Weather and climate forecasting plays a crucial role in predicting various weather-related events, such as temperature fluctuations, cyclones, hurricanes, and extreme weather conditions, helping to mitigate risks and prevent loss of life.
Training traditional weather forecasting models require collecting and processing vast amounts of data, extensive preprocessing, and significant computational resources, even on high performance computer systems making forecasting such events a complex and resource-intensive task. To address these challenges, IBM and NASA have jointly developed Prithvi Weather and Climate to enhance weather and climate predictions with greater efficiency. Esri has integrated this as a pretrained Weather and Climate Forecasting model.
This foundational model for weather and climate, is a scalable 2D vision transformer model with 2.4 billion parameters, trained on 160 variables from MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, Version-2) data. This offers advanced capabilities for forecasting atmospheric conditions, such as identifying weather patterns, and improving climate modeling. The 2D vision transformer with encoder-decoder-based architecture, enables it to effectively capture both regional and global dependencies in the input time series data, thereby enabling it to analyze large-scale meteorological datasets with high accuracy.
By utilizing this model, organizations and researchers can significantly reduce computational overhead while achieving faster, more reliable weather and climate predictions, supporting disaster preparedness, environmental monitoring, and climate resilience planning.
Model details
This model has the following characteristics:
- Input—Preprocessed MERRA-2 data with set of 20 surface, 10 vertical at 14 pressure levels and 4 static variables. MERRA-2 data is available from 1980 to the present day, at 3-hour temporal resolution. Sample data for the year 2020 is provided for access..
- Output—Multiband raster with forecasted values for each time step based on provided data of previous time steps, lead time, and input time.
- Compute—This workflow is compute intensive and a GPU with compute capability of 6.0 or higher is recommended.
- Applicable geographies—This model is expected to work well across the globe.
- Architecture—The model is a scalable 2D vision transformer adapted and modified for weather and climate data by IBM and NASA.
- Training Data—The model is trained using MERRA-2 data with a 3-hour gap from 1980 to 2019 and validated with data from one of the years in the 2020-2023 range. Input data had 4 constant static variables, 20 surface variables, and 10 vertical variables at 14 pressures levels.
Access and download the model
Download the Prithvi Weather And Climate Forecasting pretrained model from ArcGIS Living Atlas of the World. Alternatively, access the model directly from ArcGIS Pro, or use it in ArcGIS Online with the Professional or Professional Plus user type.
To download the model, complete the following steps:
- Browse to ArcGIS Living Atlas of the World.
- Sign in with your ArcGIS Online credentials.
- Search for Prithvi Weather And Climate Forecasting and open the item page from the search results.
- Click the Download button to download the model.
You can use the downloaded .dlpk file directly in ArcGIS Pro, or upload and use it in ArcGIS Enterprise. Additionally, you can fine-tune the pretrained model if necessary.
Release notes
The following are the release notes:
| Date | Description |
|---|---|
| June 2025 | First release of Prithvi Weather And Climate Forecasting |