climate forecasts involve a great deal of uncertainty, and a key aspect of forecasting at such time scales is to estimate the uncertainty in the prediction reliably. There are two sources of uncertainty in seasonal climate forecasting: the atmosphere is nowhere completely forced by conditions at the surface, but is free to vary according to its own internal dynamics; and the models used to predict the climate system are imperfect. These two sources of uncertainty are addressed by producing a set of model predictions: different initial weather conditions are used to represent the uncertainty from the internal dynamics, and different models to account for the uncertainties arising from imperfect model physics.

When seasonal climate forecasts are expressed probabilistically, it is not possible to answer simple questions such as “how often are the forecasts correct?” The simpler attributes of forecast quality, such as “accuracy” or “correctness” are not applicable to probabilistic forecasts, and instead the main attributes of interest are: reliability, which defines whether the confidence communicated in the forecasts is appropriate; resolution, which defines whether there is any usable information in the forecasts; discrimination, which defines whether the forecasts are discernibly different given different outcomes (somewhat similar to the attribute of resolution); and sharpness, which defines the level of confidence that is communicated in the forecasts (regardless of whether that level is appropriate). How these attributes are measured depends on how the forecasts are expressed. In this chapter these attributes are explained in detail, and representation by various graphical procedures and scoring metrics is described. Partly because there is more than one desirable attribute to good probabilistic forecasts, it is argued that there is no single scoring metric that can adequately summarise forecast quality, and that in many cases graphical procedures also hide important aspects of forecast quality. The aim in this chapter is to provide some guidelines for interpreting and recognising the strengths and limitations of the most important verification tools as applied to seasonal climate forecasts.

Producing a seasonal climate forecast from a dynamical model involves a great deal more than simply running the model and viewing the results. The first problem is to decide which dynamical model(s) should be run given the practical constraints of computing resources. In this chapter the pros and cons of using the more computationally intensive fully coupled models compared to atmosphereonly models are discussed. After running a dynamical model, regardless of its complexity, corrections need to be made for systematic errors because the model’s climatology and that of the observed climate are invariably different. Some simple procedures for correcting these systematic errors are assessed, but more sophisticated methods are advisable to adjust for spatial displacements of the model climate. Since the model predictions represent large spatial averages, and generally are presented as seasonal averages, downscaling may be required to make the forecast relevant for specific locations, and to provide more detailed information about the statistics of weather within the season. Commonly used spatial and temporal downscaling procedures are described. Some procedures for describing the uncertainty in the forecast are discussed (further details are provided in Chapter 9). Evidence is presented that forecasts can be improved by combining outputs from different models. Finally, the reliability of the forecast needs to be determined by verification of a historical set of forecasts.

The enthusiasm for engaging the challenges of Seasonal to Interannual Prediction, both within the disciplines of physical and social sciences and at their interface, was well demonstrated through the energetic engagement of all during the May to June 2005 NATO ASI course, upon which this book is based. Several panel sessions were held during the course, which permitted everyone to offer views within an informal setting; some, not reflected in the main body of the book, are incorporated in this chapter. Little stays stationary in such a fast-developing field, and so, to provide the most advanced position at publication, this summarising chapter has included some of the latest development to supplement the material drawn from the course presentations and the panel discussions. Additionally, a view to the future is offered so as to provide further stimulation to those interested in the fascinating field of Seasonal to Interannual Climate

Because of the inherently chaotic nature of the atmosphere, ensemble simulations are required to characterize a model’s response to the prescribed boundary forcing in probabilistic terms, particularly if the focus is on the probabilities of extreme events. At the same time, substantial computer resources are needed to produce routinely ensemble seasonal climate forecasts of sufficient size to make suitably reproducible estimates of such probabilities. We describe a method for artificially expanding the effective number of members in ensemble climate simulations on a seasonal basis, thereby reducing uncertainty in estimated probability distributions. As described here, the method involves calculating seasonal statistics using monthly values from all possible combinations of ensemble members. Under certain assumptions, this method is able to increase the effective ensemble size of an N-member M-month seasonal forecast by a factor of (asymptotically) M. One key assumption in this regard is that, aside from the effects of prescribed boundary conditions, the month-to-month values from a particular ensemble member are linear independent. This paper describes the behaviour of the ensemble expansion technique using both idealized and actual ensemble forecast data under a variety of conditions, drawing comparisons with an alternative parametric approach for ensemble expansion. A method for testing the assumption of linear independence in model simulations is also presented.

The potential for long‐range prediction of Zimbabwe summer rainfall is investigated using an analysis of variance approach. It is assumed that the variance between seasons (inter‐annual variability) is made up of two components: the climate noise (intra‐seasonal variability) and signal (any variance above the noise). The magnitude of the climate noise is estimated by computing the variance of time‐averaged rainfall data based on a statistical model whose parameters are derived from observed daily values at each station. Results from the study indicate that up to approximately 70% of the total variance in Zimbabwe summer rainfall is potentially predictable at long range. Predictability is greatest during the last half of the rainy season, January to March, and much lower during October to December. The south section of the country shows relatively more predictability than the north. Comparisons between signal and noise are discussed in the context of long‐range predictability. The natural variability (noise) is proposed as a lower limit of the standard error of the estimate for any long‐range precipitation forecast for the country.

