P. Dimitriadis, Hurst-Kolmogorov dynamics in hydroclimatic processes and in the microscale of turbulence, PhD thesis, Department of Water Resources and Environmental Engineering – National Technical University of Athens, 2017.
The high complexity and uncertainty of atmospheric dynamics has been long identified through the observation and analysis of hydroclimatic processes such as temperature, dew-point, humidity, atmospheric wind, precipitation, atmospheric pressure, river discharge and stage etc. Particularly, all these processes seem to exhibit high unpredictability due to the clustering of events, a behaviour first identified in Nature by H.E. Hurst in 1951 while working at the River Nile, although its mathematical description is attributed to A. N. Kolmogorov who developed it while studying turbulence in 1940. To give credits to both scientists this behaviour and dynamics is called Hurst-Kolmogorov (HK). In order to properly study the clustering of events as well as the stochastic behaviour of hydroclimatic processes in general we would require numerous of measurements in annual scale. Unfortunately, large lengths of high quality annual data are hardly available in observations of hydroclimatic processes. However, the microscopic processes driving and generating the hydroclimatic ones are governed by turbulent state. By studying turbulent phenomena in situ we may be able to understand certain aspects of the related macroscopic processes in field. Certain strong advantages of studying microscopic turbulent processes in situ is the recording of very long time series, the high resolution of records and the controlled environment of the laboratory. The analysis of these time series offers the opportunity of better comprehending, control and comparison of the two scientific methods through the deterministic and stochastic approach. In this thesis, we explore and further advance the second-order stochastic framework for the empirical as well as theoretical estimation of the marginal characteristic and dependence structure of a process (from small to extreme behaviour in time and state). Also, we develop and apply explicit and implicit algorithms for stochastic synthesis of mathematical processes as well as stochastic prediction of physical processes. Moreover, we analyze several turbulent processes and we estimate the Hurst parameter (H >> 0.5 for all cases) and the drop of variance with scale based on experiments in turbulent jets held at the laboratory. Additionally, we propose a stochastic model for the behaviour of a process from the micro to the macro scale that results from the maximization of entropy for both the marginal distribution and the dependence structure. Finally, we apply this model to microscale turbulent processes, as well as hydroclimatic ones extracted from thousands of stations around the globe including countless of data. The most important innovation of this thesis is that, to the Author’s knowledge, a unique framework (through modelling of common expression of both the marginal density distribution function and the second-order dependence structure) is presented that can include the simulation of the discretization effect, the statistical bias, certain aspects of the turbulent intermittent (or else fractal) behaviour (at the microscale of the dependence structure) and the long-term behaviour (at the macroscale of the dependence structure), the extreme events (at the left and right tail of the marginal distribution), as well as applications to 13 turbulent and hydroclimatic processes including experimentation and global analyses of surface stations (overall, several billions of observations). A summary of the major innovations of the thesis are: (a) the further development, and extensive application to numerous processes, of the classical second-order stochastic framework including innovative approaches to account for intermittency, discretization effects and statistical bias; (b) the further development of stochastic generation schemes such as the Sum of Autoregressive (SAR) models, e.g. AR(1) or ARMA(1,1), the Symmetric-Moving-Average (SMA) scheme in many dimensions (that can generate any process second-order dependence structure, approximate any marginal distribution to the desired level of accuracy and simulate certain aspects of the intermittent behaviour) and an explicit and implicit (pseudo) cyclo-stationary (pCSAR and pCSMA) schemes for simulating the deterministic periodicities of a process such as seasonal and diurnal; and (c) the introduction and application of an extended stochastic model (with an innovative identical expression of a four-parameter marginal distribution density function and correlation structure, i.e. g(x;C)=λ/[(1+|x/a+b|^c )]^d, with C=[λ,a,b,c,d]), that encloses a large variety of distributions (ranging from Gaussian to powered-exponential and Pareto) as well as dependence structures (such as white noise, Markov and HK), and is in agreement (in this form or through more simplified versions) with an interestingly large variety of turbulent (such as horizontal and vertical thermal jet of positively buoyancy processes using laser-induced-fluorescence techniques as well as grid-turbulence generated within a wind-tunnel), geostatistical (such as 2d rock formations), and hydroclimatic processes (such as temperature, atmospheric wind, dew-point and thus, humidity, precipitation, atmospheric pressure, river discharges and solar radiation, in a global scale, as well as a very long time series of river stage, and wave height and period). Amazingly, all examined physical processes (overall 13) exhibited long-range dependence and in particular, most (if treated properly within a robust physical and statistical framework, e.g. by adjusting the process for sampling errors as well as discretization and bias effects) with a mean long-term persistence parameter equal to H ≈ 5/6 (as in the case of isotropic grid-turbulence), and (for the processes examined in the microscale such atmospheric wind, surface temperature and dew-point, in a global scale, and a long duration discharge time series and storm event in terms of precipitation and wind) a powered-exponential behaviour with a fractal parameter close to M ≈ 1/3 (as in the case of isotropic grid-turbulence).
