Observation Log — Greater Tokyo
The world’s largest urban heat island laboratory, measured one thermometer at a time
Tokyo is not one climate. It's five, minimum, and they don't blend politely at the borders. Stand at the Shinjuku south exit at 22:30 in mid-August and the air sits on your skin at 29°C, thick and recycled through ten thousand condenser fans. Ride the Chuo Line forty minutes west to Hachioji the same night and you'll step onto the platform into 25°C air that actually moves. Same prefecture. Same evening. A different microclimate. This site exists to document that difference — with station data, survey vehicles, rain gauges, and a stubborn refusal to talk about “Tokyo weather” as if it were a single thing.
The five zones we track aren't administrative inventions. They're thermal regimes with distinct mechanics. The Center Core — the 23 special wards — runs a +2.8°C annual mean bias against the Tama Hills baseline, the strongest sustained heat island signal in Japan. The Bay Coast trades raw temperature for humidity: Yokohama and the Odaiba strip stay 1–2°C cooler in July afternoons, then pay it back in July nights when the sea breeze quits and the wet bulb climbs. The Tama Hills collect orographic rain and radiate heat away under clearer night skies. The Saitama Plains swing continental: 41.1°C at Kumagaya in 2018, −8.5°C in 1984, a 27°C annual range that makes the 23 wards look maritime. And the Boso Peninsula answers to the Pacific directly — wind, fog, delayed spring, typhoon landfalls — a climate that shares almost nothing with the capital except a commute corridor.
Why zone-level weather matters more than city-level weather
A forecast that says “Tokyo: 32°C, 70% humidity” is technically true and practically useless. It describes nobody's afternoon. In August the Center Core can carry twelve to fifteen days above 35°C while the Boso coast counts two to four. June rainfall splits nearly 40mm apart between the Center Core's 168mm and Saitama's dry-slot months. Even cherry blossom timing — the single most tracked phenological event in the country — staggers three to four days between the wards and the Tama foothills, and seven to ten days out to the Pacific side of Boso. If you're choosing where to live, where to site a data center, when to pour concrete, or when to visit, the metro-scale average is the wrong instrument.
We've spent years taking the right instrument's readings. Between 2019 and 2022 we ran a mobile survey program — a vehicle with aspirated thermometers and GPS logging — across fixed transects in all 23 wards, out through the Tama cities, up into Saitama's rice country, and around the Boso coast, on clear nights and typhoon shoulders alike. The transects repeat the same routes at the same hours, so the bias figures on this site aren't model abstractions. They're what the air actually did at 1.5 meters above the asphalt, logged at ten-second intervals.
The heat island is the headline, but the lag is the story
Everyone knows central Tokyo is hot. What's less appreciated is how the five zones drift out of phase with each other across the calendar. The bay and the Pacific don't just moderate temperature — they delay it. March mean temperatures on the Boso coast run about 2°C behind the Center Core, so plum and cherry blossom arrive late even as the city bursts pink. Then autumn reverses the debt: the ocean holds its summer warmth, and November on the Boso shore averages 1.5°C warmer than frost-prone Saitama. The Kanto plain doesn't experience seasons so much as it experiences a slow thermal wave, arriving in Shinjuku first and reaching Katsuura last.
The baiu front respects these boundaries too. When the June rain front stalls over the Kanto plain, the Omine Mountains and Chichibu Hills wring moisture out of the Pacific flow before it reaches the northern plain — the “Saitama dry slot” we document on the Saitama page and in the seasonal matrix. The same front, pushed against the Tama Mountains, produces the orographic cloudbursts that gave Hachioji and Ome 450mm in a single day during Typhoon Hagibis in 2019, while the Center Core recorded 280mm. Rain doesn't fall on Tokyo. It falls on zones.
What we actually measure
Every number on this site traces to one of four instrument families. First, the Japan Meteorological Agency's surface network: the four Tokyo-area observatories (Tokyo, Yokohama, Chiba, Saitama) with hourly records back to 1949, and the 1991–2020 normals as our baseline. Second, the AMeDAS automated stations — eight within Tokyo prefecture alone, sampling at ten-minute intervals — which resolve the micro-scale variation the flagship observatories smooth over. Third, the Tokyo Metropolitan Government's heat island monitoring program, including its 2023 report and 120 surface-temperature survey points, plus Landsat thermal imagery for the surface skin temperature that air thermometers can't see. Fourth, our own mobile transects and the forty cooperative rain gauges in Saitama that let us map the dry slot at ward-town resolution. The full accounting, including where each source falls short, is on the data sources page.
The live conditions boxes above come from the Open-Meteo forecast API, pulled fresh on every page load for our five measurement points: Tokyo (35.6895, 139.6917), Yokohama (35.4437, 139.6380), Hachioji (35.6559, 139.3234), Saitama (35.9089, 139.6234), and Chiba (35.6074, 140.1065). If the feed drops, the boxes say “Updating…” — we don't paste in static numbers and pretend they're current. That's a small thing, but it's the whole ethic of the project in one line: measured or nothing.
