Teresa Canavan

Thesis Approved: 1992

Climatology of Sea Breezes at Shearwater Airport, Nova Scotia and Prototype Expert System Design

Sea breezes are defined as thermally direct circulations resulting from unequal heating of air over adjacent land and water surfaces. These flow patterns cause changes in air quality, temperature drops and induce convection at coastal locations. Spatial and temporal dimensions place them at the upper end of the mesoscale. Theoretical investigations have revealed factors influencing sea breezes including friction, Coriolis effects, air mass stability, latitude, topography, coastline orientation, prevailing wind velocity and land-sea temperature contrast.

Investigation of ten years (1980-1989) of hourly wind velocity data for Shearwater, Nova Scotia reveals across shore energy at 12 and 24 hour cycles. At noon (AST) onshore winds occur most frequently. A gradual veering trend is observed in most frequent wind directions for the remainder of the day followed by offshore or calm winds at night. This supports existence of a land-sea breeze regime. Various criteria are specified to select sea breeze event from climatological data. Frequency, duration and strength of events are determined. Results from each method are compared and suitability of each criterion discussed.

Expert systems are computer programs that mimic the human decision making process. Reasoning techniques are encoded into these programs and they access or request necessary information to provide solutions. Most research in this field has occurred during the past decade with applications in weather forecasting constrained to the past five years.

A rule-based expert system is designed to predict occurrences of sea breezes at Shearwater, Nova Scotia. Knowledge acquisition relies on the author's five years of operational forecasting experience in this area. Current predictors used in this system include boundary layer wind velocities, land-sea temperature contrasts and cloud cover. Preliminary test results show high probability of detection percentiles and high false alarm ratios. Addition of forecast values of all variables and quantitative values from the climatological analysis is expected to improve model performance. Further testing and refinement are required to determine its merit as a forecasting tool.

Teresa is now working at the Atmospheric Science Division at Atmospheric Environment Branch, Bedford, NS. She can be reached at canavant@ns.ec.gc.ca

Back to Atmospheric Science Page.