Civil Engineer, MSc, Dr. Engineer
montchrister@gmail.com
+30-2107722860
Maintenance, upgrading and extension of the Decision Support System for the management of the Athens water resource system
Duration: October 2008–November 2011
Budget: €72 000
Project director: N. Mamassis
Principal investigator: D. Koutsoyiannis
This research project includes the maintenance, upgrading and extension of the Decision Support System that developed by NTUA for EYDAP in the framework of the research project “Updating of the supervision and management of the water resources’ system for the water supply of the Athens’ metropolitan area”. The project is consisted of the following parts: (a) Upgrading of the Data Base, (b)Upgrading and extension of hydrometeorological network, (c) upgrading of the hydrometeorological data process software, (d) upgrading and extension of the Hydronomeas software, (e) hydrological data analysis and (f) support to the preparation of the annual master plans
G. Papacharalampous, H. Tyralis, D. Koutsoyiannis, and A. Montanari, Quantification of predictive uncertainty in hydrological modelling by harnessing the wisdom of the crowd: A large-sample experiment at monthly timescale, Advances in Water Resources, 136, 103470, doi:10.1016/j.advwatres.2019.103470, 2020.
Predictive hydrological uncertainty can be quantified by using ensemble methods. If properly formulated, these methods can offer improved predictive performance by combining multiple predictions. In this work, we use 50-year-long monthly time series observed in 270 catchments in the United States to explore the performances provided by an ensemble learning post-processing methodology for issuing probabilistic hydrological predictions. This methodology allows the utilization of flexible quantile regression models for exploiting information about the hydrological model's error. Its key differences with respect to basic two-stage hydrological post-processing methodologies using the same type of regression models are that (a) instead of a single point hydrological prediction it generates a large number of “sister predictions” (yet using a single hydrological model), and that (b) it relies on the concept of combining probabilistic predictions via simple quantile averaging. A major hydrological modelling challenge is obtaining probabilistic predictions that are simultaneously reliable and associated to prediction bands that are as narrow as possible; therefore, we assess both these desired properties of the predictions by computing their coverage probabilities, average widths and average interval scores. The results confirm the usefulness of the proposed methodology and its larger robustness with respect to basic two-stage post-processing methodologies. Finally, this methodology is empirically proven to harness the “wisdom of the crowd” in terms of average interval score, i.e., the average of the individual predictions combined by this methodology scores no worse –usually better− than the average of the scores of the individual predictions.
Additional material:
Works that cite this document: View on Google Scholar or ResearchGate
G. Papacharalampous, H. Tyralis, A. Langousis, A. W. Jayawardena, B. Sivakumar, N. Mamassis, A. Montanari, and D. Koutsoyiannis, Probabilistic hydrological post-processing at scale: Why and how to apply machine-learning quantile regression algorithms, Water, doi:10.3390/w11102126, 2019.
We conduct a large-scale benchmark experiment aiming to advance the use of machine-learning quantile regression algorithms for probabilistic hydrological post-processing “at scale” within operational contexts. The experiment is set up using 34-year-long daily time series of precipitation, temperature, evapotranspiration and streamflow for 511 catchments over the contiguous United States. Point hydrological predictions are obtained using the Génie Rural à 4 paramètres Journalier (GR4J) hydrological model and exploited as predictor variables within quantile regression settings. Six machine-learning quantile regression algorithms and their equal-weight combiner are applied to predict conditional quantiles of the hydrological model errors. The individual algorithms are quantile regression, generalized random forests for quantile regression, generalized random forests for quantile regression emulating quantile regression forests, gradient boosting machine, model-based boosting with linear models as base learners and quantile regression neural networks. The conditional quantiles of the hydrological model errors are transformed to conditional quantiles of daily streamflow, which are finally assessed using proper performance scores and benchmarking. The assessment concerns various levels of predictive quantiles and central prediction intervals, while it is made both independently of the flow magnitude and conditional upon this magnitude. Key aspects of the developed methodological framework are highlighted, and practical recommendations are formulated. In technical hydro-meteorological applications, the algorithms should be applied preferably in a way that maximizes the benefits and reduces the risks from their use. This can be achieved by (i) combining algorithms (e.g., by averaging their predictions) and (ii) integrating algorithms within systematic frameworks (i.e., by using the algorithms according to their identified skills), as our large-scale results point out.
Full text: http://www.itia.ntua.gr/en/getfile/2001/1/documents/water-11-02126.pdf (6451 KB)
Works that cite this document: View on Google Scholar or ResearchGate
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Comparison of stochastic and machine learning methods for multi-step ahead forecasting of hydrological processes, Stochastic Environmental Research & Risk Assessment, doi:10.1007/s00477-018-1638-6, 2019.
Research within the field of hydrology often focuses on the statistical problem of comparing stochastic to machine learning (ML) forecasting methods. The performed comparisons are based on case studies, while a study providing large-scale results on the subject is missing. Herein, we compare 11 stochastic and 9 ML methods regarding their multi-step ahead forecasting properties by conducting 12 extensive computational experiments based on simulations. Each of these experiments uses 2000 time series generated by linear stationary stochastic processes. We conduct each simulation experiment twice; the first time using time series of 100 values and the second time using time series of 300 values. Additionally, we conduct a real-world experiment using 405 mean annual river discharge time series of 100 values. We quantify the forecasting performance of the methods using 18 metrics. The results indicate that stochastic and ML methods may produce equally useful forecasts.
Remarks:
Supplementary information: https://doi.org/10.6084/m9.figshare.7092824.v1
Additional material:
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.
Determining the geophysical properties of rocks and geological formations is of high importance in many fields such as geotechnical engineering. In this study, we investigate the second-order dependence structure of spatial (two-dimensional) processes through the statistical perspective of variance vs. scale (else known as the climacogram) instead of covariance vs. lag (e.g. autocovariance, variogram etc.) or power vs. frequency (e.g. power spectrum, scaleogram, wavelet transform etc.) which traditionally are applied. In particular, we implement a two-dimensional (visual) estimator, adjusted for bias and for unknown process mean, through the (plot of) variance of the space-averaged process vs. the spatial scale. Additionally, we attempt to link the climacogram to the type of rocks and provide evidence on stochastic similarities in certain of their characteristics, such as mineralogical composition and resolution. To this end, we investigate two-dimensional spatial images of rocks in terms of their stochastic microstructure as estimated by the climacogram. The analysis is based both on microscale and macroscale data extracted from grayscale images of rocks. Interestingly, a power-law drop of variance vs. scale (or else known as long-term persistence) is detected in all scales presenting a similar power-exponent. Furthermore, the strengths and limitations of the climacogram as a stochastic tool are discussed and compared with the traditional tool in spatial statistics, the variogram. We show that the former has considerable strengths for detecting the long-range dependence in spatial statistics.
Remarks:
Share Link: https://authors.elsevier.com/c/1YJjr7su79fMuR
Additional material:
Works that cite this document: View on Google Scholar or ResearchGate
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Univariate time series forecasting of temperature and precipitation with a focus on machine learning algorithms: a multiple-case study from Greece, Water Resources Management, 32 (15), 5207–5239, doi:10.1007/s11269-018-2155-6, 2018.
We provide contingent empirical evidence on the solutions to three problems associated with univariate time series forecasting using machine learning (ML) algorithms by conducting an extensive multiple-case study. These problems are: (a) lagged variable selection, (b) hyperparameter handling, and (c) comparison between ML and classical algorithms. The multiple-case study is composed by 50 single-case studies, which use time series of mean monthly temperature and total monthly precipitation observed in Greece. We focus on two ML algorithms, i.e. neural networks and support vector machines, while we also include four classical algorithms and a naïve benchmark in the comparisons. We apply a fixed methodology to each individual case and, subsequently, we perform a cross-case synthesis to facilitate the detection of systematic patterns. We fit the models to the deseasonalized time series. We compare the one- and multi-step ahead forecasting performance of the algorithms. Regarding the one-step ahead forecasting performance, the assessment is based on the absolute error of the forecast of the last monthly observation. For the quantification of the multi-step ahead forecasting performance we compute five metrics on the test set (last year’s monthly observations), i.e. the root mean square error, the Nash-Sutcliffe efficiency, the ratio of standard deviations, the coefficient of correlation and the index of agreement. The evidence derived by the experiments can be summarized as follows: (a) the results mostly favour using less recent lagged variables, (b) hyperparameter optimization does not necessarily lead to better forecasts, (c) the ML and classical algorithms seem to be equally competitive.
Additional material:
Works that cite this document: View on Google Scholar or ResearchGate
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Predictability of monthly temperature and precipitation using automatic time series forecasting methods, Acta Geophysica, 66 (4), 807–831, doi:10.1007/s11600-018-0120-7, 2018.
