The perils and hazards of weather forecasting

From a talk given by Julian Hunt, the former chief executive of the Meteorological Office to a Geographers' Conference in Plymouth
Click to follow
The Independent Online

During the past decade, considerable progress was made in the scientific understanding and forecasting of natural and artificial hazards, and in the technology of risk reduction. However, as always in science and technology, each step forward reveals more questions and uncertainties. An encouraging feature was the growing collaboration in the academic sphere between social scientists on the one hand and natural scientists and engineers on the other; for example, in design and operation of warning systems and in post-disaster recovery.

During the past decade, considerable progress was made in the scientific understanding and forecasting of natural and artificial hazards, and in the technology of risk reduction. However, as always in science and technology, each step forward reveals more questions and uncertainties. An encouraging feature was the growing collaboration in the academic sphere between social scientists on the one hand and natural scientists and engineers on the other; for example, in design and operation of warning systems and in post-disaster recovery.

Another significant feature was the continued relearning that government cannot do everything to reduce disaster risk. In the Netherlands, local flood-defence committees are regarded as the basic element of their democracy. The insurance industry and employers of workers in vulnerable communities (especially in developing countries) can do much to encourage self-help by communities.

In understanding and predicting natural hazards, it is instructive to divide them between primary and consequential (or secondary) hazards (eg mud slides follow massive rain). The latter may be exacerbated by artificial factors or structures (eg a dam burst, disease outbreak, etc).

The actual location and time of suddenly occurring hazards cannot generally be forecast, because they arise from the instabilities that grow very rapidly (within minutes) from low amplitude disturbances - for example, avalanches and earthquakes. However, estimations of probability (or risks) of such hazards are improving, as the events and their local environments (eg cracking in rocks) are monitored in every detail, and as the conditions for their occurrence (eg snow conditions for avalanches) are better understood and predicted.

Where primary and secondary hazards such as high winds, floods, mud slides, lava flows, drought, dust storms, high temperature and pollution, develop more slowly (over hours and months), they usually arise from disturbances in reasonably well-defined dynamic and thermodynamical systems. Consequently, these hazards can be predicted if there is sufficient information about the initial state of the system before the disturbance grows.

For example, over the past five years there has been a 40 per cent reduction in the difference between the actual and forecast trajectories of tropical cyclones (the most severe kind of storm). This improvement in accuracy has largely come from deterministic models using dynamical equations, computer calculations and data from ships, aircraft, satellites, etc. Forecasting and the dissemination of warning on secondary hazards have also improved - for example the dispersion of volcanic dust in the atmosphere following eruptions and their effect on aircraft.

However, the errors in the forecasts of other hazards, such as floods along rivers, have not measurably improved. These events (and forecast errors) are not systematically recorded, even in advanced countries, probably due to insufficient data and inadequate understanding of fast run-off from built-up and intensively farmed areas.

A striking feature of longer-range hazard prediction over the past 10 years has been the transition from statistical methods to more deterministic calculations, as the connections between different factors have been better understood, the computer modelling systems have become more comprehensive, and more data becomes available. For example, seasonal drought/flood forecasts are now based on the same computer models used in secular-decadal climate predictions. As in environment and biological sciences, pioneering work on statistical prediction plays a valuable role in showing that deterministic predictions should be possible.

Increasingly, I predict this tradition will be seen in other fields of hazard prediction, and will be used for developing strategies in hazard reduction.

Comments