STORM TRACK: July 31, 1982 (Volume 5 Issue 5)

Twentieth century meteorology is marked by many important advances and events. The nature of the upper flow and its relation to surface weather features has generally been established. Better prediction techniques have evolved as new knowledge has been added, along with the advances made in the field of electronic computers. Computers have also helped to model the atmosphere mathematically, in order that we might 'study' weather in the laboratory.

I think that it, can be safely stated that what we now know regarding the nature of the atmosphere outweighs what we don't. The acid test for this statement is the fact that we can now predict, weather events, over even a relatively small region, with more than seventy percent, reliability. Prior to the turn of the century, this was probably not true. But, it is important to remember that knowledge is cumulative. Without the dogged determination of those early meteorologists, we would have little to feed our computers today. Man has a tendency to forget faraway events. For example, when the Norwegians developed a full, three-dimensional model of the life cycle of cyclones, it was easy to speak of the 'Norwegian air mass analysis and forecast technique', without constantly mentioning the groundwork in air mass theory laid by the American meteorologists Espy and Loomis, and tested by the Scottish Admiral Fitzroy. It is the purpose of this, as with any historical record, to assure that the early contributions are not forgotten.

The first recorded measurements of the upper atmosphere were the mountain top experiments on pressure, conducted by Pascal and Perier in 1648. By 1686, enough was known about the vertical distribution of pressure (at least in the first few thousand feet), to allow Sir Edmund Halley to fit the values to a hyperbolic curve. (It's actually an exponential decrease with height, but a hyperbola is very close.) The vertical temperature distribution was first measured by Wilson and Melville at Glasgow, Scotland in 1749, utilizing a 'minimum' thermometer and a kite. Shortly thereafter (1782), hot air balloons were inventedand during the following century scientists carried out experimental kits and balloon flights to study the characteristics of the upper air.

Although many of the general features of the vertical structure of the atmosphere (atleast to 15,000') were measured, it wasn't until 1898 that serious efforts were made (via instrumented kite flights)to study the space and time distribution of upper air pressure, temperature and dewpoint.

Meteorologists' desire for upper air observations was becoming increasingly strong as the twentieth century approached. For the previous century and a half, much had been suggested concerning the upper atmosphere -- there was a lot of theory, based on few measurements. The theories might have represented important, insights, but a bit more substantiation was needed. In 1755, Leonard Euler had utilized Newton's three laws of motion to formulate equations for fluid motion. Observational and experimental meteoro- logists during the 1800 s had accumulated sufficient knowledge to show that the atmos- phere behaved as a fluid. Euler's equations could, therefore, be applied, if more about the upper air characteristics were known. In 1904, Vilhelm Bjerknes suggested a method for objective forecasting which could apply the equations of motion to yield a numerical solution for forecast values of atmospheric variables. Unfortunately, the calculations required massive amounts of data and data manipulation and, so, could not be tested at the time.

There were several major problems that remained unsolved at the time, not, the least of which was that -somehow- the upper and lower portions of the troposphere were being treated as almost separate entities. Meteorologists knew that, there were important interactions (e.g. vertical motion near the center of extratropical cyclones), but too little information was available to tie them together. It wasn't until 1901, for example, that a group headed by Leon de Bort discovered the tropopause and stratosphere. However, this lack of understanding and knowledge was soon to be substantially reduced. In 1909, balloons equipped with parachutes and continuously recording instrument packages (Marvin meteorographs) began to replace kites in the United States. The balloons occasionally reached altitudes of 50,000 ft (but more commonly 15-20,000 ft), becoming a data source for forecasters by the mid-1920's.

In Europe, more abundant surface and upper air data were also becoming available. Discoveries in many facets of meteorology, including forecasting, were increasing. In 1918, V. Bjerknes took a closer look at, the structure of extratropical cyclones. Utilizing surface data and cloud observations (along with subsequent implications from the continuity equation), he constructed an improved three-dimensional cyclone model. His results gave forecasters a clear, understandable picture of a cyclone as a 'mover of air masses', of a system wherein cold air undercuts and lifts warm at the cold front boundary -and wherein warm air rises gently as it moves into cooler air at the warm front boundary. He called the two boundaries the 'squall' and 'steering' boundaries, respectively. In 1922, J. Bjerknes and H. Solberg of Norway then described the full lifecycle of extra-tropical cyclones. These authors envisioned cyclogenesis as a small, shear-induced, horizontal wave on the interface between two different air masses. They suggested that once warm air advances -even slightly- into the colder air, the pressure near the wave falls, and the winds take on components from cold toward warm air behind the disturbance, and vice-versa in the region ahead. A cyclone is thereby generated, along with the associated cold front/warm front boundaries. It was also noted that cold fronts advance faster than warm fronts and eventually overtake them. when this occurs, the region where cold air undercuts the warm front is referred to as an 'occlusion.' As this occlusion grows, the air mixes, becomes stable and barotropic, and finally dissipates the wave through friction.

