Research

Geophysical turbulence is highly dimensional, nonlinear, and multi-scale. Scales of motion range from planetary waves with lengths in the order of 10,000 km to millimetre sized turbulence. It is chaotic, whilststill comprising of large-scale three-dimensional coherent structures with significant temporal and spatial correlations. Our ability to understand and predict such physical phenomena benefits from reducing thesesystems to their most basic building blocks. As Einstein famously put it in his 1933 lecture,“It can scarcely be denied that the supreme goal of all theory is to make the irreducible basic elements as simple and as few aspossible without having to surrender the adequate representation of a single datum of experience"[4]. This quote has commonly been paraphrased to state that a model should be as simple as possible but no simpler. One might consider reduced-order modelling as a mathematical representation of this statement. In this line of research I have adopted reduced-order modelling and machine learning approaches to develop computationally rapid and cheap simulations of geophysical flows. In the context of climate science, such models are referred to as climate emulators.

Key References - Climate Emulators

Kitsios, V., Cordier, L. & O'Kane, T., 2024, Proper orthogonal decomposition reduced order model of the global ocean, Theoretical and Computational Fluid Dynamics. https://doi.org/10.1007/s00162-024-00719-9 

Interview with reporter at Nature magazine, Carissa Wong, on machine learning in climate and weather, 16 February, 2024. https://www.nature.com/articles/d41586-024-00780-8 

Kitsios, V., O'Kane, T.J. & Newth, D., 2023, A machine learning approach to rapidly project climate responses under a multitude of net-zero emission pathways, Nature Communications Earth and Environment, Vol 4., 355. https://doi.org/10.1038/s43247-023-01011-0 

Key References - Coherent Structures

O'Kane, T.J., Kitsios, V. & Collier, M., 2021, On the semiannual formation of large scale three-dimensional vortices at the stratopause, Geophysical Review Letters, Vol. 48, e2020GL090072. doi. org/10.1029/2020GL090072

Kitsios, V., O'Kane, T.J., & Zagar, N., 2019, A reduced order representation of the Madden-Julian oscillation based on reanalyzed normal mode coherences, Journal of the Atmospheric Sciences, Vol. 76, pp. [link]

My most recent research direction has been motivated by an adage from ecologoical economics: "How can we meet the needs and desires of humans today whilst not reducing our capacity to do so into the future?" To contribute to answering this question I have focussed on developing data-driven means to learn the influence that the climate has upon human systems. The physical and socio-economic environments in which we live are intrinsically linked over a wide range of scales. Geophysical phenomena have spatial scales ranging from waves that span the entire globe, down to millimetre sized turbulence.  Likewise socio-economic and market impacts are felt on a range of spatial scales from global factors of financial instability of the world's banking sector, down to the environmental hazard risk of individual assets. Despite the inherent nonlinear multi-scale complexity of both the physical and human systems, for certain spatio-temporal scales there are causal relationships between the two. In this line of research I have been developing data driven methods (e.g. econometric, timeseries analysis, machine learning) of determining the socio-economic and financial risks associated with climate variability and change. Applications span timescales from multi-year forecasts (initial value problem) to multi-decade projections (boundary value problem). Application areas thus far have included the influence that El Niño and La Niña have on the financial markets of certain agricultural commodities, and also the influence of climate change on the intensive care unit admissions in hospitals across Australia.

Key References

Kitsios, V., Causal inference and prediction of climate-amplified food-security induced conflict, in session “Harmony in the digital age: Exploring AI-Powered Paths to Peace Building”, 19th Jeju Forum for Peace & Prosperity, Jeju, South Korea, 29-31 May, 2024. http://www.cifaljeju.org/bbs/board.php?bo_table=sub05_3&wr_id=188  http://jejuforum.or.kr/en/ https://www.youtube.com/watch?v=FR9Sei5dvAw&t=0s

Yahyaei, H.,  Kitsios, V. & De Mello, L., 2024, The Impacts of the El Niño-Southern Oscillation on Global Food Security An Implied Volatility Approach, Journal of Climate Finance, Vol. 7, 100038.  https://doi.org/10.1016/j.jclimf.2024.100038

O’Kane, T.J., Scaife, A.A., Kushnir, Y., Brookshaw, A., Buontempo, C., Carlin, D., Connell, R.K., Doblas-Reyes, F., Dunstone, N., Förster, K., Graca, A., Hobday, A.J., Kitsios, V., van der Laan, L., Lockwood, J., Merryfield, W.J., Paxian, A., Payne, M.R., Reader, M.C., Saville, G.R., Smith, D., Solaraju-Murali, B., Caltabiano, N., Carman, J., Hawkins, E., Keenlyside, N., Kumar, A., Matei, D., Pohlmann, H., Power, S., Raphael, M., Sparrow, M., & Wu, B.,  2023, Recent applications and potential of near-term (interannual to decadal) climate predictions, Frontiers in Climate, Vol. 5, 1121626. https://www.frontiersin.org/articles/10.3389/fclim.2023.1121626  

