Methodology · Updated November 2026

How MyWeatherToday gets its numbers

MyWeatherToday does not measure the weather. It shows forecasts and air-quality estimates produced by national weather services and international forecasting centres, and it tells you which one each number came from. This page explains the sources, the calculations, the limits and how to correct us.

Data sources

Data sources used by MyWeatherToday
SourceWhat we use it forCoverage and refresh
Open-MeteoForecasts: temperature, rain, wind, humidity, UV, cloud, pressure, sunrise and sunsetBlends NOAA GFS and HRRR, ECMWF, DWD ICON and others; hourly, refreshed about every 15 minutes
Copernicus Atmosphere Monitoring Service (CAMS)Air quality: PM2.5, PM10, ozone, nitrogen dioxide, sulphur dioxide, carbon monoxideGlobal model on a grid of about 45 km (about 11 km over Europe); hourly, 5-day outlook
ECMWF ERA5 reanalysisHourly weather history used for the typical-day statisticsReanalysis on a grid of about 25 km; hourly, 1940 to now
RainViewerRadar frames for the weather mapComposite of national radar networks including US NEXRAD; last two hours, every few minutes
NOAA US Climate Normals 1991–2020Climate averages on the city weather pagesNOAA 1991–2020 normals from official station records
US EPA AirNow: AQI basicsAQI categories and the sub-index formulaUS EPA AQI breakpoints and health messages
WHO global air quality guidelines (2021)PM2.5 comparison valuesWHO 2021 guidelines: 5 µg/m³ annual mean, 15 µg/m³ daily mean

Open-Meteo data is published under CC BY 4.0. We are not affiliated with any of these organisations, and we are not affiliated with The Weather Channel or any government agency.

Weather forecasts

For each place we ask Open-Meteo for the highest-resolution model that covers it (for example NOAA’s HRRR over the United States, or DWD’s ICON over Europe) for the first hours and days, and a global model for the days after that. We show the numbers as returned, converted to your units. We do not adjust or bias-correct them. Current conditions and the short-range forecast refresh about every fifteen minutes while a page is open. The “chance of rain” is the model’s probability of measurable rain in that hour or day, not the share of the day it will rain.

How the AQI is calculated

The live AQI on this site is the US AQI. For each pollutant (PM2.5, PM10, ozone, nitrogen dioxide, sulphur dioxide, carbon monoxide), the concentration is converted to a sub-index from 0 to 500 using the US EPA’s breakpoints. The highest sub-index becomes the AQI, and the pollutant that produced it is the “main pollutant”. The categories are Good (0–50), Moderate (51–100), Unhealthy for Sensitive Groups (101–150), Unhealthy (151–200), Very Unhealthy (201–300) and Hazardous (301–500).

The concentrations come from the CAMS atmospheric model, which combines satellite observations with a chemistry and weather model. They are not measurements from a monitor. Over India the model’s grid cells are about 45 kilometres wide, so a reading describes a wide area, and a monitor beside a busy road will often read higher. India’s Central Pollution Control Board publishes its own National AQI from ground stations, with different category names and breakpoints, so the two numbers will not match one-for-one.

How the yearly AQI statistics are built

The “2025 by month” tables, the city ranking and the dataset downloads for the eight Indian city pages use this method:

  1. Download hourly PM2.5 and the PM2.5 US AQI sub-index for every hour of 2025 at each city’s coordinates, from the CAMS global model through Open-Meteo.
  2. Cap the hourly AQI at 500, the top of the scale.
  3. Average by day, by month and by year. A day is counted in an AQI category by its daily-average AQI. “Unhealthy days” are days with a daily average above 150.
  4. Find the hour of the day with the highest and lowest average PM2.5 concentration.

Why PM2.5 only. In spring the model can place hundreds to thousands of micrograms per cubic metre of desert dust in PM10, which pushes the combined US AQI far past 500 (we saw values above 2,000 over Noida on 29 March 2025). PM2.5 is also the pollutant that sets the AQI on most days. Using the PM2.5 sub-index keeps the numbers inside the scale and comparable between cities. The headline live AQI on the city pages is the combined US AQI, so on a dusty day it can be higher than the PM2.5-based figures in the tables.

