How Did Japan Learn to See a Storm Coming?

On 16 February 1883, the Tokyo Meteorological Observatory began making trial weather charts from telegrams. Reports of morning conditions were being collected from observing stations around the country. The staff could bring weather from separate places onto the same sheet before those conditions became visible from Tokyo itself.^[1]

The important change was not that someone had finally thought to look at the sky. People had long watched clouds and winds because a wrong decision about tomorrow could endanger a voyage or spoil work. Some had kept detailed records; others had used instruments decades before the observatory opened.

What the telegram offered was a different reach. An observer's horizon no longer had to be the limit of the information available for a forecast. Yet extending that reach required more than a wire. Measurements had to be comparable, reports had to arrive in time, and a prediction had to travel back out to people who could act on it.

The history from nineteenth-century weather watching to the interruption and restoration of public forecasts during the Second World War shows how those connections were made—and how they could be broken without anyone ceasing to observe the weather.

At Hirado, in northwestern Kyushu, a family connected with the domain's vessels preserved instructions for reading the weather. In 1819, the lord of Hirado had material from the Yamazaki family's books copied into a compilation called Tenki-mi densho, a transmission of weather-watching knowledge. Maritime researcher Yamada Yoshihiko describes the compilation and reproduces part of its preface in modern Japanese.^[2]

The practical setting matters. Wind could help a vessel make its journey or make that journey dangerous. A useful judgment needed to come while changing the departure or preparations was still possible. The sky was therefore being read for consequences, rather than merely described as beautiful or unusual.

The instructions selected moments and features for attention. They directed an observer toward dawn and the rising sun, then toward sunset when judging the next day. One passage distinguished an evening color that changed from redness that gradually faded: it associated the first with wind or rain and the second with favorable conditions. The rule asked a watcher to notice a sequence, not simply remember that the sky had been red.

The compilation's preface also recognized a problem in learning the practice. Possessing a written text did not make the oral instruction and experience associated with it easy to acquire. A rule had to be recognized in the changing sky. The words could preserve a distinction without guaranteeing that another observer would apply it in the same way.

That is a useful starting point for understanding kanten bōki, the practice of anticipating weather from natural signs. It was a collection of ways to connect an observed condition with an expected consequence, not one unified theory that everyone in Japan accepted. The reliability of a particular rule remained a separate matter from the seriousness of the need it addressed.

Local usefulness also depended on the decision being made. A boat about to leave sheltered water needed a judgment about the conditions it would encounter. The same observation could be less consequential to someone whose work remained under cover. There was no single, context-free answer to what counted as favorable weather.

Hirado's records show people treating prediction as skilled work with something at stake. They do not require us to endorse every inherited sign. Their importance is that an expectation could be stated before the outcome—and could therefore be reconsidered when the outcome arrived.

The Hirado undertaking went beyond copying advice. According to Yamada's account, a Yamazaki retainer recorded conditions from 1819 through 1828. A two-part work compiled in 1832 compared the observations with the inherited instructions. Entries connected the appearance of the morning sky with weather later in the day.^[3]

This sequence gave memory a written counterpart. A striking success can be remembered while uneventful failures disappear. A continuing record potentially preserves both, making it possible to return to occasions that no longer stand out in recollection. The record's value lies in that opportunity for comparison, not in an assumption that its surviving compiler performed a modern statistical test.

The comparison depended on keeping the earlier appearance distinct from what followed it. A red sky and a later storm had to remain identifiable as observations, rather than merge into a recollection that the storm had looked inevitable. The compilation brought inherited instructions into contact with a dated record. Its purpose was to test what those instructions could predict, although the available account does not establish a measured success rate.

Elsewhere, weather knowledge occupied a broad field of inquiry. In 1820, Inagaki Denryū, a scholar from the area now within Saitama, copied a work called Tenmon fūusetsu. Its subjects included lunar mansions and celestial appearances alongside clouds, wind, rain, snow, dew, and fog. A record among his papers described a solar halo observed in 1837.^[4]

Those materials help explain why a modern division between astronomy and meteorology should not determine in advance what early observers were allowed to notice. The visible heavens contained phenomena with several possible meanings. Recording a halo preserved an event; deciding whether it predicted rain or signified something else required an additional interpretation.

Inagaki's teacher, Asano Hokusui, criticized attempts to read future fortune from celestial motions. That criticism belonged within the same learned world, not solely to a later generation looking back upon it. Observation, inherited explanation, and disagreement could coexist.

The two settings reveal different forms of checking. Hirado's comparison asked whether a rule matched subsequent weather. Inagaki's collected and observed phenomena supplied material that could be described and classified. Neither made the future transparent. Both left a record through which somebody other than the original observer could examine what had been noticed and what had been made of it.