We investigate the predictability of monthly temperature and precipitation by applying automatic univariate time series forecasting methods to a sample of 985 40-year long monthly temperature and 1552 40-year long monthly precipitation time series. The methods include a naïve one based on the monthly values of the last year, as well as the random walk (with drift), AutoRegressive Fractionally Integrated Moving Average (ARFIMA), exponential smoothing state space model with Box-Cox transformation, ARMA errors, Trend and Seasonal components (BATS), simple exponential smoothing, Theta and Prophet methods. Prophet is a recently introduced model inspired by the nature of time series forecasted at Facebook and has not been applied to hydrometeorological time series before, while the use of random walk, BATS, simple exponential smoothing and Theta is rare in hydrology. The methods are tested in performing multi-step ahead forecasts for the last 48 months of the data. We further investigate how different choices of handling the seasonality and non-normality affect the performance of the models. The results indicate that (a) all the examined methods apart from the naïve and random walk ones are accurate enough to be used in long-term applications, (b) monthly temperature and precipitation can be forecasted to a level of accuracy which can barely be improved using other methods, (c) the externally applied classical seasonal decomposition results mostly in better forecasts compared to the automatic seasonal decomposition used by the BATS and Prophet methods and (d) Prophet is competitive, especially when it is combined with externally applied classical seasonal decomposition
Additional material:
Works that cite this document: View on Google Scholar or ResearchGate
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, One-step ahead forecasting of geophysical processes within a purely statistical framework, Geoscience Letters, 5, 12, doi:10.1186/s40562-018-0111-1, 2018.
The simplest way to forecast geophysical processes, an engineering problem with a widely recognized challenging character, is the so-called “univariate time series forecasting” that can be implemented using stochastic or machine learning regression models within a purely statistical framework. Regression models are in general fast-implemented, in contrast to the computationally intensive Global Circulation Models, which constitute the most frequently used alternative for precipitation and temperature forecasting. For their simplicity and easy applicability, the former have been proposed as benchmarks for the latter by forecasting scientists. Herein, we assess the one-step ahead forecasting performance of 20 univariate time series forecasting methods, when applied to a large number of geophysical and simulated time series of 91 values. We use two real-world annual datasets, a dataset composed by 112 time series of precipitation and another composed by 185 time series of temperature, as well as their respective standardized datasets, to conduct several real-world experiments. We further conduct large-scale experiments using 12 simulated datasets. These datasets contain 24,000 time series in total, which are simulated using stochastic models from the families of AutoRegressive Moving Average and AutoRegressive Fractionally Integrated Moving Average. We use the frst 50, 60, 70, 80 and 90 data points for model-ftting and model-validation, and make predictions corresponding to the 51st, 61st, 71st, 81st and 91st respectively. The total number of forecasts produced herein is 2,177,520, among which 47,520 are obtained using the real-world datasets. The assessment is based on eight error metrics and accuracy statistics. The simulation experiments reveal the most and least accurate methods for long-term forecasting applications, also suggesting that the simple methods may be competitive in specifc cases. Regarding the results of the realworld experiments using the original (standardized) time series, the minimum and maximum medians of the absolute errors are found to be 68 mm (0.55) and 189 mm (1.42) respectively for precipitation, and 0.23 °C (0.33) and 1.10 °C (1.46) respectively for temperature. Since there is an absence of relevant information in the literature, the numerical results obtained using the standardized real-world datasets could be used as rough benchmarks for the one-step ahead predictability of annual precipitation and temperature
Full text: http://www.itia.ntua.gr/en/getfile/1834/1/documents/s40562-018-0111-1.pdf (3083 KB)
Works that cite this document: View on Google Scholar or ResearchGate
H. Tyralis, P. Dimitriadis, D. Koutsoyiannis, P.E. O’Connell, K. Tzouka, and T. Iliopoulou, On the long-range dependence properties of annual precipitation using a global network of instrumental measurements, Advances in Water Resources, 111, 301–318, doi:10.1016/j.advwatres.2017.11.010, 2018.
The long-range dependence (LRD) is considered an inherent property of geophysical processes, whose presence increases uncertainty. Here we examine the spatial behaviour of LRD in precipitation by regressing the Hurst parameter estimate of mean annual precipitation instrumental data which span from 1916-2015 and cover a big area of the earth’s surface on location characteristics of the instrumental data stations. Furthermore, we apply the Mann-Kendall test under the LRD assumption (MKt-LRD) to reassess the significance of observed trends. To summarize the results, the LRD is spatially clustered, it seems to depend mostly on the location of the stations, while the predictive value of the regression model is good. Thus when investigating for LRD properties we recommend that the local characteristics should be considered. The application of the MKt-LRD suggests that no significant monotonic trend appears in global precipitation, excluding the climate type D (snow) regions in which positive significant trends appear.
Remarks:
Supplementary information files are hosted at: https://doi.org/10.6084/m9.figshare.4892447.v1
Additional material:
Works that cite this document: View on Google Scholar or ResearchGate
H. Tyralis, and G. Papacharalampous, Variable selection in time series forecasting using random forests, Algorithms, 10, 114, doi:10.3390/a10040114, 2017.
Time series forecasting using machine learning algorithms has gained popularity recently. Random forest is a machine learning algorithm implemented in time series forecasting; however, most of its forecasting properties have remained unexplored. Here we focus on assessing the performance of random forests in one-step forecasting using two large datasets of short time series with the aim to suggest an optimal set of predictor variables. Furthermore, we compare its performance to benchmarking methods. The first dataset is composed by 16,000 simulated time series from a variety of Autoregressive Fractionally Integrated Moving Average (ARFIMA) models. The second dataset consists of 135 mean annual temperature time series. The highest predictive performance of RF is observed when using a low number of recent lagged predictor variables. This outcome could be useful in relevant future applications, with the prospect to achieve higher predictive accuracy.
Full text: http://www.itia.ntua.gr/en/getfile/1827/1/documents/Variable_Selection_in_Time_Series_Forecasting_Using_Random_Forests.pdf (5509 KB)
H. Tyralis, G. Karakatsanis, K. Tzouka, and N. Mamassis, Data and code for the exploratory data analysis of the electrical energy demand in the time domain in Greece, Data in Brief, 13 (700-702), doi:http://dx.doi.org/10.1016/j.energy.2017.06.074, 2017.
We present data and code for visualizing the electrical energy data and weather-,climate-related and socioeconomic variables in the time domain in Greece. The electrical energy data include hourly demand, weekly-ahead forecasted values of the demand provided by the Greek Independent Power Transmission Operator and pricing values in Greece. We also present the daily temperature in Athens and the Gross Domestic Product of Greece. The code combines the data to a single report, which includes all visualizations with combinations of all variables in multiple time scales. The data and code we reused in Tyralis et al.(2017)
Full text: http://www.itia.ntua.gr/en/getfile/1825/1/documents/DataAndCode.pdf (127 KB)
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Forecasting of geophysical processes using stochastic and machine learning algorithms, European Water, 59, 161–168, 2017.
We perform an extensive comparison between four stochastic and two machine learning (ML) forecasting algorithms by conducting a multiple-case study. The latter is composed by 50 single-case studies, which use time series of total monthly precipitation and mean monthly temperature observed in Greece. We apply a fixed methodology to each individual case and, subsequently, we perform a cross-case synthesis to facilitate the detection of systematic patterns. The stochastic algorithms include the Autoregressive order one model, an algorithm from the family of Autoregressive Fractionally Integrated Moving Average models, an Exponential Smoothing State Space algorithm and the Theta algorithm, while the ML algorithms are Neural Networks and Support Vector Machines. We also use the last observation as a Naïve benchmark in the comparisons. We apply the forecasting methods to the deseasonalized time series. We compare the one-step ahead as also the multi-step ahead forecasting properties of the algorithms. Regarding the one-step ahead forecasting properties, the assessment is based on the absolute error of the forecast of the last observation. For the comparison of the multi-step ahead forecasting properties we use five metrics applied to the test set (last twelve observations), i.e. the root mean square error, the Nash-Sutcliffe efficiency, the ratio of standard deviations, the index of agreement and the coefficient of correlation. Concerning the ML algorithms, we also perform a sensitivity analysis for time lag selection. Additionally, we compare more sophisticated ML methods as regards to the hyperparameter optimization to simple ones.
