Multivariate Sensitivity Analysis of Orographic Precipitation within an Idealized Atmospheric River Environment

Atmospheric rivers (ARs) are long, narrow corridors of strong water vapor transport typically located east of the cold front in extra-tropical cyclones. The orographic precipitation produced when water vapor transported in an AR is forced upward by topography provides a source of freshwater but can also result in flooding. The location, amount, and type of precipitation forecast to reach the surface is important to determine how much water will enter the surface and sub-surface water systems, and to prepare risk managers for potential flood threats. The environmental conditions and microphysical processes in a particular AR event may lead to a specific precipitation outcome (e.g., extreme rainfall and flooding), but do there exist other scenarios where a similar precipitation outcome could be produced? For example, are there other AR configurations that might also produce extreme rainfall and flooding? Additionally, if there are different environments and moist processes that can produce a similar total precipitation amount and intensity, will the fraction of frozen versus liquid precipitation be different?

These questions are examined using an idealized modeling framework that simulates moist, nearly neutral flow interacting with a two-dimensional bell-shaped mountain and applying a Bayesian Markov chain Monte Carlo (MCMC) algorithm. A range of microphysical parameter values associated with the snow fallspeed, riming process, and rain accretion, as well as a range of upstream wind speed and relative humidity profiles and surface potential temperatures, are examined. Results show multiple sets of parameter combinations can yield a similar total precipitation amount and intensity. Some of these parameter pairs show compensating effects, i.e., increased snow fallspeeds can mitigate the effect of increased wind speeds to yield a similar precipitation outcome. Within a narrow range of wind speeds and temperatures, any microphysical parameter value can reproduce the same precipitation intensity and amount, thus implying total precipitation is more sensitive to upstream environmental conditions. Given results thus far, it is hypothesized that scenarios where a larger liquid precipitation fraction compared to the control simulation will dominate the MCMC-generated ensemble.