Full text (37297 KB)
Our works that reference this work:
|1.||P. Dimitriadis, and D. Koutsoyiannis, Stochastic synthesis approximating any process dependence and distribution, Stochastic Environmental Research & Risk Assessment, 32 (6), 1493–1515, doi:10.1007/s00477-018-1540-2, 2018.|
|2.||E. Klousakou, M. Chalakatevaki, P. Dimitriadis, T. Iliopoulou, R. Ioannidis, G. Karakatsanis, A. Efstratiadis, N. Mamassis, R. Tomani, E. Chardavellas, and D. Koutsoyiannis, A preliminary stochastic analysis of the uncertainty of natural processes related to renewable energy resources, Advances in Geosciences, 45, 193–199, doi:10.5194/adgeo-45-193-2018, 2018.|
|3.||P. Dimitriadis, K. Tzouka, D. Koutsoyiannis, H. Tyralis, A. Kalamioti, E. Lerias, and P. Voudouris, Stochastic investigation of long-term persistence in two-dimensional images of rocks, Spatial Statistics, 29, 177–191, doi:10.1016/j.spasta.2018.11.002, 2019.|
|4.||G.-F. Sargentis, P. Dimitriadis, R. Ioannidis, T. Iliopoulou, and D. Koutsoyiannis, Stochastic evaluation of landscapes transformed by renewable energy installations and civil works, Energies, 12 (4), 2817, doi:10.3390/en12142817, 2019.|
|5.||T. Iliopoulou, and D. Koutsoyiannis, Revealing hidden persistence in maximum rainfall records, Hydrological Sciences Journal, 64 (14), 1673–1689, doi:10.1080/02626667.2019.1657578, 2019.|
|6.||G.-F. Sargentis, P. Dimitriadis, and D. Koutsoyiannis, Aesthetical issues of Leonardo Da Vinci’s and Pablo Picasso’s paintings with stochastic evaluation, Heritage, 3 (2), 283–305, doi:10.3390/heritage3020017, 2020.|
|7.||T. Iliopoulou, and D. Koutsoyiannis, Projecting the future of rainfall extremes: better classic than trendy, Journal of Hydrology, 588, doi:10.1016/j.jhydrol.2020.125005, 2020.|
|8.||G.-F. Sargentis, T. Iliopoulou, S. Sigourou, P. Dimitriadis, and D. Koutsoyiannis, Evolution of clustering quantified by a stochastic method — Case studies on natural and human social structures, Sustainability, 12 (19), 7972, doi:10.3390/su12197972, 2020.|
|9.||N. Mamassis, A. Efstratiadis, P. Dimitriadis, T. Iliopoulou, R. Ioannidis, and D. Koutsoyiannis, Water and Energy, Handbook of Water Resources Management: Discourses, Concepts and Examples, edited by J.J. Bogardi, T. Tingsanchali, K.D.W. Nandalal, J. Gupta, L. Salamé, R.R.P. van Nooijen, A.G. Kolechkina, N. Kumar, and A. Bhaduri, Chapter 20, 617–655, doi:10.1007/978-3-030-60147-8_20, Springer Nature, Switzerland, 2021.|
|10.||G.-F. Sargentis, P. Dimitriadis, T. Iliopoulou, and D. Koutsoyiannis, A stochastic view of varying styles in art paintings, Heritage, 4, 21, doi:10.3390/heritage4010021, 2021.|
|11.||P. Dimitriadis, D. Koutsoyiannis, T. Iliopoulou, and P. Papanicolaou, A global-scale investigation of stochastic similarities in marginal distribution and dependence structure of key hydrological-cycle processes, Hydrology, 8 (2), 59, doi:10.3390/hydrology8020059, 2021.|