Research

Last updated: January 2019

Our research is focused on these four inter-connected themes:
1- Reduced-order modeling of turbulent flows
2- Geophysical fluid dynamics & Climate dynamics 
3- Extreme weather events
4- Data science and machine learning

1- Reduced-order modeling of turbulent flows
We work on developing and using data-driven and physics-based reduced-order model (ROMs) for high-dimensional turbulent flows. We are particularly interested in buoyancy-driven turbulence which is important for environmental flows (e.g., atmosphere and ocean) and energy systems (e.g., wind farms and heating/cooling in buildings). We develop and use Direct Numerical Simulations (DNS) and Large-Eddy Simulations (LES) to simulate the turbulent flows, and currently focus on reduced-order modeling techniques that are based on Fluctuation-Dissipation Theorem (FDT), Green’s functions, Koopman operator theory, and Dynamic Mode Decomposition (DMD). Our goal is to find data-driven methods that can accurately extract dynamically relevant modes of high-dimensional turbulent systems and can compute ROMs for predicting the long-term and short-term spatio-temporal evolution of such systems.

Our recent work in this area includes
1.1 Applying FDT to find a ROM for a global circulation model (GCM), see Hassanzadeh & Kuang (2016) (with a collaborator from Harvard University). In this study, we identified a major challenged in applying FDT to high-dimensional, non-normal systems.
1.2 Developing a DMD-enhanced FDT method to calculate a ROM for 3D Rayleigh-Benard turbulent convection, see Khodkar & Hassanzadeh (2018). The DMD-enhanced FDT method provides a remedy for the challenge mentioned in 1.1.
1.3 Applying the Green’s function to find a ROM for GCM, see Hassanzadeh & Kuang (2016) (with a collaborator from Harvard University). This accurate ROM is used to study problems in climate dynamics, see 3.2.
1.4 Applying the Green’s function to find a ROM for 3D Rayleigh-Benard turbulent convection, see Khodkar et al. (2018) (with collaborators from Mitsubishi Electric Research Lab).
1.5 Work on finding the flow fields that maximize heat transport in 2D Rayleigh-Benard turbulent convection using a variational formulation, see Hassanzadeh et al. (2014) (with collaborators from the University of Michigan and Univeristy of New Hampshire).

Our work in this area is supported by Rice University Creative Ventures and Mitsubishi Electric Research Lab.

2- Geophysical Fluid Dynamics & Climate dynamics
We work on developing a deeper understanding of the dynamics of the climate system and in particular the large-scale atmospheric circulation. Our goal is to improve weather predictions and climate change projections, especially for extreme events. We leverage the fundamentals of geophysical fluid dynamics and develop and use hierarchies of idealized to comprehensive numerical models combined with mathematical and statistical methods and machine learning techniques to gain a deeper understanding into various phenomena of the significantly complex climate system.

Our recent work in this area includes

2.1 Work on the dynamics of the annular modes and the role of eddy feedback, see Hassanzadeh & Kuang (2018) and Me at al. (2017) (with collaborators from Columbia University and Harvard University).
2.2 Work on understanding how the midlatitude circulation responds to Arctic warming and sea ice loss, see  Hassanzadeh et al. (2014) (with collaborators from Harvard University) and Ronalds et al. (2018) (with collaborators from Colorado State University).
2.3 Some thoughts on understanding the climate system using hierarchies of models, see Jeevanjee et al. (2017) (with collaborators from Princeton, UCLA, Caltech and Stanford University).
2.4 Work on investigating the effect of persistent patterns in the extratropical circulation on the onset of El Nino, see Anderson et al. (2017) (with collaborators from Boston University and Stockholm University).

Our work in this area is supported by NSF and NASA.

3- Extreme weather events
We work on dynamics, prediction, and climate change projection of extreme weather events such as heat waves, cold spells, droughts, and floods. We are particularly interested in the role of large-scale atmospheric circulation and persistent weather patterns in causing and affecting these extreme events. We develop and use a hierarchy of climate models from idealized models of atmospheric turbulence to comprehensive global climate models (GCMs) in addition to observational data. Our goal is to better predict these events and better project their future changes by developing a deeper understanding of their dynamics. We are also interested in the societal and environmental impacts of these extreme events.

Our recent work in this area includes
3.1 Work on why Hurricane Harvey stalled, how the large-scale circulation steers hurricanes over the Gulf of Mexico, and how future hurricanes in this region are expected to move.
3.2 Work on how persistent weather patterns, known as blocking events, which can cause devastating extreme events, are potentially affected by warming in the Arctic region and sea ice loss, see Hassanzadeh et al. (2014) and Hassanzadeh & Kuang (2015) (with collaborators from Harvard University). The latter work used the tool developed in 1.3.
3.3 Work on identifying and predicting large-scale weather patterns that cause extreme events.

Our work in this area is supported by NASA, The National Academies, and Rice Houston Engagement and Recovery Effort.

4- Data science and machine learning
We use a broad range of statistical and machine learning methods to study the high-dimensional spatio-temporal data from complex nonlinear dynamical systems such as the climate system or turbulent systems.

Our recent work in this area includes
4.1 Using convolutional neural networks (CNN) to classify weather patterns, see  Chattopadhyay et al. (2018)
4.2 Using recurrent neural networks (RNN) to predict extreme events in turbulent flows

Our work in this area is supported by Microsoft AI.