Poon, E.K.W.*, Kitsios, V.*, Pilcher, D., Bellomo R. & Raman, J., 2023, Projecting future climate impact on national Australian respiratory-related intensive care unit demand, Heart Lung and Circulation, Vol. 32(1), pp 95-104, * - equal first authorship. https://doi.org/10.1016/j.hlc.2022.12.001

Kitsios, V., De Mello, L. & Matear, R., 2022, Forecasting commodity returns by exploiting climate model forecasts of the El Niño Southern Oscillation, Environmental Data Science, Vol. 1, e7, pp 1-16. doi.org/10.1017/eds.2022.6

Whitten, S., Verikios, G., Kitsios, V., Mason-D’Croz, D., Cook, S.L. & Holt, P., 2022, Exploring Climate Risk in Australia: The economic implications of a delayed transition to net zero emissions for the finance sector, CSIRO. [PDF]

Squire, D. T. Richardson, D., Risbey, J. S., Black, A. S., Kitsios, V., Matear, R. J., Monselesan, D., Moore, T. S. & Tozer, C. R., 2021, Unprecedented compound climate extremes and Australia’s 2019/2020 megafires, Nature Portfolio Journal Climate and Atmospheric Science, Vol. 4, 64. https://www.nature.com/articles/s41612-021-00220-8 

The development of a numerically simulated geophysical forecasting system is intrinsically linked to the forecast time horizon of interest. Climate change projections are predominantly a boundary value problem (BVP), in that the determination of the system is primarily due to the specified boundary conditions (e.g. carbon dioxide emissions) associated with prescribed scenarios of future economic development. The initial conditions today are largely irrelevant for the properties of the Earth a century from now. Weather (lead time days) and seasonal (lead time months) prediction can be considered to be predominantly initial value problems (IVP), in which how well one knows the initial conditions is a key factor contributing to forecast skill. In decadal prediction (lead time years), both initial and boundary conditions are arguably important for many of the spatial scales. One only has the potential to forecast physical phenomena with time scales commensurate with the forecast time horizon of interest.

We developed Climate Analysis Forecast Ensemble (CAFE) system specifically for decadal climate forecasting and reanalysis. The CAFE system manages 96 simultaneous numerical simulations of the climate system each starting from slightly different initial conditions (and certain parameters). It uses data assimilation to update (or correct) the simulations on the basis of a comprehensive network of real world ocean, atmosphere and sea-ice observations. In general, data assimilation provides the ability to modify imperfect simulations of reality with a series of incomplete and possibly noisy measurements, which ideally results in a better representation of the true system state. The CAFE system does so via the ensemble transform Kalman filter (ETKF) algorithm. We used this system to generate an ensemble reanalysis representing the past and current climate, and ensemble forecasts (and hindcasts) as predictions of the future climate. These ensemble datasets have extensive diagnostics output daily, including fully three-dimensional fields of atmospheric, oceanic, sea-ice and bio-geo-chemical variables. This means that the full probability density function of all of these quantities are available every day. 

Key References - Data Assimilation and Parameter Estimation

Kitsios, V., Sandery, P.,  O’Kane, T.J. & Fiedler, R., 2021, Ensemble Kalman filter parameter estimation of ocean optical properties for reduced biases in a coupled general circulation model, Journal of Advances in Modeling the Earth System, Vol. 13, Issue 2, e2020MS002252. dx.doi.org/10.1029/2020MS002252

O’Kane, T.J., Sandery, P.A., Kitsios, V., Sakov, P., Matear, R.J., Chamberlain, M.A., Collier, M.A., Fiedler, R., Chapman, C., Moore, T.S. & Sloyan, B., 2021, CAFE60v1: A 60-year large ensemble climate reanalysis. Part I: System design, model configuration and data assimilation, Journal of Climate, Vol. 34 (13), pp 5153-5169. doi.org/10.1175/JCLI-D-20-0974.1

O’Kane, T.J., Sandery, P.A., Kitsios, V., Sakov, P., Matear, R.J., Chamberlain, M.A., Squire, D.T., Collier, M.A., Chapman, C., Fiedler, R., Harries, D., Moore, T.S., Richardson, D., Risbey, J.S., Schroeter, B.J.E., Schroeter, S., Sloyan, B., Tozer, C., Watterson, I.G., Black, A. & Quinn, C., 2021, CAFE60v1: A 60-year large ensemble climate reanalysis. Part II: Evaluation, Journal of Climate, Vol. 34 (13), pp 5171-5194. https://doi.org/10.1175/JCLI-D-20-0518.1 .