Full text: http://www.itia.ntua.gr/en/getfile/1768/1/documents/EW_2017_59_22.pdf (1163 KB)
See also: http://www.ewra.net/ew/issue_59.htm
Works that cite this document: View on Google Scholar or ResearchGate
K. Mavroyeoryos, I. Engonopoulos, H. Tyralis, P. Dimitriadis, and D. Koutsoyiannis, Simulation of electricity demand in a remote island for optimal planning of a hybrid renewable energy system, Energy Procedia, 125, 435–442, doi:10.1016/j.egypro.2017.08.095, 2017.
Here we simulate the electrical energy demand in the remote island of Astypalaia. To this end we obtain information regarding the local socioeconomic conditions and energy demand needs. The available hourly demand load data are analyzed at various time scales (hourly, weekly, daily, seasonal). The cross-correlations between the electricity demand load and the mean daily temperature are computed. An exploratory data analysis including all variables is performed to find hidden relationships. Finally, the demand is simulated. The simulation time series will be used in the development of a framework for planning of a hybrid renewable energy system in Astypalaia.
Full text: http://www.itia.ntua.gr/en/getfile/1735/1/documents/energy_demand_procedia.pdf (1370 KB)
Works that cite this document: View on Google Scholar or ResearchGate
H. Tyralis, and D. Koutsoyiannis, On the prediction of persistent processes using the output of deterministic models, Hydrological Sciences Journal, 62 (13), 2083–2102, doi:10.1080/02626667.2017.1361535, 2017.
A problem frequently met in engineering hydrology is the forecasting of hydrologic variables conditional on their historical observations and the hindcasts and forecasts of a deterministic model. On the contrary, it is a common practice for climatologists to use the output of general circulation models (GCMs) for the prediction of climatic variables despite their inability to quantify the uncertainty of the predictions. Here we apply the well-established Bayesian Processor of Forecasts (BPF) for forecasting hydroclimatic variables using stochastic models through coupling them with GCMs. We extend the BPF to cases where long-term persistence appears, using the Hurst-Kolmogorov process (HKp, also known as fractional Gaussian noise) and we investigate analytically its properties. We apply the framework to calculate the distributions of the mean annual temperature and precipitation stochastic processes for the time period 2016-2100 in the United States of America conditional on historical observations and the respective output of GCMs.
Full text: http://www.itia.ntua.gr/en/getfile/1727/1/documents/2017HSJ_OnTthePredictionOfPersistentProcesses.pdf (3152 KB)
Additional material:
Works that cite this document: View on Google Scholar or ResearchGate
Other works that reference this work (this list might be obsolete):
1. | Kundzewicz, Z. W., Quo vadis, hydrology?, Hydrological Sciences Journal, doi:10.1080/02626667.2018.1489597, 2018. |
H. Tyralis, G. Karakatsanis, K. Tzouka, and N. Mamassis, Exploratory data analysis of the electrical energy demand in the time domain in Greece, Energy, 134 (902-918), 16 pages, doi:10.1016/j.energy.2017.06.074 0360-5442, 2017.
The electrical energy demand (EED) in Greece for the time period 2002-2016 is investigated. The aim of the study is to introduce a framework for the exploratory data analysis (EDA) of the EED in the time domain. To this end, the EED at the hourly, daily, seasonal and annual time scale along with the mean daily temperature and the Gross Domestic Product (GDP) of Greece are visualized. The forecast of the EED provided by the Greek Independent Power Transmission Operator (IPTO) is also visualized and is compared with the actual EED. Furthermore, the EED pricing system is visualized. The results of the study in general confirm and summarize the conclusions of previous relevant studies in Greece, each one treating a single topic and covering shorter and earlier time periods. Furthermore, some unexpected patterns are observed, which if not considered carefully could result to dubious models. Therefore, it is shown that the EDA of the EED in the time domain coupled with weather-, climate-related and socio-economic variables is essential for the building of a model for the short-, medium- and long-term EED forecasting, something not highlighted in the literature.
Full text: http://www.itia.ntua.gr/en/getfile/1722/1/documents/EDA_electricity_2017.pdf (3406 KB)
A. Tegos, H. Tyralis, D. Koutsoyiannis, and K. H. Hamed, An R function for the estimation of trend signifcance under the scaling hypothesis- application in PET parametric annual time series, Open Water Journal, 4 (1), 66–71, 6, 2017.
We present an R function for testing the significant trend of time series. Te function calculates trend significance using a modified Mann-Kendall test, which takes into account the well-known physical behavior of the Hurst-Kolmogorov dynamics. Te function is tested at 10 stations in Greece, with approximately 50 years of PET data with the use of a recent parametric approach. A significant downward trend was detected at two stations. Te R software is now suitable for extensive use in several fields of the scientific community, allowing a physical consistent of a trend analysis.
Full text: http://www.itia.ntua.gr/en/getfile/1703/1/documents/2017OW_An_R_FunctionForTrendSignificance.pdf (326 KB)
Additional material:
See also: http://scholarsarchive.byu.edu/openwater/vol4/iss1/6/
Works that cite this document: View on Google Scholar or ResearchGate
H. Tyralis, A. Tegos, A. Delichatsiou, N. Mamassis, and D. Koutsoyiannis, A perpetually interrupted interbasin water transfer as a modern Greek drama: Assessing the Acheloos to Pinios interbasin water transfer in the context of integrated water resources management, Open Water Journal, 4 (1), 113–128, 12, 2017.
Interbasin water transfer is a primary instrument of water resources management directly related with the integrated development of the economy, society and environment. Here we assess the project of the interbasin water transfer from the river Acheloos to the river Pinios basin which has intrigued the Greek society, the politicians and scientists for decades. Te set of criteria we apply originate from a previous study reviewing four interbasin water transfers and assessing whether an interbasin water transfer is compatible with the concept of integrated water resources management. In this respect, we assess which of the principles of the integrated water resources management the Acheloos to Pinios interbasin water transfer project does or does not satisfy. While the project meets the criteria of real surplus and deficit, of sustainability and of sound science, i.e., the criteria mostly related to the engineering part, it fails to meet the criteria of good governance and balancing of existing rights with needs, i.e., the criteria associated with social aspects of the project. Te non-fulfillment of the latter criteria is the consequence of chronic diseases of the Greek society, which become obvious in the case study
Full text: http://www.itia.ntua.gr/en/getfile/1702/1/documents/2017OW_AcheloosToPiniosInterbasinWaterTransfer.pdf (2744 KB)
See also: http://scholarsarchive.byu.edu/openwater/vol4/iss1/11/
Works that cite this document: View on Google Scholar or ResearchGate
H. Tyralis, N. Mamassis, and Y. Photis, Spatial analysis of the electrical energy demand in Greece, Energy Policy, 102 (340-352), doi:10.1016/j.enpol.2016.12.033, March 2017.
The Electrical Energy Demand (EED) of the agricultural, commercial and industrial sector in Greece, as well as its use for domestic activities, public and municipal authorities and street lighting are analysed spatially using Geographical Information System and spatial statistical methods. The analysis is performed on data which span from 2008 to 2012 and have annual temporal resolution and spatial resolution down to the NUTS (Nomenclature of Territorial Units for Statistics) level 3. The aim is to identify spatial patterns of the EED and its transformations such as the ratios of the EED to socioeconomic variables, i.e. the population, the total area, the population density and the Gross Domestic Product (GDP). Based on the analysis, Greece is divided in five regions, each one with a different development model, i.e. Attica and Thessaloniki which are two heavily populated major poles, Thessaly and Central Greece which form a connected geographical region with important agricultural and industrial sector, the islands and some coastal areas which are characterized by an important commercial sector and the rest Greek areas. The spatial patterns can provide additional information for policy decision about the electrical energy management and better representation of the regional socioeconomic conditions.
Full text: http://www.itia.ntua.gr/en/getfile/1674/1/documents/electr_GIS_R_2016.pdf (2357 KB)
H. Tyralis, N. Mamassis, and Y. Photis, Spatial Analysis of Electrical Energy Demand Patterns in Greece: Application of a GIS-based Methodological Framework, Energy Procedia, 97 (262-269), 8 pages, doi:10.1016/j.egypro.2016.10.071, November 2016.
We investigate various uses of the Electrical Energy Demand (EED) in Greece (agricultural, commercial, domestic, industrial use) and we examine their relationships with variables such as population and the Gross Domestic Product. The analysis is performed on data from the year 2012 and have spatial resolution down to the level of prefecture. We both visualize the results of the analysis and we perform spatial cluster and outlier analysis. The definition of the spatial patterns of the aforementioned variables in a GIS environment provides insight of the regional development model in Greece.
Full text: http://www.itia.ntua.gr/en/getfile/1672/1/documents/SpatialAnalyis.pdf (1019 KB)
H. Tyralis, and D. Koutsoyiannis, A Bayesian statistical model for deriving the predictive distribution of hydroclimatic variables, Climate Dynamics, 42 (11-12), 2867–2883, doi:10.1007/s00382-013-1804-y, 2014.
