A major El Niño occurred in 1888. Drought struck Brazil and Australia, resulting in considerable damage. The country most tragically affected by the 1888 El Niño-Southern Oscillation was, however, Ethiopia. Detailed contemporary descriptions of the 1888 drought and the ensuing famine are available (Pankhurst, 1966). The 1888/89 Ethiopian famine provides a background for discussing how research into El Niño-Southern Oscillation could contribute to famine early warning systems for Africa.

       Pankhurst reports that 1888 was excessively hot and dry in Ethiopia, and unsatisfactory for agriculture. On 16 November 1888 there were reports that lack of rain had caused a large proportion of the crops to perish. By 8 January 1889, in certain areas, all the crops had been burned up by the sun. The harvest failure had reduced much of the country to misery by late in February. There was no grain production at all in some usually productive areas. The resulting famine was later exacerbated by a major epidemic of cattle plague (rinder-pest) and an outbreak of locusts and caterpillars.

       The failure of the Ethiopian harvest led to a doubling of the price of provisions by November 1888. In fact, it was very difficult to obtain food at any price. People were so discouraged that they made no efforts to save themselves but merely waited to die of hunger. Others were reduced to cutting up cow skins into pieces, which they dried, ground, and made into cakes. Various "unnatural practices," including the eating of traditionally forbidden food, the abandonment or sale of children by their parents, self-enslavement, suicide, murder, and cannibalism emerged all over the country. The famine was accompanied by the outbreak of epidemics and a sharp rise in the death rate through illness. Enfeebled famine victims often lacked the stamina to resist infection. The existence of large numbers of unburied corpses led to a substantial deterioration of sanitary conditions. Smallpox, typhus, cholera, influenza, and dysentery made their appearance in many parts of the country and killed large numbers. The famine and subsequent epidemics are believed to have resulted in the death of one-third of the entire population of the country. In certain areas perhaps 80% of the total population was lost, leading to the total desertion of some previously well-populated areas.

Does the El Niño-Southern Oscillation Cause Famine?

        All countries suffer from droughts. Not all are affected by famine. Does the influence of the El Niño predispose countries such as Ethiopia to the threat of severe drought and famine? El Niño-Southern Oscillation substantially affects the climate of much of the Indian-Pacific region. Although not every area is affected severely in each episode, El Niño does lead to an increased likelihood of substantial climate anomalies in each of these areas. It tends to amplify the climate variability, impose a specific temporal pattern on droughts and heavy rainfall periods, and allows some predictability of these variations. Nicholls (1988) found that the relative variability of annual rainfall was typically one-third to one-half higher for rainfall stations in areas affected by the El Niño-Southern Oscillation, compared with stations with the same mean rainfall in areas not affected by El Niño-Southern Oscillation.

       Nicholls and Wong (1990) confirmed, on recent data and using the coefficient of variation as a measure of relative variability, that the El Niño-Southern Oscillation does amplify rainfall variability in the areas it affects, relative to elsewhere. This effect was strongest at lower latitudes and low rainfalls and so is especially relevant to the semi-arid areas of Africa. The amplification factor is substantial. The variance of annual rainfall in an area strongly affected by the Southern Oscillation might be, depending on latitude and mean rainfall, more than double that in an area with similar mean rainfall that is not influenced by the Southern Oscillation. The higher climate variability (i.e., more severe droughts and floods) in countries affected by El Niño-Southern Oscillation may provide a partial explanation of why in these countries droughts can lead to severe food shortages.

       This is not to say that the high relative climate variability in countries affected by El Niño-Southern Oscillation will necessarily lead to famine. Social, economic, and political factors may exacerbate or mitigate the potentially higher climatic hazards. A highly variable climate can provide the frequent severe droughts and floods which set the scene for potential food shortages and even famine. But social, economic, and political factors must operate to realize this potential. There are countries with high relative climate variability (e.g., Australia) where severe droughts did not lead to famine. In some other areas affected by El Niño-Southern Oscillation (most notably northern China and India), the famine-drought nexus appears to have been broken in the past few decades, at least at the national level. In other El Niño-affected countries (e.g., Ethiopia), not every drought leads to famine.

       The influence of social, economic, and political structures in determining whether climate anomalies in a particular region will lead to severe societal impacts can explain why some countries with a climate strongly affected by El Niño-Southern Oscillation do not suffer food shortages that lead to famine and why in others, the inexorability with which famine followed drought during the nineteenth and early twentieth centuries has now been broken. The conclusion that nonphysical factors may determine whether the potential health problems caused by climate anomalies will be realized in a specific country or at a specific time does not, however, invalidate the point that increased climate variability associated with the El Niño-Southern Oscillation heightens the potential for severe societal impacts. It is important, therefore, to consider whether our current knowledge would enable the development of systems for the early warning of climate variations which could increase the potential risk for famine. It is instructive to consider the case of the 1888 Ethiopian drought and famine in this respect.

