Patrick Britt

Date of Award

January 2025

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Atmospheric Sciences

First Advisor

Aaron Kennedy

Abstract

Small Unmanned Aerial Systems (sUAS) have become a fast-growing part of air traffic in the United States. This significant growth is due to the numerous uses these systems can provide, both recreationally and professionally. Due to the small nature of the sUAS, any weather, regardless of intensity, plays a vital role in flight operations and planning. In a survey of sUAS pilots, it was found that precipitation was the most significant hazard of concern, and the most time-sensitive hazard, as precipitation conditions are most often checked by operators in the immediate moments before flight. This project aimed to create, analyze, and verify the performance of precipitation forecasts to improve sUAS flight operations and planning, given the needs of sUAS operators. Forecasting of precipitation is done probabilistically by taking the fraction of precipitation in a neighborhood. Three methods are used to create such forecasts: modeling data from the High-Resolution Rapid Refresh, utilizing computer vision to generate projections of the observations, and a persistence forecast of the observations. From these three methods, probabilistic forecasts are generated for a period of twelve hours. Validation of each method is conducted over multiple degrees of freedom, including forecast hour, region, season, and neighborhood size, as well as further analysis of the effects of convection. Results show that a 100 km neighborhood radius offers the best blend of performance across all forecasting techniques. For the early forecast hours, the observational methods of optical flow and persistence always beat the HRRR in forecast skill. Skill in the observational methods over the forecast period is consistent across regions, except in the western regions of the United States, where optical flow and persistence perform similarly in skill due to orographic effects which resulted in slower-moving precipitation. The available amount of verification data was limited in some regions, which resulted in poorer HRRR verification in those regions. Switch-over times, the time in the forecast period when optical flow becomes less performant than the HRRR, were found to occur around F04 in the forecast period. This information will be used to improve precipitation forecasts for the sUAS application located at://uasforecast.com/

Download

Share

COinS