Recent publications have provided evidence that hydrological processes exhibit a scaling behaviour, also known as the Hurst phenomenon. An appropriate way to model this behaviour is to use the Hurst-Kolmogorov stochastic process. The Hurst-Kolmogorov process entails high autocorrelations even for large lags, as well as high variability even at climatic scales. A problem that, thus, arises is how to incorporate the observed past hydroclimatic data in deriving the predictive distribution of hydroclimatic processes at climatic time scales. Here with the use of Bayesian techniques we create a framework to solve the aforementioned problem. We assume that there is no prior information for the parameters of the process and use a non-informative prior distribution. We apply this method with real-world data to derive the posterior distribution of the parameters and the posterior predictive distribution of various 30-year moving average climatic variables. The marginal distributions we examine are the normal and the truncated normal (for nonnegative variables). We also compare the results with two alternative models, one that assumes independence in time and one with Markovian dependence, and the results are dramatically different. The conclusion is that this framework is appropriate for the prediction of future hydroclimatic variables conditional on the observations.
Additional material:
See also: http://dx.doi.org/10.1007/s00382-013-1804-y
Works that cite this document: View on Google Scholar or ResearchGate
H. Tyralis, D. Koutsoyiannis, and S. Kozanis, An algorithm to construct Monte Carlo confidence intervals for an arbitrary function of probability distribution parameters, Computational Statistics, 28 (4), 1501–1527, doi:10.1007/s00180-012-0364-7, 2013.
We derive a new algorithm for calculating an exact confidence interval for a parameter of location or scale family, based on a two-sided hypothesis test on the parameter of interest, using some pivotal quantities. We use this algorithm to calculate approximate confidence intervals for the parameter or a function of the parameter of one-parameter continuous distributions. After appropriate heuristic modifications of the algorithm we use it to obtain approximate confidence intervals for a parameter or a function of parameters for multi-parameter continuous distributions. The advantage of the algorithm is that it is general and gives a fast approximation of an exact confidence interval. Some asymptotic (analytical) results are shown which validate the use of the method under certain regularity conditions. In addition, numerical results of the method compare well with those obtained by other known methods of the literature on the exponential, the normal, the gamma and the Weibull distribution.
Additional material:
See also: http://dx.doi.org/10.1007/s00180-012-0364-7
Works that cite this document: View on Google Scholar or ResearchGate
Other works that reference this work (this list might be obsolete):
1. | Campos, J. N.B., F. A. Souza Filho and H. V.C. Lima, Risks and uncertainties in reservoir yield in highly variable intermittent rivers: Case of the Castanhão Reservoir in semi-arid Brazil, Hydrological Sciences Journal, 59 (6), 1184-1195, 2014. |
H. Tyralis, and D. Koutsoyiannis, Simultaneous estimation of the parameters of the Hurst-Kolmogorov stochastic process, Stochastic Environmental Research & Risk Assessment, 25 (1), 21–33, 2011.
Various methods for estimating the self-similarity parameter (Hurst parameter, H) of a Hurst-Kolmogorov stochastic process (HKp) from a time series are available. Most of them rely on some asymptotic properties of processes with Hurst-Kolmogorov behaviour and only estimate the self-similarity parameter. Here we show that the estimation of the Hurst parameter affects the estimation of the standard deviation, a fact that was not given appropriate attention in the literature. We propose the Least Squares based on Variance estimator, and we investigate numerically its performance, which we compare to the Least Squares based on Standard Deviation estimator, as well as the maximum likelihood estimator after appropriate streamlining of the latter. These three estimators rely on the structure of the HKp and estimate simultaneously its Hurst parameter and standard deviation. In addition, we test the performance of the three methods for a range of sample sizes and H values, through a simulation study and we compare it with other estimators of the literature.
Additional material:
See also: http://dx.doi.org/10.1007/s00477-010-0408-x
Works that cite this document: View on Google Scholar or ResearchGate
Other works that reference this work (this list might be obsolete):
1. | Bakker, A. M. R., and B. J. J. M. van den Hurk, Estimation of persistence and trends in geostrophic wind speed for the assessment of wind energy yields in Northwest Europe, Climate Dynamics, 39 (3-4), 767-782, 2012. |
2. | Prass, T. S., J. M. Bravo, R. T. Clarke, W. Collischonn, and S. R. C. Lopes, Comparison of forecasts of mean monthly water level in the Paraguay River, Brazil, from two fractionally differenced models, Water Resour. Res., 48, W05502, doi: 10.1029/2011WR011358, 2012. |
3. | Bakker, A., J. Coelingh and B. van den Hurk, Long-term trends in the wind supply in the Netherlands, Proceedings EWEA 2012 Annual Event, Copenhagen, Denmark, 2012. |
4. | Navarro, X., F. Porée, A. Beuchée and G. Carrault, Performance analysis of Hurst exponent estimators using surrogate-data and fractional lognormal noise models: Application to breathing signals from preterm infants, Digital Signal Processing, 10.1016/j.dsp.2013.04.007, 2013. |
5. | Serinaldi, F., L. Zunino and O. Rosso, Complexity–entropy analysis of daily stream flow time series in the continental United States, Stochastic Environmental Research and Risk Assessment, 28 (7), 1685-1708, 2014. |
6. | Szolgayova, E., G. Laaha, G. Blöschl and C. Bucher, Factors influencing long range dependence in streamflow of European rivers, Hydrological Processes, 28 (4), 1573-1586, 2014. |
7. | Serinaldi, F., and C.G. Kilsby, The importance of prewhitening in change point analysis under persistence, Stochastic Environmental Research and Risk Assessment, 10.1007/s00477-015-1041-5, 2015. |
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Error evolution patterns in multi-step ahead streamflow forecasting, 13th International Conference on Hydroinformatics (HIC 2018), Palermo, Italy, doi:10.29007/84k6, 2018.
Multi-step ahead streamflow forecasting is of practical interest. We examine the error evolution in multi-step ahead forecasting by conducting six simulation experiments. Within each of these experiments we compare the error evolution patterns created by 16 forecasting methods, when the latter are applied to 2 000 time series. Our findings suggest that the error evolution can differ significantly from the one forecasting method to the other and that some forecasting methods are more useful than others. However, the errors computed at each time step of a forecast horizon for a specific single-case study strongly depend on the case examined and can be either small or large, regardless of the used forecasting method and the time step of interest. This fact is illustrated with a comparative case study using 92 monthly time series of streamflow.
Full text: http://www.itia.ntua.gr/en/getfile/1851/1/documents/2018HIC_ErrorEvolution_pp.pdf (5650 KB)
Additional material:
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Forecasting of geophysical processes using stochastic and machine learning algorithms, 10th World Congress on Water Resources and Environment "Panta Rhei", Athens, EWRA2017_A_110904, doi:10.13140/RG.2.2.30581.27361, European Water Resources Association, Athens, 2017.
We perform an extensive comparison between four stochastic and two machine learning (ML) forecasting algorithms by conducting a multiple-case study. The latter is composed by 50 single-case studies, which use time series of total monthly precipitation and mean monthly temperature observed in Greece. We apply a fixed methodology to each individual case and, subsequently, we perform a cross-case synthesis to facilitate the detection of systematic patterns. The stochastic algorithms include the Autoregressive order one model, an algorithm from the family of Autoregressive Fractionally Integrated Moving Average models, an Exponential Smoothing State Space algorithm and the Theta algorithm, while the ML algorithms are Neural Networks and Support Vector Machines. We also use the last observation as a Naive benchmark in the comparisons. We apply the forecasting methods to the deseasonalized time series. We compare the one-step ahead as also the multi-step ahead forecasting properties of the algorithms. Regarding the one-step ahead forecasting properties, the assessment is based on the absolute error of the forecast of the last observation. For the comparison of the multi-step ahead forecasting properties we use five metrics applied to the test set (last twelve observations), i.e. the root mean square error, the Nash-Sutcliffe efficiency, the ratio of standard deviations, the index of agreement and the coefficient of correlation. Concerning the ML algorithms, we also perform a sensitivity analysis for time lag selection. Additionally, we compare more sophisticated ML methods as regards to the hyperparameter optimization to simple ones.
Full text: http://www.itia.ntua.gr/en/getfile/1717/1/documents/EWRA2017_paper.pdf (8540 KB)
Additional material:
G. Papacharalampous, H. Tyralis, A. Montanari, and D. Koutsoyiannis, Large-scale calibration of conceptual rainfall-runoff models for two-stage probabilistic hydrological post-processing, EGU General Assembly 2021, online, doi:10.5194/egusphere-egu21-18, European Geosciences Union, 2021.
