About the data
Where numbers come from and how far to trust them.
Real numbers from a weather model, checked against thermometers.
Words we use
- Usual
- The average of the same month over 1991 to 2020. (The technical word is “baseline”.)
- Warmer or cooler than usual
- This month’s number minus the usual. (The technical word is “anomaly”; we do not use it on the pages.)
- Typical daytime high / night-time low
- The average of the daily highest or lowest temperature. Not the hottest day ever.
- Model estimate
- A computer-made picture of the weather for a square about 28 km wide. Not a thermometer.
- Early figure
- The latest three months. They can change slightly when final data arrives.
How far to trust it
Thermometer record a real station reading.
Model estimate computed for a grid square about 28 km wide. Best for warmer or cooler than usual; rough for rain.
Usual the 1991 to 2020 average. Changes from usual are robust; absolute values can shift by up to about 1 °C against thermometers.
How many years do we hold?
| Item | Years |
|---|---|
| Monthly average temperature and rain, every place | January 1950 to September 2026 (from July 2026 early figures) |
| Rankings and comparisons for India | From 1954 for years, from 1957 for single months (earlier model values are too cold) |
| Monthly typical daytime high and night-time low | Where the data package carries them, to September 2026, model estimate |
| Daily values (hottest day, very hot days, days at 45 °C, warm nights) | 1981 to 2025, only for cities that passed our check, labelled model estimates |
Sources
| Source | Used for | Licence |
|---|---|---|
| ERA5 monthly averaged data on single levels, 1940 to present | Monthly average temperature and monthly rainfall for the world, India, states and cities, and the colours of the maps. A past-weather record rebuilt by a computer model from millions of real measurements. | CC BY 4.0 |
| ERA5 hourly data on single levels, 1940 to present | Typical daily highs and lows, worked out from hourly model data and, where a thermometer exists nearby, adjusted to it. | CC BY 4.0 |
| NASA GISS Surface Temperature Analysis (GISTEMP) version 4 | A cross-check of our world temperature against NASA’s own world record. The yearly figures agree closely from 1950 to 2025. | NASA open data; citation and acknowledgment requested |
| India Meteorological Department (IMD) published figures | India’s monsoon rain as a percentage of normal, and a check on the India temperature series. Only numbers that IMD prints in its own reports are used, with the document named. The Records page also copies station records from IMD documents, each named there with its station, date and link. | Not a data licence: figures as printed by IMD, no raw data files copied |
| NOAA Global Historical Climatology Network daily (GHCN-Daily) | Weather-station daily highs and lows, used to check the model and, for cities with a station within 25 km, shown as a separate thermometer block. | NOAA open data (the data page gives no separate licence sentence) |
| NASA POWER daily meteorological data | Daily highs and lows for the estimates of the hottest day, very hot days (40 °C), days at 45 °C and warm nights shown for cities that passed our check, adjusted to a nearby thermometer. Estimates, not records. | NASA open data service |
| Natural Earth state outlines | Used only to work out which grid squares belong to a state, so state averages can be calculated. No outline or border is drawn on this site. | Public domain |
| Weather data from MET Norway | The optional “Right now” line: the weather model’s value for the current hour at the chosen city, not a thermometer reading. | CC BY 4.0 |
| WMO World Weather and Climate Extremes Archive (table of records) | Three Indian entries on the Records page (48-hour and 12-month rainfall at Cherrapunji, average annual rain at Mawsynram), shown as WMO prints them. | Not a data licence: figures copied as printed, with the table named |
Links, credits and DOIs for every source are in Technical details below.
Download the data
Every page has its own Download CSV button. Pick a place for the full month-by-month file or a PDF place report.
PDF place report
A printable report for the place above: typical high and low, rain, change from usual, a 12-month chart and the method. Model estimates, not a forecast.
What we can tell you instead of daily extremes: for most places we show monthly averages and label them as estimates. Where a city passed our check you also see the hottest day, very hot days and warm nights of a typical year, each as a labelled model estimate.
