Uncertainty, weather and climate

E. Linacre and B. Geerts



‘Certainty’ is defined as ‘undoubted fact, indubitable prospect, absolute conviction, beyond possibility of doubt’ (Oxford Dictionary). In sciences, certainties are rare, conclusions doubtful, discoveries provisional and even widely accepted proofs vulnerable to change in the light of contrary new evidence. This absence of any claim to infallibility in scientific results is sometimes overlooked, especially by the media, because of the obvious success of the scientific procedure in fields like earth sciences and biomedicine, as well as in other more technological fields where much of the credit is due to engineers rather than scientists. As regards calculations, especially in earth sciences, it should always include an estimate of the possible error, the existence of some error being a basic assumption. This applies equally to weather forecasting and considerations of future climates. In fact weather and climate forecasting are perhaps computationally the most demanding yet still the least certain of all scientific trades.

Weather prediction

In the case of weather forecasting, it is customary, in Australia and the USA for instance, for the chance of rain to be mentioned as a percentage, eg 20% (1). Operationally, a 20% chance of rain means that in the long run more than a trace amount of rain falls sometimes during the forecast period (usually a day) at an arbitrary location within the forecast area, on about a fifth of the occasions when that chance is quoted. Whether indeed a forecaster allots the correct percentages, is a matter for retrospective checking. Prospectively, it is an intuitive process.

Practically, probabilistic precipitation forecasting is guided by considering a wide range of evidence. The allotted percentage depends on the degree to which all the evidence is mutually consistent. For example, prognoses of weather patterns from several numerical models can be compared with each other, and the lower the various predicted rainfall amounts, the lower the percentage probability will be. Two bits of information usually not mentioned in public weather forecasts are: (a) how much rain is expected (i.e. the model consensus); and (b) the uncertainty, or range of model estimates (i.e. the confidence to be placed on the overall consensus).

The range of model forecasts is due to the limited predictability of the behaviour of the atmosphere. Even a perfect numerical weather prediction (based on available observational input) eventually becomes meaningless, because the atmosphere is an inherently chaotic fluid. Chaos is due to non-linear processes, and since the smallest (molecular) scale cannot be resolved in macroscale weather prediction models, the non-linear feedback of unresolved motions on macroscale motions remains unknown. A non-linear process forces a change that is not simply proportional to a variable. Attempts have been made to parameterise non-linear feedbacks in terms of macroscale (resolved) variables, but errors remain which eventually overwhelm the prediction.

 Chaos in fluids such as the atmosphere is an obdurate characteristic of reality, yet numerical weather forecasts have improved over the last few decades for three reasons:

  1. More detailed, more frequent, and more accurate observations have led to better model initializations. Especially the number of remotely-sensed data, such as satellite and radar data, have dramatically increased.
  2. The rapid increase in size and speed of available computers has allowed higher resolution simulations and the use of computationally more demanding parameterisations.
  3. Our knowledge of atmospheric processes has improved, in particular the parameterisation of clouds, radiation, subgrid-scale processes, and surface interaction.

Climate prediction

As regards estimating climates some decades hence, there is again considerable uncertainty. Clearly the concentration of greenhouse gases is increasing, and all models, from the simplest one-dimensional radiative transfer model to the most complex climate model, agree that this results in higher global-mean surface temperatures. On the other hand, interactions between the atmosphere and other ‘spheres’ (mainly the biosphere and the hydrosphere), which are of negligible relevance to weather forecasting, are essential to climate prediction, and are poorly understood. The uncertainty allows for a diversity of opinions in interpreting climate-modeling results.

Given the quasi-certainty of global warming, there are pessimists who foresee disasters of various kinds and therefore counsel immediate and drastic alterations to human society. They point to the overwhelming bulk of evidence in favour of global warming from humanity’s activities, and emphasise that all climate forecasts agree on the fact of significant warming, and the narrowing range of differences. Many pessimists also have a long-standing concern for the natural environment. The pessimists also include faddists, who operate through publicity rather than through financial clout. Also people who actually enjoy predicting gloom, since this provides either the pleasure of saying ‘I told you so’ or the happiness of doom averted. All but a few climate experts seem to be pessimist in their opinion about the societal response to the changes they foresee, mainly because of a realisation of the magnitude of the uncertainty, and of the possible consequences.