Could the El Niño-Southern Oscillation Have Provided Early Warning of the 1888/89 Ethiopian Famine?

       The Southern Oscillation Index (SOI) is a commonly used indicator of the state of the El Niño-Southern Oscillation. Large negative values indicate El Niño episodes. From about May 1888 through into 1889, the SOI was strongly negative, indicating a developing El Niño. Given some knowledge of the typical behavior of the El Niño-Southern Oscillation, predictions of the 1888/89 Ethiopian drought and famine could have been made some months in advance of the worst impacts. El Niño episodes typically start around March-May and last about twelve months. If the SOI is strongly negative by around June-July, then an El Niño has usually commenced and will last into the following year. This was the pattern during the 1982/83, 1991/92, and 1994/95 El Niño episodes, as well as in 1888. Since the worst impacts of the 1888 El Niño appeared late in the year in Ethiopia, simply monitoring the SOI could have provided some advance warning of potential problems in the areas it usually affects, including parts of Ethiopia.

       Figure 1 shows that this strategy of monitoring the SOI can be used to predict the likely severity of droughts in Australia (Australian droughts, like those in parts of Africa, tend to occur during El Niño episodes). The figure plots the number of districts in the state of New South Wales, in eastern Australia, and the SOI some 9-10 months earlier. The years when much of the state is in drought have usually been preceded by periods of low values of the SOI, indicative of the start of an El Niño. Similar diagrams could be prepared, if the data were available, for those parts of Africa affected by droughts during El Niño episodes, including Ethiopia and parts of southern Africa.

       It is important to note that monitoring the SOI, as described above, can provide drought "predictions," given our current state of knowledge. However, this approach is limited by the existence of a "predictability gap" early in the calendar year. It is difficult, with this simple approach, to provide useful forecasts before about July. Predictability across the March-May interval is limited by unknown factors. Improvement of predictions across this interval would be the most importance advance for increasing the utility of the El Niño-Southern Oscillation in famine early warning systems, because it would allow forecasts of drought 12-18 months in advance. This would enable more substantial and timely interventions to reduce the deleterious effects. The remainder of this paper examines the current state of El Niño-Southern Oscillation prediction and the research necessary to improve predictability across the "predictability gap."

Predicting the El Niño-Southern Oscillation

       There has been a consistent effort recently to develop methods for predicting the El Niño-Southern Oscillation at longer lead times than is possible just through monitoring indices such as the SOI. The main focus has been on predicting the onset and evolution of El Niño episodes. Occasionally, one sees claims of the discovery of a possible El Niño "trigger" (volcanic eruptions either on land or on the ocean floor are popular). In fact, the concept of a "trigger," i.e., an identifiable event that causes an El Niño, is irrelevant in a quasi-cyclic system such as the El Niño-Southern Oscillation. It is quite feasible for an initial small anomaly in a coupled ocean-atmosphere system to grow into a major El Niño. El Niño (and La Niña) episodes thus arise naturally because of the feedback between the ocean and the atmosphere. No external "trigger" is needed. Even if such a trigger did exist, its role (which could only cause the initial anomaly) would be unimportant relative to the feedback between the tropical ocean and atmosphere which amplifies the initial anomaly.

       The key to predicting El Niño, then, is not to find an elusive "trigger" but to develop a model that either identifies the anomalies in the ocean-atmosphere system as early as possible, or includes the various feedback processes and can thus mimic the inflating of initial small anomalies and the quasi-cyclic nature of the El Niño-Southern Oscillation. Three very different types of models have been developed to predict El Niño: statistical models, ocean models forced by observed winds, and coupled ocean-atmosphere models. All three kinds of models showed some success in forecasting the El Niño of 1986-87 (Barnett et al., 1988), and of 1991-92. They were, however, less successful in predicting the onset of the 1994-95 El Niño, although it was clear by mid-1994 that an El Niño was developing.

Models for Forecasting El Niño

1. Statistical models

       There have been efforts for several decades to identify statistical precursors to El Niño episodes. These precursors are not "triggers" or "causes" of El Niño episodes. They are, rather, indicators of the relatively early stages in the ocean-atmosphere interaction leading to El Niño. The best-established precursors appear in the tropical Pacific surface wind field and the global sea level atmospheric pressure field. Skillful forecasts could be expected from this sort of model at lead times of several months, although the expected skill varies throughout the year. Little skill is evident for forecasts for the Southern Hemisphere autumn and early winter.