Probabilistic hydrological modelling methodologies often comprise two-stage post-processing schemes, thereby allowing the exploitation of the information provided by conceptual or physically-based rainfall-runoff models. They might also require issuing an ensemble of rainfall-runoff model simulations by using the rainfall-runoff model with different input data and/or different parameters. For obtaining a large number of rainfall-runoff model parameters in this regard, Bayesian schemes can be adopted; however, such schemes are accompanied by computational limitations (that are well-recognized in the literature). Therefore, in this work, we investigate the replacement of Bayesian rainfall-runoff model calibration schemes by computationally convenient non-Bayesian schemes within probabilistic hydrological modelling methodologies of the above-defined family. For our experiments, we use a methodology of this same family that is additionally characterized by the following distinguishing features: It (a) is in accordance with a theoretically consistent blueprint, (b) allows the exploitation of quantile regression algorithms (which offer larger flexibility than parametric models), and (c) has been empirically proven to harness the “wisdom of the crowd” in terms of average interval score. We also use a parsimonious conceptual rainfall-runoff model and 50-year-long monthly time series observed in 270 catchments in the United States to apply and compare 12 variants of the selected methodology. Six of these variants simulate the posterior distribution of the rainfall-runoff model parameters (conditional on the observations of a calibration period) within a Bayesian Markov chain Monte Carlo framework (first category of variants), while the other six variants use a simple computationally efficient approach instead (second category of variants). Six indicative combinations of the remaining components of the probabilistic hydrological modelling methodology (i.e., its post-processing scheme and its error model) are examined, each being used in one variant from each of the above-defined categories. In this specific context, the two large-scale calibration schemes (each representing a different “modelling culture” in our tests) are compared using proper scores and large-scale benchmarking. Overall, our findings suggest that the compared “modelling cultures” can lead to mostly equally good probabilistic predictions.
Full text: http://www.itia.ntua.gr/en/getfile/2126/2/documents/EGU21-18_presentation.pdf (1464 KB)
Additional material:
G. Papacharalampous, H. Tyralis, A. Langousis, A. W. Jayawardena, B. Sivakumar, N. Mamassis, A. Montanari, and D. Koutsoyiannis, Large-scale comparison of machine learning regression algorithms for probabilistic hydrological modelling via post-processing of point predictions, European Geosciences Union General Assembly 2019, Geophysical Research Abstracts, Vol. 21, Vienna, EGU2019-3576, European Geosciences Union, 2019.
Quantification of predictive uncertainty in hydrological modelling is often made by post-processing point hydrological predictions using regression models. We perform an extensive comparison of machine learning algorithms in obtaining quantile predictions of daily streamflow under this specific approach. The comparison is performed using a large amount of real-world data retrieved from the Catchment Attributes and MEteorology for Large sample Studies (CAMELS) dataset. Various climate types are well-represented by the examined catchments. The point predictions are obtained using the GR4J model, a lumped conceptual hydrological model comprising of four parameters, while their post-processing is made by predicting conditional quantiles of the hydrological model's errors. The latter are transformed to conditional quantiles of daily streamflow and finally assessed by using various performance metrics. The machine learning regression algorithms are also benchmarked against the quantile regression algorithm.
Full text: http://www.itia.ntua.gr/en/getfile/1943/1/documents/EGU2019-3576.pdf (33 KB)
H. Tyralis, and G. Papacharalampous, Univariate time series forecasting properties of random forests, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-1901, European Geosciences Union, 2018.
The random forests’ univariate time series forecasting properties have remained unexplored. Here we assess the performance of random forests in one-step forecasting using two large datasets of short time series with the aim to suggest an optimal set of predictor variables. Furthermore, we compare their performance to benchmarking methods. The first dataset consists of 16 000 simulated time series from a variety of Autoregressive Fractionally Integrated Moving Average (ARFIMA) models. The second dataset consists of 135 mean annual temperature time series. The random forests performed better mostly when using a few recent lagged predictor variables. A possible explanation of this result is that increasing the number of lagged variables decreases the length of the training set and simultaneously decreases the information exploited from the original time series during the model fitting phase. Furthermore, the random forests were comparable to the benchmarking methods.
Full text: http://www.itia.ntua.gr/en/getfile/1826/1/documents/EGU2018-1901_abstract.pdf (31 KB)
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, A step further from model-fitting for the assessment of the predictability of monthly temperature and precipitation, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-864, doi:10.6084/m9.figshare.7325783.v1, European Geosciences Union, 2018.
“With four parameters I can fit an elephant, and with five I can make him wiggle his trunk”,∼John von Neumann. This famous quote, literally possible as proved by Mayer et al. (2010), has been widely used to question the parsimony of a model providing a good description of the available data. Still, a significant part of the hydrological literature insists in adding parameters, trend or of other type, to models to increase their descriptive power within the concept of geophysical time series analysis and without testing their predictive ability. Herein, we move a step further from model-fitting and actually run in forecast mode several automatic univariate time series models with the aim to assess the predictability of monthly temperature and precipitation. We examine a sample of 985 monthly temperature and 1552 monthly precipitation time series, observed at stations covering a significant part of the Earth’s surface and, therefore, including various real-world process behaviours. All the time series are 40-years long with no missing values. We compare the naïve based on the monthly values of the last year, ARFIMA, exponential smoothing state space model with Box-Cox transformation, ARMA errors, Trend and Seasonal components (BATS), simple exponential smoothing (SES), Theta and Prophet forecasting methods. Prophet is a recently introduced model inspired by the nature of time series forecasted at Facebook and has not been applied to hydrometeorological time series in the past, while the use of BATS, SES and Theta is rare in hydrology. The methods are tested in performing multi-step ahead forecasts for the last 48 months of the data. The results are summarized in global scores, while their examination by group of stations leads to 5 individual scores for temperature and 6 for precipitation. The groups are formed according to the geographical vicinity of the stations. The findings suggest that all the examined models are accurate enough to be used in long-term forecasting applications. For the total of the temperature time series the use of an ARFIMA, BATS, SES, Theta or Prophet model, instead of the naïve method, leads in about 19-29% more accurate forecasts in terms of root mean square error, or even in about 30-32% more accurate forecasts specifically for the temperature time series observed in North Europe. For the total of the precipitation time series the use of all these automatic methods leads in about 21-22% better forecasts than the use of the naïve method, while for the geographical regions of North America, North Europe and East Asia these percentages are 26-29%, 22-24% and 32-38% respectively. We think that the level of the forecasting accuracy can barely be improved using other methods, as indicated by the experiments of Papacharalampous et al. (2017).
Full text:
Additional material:
G. Papacharalampous, and H. Tyralis, Large-scale assessment of random forests for data-driven hydrological modelling at monthly scale, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-1902, European Geosciences Union, 2018.
We assess the performance of random forests in modelling mean monthly streamflow based on mean monthly precipitation and potential evapotranspiration for 293 catchments. The assessment is made by computing the values of 18 metrics for the calibration and test periods, as well as by comparing these values with their respective computed for a lumped conceptual hydrological model with two parameters. The results are presented in maps and in an aggregated form. While the performance of the conceptual model is mostly similar for the two examined periods, the performance of random forests is far better for the calibration period than it is for the test period. Still, random forests perform better than the conceptual model for both periods.
Full text: http://www.itia.ntua.gr/en/getfile/1807/1/documents/EGU2018-1902.pdf (30 KB)
G. Papacharalampous, and H. Tyralis, One-step ahead forecasting of annual precipitation and temperature using univariate time series methods (solicited), European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-2298-1, European Geosciences Union, 2018.
We investigate the one-step ahead predictability of annual geophysical processes using 16 univariate time series forecasting methods. We examine two real-world datasets, a precipitation dataset and a temperature dataset, together containing 297 annual time series of 91 values. We use the first 50, 60, 70, 80 and 90 data points for model-fitting and model-validation and make predictions corresponding to the 51st, 61st, 71st, 81st and 91st respectively. The assessment of the methods’ performance is based on four error metrics and three accuracy statistics. The former are the error, absolute error, percentage error and absolute percentage error, while the latter are the median of the absolute errors, median of the absolute percentage errors and linear regression coefficient computed per category of tests.
Full text: http://www.itia.ntua.gr/en/getfile/1806/1/documents/EGU2018-2298-1.pdf (31 KB)
H. Tyralis, and A. Langousis, Modelling of rainfall maxima at different durations using max-stable processes, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-2299, European Geosciences Union, 2018.
