The Large-Scale Drivers of Precipitation in the Northeastern United States via Linear Orthogonal Decomposition
Under future climate change scenarios, water availability is expected to be altered in the northeastern United States. These changes are due to both an increasing trend in the frequency and intensity of extreme precipitation, as well as the likely increased frequency of short-duration drought events. Given the many problems that can be associated with an increase in these types of extremes, our work has sought to improve the current understanding of the large-scale environmental conditions favorable to precipitation in this region by using a novel, linear orthogonal decomposition (LOD) approach, and to quantify the relative contribution to precipitation amounts from relevant upstream drivers.
The LOD approach is able to identify a set of independent, large-scale modes that are nearly optimal for the prediction of precipitation totals by sub-sampling the time series of all predictor fields at all grid points in order to maximize the linear predictability, in the sense of multiple linear regression (MLR). Such an approach is complementary to other methods for decomposing the meteorological fields that drive precipitation, with the added advantage that the linear modes are easily combined within a single linear model. This method can be useful as another tool for model validation and can provide insight into the large-scale drivers and other relevant meteorological quantities.
Given the significant impact and damages that can accompany the likely increase in the frequency and intensity of precipitation extremes in the northeastern United States, there is a growing need to better understand the dominant, large-scale drivers of precipitation in this region. As such, a linear orthogonal decomposition (LOD) method is implemented to iteratively extract time series (based on field and geographic location) of absolute maximum correlation with precipitation. Linear modes are then projected onto the full set of 2D atmospheric fields to provide physical insight into the precipitation-generating mechanisms. The first mode is associated with vapor transport from the Atlantic seaboard, the second mode is characterized by westward vapor transport associated with extratropical cyclones, and the third mode captures vapor transport from the Gulf of Mexico during the fall and winter. However, the third mode is less robust in the spring and summer. Implementing LOD also produces similar results across multiple datasets (reanalyses and CESM1 LENS). From analyzing these results, LOD emerges as a useful and complementary tool to clustering approaches as a means for identifying large-scale modes that are associated with precipitation, and its use can be extended for a variety of purposes, including inter-dataset comparison, model validation, and understanding changes to large-scale drivers in future climate scenarios.