2. Simple dynamical ocean models forced by observed winds

       The next step in sophistication is to express ocean response to the atmosphere in terms of the physical laws that govern that response, rather than use statistical relationships (e.g., Busalacchi and O'Brien, 1981). Here the ocean physics are described in terms of a linear transport model of the equatorial Pacific, driven by observed winds, and with solid walls at the meridional (north-south) boundaries and limited vertical resolution. Such models can reproduce many observed features of El Niño episodes, even though they are linear and without thermodynamics. They do not predict sea surface temperatures (SSTs) but rather a closely related variable, the thickness of the upper layer of the ocean. A statistical decision-making process, derived from inspection of model performance over a prior record, can be used to predict the onset of an El Niño. The problem with using ocean models driven by observed winds is that the winds can change quite rapidly over a period of a month or two. So the forcing of ocean models by observed winds can only provide short-term forecasts, i.e., up to a few months. A method to predict the interaction between the atmosphere and the ocean is needed, if we are to forecast with long lead times.

3. Coupled ocean-atmosphere models

        Several relatively simple (or "low-order") coupled ocean-atmosphere models of the equatorial Pacific have been developed to predict El Niño. The best-known is that of Cane and Zebiak (1985), which uses an ocean component similar to that used by Busalacchi and O'Brien (1981) with the addition of an Ekman layer and thermodynamics. Although the dynamics are linear, the equation predicting SST evolution is nonlinear (i.e., includes interactions between the variables governing the evolution of SST anomalies and the SST anomalies themselves). The SST anomalies are dynamically forced by surface wind stress anomalies (i.e., deviations of wind intensity and direction). The atmospheric model calculates surface wind anomalies that occur in response to SST anomalies. It has steady-state dynamics and a nonlinear heating parameterization to simulate the warming of the atmosphere by latent heating associated with precipitation. The heating depends on both the SST prescribed by the oceanic component and the surface wind convergence calculated within the atmospheric model. The ocean model has very coarse vertical resolution. When the components are coupled, the atmospheric heating depends on the model SSTs which, in turn, are determined by the surface winds generated by the atmospheric model. In operation, the coupled model is forced by observed winds up to the time of forecast, and then both the atmospheric and oceanic components of the model are allowed to evolve.

        There are a variety of other low-order coupled models, with rather different parameterizations of the ocean-atmosphere interactions. Models of this type have simulated some aspects of the El Niño-Southern Oscillation, including the time scales and spatial structures of SST anomalies, and have been used in experimental forecast tests, with some success (e.g., in 1991-92). More complex coupled models have also been developed (e.g., Philander et al., 1992; Latif et al., 1993). These models consist of an ocean model coupled to an atmospheric general circulation model (GCM), and show promise in their simulations of the El Niño-Southern Oscillation. Some "hybrid" models have also been developed, consisting of an ocean GCM and a statistical atmosphere. The final word in predicting El Niño will probably come with the development of coupled models with higher spatial resolution. These would be able to resolve adequately the equatorial waves but would also include better representations of other features of the tropical ocean and atmosphere than can low-order models. Their use in a prediction mode will require the development of schemes to assimilate ocean and atmosphere observations into the model.

Will El Niño Forecast Models Be Useful for Famine Early Warning Systems?

        Current forecast systems, and expected improvements to these systems and models, can provide forecasts of climate anomalies such as droughts with some skill. However, the availability of such forecasts does not necessarily mean that they will be useful in efforts to mitigate the adverse consequences of the climate anomalies. In particular, such forecasts may not necessarily help mitigate famines, even if they are incorporated into famine early warning systems. The utility of improved El Niño forecast systems in famine early warning will depend on a number of factors:

• The accuracy of the forecasts. Coupled-model experimental forecasts of El Niño episodes have been promising, but not perfect. For instance, the "re-emergence" of the El Niño in late 1992 was not predicted (e.g., Kerr, 1992). The imperfect nature of the forecasts means that care must be taken in their application. In particular, it is crucial that the forecasts be expressed as probabilities, rather than categorical, reflecting the less-than-perfect expected skill. Equally important is that users develop decision-making skills enabling them to use probabilistic forecasts.

• Variations in impacts between El Niño episodes. The 1888 episode impacted Ethiopia severely; the 1877 episode mainly affected India and China. Variations from episode to episode continue to confuse the picture of the impacts of the phenomenon. These variations emphasize the need to express forecasts derived from the El Niño-Southern Oscillation in probabilities, and to use them appropriately.

• The way in which the forecasts are used. There is a tendency for forecast users either to dismiss forecasts or to take excessive actions based on the forecast. These approaches fail to take account of the level of skill of the forecast. No forecast system for El Niño will be perfect, so a method for "hedging" is necessary in any application of the forecasts. This would be facilitated by the use of probabilistic forecasts, but methods for the correct application of probability forecasts in decisionmaking are required.

