Lumpers v. Splitters

Lorraine Daston

  • Predicting the Weather: Victorians and the Science of Meteorology by Katharine Anderson
    Chicago, 331 pp, £31.50, July 2005, ISBN 0 226 01968 3

On the morning of 30 April 1865, Vice-Admiral Robert Fitzroy, head of the British Meteorological Department, slit his throat. Because Fitzroy had been the captain of the Beagle, which several decades earlier had carried the young Charles Darwin around the world to conduct the research that eventually bore fruit in On the Origin of Species (1859), and because he was a devout evangelical, some historians have chalked up his suicide to guilt over his unwitting complicity in the genesis of a scientific theory he detested. But it is more likely, as Katharine Anderson suggests in her engaging and enlightening study of Victorian meteorology, that the depressive Fitzroy was pushed over the edge by a quite different scientific controversy. Could the weather, and more particularly storms, be predicted? As a seaman and a scientist, Fitzroy had campaigned for a system of storm warnings, based on collations of data telegraphed to the Meteorological Department by observers from more than a dozen sites, six mornings a week. His elaborate signal code of cones and drums to alert ships to approaching gales was well enough known to figure in parodies and political cartoons in the London press. But the accuracy, and even the possibility, of his forecasts was doubted by some of the most prestigious scientific authorities, as Fitzroy was painfully aware. The Royal Society had been notably cool, intimating that his forecasts were more oracular than scientific.

‘Forecast’ was a word Fitzroy had chosen advisedly, in an attempt to steer a path between the solid ‘predictions’ of astronomers and the wild ‘prophecies’ of astrologers. Peering into the future was a crowded business in Victorian Britain, and a brisk business at that, as the sales of almanacs and ephemerides testified. Astronomy set the gold standard of futurology, with its precise predictions of eclipses and planetary positions. The discovery of Neptune in 1846, just where and when calculations indicated it would be, cemented astronomy’s reputation for near infallibility. Notwithstanding repeated attempts to correlate the weather with the phases of the moon, however, astronomical predictions turned out to be all but useless in foretelling rain or shine, hail or gale: matters of urgent interest not only to farmers and mariners, but also to any industrialist or merchant who depended on British ships to import and export goods; it was no accident that the Meteorological Department was subsumed under the Board of Trade. Need to know was the mother of belief, and almanacs offering tips on the upcoming year’s weather never wanted for customers, any more than economic modellers offering tips on next month’s stock market do nowadays. In both cases, an accuracy rate little better than chance proved no deterrent to either buyers or sellers. Mid-19th-century meteorology aspired to keep company with patrician astronomy, but couldn’t shake off the more plebeian associations of the almanacs: hence Fitzroy’s gingerly ‘forecast’, precariously positioned between science and charlatanism.

It wasn’t as if meteorology was a new science, still struggling to get its bearings. As early as the late 17th century, the Royal Society had enlisted volunteer observers to record barometer and temperature readings on a daily basis and send them to London to be tabulated and – somehow – synthesised to reveal the laws of the air. John Locke was one of scores of weather-watchers who interleaved their observations with entries in journals and commonplace book jottings. The diurnal rhythms of most of an adult life could be set by the metronome of the morning measurements of temperature, air pressure and wind direction. Even mental crisis could not shake meteorological routine. The German physicist Georg Christoph Lichtenberg, writing in 1771, might have been on the edge of despair – ‘Heart, head and all are infected, where shall I go?’ – but he still methodically noted in the margin that the barometer stood at 27’ 2’’ (Paris measurement scale) at 7.00 a.m. after a bad storm. Yet all these thousands and thousands of measurements failed to yield reliable regularities.

Part of the problem was that non-standardised instruments were being wielded by uncalibrated observers, as Lichtenberg’s careful specification of scale indicates. It was the rule rather than the exception for measurements sent in by far-flung correspondents to be incommensurable with one another. International congresses were still thrashing out these issues in the late 19th century. But even the best-organised observation networks, which handed out identical instruments to trained and conscientious observers, did not discover the equivalent of the law of universal gravitation for the atmosphere, or even rules of thumb that did better than hoary proverbs. By the end of the 18th century, some disappointed savants began to echo Galileo’s doubts as to the possibility of a genuine science of phenomena as capricious as air currents.

Nineteenth-century meteorology still depended on data gathered by networks of observers, but there were also striking innovations in scientific precept and practice during this period. Scientific instruments were increasingly manufactured rather than handmade, enabling an unprecedented uniformity, though British and American meteorologists continued to wrangle with their Continental colleagues over the relative merits of imperial and Celsius scales. However, even astronomers acknowledged that individual observers could be as variable as their instruments: observers, too, had to be calibrated. The sociological, psychological and physiological challenges of human co-ordination were tackled in notably different ways by Victorian meteorologists, who relied even more heavily on observers than their Enlightenment predecessors. The grand panjandrums at the Royal Society and Kew Observatory tended to opt for hierarchically organised networks crowned by their own institutions. Self-registering instruments that worked around the clock without complaint appealed to them, as did human observers who approximated machines as closely as possible. These would-be managers of science spoke with Bounderby-esque good cheer about human drudges who would tirelessly observe and calculate by the rule. (Parisian savants of similar views turned to students in écoles normales in order to enforce observing regimens with institutionally imposed discipline.)