The multivariate extreme value distribution (MEVD) has been used to model the dependence of rainfall block maxima at different temporal resolutions, as a means of estimating intensity-duration-frequency (IDF) curves for engineering applications. It is characterized by max-stability, which assumes that under proper renormalization, the rainfall block maxima at different temporal resolutions are extreme value distributed and the degree of their dependence remains invariant to the severity of the event. Due to these properties, and contrary to other commonly used approaches, MEVD allows for more conservative return level estimates at those durations used for model fitting. Max-stable processes are continuous extensions of MEVD, which are more flexible, and allow for extrapolation to temporal resolutions beyond those used for model fitting. Here we: 1) propose using max-stable processes to model rainfall block maxima, 2) apply the Brown-Resnick, Schlather and extremal-t models to hourly rainfall data, and 3) compare the obtained results to traditional approaches for IDF estimation. We discuss advantages and limitations regarding the use of max-stable processes in IDF estimation, and their potential use in hydrologic practice.
Full text: http://www.itia.ntua.gr/en/getfile/1805/1/documents/EGU2018-2299.pdf (32 KB)
G. Papacharalampous, and H. Tyralis, Illustrating important facts about multi-step ahead forecasting of univariate hydrological time series, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-3570, European Geosciences Union, 2018.
We present a case study using a long time series of monthly streamflow to illustrate important points regarding multi-step ahead forecasting of univariate hydrological processes. We forecast the monthly values of five discrete one-year periods based on the available past monthly values. To produce a faithful image of the underlying phenomena we implement a sufficient number of popular forecasting algorithms and compute an adequate number of metrics on the test sets. The algorithms are applied to the deseasonalized time series, while seasonality is subsequently recovered in the produced forecasts. The ranking of the methods clearly depends on the forecasting attempt and the computed metric, while the forecasting quality can be good or bad.
Full text: http://www.itia.ntua.gr/en/getfile/1804/1/documents/EGU2018-3570.pdf (31 KB)
H. Tyralis, and G. Papacharalampous, Multi-step ahead forecasting of monthly streamflow discharge time series using a variety of algorithms, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-3571-1, European Geosciences Union, 2018.
We compare a variety of algorithms in performing multi-step ahead forecasts of monthly streamflow discharge. We examine 285 time series originating from MOPEX catchments. We seasonally decompose the time series using a multiplicative model. We apply the algorithms to the deseasonalized time series and further make twelve-step ahead predictions corresponding to the last year’s monthly values of each time series. These values are not used in the fitting and validation processes. The forecasts are multiplied by the estimated seasonal component and, subsequently, they are compared with each other using an adequate number of metrics and two benchmarks. The results indicate that most of the methods perform well, in average better than the benchmark ones.
Full text: http://www.itia.ntua.gr/en/getfile/1803/1/documents/EGU2018-3571-1.pdf (32 KB)
G. Papacharalampous, H. Tyralis, and N. Mamassis, Conceptual hydrological modelling at daily scale: Aggregating results for 340 MOPEX catchments, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-3759, European Geosciences Union, 2018.
We present a large-scale model-implementing study aiming at the comparison of 3 daily conceptual hydrological models. These models comprise a different number of parameters, i.e. 4, 5 and 6 parameters. The comparison is performed for 340 MOPEX catchments, while each of the modelling approaches is assessed by computing the values of 18 metrics for the calibration and validation periods. The results are presented in maps and in an aggregated form, indicating that the models exhibit a quite similar performance, with the 6-parameter model being slightly better than the rest in terms of specific metrics. The metric values are mostly to a small extent better for the calibration set than they are for the validation set.
Full text: http://www.itia.ntua.gr/en/getfile/1802/1/documents/EGU2018-3759.pdf (33 KB)
P. Dimitriadis, H. Tyralis, T. Iliopoulou, K. Tzouka, Y. Markonis, N. Mamassis, and D. Koutsoyiannis, A climacogram estimator adjusted for timeseries length; application to key hydrometeorological processes by the Köppen-Geiger classification, European Geosciences Union General Assembly 2018, Geophysical Research Abstracts, Vol. 20, Vienna, EGU2018-17832, European Geosciences Union, 2018.
We present a climacogram estimator (variance of the scaled process vs. scale) that employs all the available information through a pooled time series estimation approach. This method does not discard time-series of short length or of high percentage of missing values; a common practice in hydrometeorology. Furthermore, we estimate and compare the second-order dependence structure (overall and classified by the Köppen-Geiger system) over the last two climatic periods (60 years) for several processes (temperature, dew-point, wind, precipitation, river discharge and atmospheric pressure) using worldwide surface stations. This analysis is performed based on the standardized climacogram, which shows numerous benefits compared to the autocorrelation and standardized power-spectrum.
Full text: http://www.itia.ntua.gr/en/getfile/1800/1/documents/EGU2018-17832.pdf (34 KB)
P. Dimitriadis, K. Tzouka, H. Tyralis, and D. Koutsoyiannis, Stochastic investigation of rock anisotropy based on the climacogram, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, EGU2017-10632-1, European Geosciences Union, 2017.
Anisotropy plays an important role on rock properties and entails valuable information for many fields of applied geology and engineering. Many methods are developed in order to detect transitions from isotropy to anisotropy but as a scale–depended effect, anisotropy also needs to be determined in multiple scales. We investigate the application of a stochastic tool, the climacogram (i.e. variance of the averaged process vs. scale) to characterize anisotropy in rocks at different length scales through image processing. The data are pictures from laboratory, specifically thin sections, and pictures of rock samples and rock formations in the field in order to examine anisotropy in nano, micro and macroscale.
Additional material:
P. Dimitriadis, T. Iliopoulou, H. Tyralis, and D. Koutsoyiannis, Identifying the dependence structure of a process through pooled timeseries analysis, IAHS Scientific Assembly 2017, Port Elizabeth, South Africa, IAHS Press, Wallingford – International Association of Hydrological Sciences, 2017.
Geophysical processes are known to exhibit significant departures from time-independence, ranging from short-range Markovian structure to Hurst-Kolmogorov behavior with large Hurst parameters. However, the identification of the dependence structure of a process is subject to many uncertainties, namely model uncertainty and estimation uncertainty particularly arising from the short length of available timeseries. Here we apply the climacogram (i.e. plot of the variance of the averaged process vs. scale) estimation method which has been shown to be the more robust and less uncertain among various stochastic metrics for the characterization of time-dependence. We further investigate the possibility of eliminating the sampling uncertainty by adequately employing all the available information through a pooled timeseries estimation approach, instead of discarding time-series of short length or of high percentage of missing values as typically performed in such tasks. We compare the merits and demerits of each approach as related to the strength of the dependence structure, the number and sample size of the available timeseries.
Full text: http://www.itia.ntua.gr/en/getfile/1770/1/documents/IAHS2017-182-1.pdf (196 KB)
H. Tyralis, P. Dimitriadis, and D. Koutsoyiannis, An extensive review and comparison of R Packages on the long-range dependence estimators, Asia Oceania Geosciences Society (AOGS) 14th Annual Meeting, Singapore, HS06-A003, doi:10.13140/RG.2.2.18837.22249, Asia Oceania Geosciences Society, 2017.
The long-range dependence (LRD) is a well-established property of climatic variables such as temperature and precipitation. A long list of estimators of the LRD parameters exist while a few comparison studies of their properties have been published. The emergence of R as one of the favourite programming languages among the hydrological community and its increasing number of packages enable the fast implementation of statistical methods in hydrological studies. Many R packages include functions for the estimation of the parameter, which characterizes the LRD, e.g. the Hurst parameter of the Hurst-Kolmogorov behaviour or the d parameter of the ARFIMA model. Here we present an extensive review of all R packages containing functions used to estimate the LRD parameter. Furthermore, we examine the properties of the implemented estimators and we perform an extended simulation experiment to compare them.
Full text: http://www.itia.ntua.gr/en/getfile/1721/1/documents/AOGS-HS06-A003presentation.pdf (1829 KB)
Additional material:
H. Tyralis, and D. Koutsoyiannis, The Bayesian Processor of Forecasts on the probabilistic forecasting of long-range dependent variables using General Circulation Models, Asia Oceania Geosciences Society (AOGS) 14th Annual Meeting, Singapore, HS20-A002, doi:10.13140/RG.2.2.15481.77922, Asia Oceania Geosciences Society, 2017.