Technical details
Can past climate data tell me next week’s weather?
No. These pages show what has already happened, as monthly and yearly averages. For today’s weather and warnings, use the India Meteorological Department (IMD) forecasts and warnings.
Why does the figure for my town differ from a weather station?
A town is shown as the model grid square nearest its centre, about 28 km wide. The model smooths out hills, the coast and the heat of a built-up city, so it can differ from a thermometer by about 0.5 to 1 degree, and more in hill, coastal and high places.
What does “usual” mean?
The average of 1991 to 2020 for the same calendar month at the same place. “2 degrees above usual” means 2 degrees warmer than that average for that month.
Why is rainfall called low confidence?
Monthly rain here comes from a computer model, and a model smooths out downpours. It follows wet and dry years reasonably, but it is a rough estimate. We do not show heavy-rain-day counts, the wettest day or dry spells, because they have not been checked against rain gauges.
Is a “hottest day” on your pages an all-time record?
No. The hottest day on our pages is the value for a typical year (the average of 1991 to 2020), not an all-time record. It is a model estimate, adjusted to a nearby thermometer, and shown only for cities that passed our check.
Why does India start in 1954?
The model reads about 0.9 degrees too cold over India in 1950 to 1953 when checked against weather stations. So India figures start in 1954. The earlier years can be switched on in the explorer, with a warning.
Can I use these figures in a report or school project?
Yes, but please say they are model estimates averaged by HowToSurvive.in and name the original sources listed above. For the underlying data, follow each source’s own terms.
How were the estimates tested?
We tested each estimate by hiding five years of the thermometer record at a time, estimating them from the remaining years, and comparing; the error figures on the page are the average miss over all hidden years.
Four things to know
- Temperature. Most figures are the average of day and night, measured 2 metres above the ground. Daily highs and lows are typical values, not records.
- Rainfall. Monthly totals in mm, from a model. They follow wet and dry years but miss heavy downpours, so they carry low confidence.
- Change from usual. We compare each month with the 1991 to 2020 average for the same calendar month at the same place.
- Estimates and records. Model estimates, thermometer records and the 1991 to 2020 usual are kept apart and labelled. Official records are kept on the Records page, apart from our estimates.
Sources, links and credits
- ERA5 monthly averaged data on single levels, 1940 to present (Copernicus Climate Change Service (C3S), ECMWF). Monthly average temperature and monthly rainfall for the world, India, states and cities, and the colours of the maps. A past-weather record rebuilt by a computer model from millions of real measurements. Licence: CC BY 4.0. We average it over areas, re-base it to 1991 to 2020 and round it. Generated using Copernicus Climate Change Service information (2026). Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains. Source page (opens in a new tab) · DOI 10.24381/cds.f17050d7 (opens in a new tab)
- ERA5 hourly data on single levels, 1940 to present (Copernicus Climate Change Service (C3S), ECMWF). Typical daily highs and lows, worked out from hourly model data and, where a thermometer exists nearby, adjusted to it. Licence: CC BY 4.0. Model estimates, not thermometer readings. Generated using Copernicus Climate Change Service information (2026). Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains. Source page (opens in a new tab) · DOI 10.24381/1cf1ad76 (opens in a new tab)
- NASA GISS Surface Temperature Analysis (GISTEMP) version 4 (NASA Goddard Institute for Space Studies). A cross-check of our world temperature against NASA’s own world record. The yearly figures agree closely from 1950 to 2025. Licence: NASA open data; citation and acknowledgment requested. GISTEMP is published against 1951 to 1980; we re-base it to 1991 to 2020 for the check. NASA's Goddard Institute for Space Studies (GISS), GISTEMP v4. Source page (opens in a new tab)