Who this is for
Two audiences, mostly. The first is residents — people deciding where in Greater Tokyo to rent or buy, who suspect (correctly) that “Tokyo” spans at least four degrees of August night temperature and thirty frost days of January difference. The thermal comfort calculus is real: a Tama Hills bedroom at 20°C on an August night versus a Center Core bedroom at 24–27°C is a sleep-quality difference the Tokyo Metropolitan Institute of Public Health measured at 23% in its 2022 survey. Bay Coast residents trade mold risk and salt corrosion for afternoon sea breezes. Saitama homeowners get cheaper land and cracked asphalt. Boso retirees get wind, fog, and quiet. Each zone is a different daily negotiation with the atmosphere, and we lay out the terms zone by zone.
The second audience is travelers and planners timing visits by humidity and rain rather than by brochure seasons. Cherry blossoms: late March in the Center Core, early April in Tama and Boso. Summer survival: the yamase days on the Bay Coast, when a cool Pacific easterly pins Yokohama at 26–28°C while the wards bake at 32–34°C. Autumn color: mid-November in the city parks, late November in Okutama. Winter clarity: January, when Mt Fuji is visible from the Tama Hills fifteen to twenty days a month against five to eight from the Center Core's blocked sightlines. The seasonal heat-rain matrix compresses all twelve months and five zones into a single grid — it's the fastest way to see why “best time to visit Tokyo” has five different correct answers.
The map above, briefly
The schematic at the top of this page is drawn to observation logic, not cartographic scale. Tokyo Bay sits at the bottom, as it does in every Tokyo resident's mental geography. The terracotta blob is the Center Core heat island, running +2.8°C against the western baseline; the two dashed isotherm arcs show the typical heat plume axis on a summer afternoon, southwest to northeast with the prevailing flow. The bay-blue strip is the sea-moderated coast. Green is the Tama Hills, wheat is the Saitama Plains, lavender is the Boso Peninsula across the water. The white circles carry annual mean temperature at 1.5 meters: 16.8°C in the core, 15.9°C on the bay, 15.2°C in the hills, 15.5°C on the plains, 15.6°C on the peninsula. The Arakawa and Tama rivers matter because they're the region's only reliable nocturnal cool-air corridors; Mt Fuji matters because its visibility is our most honest air-clarity index; the Rainbow Bridge arc matters because it crosses the sharpest thermal boundary in the metropolis, from reclaimed-land heat to open bay air in under two kilometers.
And the boundary positions themselves move year to year — new towers redirect the sea breeze, new pavement extends the core, and each typhoon season rewrites the rainfall ledger. So the map is a snapshot, not a monument. Check the live numbers. Then check the zone essays. The atmosphere over the Kanto plain is running a five-way experiment on forty million people, and it's readable if you watch the right thermometers. We do.
How the five zones were drawn
The zone boundaries on the map aren't arbitrary polygons colored for decoration. Each one follows a measurable discontinuity. The Center Core boundary tracks the dense ward fabric itself — where nighttime cooling rates fall below 0.4°C per hour after sunset, you're inside the heat island, and that line hugs the 23-ward perimeter with eerie fidelity. The Bay Coast boundary is drawn at the inland penetration limit of the summer sea breeze front, which our transects logged at eight to twelve kilometers depending on synoptic wind. The Tama Hills boundary follows the 40-meter elevation contour where nocturnal radiative cooling starts to outpace anthropogenic heat release. The Saitama Plains boundary is the frost line: where January clear-night minimums begin falling 2°C or more below the ward stations, regularly and predictably. And the Boso Peninsula defines itself — it's the land that answers to the open Pacific rather than to the enclosed bay.
Some places sit on the seams, and we say so. Minato Ward's Roppongi ridge is Center Core by density but bay-moderated by exposure, and its +2.2°C bias reflects both. Machida is Tama Hills by elevation but warmer than Hachioji because its basin traps afternoon heat. Northern Chiba cities like Ichikawa and Funabashi are administratively Chiba but thermally suburban Tokyo. These edge cases are where microclimate analysis earns its keep — the zone label is a starting hypothesis, and the station readings decide.
A short history of watching this heat
Tokyo's heat island isn't new, and neither is the habit of measuring it. JMA's Tokyo observatory has logged hourly since 1949, surviving three relocations, and the warming trend in its nighttime minimums is one of the cleanest urban climate signals anywhere on earth: roughly three times the global background rate, with almost all of the excess concentrated between sunset and dawn. The postwar rebuilding poured the concrete. The 1964 Olympics paved over the rivers. The bubble years raised the canyons that block the sea breeze. Each decade added a layer of anthropogenic heat, and the thermometer record shows every one of them. What changed recently is resolution. AMeDAS blanketed the prefecture in ten-minute sampling. XRAIN radar watches rain bands drift across the plain at 250-meter pixels. Landsat passes overhead and photographs the surface skin of the city in thermal infrared. A generation ago, “the Tokyo heat island” was a single number. Now it's a field, with structure, and the structure is what this site is about.
What we don't claim
We don't claim street-corner precision. Open-Meteo point forecasts interpolate model grids, and within any two-kilometer radius the real surface can vary by a degree or more depending on shade, pavement, and park proximity. We don't claim our +2.8°C bias figure supersedes the Tokyo Metropolitan Government's official +2.6°C — the difference is a baseline-station choice, and both are defensible; the data sources page explains exactly where the two methods diverge. We don't forecast typhoon landfalls, sell insurance advice, or certify anything. And when we get a number wrong, we say so on the corrections page, dated, with the wrong figure left visible next to the right one. Measurement is a discipline of admitting error. We'd rather keep the ledger honest than keep it tidy.