We derive the distribution of the mean annual temperature and precipitation in the USA for the time period 2016-2100 conditional on observations from the time period 1916-2015 and ensembles from the phase 5 of the Coupled Model Intercomparison Project (CMIP5). To this end, we model the mean annual temperature and precipitation with the Hurst-Kolmogorov stochastic process (HKp, also known as Fractional Gaussian noise) to represent their long-range dependence (LRD). The HKp is a suitable model for climatic variables as has thoroughly been examined in the literature, while it can produce probabilistic forecasts conditional on historical observations. To improve the forecasts using the CMIP5 model ensembles, we apply the Bayesian Processor of Forecasts (BPF), which is a well-established technique used to forecast probabilistically weather and climatic variables conditional on a deterministic model output. The BPF is a general algorithm in the sense that it can be applied to any distribution and dependence pattern of the variables. However, it has been analysed theoretically and numerically solely for independent or Markov dependent variables. Here we extend its application to LRD dependent variables. The computation of uncertainties of climate projections is a mainstream subject in the climate literature and here we show that the BPF can be a sufficient solution.
Full text: http://www.itia.ntua.gr/en/getfile/1720/1/documents/AOGS-HS20-A002presentation.pdf (1895 KB)
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Large scale simulation experiments for the assessment of one-step ahead forecasting properties of stochastic and machine learning point estimation methods, Asia Oceania Geosciences Society (AOGS) 14th Annual Meeting, Singapore, HS06-A002, doi:10.13140/RG.2.2.33273.77923, Asia Oceania Geosciences Society, 2017.
The research in geophysical sciences often focuses on the comparison between stochastic and machine learning (ML) point estimation methods for time series forecasting. The comparisons performed are usually based on case studies. The present study aims to provide generalized results regarding the one-step ahead forecasting properties of several popular forecasting methods. This problem cannot be examined analytically, mainly because of the nature of the ML methods. Therefore, we conduct large-scale computational experiments based on simulations. Regarding the methodology, we compare a total of 20 methods among which 9 ML methods. Three of the latter methods are build using a neural networks algorithm, other three using a random forests algorithm and the remaining three using a support vector machines algorithm. The stochastic methods include simple methods, models from the frequently used families of Autoregressive Moving Average (ARMA), Autoregressive Fractionally Integrated Moving Average (ARFIMA) and Exponential Smoothing models. We perform 12 simulation experiments, each of them using 2 000 simulated time series. The time series are simulated using a stochastic model from the families of ARMA and ARFIMA models. The comparative assessment of the methods is based on the error and the absolute error of the forecast of the last observation.
Full text: http://www.itia.ntua.gr/en/getfile/1719/1/documents/AOGS-HS06-A002presentation.pdf (4029 KB)
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, A set of metrics for the effective evaluation of point forecasting methods used for hydrological tasks, Asia Oceania Geosciences Society (AOGS) 14th Annual Meeting, Singapore, HS01-A001, doi:10.13140/RG.2.2.19852.00641, Asia Oceania Geosciences Society, 2017.
The selection of metrics for the evaluation of point forecasting methods can be challenging even for very experienced hydrologists. We conduct a large-scale computational experiment based on simulations to compare the information that 18 metrics proposed in the literature give about the forecasting performance. Our purpose is to provide generalized results; thus we use 2 000 simulated Autoregressive Fractionally Integrated Moving Average time series. We apply several forecasting methods and we compute the values of the metrics for each forecasting experiment. Subsequently, we measure the correlation between the values of each pair of metrics, separately for each forecasting method. Furthermore, we explore graphically the detected relationships. Finally, we propose a set of metrics that we consider to be suitable for the effective evaluation of point forecasting methods.
Full text:
H. Tyralis, P. Dimitriadis, T. Iliopoulou, K. Tzouka, and D. Koutsoyiannis, Dependence of long-term persistence properties of precipitation on spatial and regional characteristics, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, EGU2017-3711, doi:10.13140/RG.2.2.13252.83840/1, European Geosciences Union, 2017.
The long-term persistence (LTP), else known in hydrological science as the Hurst phenomenon, is a behaviour observed in geophysical processes in which wet years or dry years are clustered to respective long time periods. A common practice for evaluating the presence of the LTP is to model the geophysical time series with the Hurst-Kolmogorov process (HKp) and estimate its Hurst parameter H where high values of H indicate strong LTP. We estimate H of the mean annual precipitation using instrumental data from approximately 1 500 stations which cover a big area of the earth’s surface and span from 1916 to 2015. We regress the H estimates of all stations on their spatial and regional characteristics (i.e. their location, elevation and Köppen-Geiger climate class) using a random forest algorithm. Furthermore, we apply the Mann-Kendall test under the LTP assumption (MKt-LTP) to all time series to assess the significance of observed trends of the mean annual precipitation. To summarize the results, the LTP seems to depend mostly on the location of the stations, while the predictive value of the fitted regression model is good. Thus when investigating for LTP properties we recommend that the local characteristics should be considered. Additionally, the application of the MKt-LTP suggests that no significant monotonic trend can characterize the global precipitation. Dominant positive significant trends are observed mostly in main climate type D (snow), while in the other climate types the percentage of stations with positive significant trends was approximately equal to that of negative significant trends. Furthermore, 50% of all stations do not exhibit significant trends at all.
Full text: http://www.itia.ntua.gr/en/getfile/1695/1/documents/EGU2017-3711presentation_.pdf (1608 KB)
Additional material:
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Investigation of the effect of the hyperparameter optimization and the time lag selection in time series forecasting using machine learning algorithms, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, 19, EGU2017-3072-1, doi:10.13140/RG.2.2.20560.92165/1, European Geosciences Union, 2017.
The hyperparameter optimization and the time lag selection are considered to be of great importance in time series forecasting using machine learning (ML) algorithms. To investigate their effect on the ML forecasting performance we conduct several large-scale simulation experiments. Within each of the latter we compare 12 methods on 2 000 simulated time series from the family of Autoregressive Fractionally Integrated Moving Average (ARFIMA) models. The methods are defined by the set {ML algorithm, hyperparameter selection procedure, time lags}. We compare three ML algorithms, i.e. Neural Networks (NN), Random Forests (RF) and Support Vector Machines (SVM), two procedures for hyperparameter selection i.e. predefined hyperparameters or defined after optimization and two regression matrices (using time lag 1 or 1, …, 21). After splitting each simulated time series into a fitting and a testing set, we fit the models to the former set and compare their performance on the latter one. We quantify the methods’ performance using several metrics proposed in the literature and benchmark methods. Furthermore, we conduct a sensitivity analysis on the length of the fitting set to examine how it affects the robustness of our results. The findings indicate that the hyperparameter optimization mostly has a small effect on the forecasting performance. This is particularly important, because the hyperparameter optimization is computationally intensive. On the other hand, the time lag selection seems to mostly significantly affect the methods performance when using the NN algorithm, while we observe a similar behaviour for the RF algorithm albeit to a smaller extent.
Full text: http://www.itia.ntua.gr/en/getfile/1693/1/documents/EGU2017-3072presentation.pdf (1731 KB)
Additional material:
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Multi-step ahead streamflow forecasting for the operation of hydropower reservoirs, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, 19, EGU2017-3069, doi:10.13140/RG.2.2.27271.80801, European Geosciences Union, 2017.
Multi-step ahead forecasting is of practical interest for the operation of hydropower reservoirs.We conduct several large scale simulation experiments using both streamflow data and simulated time series to provide generalized results concerning the variation over time of the error values in multi-step ahead forecasting. In more detail, we apply several popular forecasting methods to each time series as explained subsequently. Each time series is split into a fitting and a testing set. We fit the models to the former set and we test their forecasting performance in the latter set. Lastly, we compute the error and the absolute error at each time step of the forecast horizon for each test and carry out a statistical analysis on the formed data sets. Furthermore, we perform a sensitivity analysis on the length of the fitting set to examine how it affects the results.
Remarks:
See preprint in http://doi.org/10.20944/preprints201710.0129.v1
Full text: http://www.itia.ntua.gr/en/getfile/1692/1/documents/EGU2017-3069presentation.pdf (3930 KB)
Additional material:
G. Papacharalampous, H. Tyralis, and D. Koutsoyiannis, Comparison between stochastic and machine learning methods for hydrological multi-step ahead forecasting: All forecasts are wrong!, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, 19, EGU2017-3068-2, doi:10.13140/RG.2.2.17205.47848, European Geosciences Union, 2017.
