GRAFS (GRidded Atmospheric Forecast System)
GRAFS (GRidded Atmospheric Forecast System) is a software system that is designed to address the difficult problem of producing high quality weather forecasts at locations where observations are not available. Accurate forecasts at remote locations are used to drive many user–specific applications such as road temperature forecasts along an entire roadway, or soil temperature forecasts for agriculture.
GRAFS combines numerical weather prediction (NWP) model forecasts, statistically refined point forecasts, and climatology to interpolate forecast data to a high resolution grid. Applications using GRAFS typically use 4–km grid cell resolution. Higher resolution grids may be appropriate in regions of rapidly changing topography.
To generate these forecasts, GRAFS starts with model data from a NWP model(s) and then downscales those data onto the target grid using a sophisticated climatological difference interpolation scheme. Finally, this downscaled forecast is corrected to match a statistically optimized forecast generated at a set of observation points.
GRAFS technology is ideality suited for downstream systems and users that require precision location–based forecasts. GRAFS is a licensed technology of the UCAR Foundation.
This funding from New York Power Authority is their cost share for the DOE Solar Power Forecasting project. Deliverables are the same as for the DOE project.
We are using these funds to stand up a gridded forecasting system for CONUS. Since it is connected to DOE Solar, the first forecasts will be for solar irradiance.
The Goal is to start a new OpenSource gridded forecasting system that could be used for a multitude of clients in the future
GRAFS - Motivation
- Gridded forecasting system: synthesis of numerical, statistical, and human weather forecasts on a regular grid
- Last mile in information pipeline from forecasters to users
- Problem: current gridded forecasting systems are inadequate, proprietary, or too specialized
- Need: A community gridded forecasting system in which many different interpolation, blending, and correction methods can be tested under a common framework
GRAFS - Vision
- Produce a gridded forecast product that blends information from multiple numerical models with smart artificial intelligence techniques.
- This product will be modular and customizable, allowing a variety of input sources, blending and optimization algorithms, and output formats.
- Make this product widely accessible and applicable for a broad community – work toward joint development and OpenSource software.
- Implement a version that can also contain licensed software and proprietary data.