Spatial patterns of inter-annual summer rainfall variability over Zimbabwe are investigated using principal component analysis for the period 1920-96. Inter-annual rainfall anomalies for all years and stratified according to warm El Niño - Southern Oscillation (ENSO) events are used in the study. The varimax rotated PC loadings subdivide the country into three regions, north, southwest and southeast. Because of the high correlation between two of the regions, only two regions, north and south, consistent with the underlying physical processes are retained. The two leading PCs are consistent with the main rain-producing synoptic weather systems affecting the country during austral summer. Inter-comparison of the rainfall anomaly time series from the two regions reveals that the major controlling factors influencing inter-annual rainfall variability are possibly of a global or hemispheric nature. The spatially coherent regions have practical use in the development and application of long-range seasonal climate forecasts and climate change studies. The file associated with this record contains pages 430-31 of this article.

This review paper presents an assessment of the current state of knowledge and capability in seasonal climate prediction at the end of the 20th century. The discussion covers the full range of issues involved in climate forecasting, including (1) the theory and empirical evidence for predictability; (2) predictions of surface boundary conditions, such as sea surface temperatures (SSTs) that drive the predictable part of the climate; (3) predictions of the climate; and (4) a brief consideration of the application of climate forecasts. Within this context, the research of the coming decades that seeks to address shortcomings in each area is described.

Sea‐surface temperature (SST) variations of the oceans surrounding southern Africa are associated with seasonal rainfall variability, especially during austral summer when the tropical atmospheric circulation is dominant over the region. Because of instabilities in the linear association between summer rainfall over southern Africa and SSTs of the tropical Indian Ocean, the skilful prediction of seasonal rainfall may best be achieved using physically based models. A two‐tiered retro‐active forecast procedure for the December–February (DJF) season is employed over a 10‐year period starting from 1987/1988. Rainfall forecasts are produced for a number of homogeneous regions over part of southern Africa. Categorized (below‐normal, near‐normal and above‐normal) statistical DJF rainfall predictions are made for the region to form the baseline skill level that has to be outscored by more elaborate methods involving general circulation models (GCMs). The GCM used here is the Centre for Ocean–Land–Atmosphere Studies (COLA) T30, with predicted global SST fields as boundary forcing and initial conditions derived from the National Centres for Environmental Prediction (NCEP) reanalysis data. Bias‐corrected GCM simulations of circulation and moisture at certain standard pressure levels are downscaled to produce rainfall forecasts at the regional level using the perfect prognosis approach. In the two‐tiered forecasting system, SST predictions for the global oceans are made first. SST anomalies of the equatorial Pacific (NIÑO3.4) and Indian oceans are predicted skilfully at 1‐ and 3‐month lead‐times using a statistical model. These retro‐active SST forecasts are accurate for pre‐1990 conditions, but predictability seems to have weakened during the 1990s. Skilful multi‐tiered rainfall forecasts are obtained when the amplitudes of large events in the global oceans (such as El Niño and La Niña episodes) are described adequately by the predicted SST fields. GCM simulations using persisted August SST anomalies instead of forecast SSTs produce skill levels similar to those of the baseline for longer lead‐times. Given high‐skill SST forecasts, the scheme has the potential to provide climate forecasts that outscore the baseline skill level substantially.

The skill of global-scale sea surface temperature forecasts using a statistically based linear forecasting technique is investigated. Canonical variates are used to make monthly sea surface temperature anomaly forecasts using evolutionary and steady-state features of antecedent sea surface temperatures as predictors. Levels of forecast skill are investigated over several months' lead time by comparing the model performance with a simple forecast strategy involving the persistence of sea surface temperature anomalies. Forecast skill is investigated over an independent test period of 18 yr (1982/83–1999/2000), for which the model training period was updated after every 3 yr. Forecasts for the equatorial Pacific Ocean are a significant improvement over a strategy of random guessing, and outscore forecasts of persisted anomalies beyond lead times of about one season during the development stages of the El Niño–Southern Oscillation phenomenon, but only outscore forecasts of persisted anomalies beyond 6 months' lead time during its most intense phase. Model predictions of the tropical Indian Ocean outscore persistence during the second half of the boreal winter, that is, from about December or January, with maximum skill during the March–May spring season, but poor skill during the autumn months from September to November. Some loss in predictability of the equatorial Pacific and Indian Oceans is evident during the early and mid-1990s, but forecasts appear to have improved in the last few years. The tropical Atlantic Ocean forecast skill has generally been poor. There is little evidence of forecast skill over the midlatitudes in any of the oceans. However, during the spring months significant skill has been found over the Indian Ocean as far south as 20°S and over the southern North Atlantic as far north as 30°N, both of which outscore persistence beyond a lead time of less than about one season.

In light of strong encouragement for disaster managers to use climate services for flood preparation, we question whether seasonal rainfall forecasts should indeed be used as indicators of the likelihood of flooding. Here, we investigate the primary indicators of flooding at the seasonal timescale across sub-Saharan Africa. Given the sparsity of hydrological observations, we input bias-corrected reanalysis rainfall into the Global Flood Awareness System to identify seasonal indicators of floodiness. Results demonstrate that in some regions of western, central, and eastern Africa with typically wet climates, even a perfect tercile forecast of seasonal total rainfall would provide little to no indication of the seasonal likelihood of flooding. The number of extreme events within a season shows the highest correlations with floodiness consistently across regions. Otherwise, results vary across climate regimes: floodiness in arid regions in southern and eastern Africa shows the strongest correlations with seasonal average soil moisture and seasonal total rainfall. Floodiness in wetter climates of western and central Africa and Madagascar shows the strongest relationship with measures of the intensity of seasonal rainfall. Measures of rainfall patterns, such as the length of dry spells, are least related to seasonal floodiness across the continent. Ultimately, identifying the drivers of seasonal flooding can be used to improve forecast information for flood preparedness and to avoid misleading decision-makers.