«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
A beginning to long-range weather prediction can be attributed to Sir Joseph Norman Lockyer, a talented British astronomer. It was he who discovered helium in the sun’s atmosphere in 1868, 27 years before that element was found on the Earth. A prolific writer, he founded the science periodical Nature in 1869 and edited it for more than 50 years. Lockyer was convinced that solar activity had an effect on the world’s weather and climatic changes. The pages of Nature carried many articles concerning the influence of the sun on tropical agriculture. Much of his work from 1868 lay in obtaining weather and climatic data from across the world to be collated with his observations of the sun. He thought that the number and size of sunspots was related to the amount of rainfall on Earth. His son James published a paper jointly with Sir Norman in 1900 “On solar changes of temperature and variations of rainfall in the region surrounding the Indian Ocean”. Work on the solar influence on the worlds weather systems continued to be a major theme of his research.
Serious scientifically-based efforts at seasonal prediction continued in the early 20th century with attempts to predict the onset and intensity of the Indian monsoon. At that time, the monsoon was believed to occur independently of other weather patterns such as El Niño, the recurrent warming of the Pacific Ocean, which we now know produces catastrophic and disparate effects worldwide: torrential rains, river flooding, landslides, severe droughts, and wildfires. While scientists in South America were busy documenting the local effects of El Niño, Sir Gilbert Walker was on assignment in India, studying monsoons. A British scientist, Walker, who was the head of the Indian Meteorological Service, had been asked in 1904 to try to predict the vagaries of India’s monsoons after an 1899 famine caused by Seasonal prediction 131 monsoon failure. If the rain between June and September is significantly below normal, there can be drought, crops can fail and widespread famine and starvation can follow. This was the case in 1899-1900. Walker has been credited with being the first to note that weather is a phenomenon with global-scale influences.
Walker was convinced that the monsoon changes were in some way tied to global weather. He associated some patterns of rainfall in South America with changes in ocean temperatures. A connection between pressures at stations on the eastern and western sides of the Pacific, between Tahiti in French Polynesia and Darwin, Australia was found. He noticed that pressure rises in the east were associated with falls in the west, and vice versa — he called this the “Southern Oscillation”. In addition he realized that the Asian monsoons were often linked to drought in Australia, Indonesia, India, and parts of Africa. He claimed a connection between the Indian monsoons and mild winters in western Canada. Walker was convinced that all these events were part of the same phenomenon.
Walker noted that the random failure of the monsoons in India often coincided with low pressure over Tahiti, high pressure over Darwin, and relaxed trade winds over the Pacific. He was publicly criticized for suggesting that climatic conditions over such widely separated regions of the globe could be linked. His colleagues were skeptical of theories that gave a simple, single explanation for worldwide weather patterns, and in fact he was unable to translate his ideas into a scheme to predict the nature of the monsoons.
However Walker did predict that whatever was causing the connection in weather patterns would become clear once wind patterns above ground level, which were not routinely being observed at that time, were included.
He was right.
Walker’s results fell into oblivion until Jacob Bjerknes, in 1960, started to study the causes behind El Niño. In the 1970s and 1980s, the groundwork was laid for significant advances in the science. A system of measurement of the oceans started to be established. This included tidal gauges on islands in the tropical Pacific, instruments deployed by merchant ships to measure temperatures to 500 m below the surface, and — later — satellites measuring sea level using altimeters. It became clear that the oceans could and did force the atmosphere into systematic weather patterns — and vice versa. In the late 1960s to early 1970s wind-driven “Kelvin” waves in the oceans were predicted theoretically. These are waves trapped in the equatorial belt.
They have a scale north to south of 300 km or so, but an east to west wavelength of thousands of kilometres. Observations in the mid-to-late 1970s verified the theory. Numerical modelling soon advanced to the stage where these waves were being successfully modelled.
132 Chapter 11In 1982/83, the most intense El Niño in the instrument record to that time occurred — the strongest in 300 years. The resulting collapse of fishing off the shores of Ecuador and Peru, widespread flooding, disease, famine and more resulted in a combined worldwide bill estimated at US$ 20-30 billion.
And this El Niño had been raging for months before it was finally and convincingly recognized as an El Niño! In 1982, the eruption of Mexico’s El Chichón volcano pumped at least ten times as much ash into the stratosphere as had Mount St. Helens in 1980. The volcanic dirt in the atmosphere confused instruments on the satellites, and incorrect sea-surface temperatures were being reported. While meteorologists had their suspicions that something serious was indeed happening in the Pacific, many oceanographers were not convinced.
Oceanographers and meteorologists were determined not to be caught out again. Under the leadership of Adrian Gill, eminent scientist and author, they developed a scientific programme “Tropical Ocean–Global Atmosphere” (TOGA), implemented as part of the World Climate Research Programme of WMO. TOGA started in 1985 and was completed in 1995.
This highly successful ten-year international research effort produced fundamental new knowledge of the processes that couple the tropical Pacific Ocean to the global atmosphere. It ultimately led to the successful prediction capability for the El Niño phenomenon. The programme developed and implemented a tropical Pacific Observing System to monitor the state of the tropical Pacific Ocean, providing real-time records of the evolution of El Niño events.
The centrepiece of this observing system was the Tropical Atmosphere Ocean (TAO) array, with 68 moored buoys spanning the tropical Pacific, measuring sea surface temperature, surface winds and the thermal structure of the upper ocean. TOGA also conducted an unprecedented international field campaign TOGA Coupled Ocean-Atmosphere Response Experiment (TOGA COARE) in 1992–93 to quantify air-sea interaction processes in the tropical western Pacific Ocean.
