## General information

### How would you differentiate Google's Weather API from specialized weather data vendors?

Google is integrating technology from our reputable weather company acquisition
with advancements from Google DeepMind's
[MetNet](https://research.google/blog/metnet-3-a-state-of-the-art-neural-weather-model-available-in-google-products/)
and [WeatherNext](https://deepmind.google/technologies/weathernext/) AI models
(2025). This leverages the proven reliability of the Google Maps Platform.

### What is the difference between weather data available in Earth Engine and BigQuery and Google's Weather API?

Google's weather product suite offers a comprehensive range of solutions for
diverse needs. WeatherNext models (namely
[GraphCast](https://deepmind.google/discover/blog/graphcast-ai-model-for-faster-and-more-accurate-global-weather-forecasting/)
and
[GenCast](https://deepmind.google/discover/blog/gencast-predicts-weather-with-sota-accuracy-and-faster-inference-time-on-tpu))
provide raw, AI-based model forecast outputs including parameters ideal for
research, modeling, and analysis, made available using open-source code and
historical or current weather datasets using Earth Engine and BigQuery APIs.

Conversely, the Google Maps Platform Weather API delivers processed weather data
for current conditions, hourly forecasts, and daily forecasts by combining AI
and traditional forecasting systems in a developer-friendly format that
seamlessly integrates into apps and web services. This dual approach caters to
both specialized research needs and broad developer accessibility.

### Why is Google not relying only on measurements from weather stations?

Google uses a variety of [input
sources](https://developers.google.com/maps/documentation/weather/faq#how_does_the_weather_api_produce_a_forecast_output)
for its weather model, including observations from weather stations, numerical
weather prediction models, and weather AI models. The reason for using these
various input sources is that observations from weather stations only provide
information about the weather at the location of the station at the time of
measurement. Weather station observations can be highly accurate for that
precise spot, but weather phenomena such as precipitation, especially from
showers or thunderstorms, is often localized and can vary significantly over a
short distance.

For example, a heavy downpour might be occurring right at the station's
location, while a few miles away, there's only a light drizzle or no rain at
all. Because a single station reading might not be representative of a wider
area, models are necessary to provide additional information in between
stations.

Additionally, weather stations typically provide hourly or half-hourly readings.
Even if those are provided in near-real time, there can still be a slight delay
between the actual precipitation events and its reporting. Models also assist in
getting the current conditions as close to the experienced reality at the
queried location as possible.

## API usage

### What is the Weather API coverage?

The Weather API supports all countries globally except [prohibited
territories](https://cloud.google.com/maps-platform/terms/maps-prohibited-territories).
For Japan and South Korea, weather data is provided by [Weathernews
Inc.](https://weathernews.com/weather/), an officially licensed meteorological
service operator.

For locations in Indonesia, the Weather API is available for internal usage
only.

> [!NOTE]
> **Note:** Weather forecasts are inherently subject to meteorological variability and may contain inaccuracies or discrepancies. Forecast data for Japan and South Korea is provided in partnership with Weathernews, Inc. Google is not responsible for the data accuracy or service-level agreement of update frequency for weather data returned for locations in Japan and Korea.

### What is the spatial granularity of the Google Weather API?

Data resolution is such that forecasts are created within a few kilometers for
any location globally.

### What is the update frequency for the Weather API?

| Endpoint | Current conditions | Hourly forecast | Daily forecast | Hourly history |
| **Refresh rate** | Calculated in real time. | Refreshed hourly. | Refreshed hourly. | Twice a day (7 AM and 7 PM Pacific Standard Time). |
|---|---|---|---|---|

- **Short-range precipitation (first 12 hours of the forecast):** Refreshed every 2 to 15 minutes using radar and high-resolution nowcasting models.
- **Non-precipitation \& extended precipitation horizon (up to 10 days):** Refreshed hourly to ensure the most up-to-date forecast guidance.

### Does the Weather API provide nowcast precipitation forecasting?

Google doesn't provide nowcast forecasting (minute forecast) in the API, but it
does provide current conditions, hourly forecasts, and daily forecasts.

### Does the Weather API provide separate UVA and UVB values?

UV Index is a single index which combines both UVA and UVB. We don't provide a
breakdown by component.

### How does the Weather API produce a forecast output?

As shown in the diagram below, Weather API forecasts are created from an
internal forecasting system that utilizes weather models and observations from
[global weather agencies](https://support.google.com/websearch/answer/13687874)
as inputs. Data from public ground-truth weather sensors are used to improve the
forecast model.

![Weather API forecast model diagram](https://developers.google.com/static/maps/documentation/weather/images/weather_forecast_model.png)

### What is the accuracy of the Google Weather forecast, for different regions and forecast horizons, compared with forecasts from leading government weather agencies?

The figures in **Table 1** compare [Root Mean Squared Error
(RMSE)](https://en.wikipedia.org/wiki/Root_mean_square_deviation) for
temperature and wind speed for a 240 hour (10 day) forecast horizon between
Google weather data and global and regional weather models from government
agencies across 11 months (15 August 2024 through 1 July 2025).

Globally, the Google weather forecast outperforms weather forecasts from leading
government agencies. For Europe and North America, the Google weather data has
lower error than the most accurate regional government models for at least the
later portion of their forecast horizons. The excellent performance of Google
weather data for longer forecast horizons is particularly notable, which can be
the most challenging to accurately predict.

**Table 1** lists the average RMSE values for the entire forecast horizons
relevant for each region model combination. A lower RMSE value indicates better
performance. The Google weather forecasts have the lowest, or among the lowest
average RMSE values, for nearly all parameters, regions, and forecast horizon
ranges.

