UC Davis

Objective identification of low-pressure systems (LPSs) in large-scale datasets, such as reanalyses and Earth system model outputs, is essential for understanding their physical characteristics, evaluating their representation in numerical models, and assessing their historical variability and projected future changes. Among all types of LPSs, tropical cyclones (TCs) have received the greatest attention because of their destructive impacts throughout their life cycles. Despite substantial advances in objective feature tracking, challenges remain in accurately identifying TC features in climate datasets, which can lead to biases in TC climatology and model evaluation. Contemporary TC research spans the entire TC life cycle—from precursor disturbances and genesis to post-tropical evolution. To facilitate such research using reanalyses and Earth system model outputs, an objective framework capable of tracking and classifying TCs throughout all stages of their life cycles is required.To address these challenges, Chapter 2 of this dissertation presents the System for Classification of Low-Pressure Systems (SyCLoPS), an objective LPS tracking and classification framework built upon the TempestExtremes (TE) software. Unlike previous tracking algorithms, SyCLoPS first identifies every detectable LPS signal at a prescribed temporal frequency using fundamental atmospheric fields and subsequently classifies each detected system at both the node and track levels. The classification criteria are based on physically motivated definitions developed using ERA5 and are further refined with machine-learning techniques. Evaluation demonstrates that SyCLoPS improves tropical cyclone detection relative to the previous TE-based algorithm, increasing the hit rate (HR) by 2% while reducing the false alarm rate (FAR) by 6%. SyCLoPS also achieves 95% agreement with the subjective classifications contained in the International Best Track Archive for Climate Stewardship (IBTrACS). Furthermore, the framework is applicable to a broad range of LPS types, including monsoonal systems, polar lows, subtropical storms, extratropical cyclones, and thermal lows.Chapters 3 and 4 demonstrate the application of SyCLoPS to advance the understanding of TC representation in both climate models and atmospheric reanalyses. Chapter 3 applies SyCLoPS to historical HighResMIP simulations to investigate the coupled evolution of TC intensity and size throughout the storm life cycle. By introducing an objective TC size retrieval algorithm that estimates TC quadrantal wind radii comparable to those in IBTrACS, the study enables the first direct comparison of TC size between climate models and observations. The results show that, despite increasing model resolution, most HighResMIP models and ERA5 remain unable to reproduce the observed intensity–size relationship, with simulated storms generally being too large for a given intensity and failing to capture the population of compact, intense TCs. Interestingly, the multi-model ensemble reproduces some aspects of TC climatology, as well as differences in intensity and size characteristics between ENSO phases, comparably to—and in some cases, better than—ERA5. Motivated by these findings, Chapter 4 evaluates the newly released JRA-3Q reanalysis and shows that it substantially outperforms ERA5 in TC representation, achieving an approximately 90% HR while reducing the FAR to only 6%. Error attribution analyses indicate that JRA-3Q's superior performance primarily results from its advanced TC bogussing scheme, whereas the degradation of ERA5 in earlier decades is largely attributable to increasing model error under sparser observational constraints. Together, these studies demonstrate the robustness and adaptability of SyCLoPS across diverse datasets while providing new insights into the strengths and limitations of current climate models and reanalyses in representing TCs.