Contrariwise, there are optimists who believe that predicted changes are small, often beneficial, and that human society and global ecosystems will adapt to the changes. The optimists believe that the survival of human society thus far demonstrates its ability to respond successfully to challenges such as global warming, so long as we keep on an even keel. They are true conservatives, and tend to operate politically by covert influence and appeals to the self-interest of groups privileged in the present society. Optimists emphasize the contradictory evidence provided by various climate models, in particular the disagreement about the magnitude, the timing, and the regional distribution of the warming. They point to the degree of uncertainty of some postulated feedback mechanisms, such as those related to the hydrological cycle. Both groups, the pessimists and the optimists, claim to be the more realistic.

Unfortunately, society has few options in the face of the possibility that the pessimists might be right. One choice for the government is to ignore them and carry on with business as usual. This is advocated naturally by powerful interests with a large stake in the status quo. Another option is to adopt some of the easier changes proposed by the pessimists, postponing the more difficult or expensive ones till the uncertainty is less. A delay in making changes, should the optimists prove wrong after all, would then make any remedy more difficult or impossible. Promising to change only where there is complete certainty is not an option, or, rather, amounts to the first option of pressing on, regardless of gloomy conjectures.

A common response to dilemmas of this kind is to invoke the precautionary principle. There is no agreed definition of what this really means, but perhaps it implies something like the following - in the face of uncertainty act cautiously, keep as many options open as possible by avoiding irreversible actions, and take whatever steps are possible without undue penalty.

Added justification for caution is provided by a further worry shared by many experts, regarding climate stability. There is an as-of-yet-unquantified risk that at some point we may unwittingly cross some threshold of change which would trigger sudden runaway processes, such as the switchings which have occurred in the past between the Ice Ages. The repeated flip-flopping between Ice Ages and the periods between, and back again, with two alternative patterns of global winds and ocean flows (and hence of temperatures), implies some undetermined instability. There has been a suggestion that the switching from one regime to the other may be due to stopping and starting of a vast marine circulation involving subsidence of the ocean in the north Atlantic, but that is still conjecture. Rapid climate changes appear to have occurred, for instance at the end of the last Ice Age, possibly because of a rather transient switching of the regimes. Likewise on a shorter time scale, some records of this century’s annual rainfalls or temperatures show alternating stretches of years with particular cooling or warming trends, again suggesting that a switching between regimes occurs when some threshold in crossed. It is possible that more than two semi-stable atmosphere-ocean circulation systems exist. Many experts share the concern that gradual global warming may eventually go beyond some threshold into an irretrievable situation. Global mean temperatures have varied widely and frequently during the last few million years, but during this period the global mean temperature appears to never have been as high as they project it will be in the mid 21st century. The uncertainty of climate stability is a key reason for the pessimism of many climate scientists.

A realization of our ignorance concerning those matters is not the only reason for atmospheric scientists being modest about their ability to predict the future. Confidence was recently jolted by the discovery of the Antarctic ozone hole, whose existence and extent were quite unrealised for years by the whole scientific community, despite overlooked measurements which already demonstrated it. There may be similar important factors, operative but still unappreciated, influencing the shape of the future. We are uncertain.

So what does this all mean in practice? In view of the lack of certainty in atmospheric sciences and the possibly draconian effect of climate change on mankind, we offer the following advise:

  1. In deciding how to react to strong but inconclusive evidence that humanity is indeed altering the climatic environment, do not wait for certainty; it will never come.
  2. Do not compare the uncertain advantages of a course of action, such as reducing carbon-dioxide emissions, against the uncertain disadvantages of that course, considered on its own. Instead, compare the likely outcomes of that course against those of alternative actions (such as doing nothing about the present CO2 emission increase), even though the various outcomes are each uncertain to some degree. (There may be some canceling of uncertainties when like is compared with like.)
  3. Be aware of the motives of parties involved in the current climate change debate.
  4. When the stakes are high, err on the side of caution, and guard against the worst.



(1) Vislocky, R.L., J.M. Fritsch, and S.N. DiRienzo, 1995: Operational omission and misuse of numerical precipitation probability expressions. Bull. Amer. Meteor. Soc., 76,49-52.