Click an image to view an enlarged version.

|   | Temperature | Wind speed |
|---|---|---|
| Global | [![](https://developers.google.com/static/maps/documentation/weather/images/global_tempRMSE.png)](https://developers.google.com/static/maps/documentation/weather/images/global_tempRMSE.png) | [![](https://developers.google.com/static/maps/documentation/weather/images/global_windspeedRMSE.png)](https://developers.google.com/static/maps/documentation/weather/images/global_windspeedRMSE.png) |
| North America | [![](https://developers.google.com/static/maps/documentation/weather/images/NA_tempRMSE.png)](https://developers.google.com/static/maps/documentation/weather/images/NA_tempRMSE.png) | [![](https://developers.google.com/static/maps/documentation/weather/images/NA_windspeedRMSE.png)](https://developers.google.com/static/maps/documentation/weather/images/NA_windspeedRMSE.png) |
| Europe | [![](https://developers.google.com/static/maps/documentation/weather/images/europe_tempRMSE.png)](https://developers.google.com/static/maps/documentation/weather/images/europe_tempRMSE.png) | [![](https://developers.google.com/static/maps/documentation/weather/images/europe_windspeedRMSE.png)](https://developers.google.com/static/maps/documentation/weather/images/europe_windspeedRMSE.png) |

**Table 1.** Average RMSE values for relevant for each regional model
combination

**Table 2** lists the average RMSE values for relevant forecast horizons for
each model (Google values in bold). Parameters (temperature, wind) and forecast
horizons for which Google uniquely has the lowest average RMSE are highlighted
in dark green. Parameters and forecast horizons for which Google has the lowest
average RMSE along with (an)other model(s) are highlighted in light green.

![Average RMSE values for relevant forecast
horizons](https://developers.google.com/static/maps/documentation/weather/images/faq_table.png) **Table 2.** Average
RMSE values for relevant forecast horizons per model

### Does the historical data endpoint reflect actual weather data (i.e., verified by local measurements)?

Historical weather data is primarily modeled output. While it incorporates
observations from [global weather
agencies](https://support.google.com/websearch/answer/13687874) and numerical
weather prediction (NWP) models, these are assimilated into the Google Weather
model to create a complete and consistent historical record. Actual station
measurements are considered "ground truth" at their specific locations, but
their limited global distribution makes them insufficient for a complete
worldwide picture.

### What do the `weatherCondition`, `precipitation.probability.percent`, `precipitation.qpf.quantity` fields mean, and how are they related?

These fields provide distinct but related insights into precipitation and
overall weather.

- **`weatherCondition`:** This field provides a general, intuitive description of the overall weather conditions in the queried area, taking various parameters into account for easy understanding.
- **`precipitation.probability.percent` (PoP):** This represents the probability of precipitation. The likelihood that precipitation will occur at the location of the forecast during the defined time period (typically hourly).
- **`precipitation.qpf.quantity` (QPF):** This stands for quantitative precipitation forecast and indicates the expected amount of precipitation, measured as a depth (for example, millimeters or inches). This value represents how much precipitation is expected if it does occur within the specified time and location.

> [!NOTE]
> **Note:** While both relate to precipitation, PoP (probability) tells you if it's likely to rain, while QPF (quantity) tells you how much might fall. A non-zero PoP typically implies a non-zero QPF, but they don't always scale directly. For example, a low PoP can coincide with a high QPF (such as a small chance of a very heavy thunderstorm), or vice-versa.

### How can I determine whether or not it's raining at a certain location?

We recommend using the `weatherCondition` field in order to understand whether
it is raining or not at a certain location. Multiple conditions can indicate
that it is raining (RAIN, LIGHT RAIN, RAIN SHOWERS, HEAVY RAIN, WIND AND RAIN,
THUNDERSTORM, HEAVY THUNDERSTORM, THUNDERSHOWER, SCATTERED SHOWERS, LIGHT
THUNDERSTORM RAIN, SCATTERED THUNDERSTORMS, RAIN AND SNOW). These conditions are
designed to capture all instances of light, strong, continuous or scattered
rainfall and also differentiate rain from snow. Depending on your specific use
case --- for example, if your use case only considers heavy rainfall --- you may
choose to use only a relevant subset of these conditions.

### Are the "current conditions" equivalent to the observations at weather station locations? How are "current conditions" determined in locations without stations?

Our "current conditions" provide the most up-to-date weather information by
combining various data sources, but they are not strictly equivalent to direct
station observations in all cases.

For `precipitation.probability.percent` and `precipitation.qpf.quantity`
(accumulated over the past hour), the value presented in the `currentConditions`
response is always derived from the most recent forecast. Probability itself is
a modeled, rather than directly observed, quantity.

This approach allows us to provide comprehensive "current conditions" even in
areas without direct sensor coverage, ensuring the best available information at
all times.

## API limits and access

### Is there a rate limit on the API?

For the Weather API, there is a default rate limit of 6,000 queries per minute.

### Can I access the data in bulk?

Bulk data is not available. You can query Weather API within the quota (6,000
QPM) and respect caching terms outlined in the [Terms of
Service](https://cloud.google.com/maps-platform/terms/maps-service-terms).

### Do I have to enable billing?

A valid billing account is required to use the Weather API. See [Enable
billing](https://developers.google.com/maps/documentation/weather/cloud-setup#billing) to set up your project
with a billing account.

### What happens if I query a location that is not supported?

If the latitude and longitude are not covered in the supported country list, the
response will return a 404 error code with the message, "Information is
unavailable for this location. Please try a different location."