Back now to the Centre’s role in seasonal prediction. In retrospect probably unwisely, Bengtsson’s strategic “Ten-year Plan 1985–94” had been prepared entirely without the involvement of Council, and was presented to Council in November 1984 in the form of a glossy full-colour 58-page brochure. While Bengtsson’s intention in presenting the Plan in this way was to convince the Council of the merits of the various proposals, perhaps the impression was given that Council should adopt the Director’s Plan in its entirety, or not at all. And, to be fair, the Member States represented on Council would have to provide the resources required to bring any such plan Seasonal prediction 133 to fruition. The Council discussion on the Plan was mixed. While some delegates welcomed it, Council as a body was clearly not inclined to adopt the Plan as presented.
Bengtsson agreed to develop a document, with practical proposals, for the following Council session, with input especially from the Council’s Scientific Advisory Committee. He noted that there was support for the proposed research into long-range forecasting.
The “Ten-year Plan 1985–94” evolved into the “ECMWF Long-term Strategy 1987–1996”, a — shall we say — more cautious, or less ambitious document. Reading the two side by side, the Strategy makes a slightly depressing read. Council adopted it in May 1986. Reference to forecasting
beyond the medium range was restricted to two somewhat repetitious sentences in Section 4 “The Programme of Research”:
It will be necessary to carry out extended integrations to study systematic model error and as an aid in predictability research
Extended range integrations will be required to assess not only systematic model errors but as an aid to research into atmospheric predictability
and one sentence under “Operational Aspects”:
The forecasting scheme and the range of dissemination products will be enhanced to include... should Council so decide, forecasts in the extended range.
Clearly the Council was not at the time in favour of the Centre becoming involved in seasonal prediction.
The definition of “medium-range” just adopted by Council: “the time scale beyond a few days in which the initial conditions are still crucially important” would appear to have allowed seasonal prediction; the initial conditions of the ocean are of crucial importance. However the “politics” of seasonal prediction highlights a continuing dichotomy. While for most Member States the ECMWF products are essential for their work, there is a continuing risk that the work of the Centre can overlap the activities of the National Meteorological Services. The Directors — and staff — of these services can understandably feel uncomfortable if they see the Centre encroaching so to speak on their territory. A division of responsibility between the Centre and its Member States needs to be maintained.
Finland voiced disappointment that the text relating to extended-range forecasting had now been removed almost totally. This was work that was clearly beyond the individual capability of the smaller Member States; it
134 Chapter 11hoped that by the end of the ten-year period, the Centre would be able to provide its Member States with extended-range forecasts.
It would be fair to say that privately Bengtsson was extremely unhappy with the Council’s de facto rejection of his plan for the future work of the Centre. Not only seasonal prediction but also other aspects such as wave prediction — see later — were weakened or entirely removed before the strategy document was adopted. He noted later that “this made me begin to realise the inertia of established institutions such as the National Meteorological Services, and the fragility of international organisations”.
Bengtsson knew that the observation system, the computers, and most of all the science, had advanced sufficiently in the recent years. It was time for a serious scientifically-based programme to begin at the Centre.
In any event, once the possibility of long-term prediction by the Centre had been raised, Bengtsson was not going to let it go away. He was convinced of the merit of such prediction. In spite of Council’s reaction to his proposal, Bengtsson informed Council that in the future “extended-range prediction would form an inherent part of the Centre’s research programme”. In his Four-year Programmes presented to Council after this, extended-range prediction was consistently mentioned.
Already in the early 1980s, Aksel Wiin-Nielsen and Ulrich Cubasch had made extended-range experimental model “predictions” looking at the impact of the Sea Surface Temperature (SST) on the tropical circulation. An intense El Niño in 1982–83 provided a test case for further work in 1984.
Two forecasts were run to 50 days. One had the normal SST, the second the El Niño anomalous SST. For the first ten days, there was little difference between the forecasts. In the later stages, after 15 days or so, the second forecast was measurably better. Thus use of the correct SST had correctly modified the forecast. However compared to other models, for example that being run at the UK Met Office by Tim Palmer and his colleagues, this early ECMWF effort was relatively — even “spectacularly!” — unsuccessful.
As we saw in Chapter 10, Bengtsson recruited Palmer from the Met Office; he had a strong research interest in extending forecasts to the seasonal scale. Later, Palmer and his colleague Cedo Brankovic did much work quantifying the impact of the ocean on atmospheric seasonal predictability using an improved version of the model.
Work done elsewhere was now showing the advantage of coupling the atmospheric model to a model of the world’s oceans. For example, in 1986/87 a coupled model had predicted an “El Niño Southern Oscillation” (ENSO) — a climate oscillation with a worldwide impact. Palmer was feeling frustrated; he believed that the Centre should be in the forefront of these Seasonal prediction 135 exciting developments. He recalled later that “the politics of the situation in the mid-1980s meant that the Centre became fully involved in seasonal forecasting with coupled general circulation models relatively late. However we caught up, and became one of the leaders in the field.” Also in the 1980s, Stefano Tibaldi and his colleagues, with an uncontroversial and straightforward extension of medium-range work, carried out a programme of 30-day integrations. The programme used some ensemble ideas. It was not until about 20 years later, in 2004, that 30-day forecasts became part of the Centre’s operational work.
In 1990, the Köberstiftung in Germany awarded Lennart Bengtsson, Bert Bolin and Klaus Hasselman the prestigious Förderpreis for their work relating to short-term climatic changes. Using his and Hasselman’s funds, Bengtsson hired a young, active scientist, Tim Stockdale, from Oxford University.
Bengtsson left the Centre at the end of 1990, but on becoming Director, David Burridge gave his full support to the Centre’s involvement in the field. This was in spite of the opinions of some senior staff that the Centre had been successful in large part because it had focussed strongly on its main task of medium-range prediction.