by Sarah Dry
Just as Hann saw climatology as a helpmeet for meteorology, so others argued that physics was a needed stiffener that could transform the science of the atmosphere into a true science. The scales to which these disciplines directed their energies were largely as distinct as their methods. While climatologists set out to encompass the globe with imperial maps keyed to resource production and extraction, meteorologists focused instead on devising physical theories that could be regional, local, or even, in the case of clouds, hyperlocal in nature.
As these differences between meteorology, climatology, and an incipient physical geology such as informed the ice age debates indicate, the concept of change was itself unstable from the middle decades of the nineteenth century through to its close. Which kind of changes could be looked for, by whom, and using which tools had become very public and very controversial questions by the end of the nineteenth century. What it meant to be a science—how much it could be a function of data gathering and how much it required theory—was a primary question from which everything else, including what even counted as data, emerged.
These disciplinary anxieties formed the backdrop to Walker’s conundrum. How was he to escape the confines of Schuster’s meteorological “museum,” full of musty and disconnected facts? What Walker realized, thanks in part to the work of those who had come before him, was that solving the mystery of the monsoons would require two things. First, he would need to shift the scale of inquiry from local, regional, or even pan-regional studies to truly global surveys of world weather. Second, and just as importantly, Walker realized that he needed to ditch cycle-hunting in favor of something with a qualitatively different mathematical basis. He saw himself not as a hunter in search of one charismatic meteorological megafauna—the singular “link” between one cycle and another—but as a surveyor charting the landscape of the weather itself.
Here is where Walker’s ignorance of the weather may have been his greatest asset. Without any pre-existing assumptions about which aspects of the atmosphere might have the most bearing on the monsoons to guide him, he realized he needed a tool with which to evaluate all factors and to help him determine which, if any, were the most important. The tool he had was statistics. Specifically, he developed a technique for calculating what he called the reliability of the correlation coefficient between two factors. What this meant was that he had a device by which to sift the vast mountain of data. Before Walker’s innovation, the best tool cycle-hunters had was visual. They plotted charts comparing one signal against another (such as barometric pressure against sunspot appearances) and looked at the resulting curves to see if any pattern emerged—either an especially close fit or an especially poor fit, evidence perhaps of an inverse relation. Walker realized he could sharpen a device developed by statistician Karl Pearson, called a correlation coefficient, to sift through numbers statistically. Pearson’s correlation coefficient was a tool for identifying patterns—degrees of correlation—that linked two sets of data. This was a very helpful tool for sorting the vast reams of statistics which came flooding into Walker’s office.
A problem arose when it came to looking for real patterns in weather data: Pearson’s correlation coefficient tool was sometimes too good at finding patterns. When comparing two sets of random data, there is always a certain likelihood that you will find a relationship between them. The same is true when comparing real data, such as, for example, barometric pressure in different parts of the world. Pearson’s tool was unable to distinguish between the real correlations—that is, those that indicated underlying physical connections—and those that arise purely as a function of the quantity of data being compared. When comparing dozens, or even hundreds of data sets, as Walker was, the chance of false positives is large. Walker’s tool provided a measure of just how much correlation was required to balance out the likelihood of false positives in large data sets.
By applying his criterion of reliability to Pearson’s correlation coefficient, Walker was able to generate a quantitative measure of the likelihood that a correlation between two series of numbers was not due to chance. Instead of eyeballing a series of curves, Walker could rank relationships numerically and identify which were statistically robust and therefore likely to reflect something happening in the real world and those that were less strong and therefore more likely to be merely random. His technique was both more accurate and vastly more efficient at sorting through the huge data sets with which he was confronted than that of his predecessors. It was, as Napier Shaw, another leading researcher, recognized, a “kind of searchlight for sweeping the meteorological horizon from some selected point. The principal features of the otherwise invisible landscape, which in this case extends over the whole globe, can thus be located.”36
The landscape of Walker’s investigation was global. This was, again, as much a function of Walker’s ignorance as it was a calculated decision. Without a clear sense of where to shine his spotlight, he needed to shine it everywhere. Any correlation he found was, as Napier Shaw put it, a “very sensitive plant, it is much easier to kill one than to make one; whatever happens in the way of accidental errors, it must suffer.”37 That was the idea. If Walker were to find real relationships in the vast sea of data, he had to be merciless with any putative links. Only the strongest, most statistically resilient could be allowed to survive. These could lead the way for scientists who had a physical theory of the circulation of air, wind, and rain to explain what Walker had merely indicated.
