WeatherNext 3 Data Access: The Routes Into the Forecast DataData access
Reading about the model is free; using its numbers in a pipeline is a data problem. This page maps the channels that carry forecast fields, what each one is suited to, what access does not include, and the shortest honest path from a prototype to something you can run daily. For the API layer specifically, see the API guide; for what can be exported, see the download guide.
7-day free trial, then $19/mo · cancel anytime · compare plans
What access actually means here
Three facts that separate the documentation from the data.
Docs are open The research write-up, model documentation and parts of the family's code are public and free to read - the knowledge layer costs nothing.
The research write-up, model documentation and parts of the family's code are public and free to read - the knowledge layer costs nothing.
Fields are metered Forecast fields reach teams through Google Cloud data products, billed on standard cloud terms: queries, storage and requests.
Forecast fields reach teams through Google Cloud data products, billed on standard cloud terms: queries, storage and requests.
Code is partial Earlier generations and the cyclone models are published in public repositories. Treat each repository's scope as its own claim.
Earlier generations and the cyclone models are published in public repositories. Treat each repository's scope as its own claim.
The four routes into the data
Each channel answers a different question. Most teams end up using two of them.
BigQuery
SQL over large forecast tables. Best when your analysis already lives in a warehouse and you want to join weather to your own data.
Earth Engine
Gridded geospatial work in the browser or a notebook. Best for mapping, overlays and region-level studies without moving files around.
Cloud Storage
Bulk files for pipelines, archives and model input. Best when you want to own the copy and control the schema yourself.
Weather API
Request-based access for products and apps. Best when a service needs a forecast on demand rather than a dataset.
A realistic path for a team
Work in this order and the first useful result arrives before the budget conversation does.
Name the decision first
Write down the decision the data will change - dispatch, staffing, routing, risk. Weather data without a decision attached turns into an expensive dashboard.
Start with one variable, one region
Pick the variable closest to your decision and a region you can verify. Expansion is cheap later; a weak foundation is not.
Prototype in a notebook
Query the data interactively before building anything scheduled. Most of the design work happens in the first afternoon, not in the pipeline.
Budget for usage, not a license
Costs come from queries, storage and requests on standard cloud terms. Estimate them early with a rough volume per day.
Keep an independent check
Compare your pipeline output with observations or an agency forecast for the first weeks. It is the cheapest way to catch a unit or timezone error.
What access does not include
Four misunderstandings that cost teams time when they assume more than the channels provide.
No model license for sale
Nobody is selling you the model. You are paying for queries, storage, interfaces or support - a service layer around the forecast.
No single schema across sources
Different channels name and shape fields differently. Budget an afternoon for mapping units, time zones and naming before comparing anything.
No substitute for warnings
Data access does not transfer responsibility. National meteorological agencies remain the source for safety warnings wherever the numbers come from.
History is a separate question
Forecast access is not a reanalysis archive. If you need years of past conditions for training or backtesting, plan that as a distinct data project.
WeatherNext 3 data access questions
Short answers about channels, cost and what to expect on the first build.
Try the data before you build the pipeline
Hourly forecasts and local detail in one workspace, so you can judge the numbers before committing engineering time. Seven-day free trial, cancel anytime. Background: the API guide and the cost breakdown.
Related guides
The pages that go with this one, chosen by what each one answers.