Sandery, P., O'Kane, T.J., Kitsios, V. & Sakov, P., 2020, State estimation of the climate system with the EnKF using variants of coupled data assimilation, Monthly Weather Review, Vol. 148, pp 2411-2431. doi.org/10.1175/MWR-D-18-0443.1

Quinn, C., O’Kane, T. J. & Kitsios, V., 2020, Application of local attractor dimension to reduced space strongly coupled data assimilation for chaotic multiscale systems, Nonlinear Processes in Geophysics, Vol. 27, pp 51-74. [link]

Key References - Ensemble Forecasting

Hermanson, L., Smith, D., Seabrook, M., Bilbao, R., Doblas-Reyes, F., Tourigny, E., Lapin, V., Kharin, V., Merryfield, W., Sospedra-Alfonso, R., Athanasiadis, P., Nicoli, D., Gualdi, S., Dunstone, N., Eade, R., Scaife, A., Collier, M., O’Kane, T., Kitsios, V., Sandery, P., Pankatz, K., Pohlmann, H., Muller, W., Kataoka, T., Tatebe, H., Ishii, M., Imada, Y., Kruschke, T.,  Koenigk, T., Karami, M., Yang, S. Tian, T., Zhang, L., Delworth, T., Yang, X., Zeng, F., Wang, Y., Counillon, F., Keenlyside, N., & Bethke, I., Lean, J., Luterbacher, J., Kolli, R. & Kumar, A., 2022, WMO Global Annual to Decadal Climate Update: A prediction for 2021-2025, Bulletin of the American Meteorological Society, Vol. 103(4), E1117-E1129, doi.org/10.1175/BAMS-D-20-0311.1

Collier, M., O’Kane, T.J., Kitsios, V. & Sandery, P.A., 2022, CSIRO CAFE-60 Submissions to the World Meteorological Organisation Operational Decadal Forecasts and the International Multi-Model Data Exchange, Journal of Southern Hemisphere Earth Systems Science, doi.org/10.1071/ES21024

Watterson, I.G., O’Kane, T.J., Kitsios, V. & Chamberlain, M.A., 2021, Australian rainfall anomalies and Indo-Pacific Driver Indices: links and skill in two-year forecasts, Journal of Southern Hemisphere Earth Systems Science, Vol. 71 (3), pp 303-319. https://www.publish.csiro.au/es/ES21008 

O’Kane, T.J., Squire, D. T., Sandery, P.A., Kitsios, V., Matear, R.J., Moore, T.S., Risbey, J.S. & Watterson, I., 2020, Enhanced ENSO prediction via augmentation of multi-model ensembles with initial thermocline perturbations, Journal of Climate, Vol. 33, pp 2281-2293. [link]

In geophysical and engineering flows it is not possible to resolve all of the scales of motion, so one must resolve the large eddies are explicitly on a computational grid, and parameterise the interactions with the unresolved subgrid-scales. If these subgrid interactions are not properly parameterised, then an increase in resolution will not necessarily increase the accuracy of the resolved scales, hence the dependence on resolution. All simulation codes to date including the most sophisticated general circulation models suffer from this problem. This has wide ranging implications for geophysical research and operational activities, including weather / decadal / climate prediction. I solved the resolution dependence problem via the development of stochastic subgrid turbulence parameterisations. The model coefficients are determined from the subgrid statistics of higher resolution reference simulations. For regimes in which the kinetic energy is a simple function of scale, so too are the subgrid coefficients. I derived scaling laws for the subgrid coefficients, such that high resolution reference simulations are no longer required. I also proposed a unification of these scaling laws for the atmosphere and ocean on the basis of the enstrophy flux, Rossby wavenumber, and wavenumber of the energy containing scales. I have subsequently also successfully applied this technique to three-dimensional turbulent boundary layer flows, and geophysical flows with realistic topography.