Limits. Noida and Gurgaon fall in the same model grid cell, so the model returns identical numbers for both. Model values are not monitor readings and tend to differ from them. The data were generated on 2026-11-18. Download them: aqi-india-2025.json. To regenerate them, run the script scripts/fetch-geo-data.mjs in the site’s repository.

How the typical-day statistics are built

The “typical day” tables on the hourly pages for NYC, Chicago, Austin and Ann Arbor use ERA5 hourly reanalysis for 2025 through Open-Meteo. For each city we take the temperature at 2 metres, wind speed at 10 metres and precipitation for every hour, group them by local hour of day, and average them. Summer is June to August and winter is December to February. A “rainy hour” is an hour with at least 0.1 mm of precipitation. Reanalysis grid cells are about 25 kilometres wide, so values describe the area around the city rather than one weather station. Download: hourly-us-2025.json.

Accuracy and limits

  • Forecast skill falls with time. Per NOAA, a 5-day forecast is right about 90% of the time, a 7-day about 80%, and a 10-day or longer about 50%. Read days 8 to 14 as a trend.
  • Models describe areas, not points. A sea breeze, a thunderstorm cell or a hill can make your street different from the forecast for your grid cell.
  • Radar sees what has already formed. It cannot see rain that has not started, and coverage is thin in some regions.
  • We are not a warning service. For watches, warnings and the monsoon bulletin use the US National Weather Service or the India Meteorological Department, or your own national service.
  • Health advice is general. It follows the US EPA’s and WHO’s guidance and is not medical advice.

Corrections and updates

If a number, a source or a statement on any page is wrong, tell us on the contact page with the page address and what you expected. We check every report, correct confirmed errors, and record material changes in the change log below. Pages show the date they were last reviewed. The statistics describe one calendar year and are regenerated when a new year of data is complete.

How to cite this site

Cite the page, not just the site: MyWeatherToday (Codologies Pvt Ltd). “Page title.” https://myweathertoday.com/path. Accessed day month year. The yearly AQI and typical-day datasets are free to reuse under CC BY 4.0 with attribution to MyWeatherToday and the underlying sources (CAMS and ERA5, via Open-Meteo). A plain-text summary of the site for AI assistants is at /llms.txt, with full detail at /llms-full.txt.

Change log

  • 2026-11-18. Published the 2025 AQI statistics for eight Indian cities and the 2025 typical-day statistics for four US cities, with downloadable data. Added this methodology page. Corrected the AQI grid-size statement: CAMS cells over India are about 45 km wide, not 10 km.

Questions

Where does MyWeatherToday get its weather data?

From national and international weather models (NOAA GFS and HRRR, ECMWF, DWD ICON, Météo-France and others) delivered through Open-Meteo, which is published under CC BY 4.0. The highest-resolution model available for each place is used for the first days, then global models for the longer range. Air quality comes from the Copernicus Atmosphere Monitoring Service (CAMS) and radar from RainViewer.

Is the AQI on MyWeatherToday official?

No. It is the US AQI calculated from a CAMS atmospheric model, not from a government monitoring station. India’s CPCB and the US EPA publish official AQI readings from ground monitors, and those can differ from the model by a wide margin at a specific location. Use this site to compare cities and to see the pattern across the day and the year, and the official monitors for a particular street.

How accurate are the forecasts?

Per NOAA, a five-day forecast is right about 90% of the time, a seven-day forecast about 80%, and a ten-day or longer forecast about half the time. The next few hours and tomorrow are the most reliable parts of any forecast.

How do I cite MyWeatherToday?

Use the page title, publisher and date: “MyWeatherToday (Codologies Pvt Ltd). [Page title]. https://myweathertoday.com/[path]. Accessed [date].” Our yearly AQI and hourly-pattern datasets are free to reuse under CC BY 4.0 with attribution to MyWeatherToday and the underlying sources, and can be downloaded as JSON.

How do I report a mistake?

Use the contact page. We check every report, correct confirmed errors, and note material corrections in the change log at the bottom of this page.