Instruments added another way to describe conditions. Historians of climate Masumi Zaiki and Takehiko Mikami identify temperature and pressure observations made three times daily at Nagasakiya, the Dutch delegation's lodging in Edo's Nihonbashi district, between 1825 and 1828. The shogunate's astronomical office kept another series between 1839 and 1855.^[5]

These measurements predated the Tokyo observatory. Their existence prevents its founding from becoming the moment Japan supposedly exchanged vague impressions for numbers. Repeated instrumental observation was already possible at particular sites, among people with access to equipment and the knowledge needed to use it.

A number supplied a different kind of comparison from a remembered impression of heat or an evocative description of a cloud. Readings could be placed in a series and compared with earlier readings. But a thermometer in one location still measured conditions there. Writing its reading in a book did not tell someone elsewhere what was happening soon enough to change a decision.

The historical research also identifies a difficulty hidden by the apparent exactness of old figures. Instruments used different units, stood at different elevations, and did not always incorporate the same corrections. Zaiki and Mikami explain why such observations need adjustment before researchers combine them into a long climatic record. Precision in writing a number is not the same as comparability between measurements.

That distinction mattered to the developing forecasting problem too. A network would require observations from separate places to be interpretable together. Otherwise, a difference on paper might reflect the observing arrangement rather than a meaningful difference in the atmosphere. Equipment, reporting practice, and interpretation had to work in relation to one another.

The observatory established in Tokyo in 1875 gave continuing official observation an institutional home. Its meteorological readings began on 5 June, a few days after the start of its other observational services. The Japan Meteorological Agency dates its institutional ancestry through this establishment while acknowledging earlier observation elsewhere.^[6]

The first date thus tells us when a particular continuing organization began, not when useful knowledge of weather first existed. The institution could preserve staff responsibilities, equipment, and records beyond one observer's private undertaking. Turning those resources into forewarning over a larger area required the next step: moving recent observations faster than an ordinary written report could travel.

Erwin Knipping, a German specialist employed by the Home Ministry in 1882 to organize storm warnings, worked on the collection of reports. The resulting telegram service enabled the Tokyo observatory to begin its trial charts in February 1883. Daily printed charts followed in March, and the observatory issued its first storm warning on 26 May.^[1]

The telegram altered the time available for interpretation. A weather report delivered long after the event could still improve a historical record. For a warning, its usefulness depended on arriving while the reported conditions could help explain what might happen next. Speed changed the task that the same observation could perform.

The reports also changed the observer's effective position. Someone standing in Tokyo could not see the weather across the archipelago. Staff receiving observations from elsewhere could nevertheless compare those conditions with the local situation. The chart assembled a field of evidence that no single person had witnessed from one viewpoint.

Producing that field required skilled work after transmission. Staff entered the reported observations on a map and drew lines joining places with equal pressure.^[7] Separate numerical readings became an intelligible spatial pattern through their work at the desk. The telegram supplied observations; the staff still had to decide how to connect and interpret them.

Tokyo was not the only place where an official warning had been attempted. In 1880, Hokkaido's development administration had begun displaying storm signals for coastal fishers in Sapporo and Otaru. That service was interrupted when the administration was abolished in 1882, before incorporation into the later Tokyo-based signal arrangements.^[8]

The interruption is as instructive as the beginning. A useful warning practice could lose its administrative support. The need of coastal users did not disappear when the office changed. Maintaining the service required somebody to inherit its work, rather than merely preserve the idea that warnings were desirable.

In May 1884, the weather telegrams became possible three times a day. Regular forecasts followed on 1 June. More frequent incoming reports supplied opportunities to revise the picture instead of waiting until the following morning. The relationship between transmission and forecasting was therefore not simply a wire being installed once. The rhythm of information mattered.

These arrangements did not eliminate the value of looking outside. They made looking elsewhere possible through other observers. That collective reach was the decisive addition to a practice formerly constrained more tightly by local signs and the time required for news to travel.

Gathering information inward was only half the task. A storm warning kept inside the observatory could improve an official file while leaving a departing boat no better informed.

In 1883, standard storm-signal forms were established, followed by the installation of signal poles, including one at Minami-Shinagawa. They gave a warning a public form beyond the written telegram.^[8] To use such a signal, people needed both to see it and to understand the convention it represented. Raising an object was a means of communicating an official judgment, not evidence that every potential user had received it.

Regular forecasts also acquired everyday urban addresses. The National Diet Library records their display at Tokyo police boxes from June 1884. From June 1888, daily weather reports appeared in the official gazette, the Kanpō. Newspapers obtained observatory announcements directly or through news agencies and placed predictions alongside practical information such as train and ship schedules.^[9]

That company on the page is revealing. The forecast was becoming something to consult when arranging activity. It entered a stream of information readers already used to decide where to go and when. Making it useful did not necessarily require teaching each reader how to draw a pressure chart.

