Improved capability for automated production of first-guess ceiling and visibility (C&V) forecasts can beneficially support the human forecaster, who is faced with growing requirements for geographic coverage and forecast accuracy.

This paper outlines progress toward development of an automated system for production of 1-10 hr C&V forecasts for the continental U.S. The approach taken within the system is to gather the existing operational forecast resources available (e.g., RUC model, NWS LAMP, METAR observations) and synthesize from these a resultant forecast of improved skill. The system makes use of an internal observations-based ruleset forecast method as an additional input, and is easily adaptable toward use of future forecast inputs as they become available.

Two methods of forecast synthesis are being explored - an in-house agile selection approach driven by real-time tracking of the skill of each input forecast module, and a published weighted majority voting scheme which seeks an optimal weighted blend among forecast inputs. Both methods provide location-specific selection criteria and dynamically adjust as conditions and input forecast module performance change.

The paper describes forecast system architecture, early results comparing the forecast skill achieved by the two synthesis methods used, and lessons learned through testing.