And what he had helped reveal was this: There was something called world weather. It consisted in large regions of alternating high and low pressure that spanned the globe and changed with the seasons. There had been theories before of what had been called the general circulation of the atmosphere, dating back to Hadley’s theory of the trade winds in the eighteenth century. More recently, a spate of work done during the 1880s and 1890s, drawing on the same sorts of telegraphic correspondence networks that Walker did, had begun to pick out a series of such oscillatory, or seesaw, relationships between areas of characteristically high or low pressure. These papers, many of them by cosmic physicists, blended the tools and approaches of the cycle-hunters with those of the physicist accustomed to thinking about physical connections between matter. They generated maps, often of pressure but also of temperature, which demonstrated intriguing, even astonishing, connections between distant parts of the earth’s atmosphere. The term oscillation was used early on to describe the inverse relationship between pressure in different parts of the globe that many of these studies found. Léon Teisserenc de Bort, an architect of universal cloud studies, had shown that there was a relationship between the average pressure in Europe and that in certain “centres d’action” in Iceland, the Azores, and Siberia. Henry Blanford had done similar work for the Southern Hemisphere, showing that pressure in India, Siberia, and Mauritius was linked. H. H. Hildebrandsson, a round-faced Swede, had gone much further with his monumental series of five memoirs presenting a ten-year run of average monthly pressure data from no fewer than sixty-eight locations from all quarters of the globe. He used this data to push even further from these still hemispheric centers of action to suggest that there were what he called “intimate relations” between all of the centers of action on the globe.38 And finally, Hildebrandsson and de Bort’s Cloud Atlas of 1896 had shown that it was possible to leave behind what Julius von Hann called “church steeple politics” in meteorology (what could be seen from a church tower) and move toward ambitious, global projects.39 Clouds, to state the obvious, obeyed borders not at all, so any project to map them had to be similarly wide-ranging.
* * *
Here, then, was the landscape upon which Walker could shine his searchlight. From a tradition that stretched centuries into the past, which had built up interest in and knowledge of storms, to more recent attempts to collect and compare data at hemispheric scales, Walker had managed to arrive at precisely the right moment to submit a truly global data set
to scrutiny. Like Blanford, Teisserenc de Bort, and Hildebrandsson, Walker found evidence of oscillations in the pressure data he’d collected. But while they had been limited by their visual techniques to making vague statements about the nature and degree of these connections, Walker’s correlation coefficients allowed him to eliminate those connections that were less meaningful. He found 400 significant relationships—correlation coefficients worth paying attention to.40 Subtracting the spurious connections left him with “three big swayings,” or inverted relationships between pressure. The biggest was between the Pacific and Indian Oceans. This Walker named the Southern Oscillation. Two smaller swayings, between Iceland and the Azores and between parts of the North Pacific, he named the North Atlantic Oscillation and North Pacific Oscillation.41 In these locations, pressure existed in inverse relations. When the barometric pressure rose in Iceland, it seemed to fall in the Azores, and vice versa.
One of the first questions to which he put his correlation coefficients was that of sunspots. In a 1923 paper, he demonstrated that there were no meaningful correlations between the eleven-year sunspot cycle and that of the monsoons.42 He seemed to recognize the discomfort, and even disappointment, he may have caused. It was natural, he acknowledged, “after long ages of belief in the control of our affairs by the heavenly bodies,” to believe in natural cycles. But the urgent need for good monsoon forecasts, and the terrible suffering the famines had caused him, had nevertheless driven him to “replace instinct by valid quantitative criteria.”43 Eliot’s gamble in hiring Walker had paid off. Sort of. For even as Walker brought the edges of meteorology and empire to their ultimate endpoint—the entire earth—his achievement also represented a retrenchment and scaling back of ambitions. In gaining world weather, he had sacrificed the cosmos. Taking away the hope that secret cycles might unlock the monsoon was just about acceptable if, in return, Walker could offer something better.
That something better was, of course, his original goal of predicting the monsoons. The monsoon forecasts, which had begun in the 1880s and had been suspended in 1902 following the disastrous famines, had been reinstated on the basis of Walker’s findings. Walker’s predecessor, Eliot, had emphasized how dangerous “the striving after perfection in short-period forecasts was.”44 There was too much imperfect information and experience with failure to treat forecasts as anything other than probabilities. But Eliot’s cautious words were hard to hear against the background of famine and economic imperatives, and the government pressured Walker to publish forecasts once again. Walker was the first to be cautious, and even critical, of the forecasts, which he emphasized were only as good as the correlation coefficients he was able to find. These varied from year to year, sometimes dramatically. He urged that forecasts be issued only with strong provisos. Foreshadows, rather than forecasts, would, he thought, be a more appropriate, modest name for them.45 But the stronger term had stuck, and the tendency, or desire, for these pronouncements to be powerfully predictive was as great as it had ever been. There were some successes in prediction, but it seemed there were just as many failures, and it was embarrassing, after so much time and expense, and in the face of such evident need, that professional meteorology was often unable to offer better prognostications. The fear of making inaccurate predictions could lead to an absurd state of affairs where experts were less adept at forecasting than simple folk. It was unfortunate, commented one writer, Charles Daubeny, when “the untutored peasant sometimes would seem to possess an intuitive insight, whilst the philosopher, although he may plume himself on his acquaintance with the general laws of atmospheric phenomena, is often at a loss to unravel the entangled skein of effects connected with it which daily observation brings before him.” Meteorologists were damned if they did and damned if they didn’t. A bad prediction could tarnish their name, while too much reserve was also unacceptable. Daubeny continued that while charlatans had no compunction about making predictions, “a Herschel or an Arago declare themselves incompetent to anticipate what may chance to supervene within the space of the next four-and-twenty hours.”46
The ironic fact of the matter was that it was easier to use the monsoon to predict what would happen elsewhere in the world than it was to predict the rains themselves.47 Why that was the case, Walker the mathematician was unable to say. Luckily, though the monsoon continued to flummox and tantalize hundreds of millions of Indian farmers and those who relied upon their grain, no more famines occurred on the scale of the terrible death and suffering that had preceded Walker’s arrival. Changes in economic and social policy on the part of the British, and a run of good monsoon years, were to thank for that.
