Variability of Hydrological Excitation on Polar Motion Under Extreme ENSO Climate Conditions

Polar motion (PM) is a fundamental geodetic observable reflecting the global mass redistribution within the Earth system. As a major component of Earth mass changes, the variation of Terrestrial Water Storage (TWS) plays a crucial role in exciting PM, particularly on seasonal to decadal timescales. With the intensification of global climate change in recent years, understanding the coupling between TWS variability and PM has become increasingly important, especially under the influence of large-scale climate fluctuations like El Niño-Southern Oscillation (ENSO). However, quantifying these processes via traditional excitation functions remains challenging due to highly non-linear feedbacks between hydrological signals and Earth rotation.

In this study, we employ a Gated Recurrent Unit (GRU) machine learning framework to perform ablation studies, integrating HAM derived from the LSDM model as physical constraints to isolate the hydrological contribution to PM variability. We investigate the variability of TWS excitation on PM from 2014 to 2023, a period encompassing the 2015-2016 extreme El Niño and the 2020-2023 triple-dip La Niña. Under neutral conditions, the inclusion of HAM significantly reduces the PM prediction Mean Absolute Error (MAE) by 27.7% in  and 49.8% in , primarily by eliminating the recurrent bimodal error structure observed in the non-HAM solution. Spatiotemporal analysis revels that as boreal spring transits to summer, the distribution of errors coincide with coherent seasonal soil moisture depletion across mid-latitude Eurasia and North America (NA). This widespread mass deficit generates east-westward excitation vector consistent with the observed bias in the non-HAM solution, confirming that mid-latitude hydrological redistribution is the primary driver of seasonal PM excitation.

However, the contribution of HAM to PM excitatiin exhibits strong phase-dependence characteristic, which is pronounced during the ENSO developing phases and diminish significantly in the mature phases. During the peak of the 2015-2016 El Niño and the termination of the 2022 La Niña, the HAM-induced improvement in  sharply degrades to negligible (2.27%) or even negative values (-17.46%). We identify a dipole cancellation mechanism responsible for this degradation. Extreme ENSO events induce opposing precipitation anomalies in NA and East Asia. The conflicting excitation vectors neutralize the dominant hydrological signal, causing the HAM vector to lose its directionality and decoupling the hydrological signal from the linear logic of the prediction model. Our findings reveal that strong climate disturbances can disrupt conventional hydrological excitation patterns through spatial dipole cancellations. Our study not only quantifies the variable impact of TWS on Earth rotation but also highlights the necessity of considering non-linear climate-hydrology interactions in high-precision geodetic modeling.