Research Interests

My research interests focus on the growth and decay of Rossby waves, the maintenance of storm tracks, as well as how they impact extreme weather events and how they respond to global warming in terms of both the underlying dynamical mechanisms and future projections. To achieve these research goals, I use simple conceptual models, such as a two-layer QG model and barotropical spectral model, to gain fundamental physical insights. I also look at the output of coupled climate models, such as CMIP models, to examine the impacts of climate change on storm tracks and extreme weather.

Selected Highlights:

a. Life Cycles of Baroclinic Waves in the Framework of Local Finite-Amplitude Wave Activity (LWA), Part I (in the Southern Hemisphere)

It is well known that synoptic eddies embedded in a westerly flow undergo downstream developments due to their dispersive nature. We examine the finite-amplitude aspects of downstream development with the budget of local wave activity (LWA), including explicit contributions from diabatic heating. LWA captures individual troughs/ridges as well as the envelope wave packet, and its column budget affords simplified interpretations. In the LWA framework, (linear) downstream development demonstrated in previous analyses is represented by the LWA fluxes associated with the reference zonal flow plus those induced by the radiation of Rossby waves. In addition, convergence of nonlinear advective LWA flux, baroclinic sources at the lower boundary, meridional redistribution by eddy momentum flux, and diabatic sources and sinks complete the column budget of LWA. When applied to the life cycles of troughs within coherent wave packets in the Southern Hemisphere, the LWA budget reveals that individual troughs grow mainly through downstream development, convergence of nonlinear advective flux by eddies, and diabatic heating. Downstream development and divergence of nonlinear flux also dominate trough decay. Contributions from nonlinear advective eddy flux are large in the presence of a strong ridge either immediately upstream or downstream of the trough. Furthermore, anticyclonic components of advective LWA fluxes associated with the upstream or downstream ridge transfer LWA into or out of the trough. Diabatic contributions are significant when the heating exhibits a tilted vertical structure that gives rise to enhanced vertical gradient in heating. Besides certain level of consistency with the EKE budget analysis, one of the most important advantages of LWA budget is that it can distinguish between linear and nonlinear dispersion of wave activity, and thus separate the nonlinear physical process from the linear process.

b. Life Cycles of Baroclinic Waves in the Framework of Local Finite-Amplitude Wave Activity (LWA), Part II (in moist two-layer QG model)

Using the dynamical diagnostic method of local finite-amplitude wave activity (LWA) and a moist two-layer quasi-geostrophic (QG) model, we investigate the impacts of diabatic heating on the mean states, the mean LWA budget, as well as statistical features and life cycles of baroclinic waves. Despite the overall positive LWA tendency due to diabatic heating, the total generation (baroclinic plus diabatic) of LWA decreases due to the impact of precipitation reducing the baroclinicity of the mean state, and thus eddies are weaker in the moist model compared to the dry model forced with the same radiative forcing. On average, troughs are stronger than ridges in the dry model. However, this ceases to hold in the moist model due to the positive contributions of diabatic heating to the growth of ridges. Diabatic heating is found to generate negative potential vorticity in the upper layer and to be the dominant term for the growth of many ridges, consistent with what we found based on reanalysis data. However, unlike the result of reanalysis data that diabatic heating can be the dominant term responsible for the growth of some troughs, troughs in the moist two-layer QG model are unlikely to grow mainly due to diabatic heating, largely because of the lack of vertical structure in the diabatic heating profile in the two-layer model. Nevertheless, diabatic heating can still contribute to the growth of some troughs due to the non-local nature of its impact under the LWA framework.

The impacts of diabatic heating on troughs and ridges are examined in the nonlinear regime using the framework of local wave activity. As the moisture content in the moist two-layer idealized quasi-geostrophic model increases, the instability of the mean state decreases. Thus, the mean wave activity weakens despite the positive contribution of diabatic heating. Latent heating generates anticyclonic wave activity but decreases cyclonic wave activity in the upper layer of the moist model, reducing the asymmetry between troughs and ridges. Diabatic heating can be the dominant term for the growth of ridges in the upper layer, but not for troughs. Moreover, the nonlocal effect of precipitation on baroclinic waves is demonstrated in a mathematically explicit and exact way since LWA is nonlocal in latitude.

c. Local Finite-Amplitude Wave Activity of Water Vapor as a Diagnostic of Atmospheric River Events

We generalize the formalism of local wave activity of quasi-geostrophic potential vorticity to water vapor, defined as LWA-V, and derive the LWA-V budget equation. LWA-V measures the waviness of moisture contours from the hypothetical zonal symmetry in the eddy-free atmosphere. It delineates well the northward moisture intrusion and the filamentary feature of AR events. In climatology, the LWA-V budget is maintained by the LWA-V tendency due to moisture flux convergence, as the local gain, and the LWA-V tendency due to precipitation and evaporation, as the net loss. Compared with the traditional moisture budget, the climatological LWA-V and LWA-V budget terms are more consistent in structure and are able to delineate the climatological AR. Moisture flux convergence is the dominant process for the intensification and movement of the LWA-V center associated with the AR event while the combination effect of evaporation and precipitation is mainly a sink of LWA-V during the entire life cycle of the AR, which is consistent with traditional moisture budget analysis. Furthermore, utilizing the Lagrangian aspect of LWA-V, we demonstrate, in a quantitative way, that the original latitudes of the largest precipitable water of the AR event are in the tropical region, much larger than the climatological mean value.

d. Predictability of El Niño with the Conditional Nonlinear Optimal Perturbation (CNOP) Approach

The conditional nonlinear optimal perturbation (CNOP) approach is developed to find the optimal initial perturbation or parameter perturbation of a model, or the optimal combined mode of initial perturbations and model parameter perturbations, within given constrains. Mathematically, it solves a nonlinear constrained maximization problem by introducing Lagrangian multipliers to the first-order variational of the objective function. The CNOP of initial perturbations is a natural generalization of linear singular vector to nonlinear regime and has been used in stability, sensitivity, and predictability studies. The optimal mode will have the largest nonlinear evolution at a prediction time, and thus provide an estimation of the upper bound of the predictability limit. CNOP can also be used to determine the relative roles of initial errors and parameter errors in model prediction errors of various weather and climate events, such as Typhoon, ENSO, and ocean’s thermocline circulations. For example, we find that the parameter errors play only a trivial role in the occurrence of “spring predictability barrier” for El Niño events, indicating that the forecast skill of ENSO could be greatly increased by improving the initialization of the forecast model. In another work, we demonstrate that, with the Zebiak-Cane model, the CNOP-type precursors are highly likely to evolve into El Niño events and tend to exhibit the peak-phase locking of El Niño at the correct time (i.e., end of the year). Furthermore, by contrasting to the evolution of El Niño events from the same CNOP-type precursors in the linearized Zebiak-Cane model, we reveal that nonlinearities (e.g., nonlinear temperature advections, as well as the induced anomalous SST differences and westerlies) play a crucial role in leading El Niño events to peak at the end of the year.