If Walker had failed in his primary objective of predicting the monsoon through statistical means, he had also failed to provide any physical explanation for the discovery he’d made. It was, in a way, like throwing a boomerang without understanding the physics beneath it. In that case, his lack of knowledge hadn’t kept him from excelling, but he’d nevertheless been driven to try to describe precisely how the device worked. Though he’d embraced his task in India using the most effective means that were available to him, he never forgot what he’d lost in the bargain. In a 1918 lecture to the Fifth Indian Science Congress, he emphasized how important it was to have a grasp of the fundamental principles driving phenomena under study. “What is wanted in life,” he urged the students, “is ability to apply principles to the actual causes that arise . . . When Pasteur as a chemist was asked to find a remedy for the pest that was ruining the French silk industry, he knew absolutely nothing of silkworms; yet he solved the problem, and it was general understanding of Nature’s methods that brought him success.”48 Walker knew better than anyone that that general physical understanding was precisely what was missing from the world weather he had discovered.
Just as Walker had failed to find a way to predict the monsoon, the larger project of melding meteorology with astronomy that went by the term cosmical physics had, by the end of World War I, largely faded from view. In its place was a new branch of meteorology. Instead of trying to link the heavens and the earth, as cosmic physicists had, this new type of meteorologist tried to link the lower atmosphere, to which most meteorological measurements had long been confined, with the upper, which was becoming gradually more accessible. Charles Piazzi Smyth’s expedition to Tenerife was an early example of the push to establish mountaintop observatories that, in addition to offering better views of the stars, made it possible to take readings of the upper air. Mountains had obvious drawbacks when it came to tracking the movements of a free-flowing atmosphere. After a series of spectacular and dangerous balloon ascents into the upper atmosphere, notably by English meteorologist James Glaisher, researchers sought safer ways of taking readings of the air high overhead. One way was to observe the motions of clouds, as the organizers of the International Cloud Atlas understood. But these observations could only reveal so much about the atmosphere. More precise data would require sending instruments themselves into the skies. Kites and unmanned balloons soon became the prime instruments for plumbing the ocean of air. In the late 1890s, Teisserenc de Bort, who had retired from his post as director of the Central Meteorological Office of France, established a meteorological field station at Trappes, southwest of Paris. There, he pioneered techniques for launching the large and delicate balloons needed to reach the upper atmosphere, using a large hangar set on a rotating platform, which could protect the balloon from ground winds until it was safely launched. Using this apparatus, and a self-registering device to record temperature, pressure, and moisture, Teisserenc de Bort carried out dozens of soundings in the years around 1900. The traces recovered from the self-registering device—which scratched its readings into lampblack that was impervious to the damp conditions—revealed a new aspect to the atmosphere. The temperature of the atmosphere fell in a uniform manner until the balloon reached some eight kilometers high, at which point it stopped decreasing. In 1902, Teisserenc de Bort named this region of the
upper atmosphere the stratosphere, and coined a new phrase for the layer closest to the earth, the troposphere.49
Walker himself was well aware of the need to understand the upper atmosphere better. “I think the relationships of world weather are so complex that our only chance of explaining them is to accumulate the facts empirically,” he wrote at the end of his life, “and there is a strong presumption that when we have data of the pressure and temperature at 10 and 20 km, we shall find a number of new relations that are of vital importance.”50 During his tenure as Director-General, he established an upper-air observatory in the northern plains of India, at Agra. Starting in 1914, a ten-year experimental program was carried out. Among other things, the balloons sent up by Walker and his men showed that the stratosphere—the zone of constant temperature—started much higher in the atmosphere above India than it did in Europe.51
Walker left India in 1924 after twenty years of service. His achievements (including helping hire increasing numbers of Indians into the Met Office) were lauded, he was awarded a knighthood, and he took up a position as professor of meteorology at Imperial College. He soon joined the Imperial College Gliding Club. And though he complained that his reflexes were not sharp enough for successful gliding, he accompanied the younger gliders on several expeditions in the South Downs. He sometimes took his boomerang with him and sent the device flying far above him in the gentle air of southern England before it began its perfect return, vibrating cleanly through invisible turbulence before coming to rest in his long, elegant fingers.