• Multiple sources of forecasts. Many organizations and individuals are now issuing El Niño forecasts. How can a user decide which is the "best" forecast?

• Confounding effects of climate change. Drought in Ethiopia in the past was clearly related to El Niño episodes. In recent decades this relationship has been confounded by the downward trend in rainfall in this area. A similar effect has occurred in parts of Australia. Are these trends a temporary aberration, or will they continue to confound the El Niño-drought relationships? In recent decades the SOI has been low, while global temperatures have increased. Is there a causal relationship? If so, how do we design decision-making systems to take account of both El Niño-related climate variability and climate change?

Phase-locking to the annual cycle

       Extended periods of drought or extensive rains do not occur randomly in time, in relation to the annual cycle, in areas affected by the El Niño-Southern Oscillation phenomenon, and rainfall fluctuations associated with it tend to be "phase-locked" with the annual cycle, i.e., they start and finish around the same time in each event. El Niño events usually start early in the calendar year and finish early in the following year. The droughts associated with such events also tend to occur around the same time of year in each event, although the preferred time of year varies geographically. The phase-locking means that a "timetable" for prediction for famine early warning can be developed. In Australia, for example, if widespread drought associated with an El Niño has become established by mid-year, it is reasonable to expect, based on historical data, that the drought will continue until at least the end of the year. If, however, there is no sign of drought or El Niño at mid-year, then widespread drought is unlikely to develop in the next six months.

Biennial cycle

       This phase-locking is related to a biennial cycle which is a fundamental element of El Niño-Southern Oscillation variability. There is also a lower frequency variation, but it is the biennial mode which captures the major features associated with El Niño-Southern Oscillation episodes. The biennial cycle is observed over the equatorial Pacific and Indian Oceans and is tightly phase-locked with the annual cycle. It varies in amplitude from cycle to cycle and sometimes changes phase. The biennial mode means that El Niño events are often preceded and/or followed by the wetter conditions associated with La Niña events. The change from El Niño-related drought to La Niño-related wet conditions can be rapid. These rapid changes can lead to further problems. For instance, a rapid onset of wet conditions, after an extended El Niño drought, can lead to severe soil erosion and complicate planting. Further research is needed to identify the impact of the biennial and phase-locked nature of El Niño on Africa.

How Could the Results of El Niño-Southern Oscillation Research Contribute to FEWS?

        Despite the problems mentioned earlier, increased skill of forecasts of El Niño episodes, through improved coupled models, should be useful for early warning systems. It is certain that, during an El Niño, at least some parts of Africa will be severely impacted. Some aspects of these impacts are predictable with current models (statistical and dynamical). Thought needs to be given to the appropriate institutional arrangements (national and international) to enhance the skill of these models, and to ensure that the forecasts are used to reduce the likelihood of climate-related famine. With appropriate arrangements, and current models, much of the anguish of the 1888-89 Ethiopian famine might have been avoided. With improved institutional arrangements and models, the devastating impacts of El Niño can be minimized. Unless such arrangements are instituted, future El Niño events might again result in disastrous consequences for Africa.

References

Barnett, T., N. Graham, M. Cane, S. Zebiak, S. Dolan, J. O'Brien, and D. Legler, 1988: On the prediction of the El Niño of 1986-1987. Science, 241, 192.

Busalacchi, A.J., and O'Brien, J.J., 1981: Interannual variability of the equatorial Pacific in the 1960s. Journal of Geophysical Research 86, 10901.

Cane, M.A. and Zebiak, S.E., 1985: A theory for El Niño-Southern Oscillation. Science, 228, 1085.

Latif, M., Sterl, A., Maier-Reimer, E., and Junge, M.M., 1993: Climate variability in a coupled GCM: Part I: The tropical Pacific. Journal of Climate, 6, 5.

Nicholls, N., 1988: El Niño-Southern Oscillation and rainfall variability. Journal of Climate, 1, 418-421.

Nicholls, N., and Wong, K., 1990: Dependence of rainfall variability on mean rainfall, latitude, and the Southern Oscillation. Journal of Climate, 3, 163-170.

Pankhurst, R., 1966: The great Ethiopian famine of 1888-1892. Journal of the History of Medicine and Allied Sciences, 21, 95-124 & 271-294.

Philander, S.G.H., Pacanowski, R.C., Lau, N.C., and Nath, M.J., 1992: A simulation of the Southern Oscillation with a global atmospheric GCM coupled to a high-resolution tropical Pacific Ocean. Journal of Climate, 5, 308.

Main IJAS Page
Table of Contents (Vol. 2)