The new balloon data showed that, the upper westerlies were not truly zonal but flowed in rather smooth, wave-like paths, swinging first north then south in their journey about the globe. Typically, four or five such large waves (shaped like sine waves) were found around the hemisphere. These so-called 'long waves' were first investigated and described by Carl-Gustav Rossby and collaborators in the late 1930's. Rossby also found that, superimposed on these long waves, were smaller disturbances, called 'short waves', which traveled rapidly through the slowly moving long wave pattern.

The relation between the upper tropospheric flow and the lower tropospheric cyclones was formulated in a paper by Bjerknes and Holmbs in l944. Upper level shortwaves were postulated to form as a result of vertical motions induced by frontal waves (Earlier formed shortwaves might even help induce new frontal waves). Utilizing the gradient wind equation, the authors demonstrated the existence of divergence ahead of and con- vergence behind the traveling upper waves. They also found convergence ahead of and divergence behind the associated surface cyclone. Thus, the upper flow could provide for depletion of air in front of the surface system and accumulation behind. This, in turn, helps move the surface cyclone along. Also, should an upper shortwave be moving more rapidly than the surface system (which occurs quite frequently), there will be a time when the two will be 'in phase,' and the upper wave will serve to suddenly intensify, or deepen, the surface system.

But what about that fellow, Vilhelm Bjerknes, who had the idea for objective weather forecasts back in 1904? As we noted, he was unable to try out his idea because of the massive number of calculations involved. In 1922, Lewis Richardson carried out an objective forecast by hand for an area roughly equal to a third of the United States. Richardson was very innovative -- he wrote the first textbook on dynamic meteorology and designed a scheme of 'finite differencing' to help with his forecast calculations. Nevertheless, the computations for a several hour forecast took him weeks to complete. His results showed that it would require one hundred or more efficient people, calculating full-time, just to pace the weather in an objective forecast for an area the size of the United States.

Even as Richardson fought his way through the unwieldy pile of calculations, help was on thehorizon in the form of rapid automatic calculators. Actually the first mechanical adding machine had been devised in 1642 by our old friend Blaise Pascal, and in the 1670's, Leibniz had provided a device that could multiply as well as add and subtract. During the period 1700-1900, mechanical calculating machines became larger and more complex. But strictly mechanical calculators were not the answer; being both bulky and slow. Less than twenty years after Richardson's first calculations, Howard Liken of Harvard created an electromagnetic computer by connecting a series of several dozen adding machines which were controlled by coded instructions on a single punched paper tape. The first fully electronic computer followed six years later (ENIAC, 1946), and in 1948, Jule Charney began working on the problem of numerical weather prediction utilizing ENIAC. Although the first machine prediction ran in 1949, it took three more years of computer development, work before the machines could even pace the weather, with just a simple barotropic model. It wasn't until 1955 that the first routinely available computer forecasts were distributed to field forecasters. Progress since then has come quickly, with the rapid advances in electronics and computer science that have occurred over the past twenty years. By 1962, enough 'computer power' had developed to allow the running of George Cressman's three-level baroclinic model, which was upgraded to a six-layer model in 1966. Most recently, the Limited Fine Mesh (LFM) model made its appearance in 1971.

The modern forecaster can now study an approximation of the general circulation pattern as it, will appear at future times out to several days. The LFM model output furnishes twelve hour interval forecasts of long and short wave patterns aloft, moisture aloft, surface weather patterns and estimates of broad rain areas, to name a few. Because the equations used to generate these results are approximations (often necessary due to lack of knowledge or lack of computer time), and because the computational techniques represent further approximations (necessary due to time restrictions and -occasionally- solvability), the forecaster must use the output as general guidance. However, for a successful prediction, he must supplement this information with other skills, including analysis of actual data, and knowledge and conceptual grasp of the behavior traits of weather systems, analysis of satellite imagery, etc. The thunderstorm forecaster must go a step further, because thunderstorms are quite literally below the resolution of the large scale data. A special knowledge of the thunderstorm's favored environment must be combined with whatever inferences can be made from all data sources to generate these 'sub-scale' predictions. It is the specific question of 'what is known about the thunderstorm's environment, and how do we know it', that is next on the agenda for discussion.