The Salient platform provides access to gridded data for all model outputs (and historical data) on a unified grid with global extent.

• Salient gemAI proprietary generative AI model that produces daily ensemble trajectories
• Salient Blend proprietary ****multi-model blend of our proprietary climatology, AI models with properly calibrated dynamical (GEFS and ECMWF) models
• Consolidated Access to government models (NOAA GEFS + GFS, ECMWF ENS extended range + SEAS5)
• Direct Delivery Short-term Forecasts (GEFS v12 and EC ENS 15 medium range) available for point locations and/or gridded data shortly after they are released publicly and delivered directly to a cloud storage bucket (available for additional license fee).
• Salient’s probabilistic climatological forecasts provide full probabilistic distributions that characterize the uncertainty around the calculated climatological mean

Figure 1: Map of timescales to available models

Salient gemAI v2

Salient gemAI is a proprietary vision transformer model that generates natively calibrated, probabilistic daily ensemble forecasts (200 members) up to 126 days ahead. It maintains spatial and temporal covariance across variables enabling better uncertainty quantification of multivariate risk and tail events. It is an independent model, initialized with with atmospheric and ocean conditions from ECMWF HRES . It is natively trained on CRPS (continuous ranked probability score) and ECMWF ERA5 historical data.

Model Name

API model: gem

Spatial Resolution

The Salient forecast has spatial coverage for North America and provides forecasts for latitude-longitude grid points at 0.25 x 0.25 degrees (25 km) resolution. Spatial debiasing is available to downscale 0.25deg ERA5-style data to a specific location within a grid cell at a granularity of 100m.

Salient gemAI Temporal Resolution and Update Frequency

Salient Blend

The Salient Blend model is a multi-model blend of our proprietary climatology, AI models (natively probabilistic and calibrated by design) with properly calibrated dynamical (GEFS and ECMWF) models.

Our blending algorithm objectively assesses which models are most skillful for particular variables, regions, seasons, and lead times. The blending model also dynamically weights the input models based on current climate conditions, which allows the blend forecast to capture signals from each input.

Then it uses that information to appropriately apply grid-point-by-grid-point model weights to combine the AI and NWP models resulting in a 'Salient Blend' forecast that is superior to the individual model forecasts by accounting for each model's varying spatial and temporal strengths.

Model Name