Key References - Atmosphere

Kitsios, V. & Frederiksen, J.S., 2019, Subgrid parameterizations of eddy-eddy, eddy-meanfield, eddy-topographic, meanfield-meanfield and meanfield-topographic interactions in atmospheric models, Journal of the Atmospheric Sciences, Vol. 76, pp 457-477. [link]

Frederiksen, J.S., Dix, M.R., Osbrough, S. L. & Kitsios, V., 2015, Subgrid parameterisations for primitive equation atmospheric models, ANZIAM J., Vol. 56, pp C83-C100. [link]

Kitsios, V., Frederiksen, J.S. & Zidikheri, M.J., 2012, Subgrid model with scaling laws for atmospheric simulations, Journal of the Atmospheric Sciences, 69, pp 1427-1445.  [link] [animation]

Kitsios, V., Frederiksen, J.S. & Zidikheri, M.J., 2011, Subgrid parameterizations for high resolution atmospheric flows, ANZIAM J., Vol. 52, C271–C286. [link]

Key References - Ocean

Kitsios, V., Frederiksen, J.S. & O’Kane, T.J., 2023, Subgrid parameterization of eddy, meanfield and topographic interactions in simulations of an idealized Antarctic Circumpolar Current, Journal of Advances in the Modelling the Earth Systems, Vol. 15, e2022MS003412. https://doi.org/10.1029/2022MS003412 

Kitsios,V., Frederiksen, J.S. & Zidikheri, M.J., 2014, Scaling laws for parameterisations of subgrid interactions in simulations of oceanic circulations, Philosophical Transactions of the Royal Society A, Vol. 372, 20130285.  [link] [animation]

Kitsios, V., Frederiksen, J.S. & Zidikheri, M.J., 2013, Scaling laws for parameterisations of subgrid eddy-eddy interactions in simulations of oceanic circulations, Ocean Modelling, 68, pp 88-105.  [link]

Kitsios, V., Frederiksen, J.S. & Zidikheri, M.J., 2013, Subgrid parameterisation of the eddy-meanfield interactions in a baroclinic quasi-geostrophic ocean, ANZIAM J., Vol. 54, C394-C410. [link]

Key References - Wall Bounded Turbulence

Kitsios, V., Sillero, J.A., Frederiksen, J.S. & Soria, J., 2017, Scale and Reynolds number dependence of stochastic subgrid energy transfer in turbulent channel flow, Computers and Fluids, Vol. 151, pp 132-143. [link]

Kitsios, V., Sillero, J.A., Frederiksen, J.S. & Soria, J., 2015, Proposed stochastic parameterisations of subgrid turbulence in large eddy simulations of turbulent channel flow, Journal of Turbulence, Vol. 16, pp 729-741. [link]

Kitsios, V., Sillero, J.A., Soria, J. &  Frederiksen, J.S., 2014, Stochastic self-energy subgrid model for the large eddy simulation of turbulent channel flow, Journal of Physics, Vol. 506, 012001, 15pp.  [link] [animation]

Key References - Unification and Reviews

Frederiksen, J.S., Kitsios, V. & O'Kane, T., 2024, Statistical Dynamics and Subgrid Modelling of Turbulence: From Isotropic to Inhomogeneous, Atmosphere, Vol. 15, 921. https://doi.org/10.3390/atmos15080921 

Berner, J., Achatz, U., Lauriane, B., De La Camara, A., Christensen, H. Colangeli, M., Coleman, D., Crommelin, D., Dolaptchiev, S.I., Franzke, C.L.E., Friederichs, P., Imkeller, P., Jarvinen, H., Juricke, S., Kitsios, V., Lott, F., Lucarini, V., Mahajan, S., Palmer, T.N., Penland, C., Sakradzija, M., Von Storch, J.-S., Weisheimer, A., Weniger, M., Williams, P.D. &  Yano, J.-I., 2017, Stochastic Parameterization: Towards a new view of Weather and Climate Models, Bulletin of the American Meteorological Society, Vol. 98, pp 565–588 . [link]

Frederiksen, J.S., Kitsios, V., O'Kane, T.J. & Zidikheri, M.J., 2017, Stochastic subgrid modelling for geophysical and three-dimensional turbulence, Nonlinear and Stochastic Climate Dynamics, Cambridge University Press, pp 241-275. [link]

Kitsios, V., Frederiksen, J.S. & Zidikheri, M.J., 2016, Theoretical comparison of subgrid turbulence in atmospheric and oceanic quasi-geostrophic models, Nonlinear Processes in Geophysics, Vol. 23, pp 95-105. [link]

To understand the fundamental physics of separated flows geometrical effects such as those in an aerofoil must be removed. The most appropriate canonical flow is the self-similar APG TBL, in which the scaled statistics are independent of streamwise position. The difficulty the desired flow field is known, but not the required boundary conditions (BC). I derived the BCs by assuming a potential flow farfield, with corrections made accounting for the boundary layer growth. From an analysis of the statistics I developed a physical model of how the mean field, Reynolds stresses, and vortex structures changes from being less like a boundary layer and more like a free shear layer as one approaches separation. Such a data set is ideal for further interrogation, and development of subgrid turbulence models. The image is of instantaneous turbulent structures for a zero pressure gradient TBL (left, green) and the APG TBL (right, red). The mean flow direction is into the page.