At the opposite extreme were networks of amateurs like the one organised by George James Symons, one of Fitzroy’s assistants at the Meteorological Department, who recruited thousands of observers – ‘from peer to peasant’ – to the British Rainfall Organisation. Fitzroy’s nightmare had been the storm at sea; Symons’s was drought. During the 1850s Britons had seen deadly instances of both. Symons didn’t just wheedle and cajole his volunteers into submitting their observations on circulated forms; he also persuaded most of them to support the enterprise financially. There was no question here of some scientific foreman from Kew lording it over machine-like labourers. Symons’s network was dependent on the goodwill of his volunteers, and he cultivated it accordingly. Rules for making observations were up for discussion; attempts to secure uniformity were diplomatically couched as ‘suggestions’. He touted the BRO as a microcosm of England at its best: independent, public-spirited and as reliable as the weather was flighty.

These were contrasting visions of how to make science collective, and yet they shared a Victorian yearning to dissolve the egotistical individual into an ideal community. ‘L’art c’est moi, la science c’est nous,’ as the French physiologist Claude Bernard put it. Images of anonymous medieval masons building a cathedral or social insects sacrificing themselves for the good of the hive form a refrain in mid 19th-century writing about the scientific life. Fitzroy compared himself to an ant carrying its mite to the hoard. All the sciences, from astronomy to zoology, conjured with these dreams of a common undertaking to which the individual must humbly submit – in diametrical opposition to the romance of the solitary genius working in splendid isolation. But in meteorology the pressure to collaborate and co-ordinate was unusually intense. The objects of the science dwarfed the human scale of local experience. A storm front, like the geographical distribution of species, could be discerned only by multiple and dispersed observers, who pieced together their local observations into a vast jigsaw puzzle. Even phenomena that could be grasped by a single observer, like clouds, had to be studied globally. The British meteorologist Ralph Abercromby sailed twice around the world at his own expense to persuade himself and his colleagues that the main types of clouds to be found in Europe (cirrus, cumulus, stratus, nimbus) existed all over the world and so could legitimately form the basis of a universal classification system. The precondition for global science was universal nature – or was it the other way around?

Yet nothing was more local, more pungently evocative of place than weather. Abercromby admitted that the tropical cumulus cloud was irradiated by a light never seen in England and that the same fleecy cirrocumulus that presaged fair weather in England and France augured heavy rain in Italy. Studied close up by local observers, nature looked motley, however hard leading scientists in London and Paris pressed for standardised data that could be distilled into generalisations about weather everywhere and always. The relationships between London meteorologists and the scattered observers who reported to them were multiply strained: certified scientists v. amateurs, imperious organisers v. labourers, snobbish Londoners v. sullen provincials, lumpers v. splitters. The observatory-based scientists sought the natural laws of the entire atmosphere; their remote correspondents immersed themselves in the particularities of their own villages and boroughs. Most of the observatory scientists, however, acknowledged and respected the expertise of local observers. Divination by subtle signs, from a sudden stilling of birdsong to throbbing bunions, often outperformed elaborate collations and calculations in short-range forecasting. What in other sciences might have been dismissed as vulgar errors were dignified as lay wisdom by meteorologists.

The directors of observatories and meteorological services envied the local observers their forecasting record, but they also coveted the ease and speed with which a practised weather-watcher, whether sailor or farmer, could read natural signs and draw inferences from them. With the possible exception of astronomy, no Victorian science involved more tedium than official meteorology. There were data to collate and tabulate, means and averages to compute and a potentially infinite number of correlations to test. If weather didn’t follow the phases or perturbations of the moon, what about the sunspot cycle? Theories were advanced about global movements of air currents, such as the German meteorologist Heinrich Dove’s storm model of collisions between polar and tropical streams, translated into English as The Law of Storms, a gesture towards the holy grail of all meteorologists awaiting their Newton. But British meteorologists were for the most part thrice shy of what they branded as airy (in every sense) speculation, having been burned more than once by grand theories such as the lunar one. Although the human computers (many of them women) in astronomy might have crunched just as many numbers, at least they had something to show for their efforts. Tables in ephemerides predicted the ebb and flow of tides, the position of planets and the dates and paths of eclipses. Tables in meteorology just made the head swim. No wonder even the haughty meteorologists at Kew yearned for a kind of physiognomy of the weather, to be seized at a glance.

