From Hadley Cells to Heat Extremes: Novel Physical and Statistical Frameworks for Evaluating Climate Models

Citation

Khapikova, Polina (2025) From Hadley Cells to Heat Extremes: Novel Physical and Statistical Frameworks for Evaluating Climate Models. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/px1t-j224. https://resolver.caltech.edu/CaltechTHESIS:06032025-003651371

Abstract

As climate models become increasingly central to projecting future climate impacts and guiding adaptation strategies, understanding the sources of intermodel spread is critical. This thesis investigates key drivers of model uncertainty through a process-based lens grounded in fundamental climate physics. It focuses on two major contributors to model spread: atmospheric circulation and land-surface processes, which together shape regional hydrological extremes and human-relevant climate outcomes.

First, I develop a zonal-mean energetic framework to explain variations in the meridional extent of the Hadley Circulation (HC) across seasonal, interannual, and multidecadal timescales. I show that changes in tropical net energy input and eddy energy export can be used to explain HC migrations in a variety of contexts. For example, this framework can be used to explain the larger migrations of the ascending HC branch relative to the descending branch over the seasonal cycle, and the contrasting HC migration observed during El Nino events and in response to global warming. This approach provides a physically grounded method for interpreting both observed short-term variability and projected future changes in the tropical atmospheric circulation.

Next, I address uncertainty in terrestrial water storage (TWS) trends across CMIP6 models. To circumvent the limitations of short observational records, I introduce a novel method that compares the seasonal cycle of TWS using empirical orthogonal functions derived from GRACE satellite data. This approach enables the ranking of models based on their fidelity to observed seasonal variability. Models with stronger agreement with GRACE observations exhibit more consistent and often larger trends in TWS, and share features such as deeper soils and spatially distinct precipitation patterns. These insights can guide model selection and development efforts.

Finally, I explore model spread in the temperature-humidity combinations that produce extreme wet-bulb temperatures (WBTs), a key metric of human heat stress. While WBT projections are relatively consistent across models, the partitioning between temperature and humidity—quantified using the "stickiness" metric—varies substantially. I show that these differences lead to large uncertainty in the physiological impacts of heat stress at a given WBT. This work underscores the importance of accurately capturing surface energy partitioning to constrain present and future heat stress risks.

Together, the chapters of this thesis offer several innovative approaches to evaluating uncertainty in climate projections. By applying simple theory, observational constraints, and novel metrics, this work demonstrates how physically grounded methods can identify more reliable models and improve the accuracy of future climate projections---particularly for hydrological extremes and human-centered impacts.

Thesis Files