This research was undertaken at Monash University, and in conjunction with the Universidad de Polytechnica de Madrid, where I developed the world’s largest parallel direct numerical simulation (DNS) of an adverse pressure gradient (APG) turbulent boundary layer (TBL), consisting of 6 billion grid points and parallelised over 32768 cores. The code is written in Fortran, parallelised using both OpenMP and MPI, with parallel I/O enabled using the HDF5 library. 

Key References

Senthil, S., Kitsios, V., Sekimoto, A., Atkinson, C. & Soria, J., 2020, Analysis of the factors contributing to the skin friction coefficient in adverse pressure gradient turbulent boundary layers and their variation with the pressure gradient, International Journal of Heat and Fluid Flow, Vol. 80, pp 108531-13. [link]

Kitsios, V., Sekimoto, A., Atkinson, C., Sillero, J.A., Borrell, G., Gungor, A.G., Jiménez, J. & Soria, J., 2017, Direct numerical simulation of a self-similar adverse pressure gradient turbulent boundary layer at the verge of separation, Journal of Fluid Mechanics, Vol. 829, pp 392-419. [link]

Kitsios, V., Atkinson, C., Sillero, J.A., Borrell, G., Gungor, A.G., Jiménez, J. & Soria, J., 2017, Direct numerical simulation of a self-similar adverse pressure gradient turbulent boundary layer, International Journal of Heat and Fluid Flow, Vol. 61, Part A, pp 129-136. [link]

Buchner, A.-J., Lozano-Duran, A., Kitsios, V., Atkinson, C. &  Soria, J., 2016, Local topology via the invariants of the velocity gradient tensor within vortex clusters and intense Reynolds stress structures in turbulent channel flow, Journal of Physics, Vol. 708, 012005, 14pp. [link]

Soria, J., Kitsios, V. & Atkinson, C., 2016, On the identification of intense Reynolds stress structures in wall-bounded flows using information-limited two-dimensional planar data, European Journal of Mechanics B / Fluids, Vol. 55, pp 279-285 [link]

Real time data assimilation and control of aerofoil flows can delay separation, and hence improve the energy generation of wind turbines, or reduce the fuel consumption of aircraft. To undertake the data assimilation in real-time the order and complexity of the numerical simulation must be minimised. One approach is to develop a reduced order model of the full physical system to represent the fluctuations of a given state and the transition between states (i.e. from controlled to uncontrolled). To this end I: 

• generated a reference database via the large eddy simulation of a canonical aerofoil flow (instantaneouos vortex structures illustrated below);

• calculated empirical orthogonal functions (EOF) from this reference database and developed an associated reduced order model capturing the fluctuating physics; and

• calculated linear stability modes to determining the sensitivity of the turbulent flows to perturbation (i.e. control force) and transition between states .

The above research was undertaken during my prizing winning collaborative Cotutelle PhD with The University of Melbourne (Australia) and the Université de Poitiers (France), on the numerical simulation and model reduction of aerofoil flows. This project was also in conjunction with Stanford University (USA) and Monash University (Australia).

Key References

Kitsios, V., Cordier, L., Bonnet, J.-P., Ooi, A. & Soria, J., 2011, On the coherent structures and stability properties of a leading edge separated aerofoil with turbulent recirculation, Journal of Fluid Mechanics, Vol. 683, pp 395-416.  [link] 

Kitsios, V., Cordier, L., Bonnet, J.-P., Ooi, A. & Soria, J., 2010, Development of a nonlinear eddy viscosity closure for the stability analysis of a turbulent channel flow, Journal of Fluid Mechanics, Vol. 664, pp 74-107. [link]

Kitsios, V., 2010, Recovery of fluid mechanical modes in unsteady separated flows, PhD Thesis, The University of Melbourne.  [link] 

Kitsios, V., Rodríguez, D., Theofilis, V., Ooi, A. & Soria, J., 2009, BiGlobal stability analysis in curvilinear coordinates for massively separated lifting bodies, Journal of Computational Physics, Vol. 228, pp 7181-7196. [link] [animation]

Mathis, R., Duke, D., Kitsios, V. & Soria, J., 2008, Use of Zero-Net-Mass-Flow for separation control in diffusing S-duct, Experimental Thermal Fluid Science, Vol. 33, pp 169-172. [link]