Weather Prediction Machine Learning Optimization


Figure 1.  DICast system diagram.
Figure 1.  DICast system diagram.
Figure 2.  GRAFS conceptual data processing flow diagram.
Figure 2.  GRAFS conceptual data processing flow diagram.

RAL is a leader in the development of intelligent weather prediction systems that blend data from numerical weather prediction models, statistical datasets, real time observations, and human intelligence to optimize forecasts at user–defined locations. The Dynamic Integrated Forecast System (DICast®) and the GRidded Atmospheric Forecast System (GRAFS) are examples of such technology (Figures 1 and 2). DICast® is currently being used by three of the nation's largest commercial weather service companies. Applications of this technology continue to expand as there is a growing desire in industry to have fine–tuned forecasts for specific user–defined locations. This trend is clear in the energy, transportation, agriculture, and location–based service industries. RAL's expertise in meteorology, engineering, and applied mathematics and statistics is being utilized to address society's growing need for accurate weather information.


During this year the improvements in DICast's probability of precipitation forecasts have been significant with the addition of a new Gradient Boosted Regression Trees machine learning technique. This technique has improved the precipitation probability Brier scores on the order of 10-20% over the previous method. Improvements in precipitation probability using this method have been shown over the full range of forecast extents and accumulation periods. The method is also more agile as it uses a recent history of forecasts and observations to perform the machine learning training.

As updates to numerical weather prediction models continue, DICast was also updated to take advantage to model improvements. This year, DICast systems started using native-hourly forecasts in the short term (out to 48 hours) from both the GFS and GEM models. DICast was also adapted to incorporate NWP output from a commercial weather provider partner.

DICast’s impact on renewable energy forecasting has led to its use in other renewable energy projects. In particular, a new project with the Kuwait Institute of Scientific Research (KISR) started using DICast as the core forecast engine for a combined wind and solar forecasting system. This system will combine output from global numerical weather prediction models and a high-resolution version of WRF to produce custom forecasts for an extreme desert climate environment.     

Figure 3.  Sample image of Global Horizontal Irradiance forecast over the continental US produced by GRAFS.
Figure 3.  Sample image of Global Horizontal Irradiance forecast over the continental US produced by GRAFS.

RAL continued the effort to develop a new gridded forecast system (GRAFS) that is open to the university community for research (Figure 3). This system is modular in nature, allowing choices in base numerical weather prediction models, as well as consensus forecasting techniques. This system was first used in solar energy forecasting, allowing utilities to assess production of distributed solar power, but has now been extended to other weather applications as well. 

These RAL forecast systems also continuing to push the envelope of advanced weather forecasting in the transportation sector.  The Maintenance Decision Support System (MDSS) was adapted from its original focus on roadways to be used as a Runway Decision Support System for Denver International Airport (DIA). The system generates tuned weather forecasts and treatment recommendations for the runways at DIA. In addition, DICast and a weather-tuned version of GRAFS form the backend weather engine used in both the FHWA and Colorado Pikalert Hazard Assessment forecast systems. Finally, RAL teamed with the Colorado Department of Transportation (CDOT) to produce forecasts of estimated travel time over Colorado highways using advanced machine learning techniques. These travel time forecasts take weather into account when they are being generated and provide the travelling public important tactical information as they are in-route.


Areas of development for the next fiscal year include:

  • Extend machine learning techniques to other variables produced by DICast
  • Expand DICast to produce aviation-related forecast variables and use new model sources
  • Expand GRAFS to include other consensus blending methods and other variables
  • Make improvements related to road temperature and precipitation forecasts in the MDSS