Monday, 29 January 2024
Hall E (The Baltimore Convention Center)
The Pigeon River Basin (PRB) in the Southern Appalachian Mountains, USA, characterized by its complex orography and terrain-driven weather, represents a typical flash flood- and landslide-prone area with a challenging-to-observe catchment hydrology. To enable an understanding of the distribution and life cycle of storm events over this region, the Duke Great Smoky Mountains Rain Gauge Network (GSMRGN) provides high temporal resolution precipitation measurements across the PRB. With over a decade of continuous observations, the GSMRGN offers a unique data record of orographic precipitation fields. As such, it stands as a strong foundation for developing interpolation techniques over sparse point-measurements in high-terrain regions. If accurate, derived interpolated datasets provide much-needed validation reference for remotely sensed products which are typically used in mountainous regions but known to be of low-quality. Satellite-based QPEs, such as Integrated Multi-satellitE Retrievals for Global Precipitation Measurement and the Climate Prediction Center morphing, are often too coarse resolution for detailed storm structure. The present study utilizes the GSMRGN precipitation measurements to contribute towards mitigating this problem. The work considers individual precipitation events during the period of July 2009 to October 2021 to find whether elevation, ridge line, direction of storm propagation, and environmental properties affect the spatial correlation of gauge-measured precipitation rate across the PRB. The direction of propagation was found to play a dominant role. Classified by their propagation course, the measurements at each gauge are analyzed for skewness, power-transformed, and kriged. Spatial and elevation-based trends on the shape and scale parameters of the observed distributions are used to invert transformed data into raw precipitation values. Inverse distance squared weighting is applied to the gauge measurements and combined with the kriging. It is found that trends in the shape and scale parameters of the distributions are related to the size of the rain gauge tipping bucket rather than the physical processes. To eliminate this problem, the analyses are limited to events with 2 mm or greater rain rates, twice the largest tipping size of the rain gauges. To construct the final product, the generated QPEs are validated using the leave-one-out method.