The different media fitted different encounters. A person could consult a posted notice, read a forecast alongside the day's transport information, or recognize a signal without holding a printed sheet. None required the person to visit the observatory. Their usefulness depended on incorporating the announcement into places and routines beyond the forecasters' direct supervision.

There was also a problem of scale. The first regular national forecast expressed changing conditions and a predominance of rain in a single broad statement.^[8] Its geographical reach should not be confused with a precise account of each locality. A reader still had to relate the information to the place and activity that mattered to that reader.

Wireless communication opened another route in 1925, when the Central Meteorological Observatory began wireless reports and radio broadcasting developed in Tokyo. Weather information could reach receivers without waiting for a fresh printed sheet.^[10] It still depended on equipment, reception, and the timing of the broadcast. Radio changed the last stage of delivery; it did not make an audience automatic.

A forecast therefore became public through repeated acts of translation: observations into an analysis, an analysis into an announcement, and an announcement into a form accessible outside the office. At each step, something had to be selected and understood. The network's success could not be measured only at the point where a telegram arrived.

The expansion of observation did not guarantee a corresponding expansion of public access. In wartime, the relationship could move in the opposite direction.

Japan's meteorological organization was increasingly integrated with military requirements during the 1930s. By 1939, the weather offices had been brought under national administration. With the Pacific War, observation and forecasting continued while their products became subject to secrecy.^[10]

The distinction acquired explicit rules. Newspaper publication and public display of weather maps were prohibited in November 1941. On 8 December, the day Japan entered war with the United States and Britain, weather-reporting controls were ordered into effect from six in the evening.^[11]

The sky remained visible. What could be withheld was the assembled knowledge that extended beyond any one observer's horizon. An individual could watch the local wind without gaining access to the network of reports and interpretations on which official forecasting rested. Control operated at the points where information was combined and redistributed.

Matsuyama's weather office supplies a concrete institutional example. Its history describes continued observation during a period when public forecasts were restricted, and recalls the building being painted black to reduce its visibility to air attack.^[12] The office remained a working resource while its relation to people outside changed.

The same information had several potential users. Weather relevant to fishing, travel, or farming could also matter to military operations. Broad dissemination allowed an announcement to travel beyond its intended civilian audience. Wartime policy addressed that possibility by restricting dissemination, not by discovering a different atmosphere reserved for military purposes.

This changes how the history of forecasting should be told. The progression from local observation to an organized network was real. Yet a more centralized, technically capable organization could direct its work toward a narrower authorized audience. Improved collection and restricted release were entirely compatible.

Nor did a ban on reporting remove the public's weather-dependent decisions. People still had to work, travel, and protect what they could. The policy altered the information available for those decisions. It therefore interrupted a practical relationship established through the earlier signal poles, notices, newspapers, and broadcasts.

The wartime case is not an incidental detour in an otherwise automatic history of progress. It tests the central claim. Seeing a storm coming depended partly on other people's observations; benefiting from that collective sight depended on permission to receive what the system had learned.

Weather-reporting controls were lifted on 21 August 1945. Radio forecasts resumed the following day, after an interruption of three years and eight months. In February 1946, the Central Meteorological Observatory opened a weather consultation office to answer visitors and telephone inquiries.^[13]

The consultation office added something to broadcast transmission. A prepared announcement addressed many listeners at once. A question from a visitor or caller could reveal what that person needed to understand. The exchange created a place for the public to approach the institution, rather than simply wait for its next message.

The returning broadcasts did not require Japan to discover forecasting anew. They reopened a channel from accumulated expertise to everyday decisions. That distinction shows why the service cannot be identified with its instruments alone. Thermometers, trained staff, and observing stations were necessary resources, but access to their combined work had a history of its own.

The consultation counter also exposed the distance between a forecast and a decision. A general announcement could reach many people while leaving an individual unsure what it meant for a particular undertaking. Providing someone to answer questions acknowledged that reception was not the end of the work. The institution had to remain approachable after its announcement went out.

None of these steps guaranteed certainty. A surviving record can expose an earlier mistake; a broad forecast can be difficult to apply locally; a warning can arrive too late for the decision it was meant to assist. Those possibilities do not make prediction pointless. They identify where the work of making it useful has to occur.

Japan learned to see storms coming by extending observation beyond an individual's memory and horizon, then building ways to turn those observations into timely judgments. The extension was technical and institutional at once. Its benefits depended on keeping the route open all the way to the people whose next action could still change.

A storm does not wait for a report to be completed. The historical achievement was to make information arrive while there was still a future in which to use it.