Bridging the Scale Gap: Leveraging EOF and Non-Parametric Correlation to Connect Meteorological Fields and Precipitation Isotopes

In the fields of atmospheric and climate science, there is growing use of machine learning and global circulation models. These approaches are becoming increasingly sophisticated with the availability of extensive ground-based and remotely sensed datasets. However, both approaches rely on the availability of large spatial and temporal datasets. For over half a century, stable isotopes of oxygen and hydrogen have been used as robust proxies for understanding hydrometeorological processes, acting as conservative tracers of land-ocean-atmosphere interactions. However, these isotopic measurements are non-continuous and highly discreet. Although satellites such as ACE, TES, Aura, and SCIMACHY do measure the isotopic composition of atmospheric vapour, they carry high uncertainties, making them less reliable. Therefore, despite their promise, these approaches are not readily applicable for deciphering local hydrometeorological processes, primarily due to limited data availability and relatively coarser spatial resolution.

Here, we introduce a simple yet robust approach to link meteorological and atmospheric data with discreet and limited isotopic measurements, aiming to understand how large-scale ocean-atmosphere processes govern local hydrometeorology. We employed Empirical Orthogonal Function (EOF) to identify prominent oceanic and atmospheric patterns over large spatial domains and to reduce dimensionality, thus converting the 3D climate datasets (e.g., ECMWF reanalysis) into 2D representations. We then applied non-parametric correlation technique, specifically Spearman‘s rank correlation, to link the meteorological data with localized, discreet precipitation isotope measurements. Adopting a non-parametric correlation avoids strict assumptions about data distributions. This approach offers significant benefits over traditional and more complex, modern methods by handling non-linearity and spatial heterogeneity. It also provides an effective means of identifying and interpreting local hydroclimatic processes and their linkages to broader atmospheric and oceanic drivers, thereby bridging the gap between large-scale atmospheric factors and local hydrological responses. Consequently, it offers deeper insight into the complex interplay among numerous processes operating at varied spatiotemporal scales.

Our preliminary findings quantitatively highlight the roles of sea surface temperature gradient between the eastern Indian Ocean and the South China Sea, pressure, potential vorticity, boundary layer height, vertical transport, wind speeds, and specific humidity in driving precipitation isotope variability in the Malaya peninsula. These linkages were previously unknown or qualitatively estimated by traditional methods, highlighting the value of this synergistic approach in bridging the spatial data disparities and improving our understanding of the regional drivers in the local hydrological cycle.