- India Meteorological Department (IMD) published figures (India Meteorological Department). India’s monsoon rain as a percentage of normal, and a check on the India temperature series. Only numbers that IMD prints in its own reports are used, with the document named. The Records page also copies station records from IMD documents, each named there with its station, date and link. Licence: Not a data licence: figures as printed by IMD, no raw data files copied. The documents used are listed under the monsoon figures in the explorer’s “How we know” section. Source: India Meteorological Department. IMD has not endorsed this page. Source page (opens in a new tab)
- NOAA Global Historical Climatology Network daily (GHCN-Daily) (NOAA National Centers for Environmental Information). Weather-station daily highs and lows, used to check the model and, for cities with a station within 25 km, shown as a separate thermometer block. Licence: NOAA open data (the data page gives no separate licence sentence). Stations have gaps and are not homogenised; years with too few days are left out. Weather-station records: NOAA National Centers for Environmental Information, Global Historical Climatology Network Daily (GHCN-Daily). Source page (opens in a new tab)
- NASA POWER daily meteorological data (NASA Langley Research Center, POWER project). Daily highs and lows for the estimates of the hottest day, very hot days (40 °C), days at 45 °C and warm nights shown for cities that passed our check, adjusted to a nearby thermometer. Estimates, not records. Licence: NASA open data service. Squares about 50 km wide. Daily rain from this source is no longer shown. Daily satellite and model data from NASA POWER. Source page (opens in a new tab)
- Natural Earth state outlines (Natural Earth). Used only to work out which grid squares belong to a state, so state averages can be calculated. No outline or border is drawn on this site. Licence: Public domain. Grid squares are not a political map. Source page (opens in a new tab)
- Weather data from MET Norway (Norwegian Meteorological Institute). The optional “Right now” line: the weather model’s value for the current hour at the chosen city, not a thermometer reading. Licence: CC BY 4.0. Weather data from MET Norway (api.met.no). Source page (opens in a new tab)
- WMO World Weather and Climate Extremes Archive (table of records) (World Meteorological Organization). Three Indian entries on the Records page (48-hour and 12-month rainfall at Cherrapunji, average annual rain at Mawsynram), shown as WMO prints them. Licence: Not a data licence: figures copied as printed, with the table named. Source: World Meteorological Organization, Records of Weather and Climate Extremes, table as of 31 July 2025. WMO has not endorsed this page. Source page (opens in a new tab)
Our own figures: the averages, changes from usual and CSV files that HowToSurvive.in calculates from these sources are shared under CC BY 4.0 (opens in a new tab), with credit to HowToSurvive.in. The source data stays under the terms of its providers.
How to read the figures
- Daytime high is not the all-time hottest day.
- Typical daily highs and extreme temperature records are different measures. A typical daytime high is never “the highest ever”.
- “Warmer than usual” means warmer than a reference period.
- We use 1991 to 2020 for every comparison, month by month, at the same place.
- Rainfall can vary within a city.
- The gridded estimate is for a square about 28 km wide and may miss the heaviest local downpours.
Detailed methods and limits
Estimate, thermometer or usual?
Every number on the Climate pages belongs to one of three kinds, and we try to say which.
| Kind | What it is | Can it be called a record? |
|---|---|---|
| Model estimate | A value from a reanalysis (a past-weather record rebuilt by a computer model) for a grid square. Most of what you see. | No. It smooths hills, coast and city heat. |
| Thermometer record | A reading from a weather station, shown only where a station is within 25 km and has enough years. | Only as “the highest in this station’s data since the year shown”, never “of all time”. |
| Usual (1991 to 2020) | The reference average that other figures are compared with. | No. It is an average, not an extreme. |
Official records need a verified station, date, unit and source. The Records page lists only those we could check in IMD and WMO documents, and we never compare a record with a grid estimate.
How far off can the figures be?
Figures are computer-model estimates for the area around a place. They can differ from a weather station by 1 to 3 °C and rainfall by 20 to 50 percent, more in hills, on coasts and in Ladakh.