This is where maps came in. Nineteenth-century science was remarkable for innovations in visualisation techniques, from photography to the pie chart. To call these images mere displays of data is to overlook their role in discovering and crystallising new objects of scientific inquiry. They were as important for scientific ontology as they were for data management. The great German naturalist Alexander von Humboldt brilliantly synthesised the local with the global in artfully coloured and coded maps that showed the distribution of characteristic forms of vegetation (e.g. pines v. palms), thus creating what Humboldt called ‘landscape physiognomies’: typical combinations of climate, topography, flora and fauna that could be sized up in one sweeping glance by the seasoned traveller. Humboldtian maps turned indigestible tables of numbers into gestalts, as easily recognisable as a familiar face. They also displayed regularities and connections invisible at the local level. By plotting lines of constant temperature (isotherms) and air pressure (isobars), Humboldt discovered the Gulf Stream and revealed patterns previously unsuspected by less peripatetic meteorologists. When Dove was awarded the Royal Society’s Copley Medal, it was for his maps of global temperature distribution rather than his theory of storms.

Not all meteorological experiments in reducing mountains of data to a coup d’oeil were as perspicuous as Humboldtian maps. An 1864 attempt to assess the accuracy of Fitzroy’s forecasts used concentric circle diagrams: the outer circle showed the wind direction forecast two days ahead, an inner circle the direction one day ahead, and a pointer in the innermost circle the actual observed direction. It takes a fair amount of squinting to make them out, and their import is unclear. The problem in this instance lay not only in the diagrams; what was meant by accuracy was foggy too. Wind direction varies continuously around the compass rose: how precise did the forecasts need to be – south-east, or 123 degrees? Moreover, where was the relevant measurement to be taken? A Royal Society evaluation of the Meteorological Department’s forecasts conducted shortly after Fitzroy’s death came up with only a 22 per cent success rate based on comparison with observations at specific points along the coast; a more modern evaluation of the same data, which grouped coastal points geographically, boosted the grade to 76 per cent. These were conceptual issues that no attempt to make sense of data – whether presented in tables or maps or statistical curves – could dodge. There was no such thing as accuracy tout court, only accuracy good enough for this or that purpose.

Statistics and maps were both tools to integrate and winnow the meteorological data that poured in every day, but they were tools that pruned and planed to different ends. In the ingeniously constructed map, connections leapt to the eye; inferences were swift and almost intuitive. Map-making and map-reading honed acuity and judgment. Statistical and probabilistic methods, in contrast, worked by rules and calculation. In the mid-19th century, statisticians astounded the reading public with accounts of how phenomena that seemed entirely haphazard when viewed one by one – robberies, dead letters, deaths – were as regular as clockwork when studied en masse. Every year the post office reported about the same number of dead letters, the police about the same number of robberies, parishes about the same number of births and deaths. Meteorologists hoped that statistical methods could perform the same sort of magic on their unruly data. Order might yet emerge from the chaos of numbers, as it had in demography. But there was a price to be paid for statistical rigour: mechanical rules replaced lightning inferences; objectivity elbowed out insight. When the Royal Society cast a sceptical eye over Fitzroy’s forecasts after his death, they complained that his way of working was unscientific: ‘No notes or calculations are made. The operation takes about a half an hour and is conducted mentally.’ This seemed to them suspiciously like the indulgence of whim, just the sort of aberration statistics was meant to forestall. Yet Victorian statistical methods failed to yield an integration comparable to Fitzroy’s ‘windstars’, another compact visualisation of a sprawl of data.

The impulse to connect went beyond integrating the temperature and air pressure readings for Great Britain. British meteorologists sought to weave exotic weather into the empire’s worldwide tapestry. After the celebrated fickleness of weather in the British Isles, India was a salutary surprise: its weather was severe, but regularly so. Wet and dry seasons succeeded one another at predictable intervals; even monsoons arrived on cue. However disorderly Indian society might have been in the eyes of British observers, its climate, though strenuous for Europeans, seemed quite law-abiding. Humboldt, who pined for the tropics after his return to grey, chilly Berlin, remarked that the luxuriant foliage and teeming wildlife around the equator were counterbalanced by meteorological equanimity. But even in the tropics the natural laws of the atmosphere eluded meteorologists.

Anderson’s portrait of Victorian meteorology is delicately and deftly drawn, enlivened by well-chosen examples and illustrations. As one might expect of a study of the British and their weather, it penetrates into the oddest crannies of the culture, from insurance companies to Hardy’s novels. Anderson’s focus, like that of the local observers she describes, is on the telling detail, the texture of a particular place and time, the density of interconnections that only thorough acquaintance with an epoch can expose. This miniaturism can occasionally seem like tunnel vision. It is somewhat disconcerting, in a book set in an age of international scientific congresses and collaborations, to find so few non-English works listed in the bibliography. Nineteenth-century meteorology might have been organised in national institutions, but it had international ambitions, as well as international rivalries, which goaded European and American scientists to pay close attention to each other’s activities. Yet, even if Victorian meteorology was not as insular as Anderson’s account might suggest, it did have sufficient variety and distinctiveness to make a world unto itself.