The national India series is also model-based (the ERA5 average of all land squares inside India). Its average level, about 23.9 °C for 1991 to 2020, is roughly 1.3 °C below the all-India mean temperature that the India Meteorological Department publishes, because the squares include the cold high Himalaya and the model reads lower than thermometers on average. Year-to-year changes and the long-term rise follow the published national anomalies closely, so the trend wording on this site is unchanged; use the level as a model figure, not as IMD’s own number.
The 1991 to 2020 usual
The World Meteorological Organization’s current standard period for “normal” climate is 1991 to 2020. We use it for every comparison, calendar month by calendar month. NASA’s world record is published against 1951 to 1980; for the cross-check we re-base it to 1991 to 2020 ourselves.
Early figures and the 1954 start for India
- The latest three months come from an early release of the model and can change slightly when the final version replaces them (this applies from July 2026 now). They are marked “early” in the explorer.
- For India, the model reads about 0.9 degrees too cold in 1950 to 1953 when checked against 73 weather stations, so India figures start in 1954. Single months from 1954 to 1956 can still be off by 1.5 degrees or more. The earlier years can be switched on in the explorer.
- A few single months are flagged by hand where we know the model is off, for example Delhi in May 1956.
How fine is the picture?
- A city is the model grid square nearest its centre, about 28 km wide. A state is the average of the squares inside it, so small states and islands rest on few squares.
- The India temperature map uses squares about 55 km wide and the India rain map squares about 110 km wide. Outside India both maps use squares about 222 km wide, so they show broad patterns, not single towns.
- The map shows grid squares only. Boundaries are not drawn, and the squares are not a political map.
- Our checks against weather stations cover India only. Treat places elsewhere as model estimates.
- The data covers 181 places: the world, India, 36 states and union territories and 143 cities.
Rainfall and its limits
The model’s rain is a smooth picture. At the three cities we checked, it counted 2 to 9 times fewer heavy-rain days than rain gauges. For that reason this site does not show heavy-rain-day counts, the wettest day or dry spells. They will return only after daily rain has been checked against gauges. Monthly totals and the monsoon figures published by IMD remain, with their labels.
What is not shown
We would rather say what we cannot tell you than guess. Not shown at present: heatwave days, heat index, days at 35 °C, official records for most states and places, heavy-rain-day counts, dry spells, and district-level figures (the data is by state and city). Very hot days, days at 45 °C, the hottest day of a typical year and warm nights appear only for the cities that passed our check, as labelled model estimates with an error band. We tested each estimate by hiding five years of the thermometer record at a time, estimating them from the remaining years, and comparing; the error figures on the page are the average miss over all hidden years. Climate figures here are history. They are never a weather alert; for warnings use IMD and the National Disaster Management Authority.
Calculation examples
Change from usual, real figures. India averaged 23.7 °C in 2025. The usual (1991 to 2020 average) is 23.6 °C. So the change is 23.7 − 23.6 = +0.1 °C (model estimate, rounded).
Typical daytime high, simple numbers (not a real place). If the daily highs of a 30-day month add up to 1,050 degrees, the typical daytime high of that month is 1,050 ÷ 30 = 35.0 °C. The hottest single day of that month may be several degrees more, which is why a typical high is not a record.
Change log
- 4 November 2026. Records and rankings page added. It lists only records that we opened and checked in IMD and WMO documents, and keeps our own model rankings apart. Everything we could not verify says “Not available”.
- 4 November 2026. Data sources and methods and Climate safety pages added. Heavy-rain-day, very-heavy-rain-day, wettest-day and dry-spell figures removed from the Extreme days section, because daily rain from the model has not been checked against rain gauges. Hot-day estimates stay, labelled as estimates. The line “Next update about …” was replaced by “Refreshed about every three months”.
- 8 November 2026. Data package built, with data to September 2026.
Reports and downloads
Every page has its own Download CSV button. The Download tab above gives the full month-by-month file for any place, and a printable PDF place report (the old Reports page).
Past climate, not today’s weather forecast. Figures are model estimates, not weather-station records unless a page says so. See Data sources and methods. Pictures on this page are AI-generated editorial illustrations, not photographs.