Machine learning (ML) is considered to be a promising approach to hydrological processes forecasting. We conduct a comparison between several stochastic and ML point estimation methods by performing large-scale computational experiments based on simulations. The purpose is to provide generalized results, while the respective comparisons in the literature are usually based on case studies. The stochastic methods used include simple methods, models from the frequently used families of Autoregressive Moving Average (ARMA), Autoregressive Fractionally Integrated Moving Average (ARFIMA) and Exponential Smoothing models. The ML methods used are Random Forests (RF), Support Vector Machines (SVM) and Neural Networks (NN). The comparison refers to the multi-step ahead forecasting properties of the methods. A total of 20 methods are used, among which 9 are the ML methods. 12 simulation experiments are performed, while each of them uses 2 000 simulated time series of 310 observations. The time series are simulated using stochastic processes from the families of ARMA and ARFIMA models. Each time series is split into a fitting (first 300 observations) and a testing set (last 10 observations). The comparative assessment of the methods is based on 18 metrics, that quantify the methods’ performance according to several criteria related to the accurate forecasting of the testing set, the capturing of its variation and the correlation between the testing and forecasted values. The most important outcome of this study is that there is not a uniformly better or worse method. However, there are methods that are regularly better or worse than others with respect to specific metrics. It appears that, although a general ranking of the methods is not possible, their classification based on their similar or contrasting performance in the various metrics is possible to some extent. Another important conclusion is that more sophisticated methods do not necessarily provide better forecasts compared to simpler methods. It is pointed out that the ML methods do not differ dramatically from the stochastic methods, while it is interesting that the NN, RF and SVM algorithms used in this study offer potentially very good performance in terms of accuracy. It should be noted that, although this study focuses on hydrological processes, the results are of general scientific interest. Another important point in this study is the use of several methods and metrics. Using fewer methods and fewer metrics would have led to a very different overall picture, particularly if those fewer metrics corresponded to fewer criteria. For this reason, we consider that the proposed methodology is appropriate for the evaluation of forecasting methods.
Full text: http://www.itia.ntua.gr/en/getfile/1691/1/documents/EGU2017-3068presentation.pdf (1804 KB)
Additional material:
V. Daniil, G. Pouliasis, E. Zacharopoulou, E. Demetriou, G. Manou, M. Chalakatevaki, I. Parara, C. Georganta, P. Stamou, S. Karali, E. Hadjimitsis, G. Koudouris, E. Moschos, D. Roussis, K. Papoulakos, A. Koskinas, G. Pollakis, N. Gournari, K. Sakellari, Y. Moustakis, N. Mamassis, A. Efstratiadis, H. Tyralis, P. Dimitriadis, T. Iliopoulou, G. Karakatsanis, K. Tzouka, I. Deligiannis, V. Tsoukala, P. Papanicolaou, and D. Koutsoyiannis, The uncertainty of atmospheric processes in planning a hybrid renewable energy system for a non-connected island, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, EGU2017-16781-4, doi:10.13140/RG.2.2.29610.62406, European Geosciences Union, 2017.
Non-connected islands to the electric gird are often depending on oil-fueled power plants with high unit cost. A hybrid energy system with renewable resources such as wind and solar plants could reduce this cost and also offer more environmental friendly solutions. However, atmospheric processes are characterized by high uncertainty that does not permit harvesting and utilizing full of their potential. Therefore, a more sophisticated framework that somehow incorporates this uncertainty could improve the performance of the system. In this context, we describe several stochastic and financial aspects of this framework. Particularly, we investigate the cross-correlation between several atmospheric processes and the energy demand, the possibility of mixing renewable resources with the conventional ones and in what degree of reliability, and critical financial subsystems such as weather derivatives. A pilot application of the above framework is also presented for a remote island in the Aegean Sea.
Full text: http://www.itia.ntua.gr/en/getfile/1689/1/documents/EGU2017oral_16781_final.pdf (3038 KB)
Additional material:
E. Hadjimitsis, E. Demetriou, K. Sakellari, H. Tyralis, P. Dimitriadis, T. Iliopoulou, and D. Koutsoyiannis, Investigation of the stochastic nature of temperature and humidity for energy management, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, 19, EGU2017-10164-5, European Geosciences Union, 2017.
Atmospheric temperature and dew point, in addition to their role in atmospheric processes, influence the management of energy systems since they highly affect the energy demand and production. Both temperature and humidity depend on the climate conditions and geographical location. In this context, we analyze numerous of observations around the globe and we investigate the long-term behaviour and periodicities of the temperature and humidity processes. Also, we present and apply a parsimonious stochastic double-cyclostationary model for these processes to an island in the Aegean Sea and investigate their link to energy management.
Additional material:
E. Moschos, G. Manou, C. Georganta, P. Dimitriadis, T. Iliopoulou, H. Tyralis, D. Koutsoyiannis, and V. Tsoukala, Investigation of the stochastic nature of wave processes for renewable resources management: a pilot application in a remote island in the Aegean sea, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, EGU2017-10225-3, doi:10.13140/RG.2.2.30226.66245, European Geosciences Union, 2017.
The large energy potential of ocean dynamics is not yet being efficiently harvested mostly due to several technological and financial drawbacks. Nevertheless, modern renewable energy systems include wave and tidal energy in cases of nearshore locations. Although the variability of tidal waves can be adequately predictable, wind-generated waves entail a much larger uncertainty due to their dependence to the wind process. Recent research has shown, through estimation of the wave energy potential in coastal areas of the Aegean Sea, that installation of wave energy converters in nearshore locations could be an applicable scenario, assisting the electrical network of Greek islands. In this context, we analyze numerous of observations and we investigate the long-term behaviour of wave height and wave period processes. Additionally, we examine the case of a remote island in the Aegean sea, by estimating the local wave climate through past analysis data and numerical methods, and subsequently applying a parsimonious stochastic model to a theoretical scenario of wave energy production.
Full text: http://www.itia.ntua.gr/en/getfile/1685/1/documents/EGU2017-10225-3_poster.pdf (3588 KB)
Additional material:
A. Koskinas, E. Zacharopoulou, G. Pouliasis, I. Engonopoulos, K. Mavroyeoryos, I. Deligiannis, G. Karakatsanis, P. Dimitriadis, T. Iliopoulou, D. Koutsoyiannis, and H. Tyralis, Simulation of electricity demand in a remote island for optimal planning of a hybrid renewable energy system, European Geosciences Union General Assembly 2017, Geophysical Research Abstracts, Vol. 19, Vienna, 19, EGU2017-10495-4, doi:10.13140/RG.2.2.10529.81767, European Geosciences Union, 2017.
We simulate the electrical energy demand in the remote island of Astypalaia. To this end we first obtain information regarding the local socioeconomic conditions and energy demand. Secondly, the available hourly demand data are analysed at various time scales (hourly, weekly, daily, seasonal). The cross-correlations between the electrical energy demand and the mean daily temperature as well as other climatic variables for the same time period are computed. Also, we investigate the cross-correlation between those climatic variables and other variables related to renewable energy resources from numerous observations around the globe in order to assess the impact of each one to a hybrid renewable energy system. An exploratory data analysis including all variables is performed with the purpose to find hidden relationships. Finally, the demand is simulated considering all the periodicities found in the analysis. The simulation time series will be used in the development of a framework for planning of a hybrid renewable energy system in Astypalaia.
Full text: http://www.itia.ntua.gr/en/getfile/1684/2/documents/EGU2017_CrossCorr-EnergyDemand.pdf (2668 KB)
Additional material:
G. Karakatsanis, H. Tyralis, and K. Tzouka, Entropy, pricing and productivity of pumped-storage, European Geosciences Union General Assembly 2016, Geophysical Research Abstracts, Vol. 18, Vienna, European Geosciences Union, 2016.
Pumped-storage constitutes today a mature method of bulk electricity storage in the form of hydropower. This bulk electricity storability upgrades the economic value of hydropower as it may mitigate -or even neutralize- stochastic effects deriving from various geophysical and socioeconomic factors, which produce numerous load balance inefficiencies due to increased uncertainty. Pumped-storage further holds a key role for unifying intermittent renewable (i.e. wind, solar) units with controllable non-renewable (i.e. nuclear, coal) fuel electricity generation plants into integrated energy systems. We develop a set of indicators for the measurement of performance of pumped-storage, in terms of the latter's energy and financial contribution to the energy system. More specifically, we use the concept of entropy in order to examine: (1) the statistical features -and correlations- of the energy system's intermittent components and (2) the statistical features of electricity demand prediction deviations. In this way, the macroeconomics of pumped-storage emerges naturally from its statistical features (Karakatsanis et al. 2014). In addition, these findings are combined to actual daily loads. Hence, not only the amount of energy harvested from the pumped-storage component is expected to be important, but the harvesting time as well, as the intraday price of electricity varies significantly. Additionally, the structure of the pumped-storage market proves to be a significant factor as well for the system's energy and financial performance (Paine et al. 2014). According to the above, we aim at postulating a set of general rules on the productivity of pumped-storage for (integrated) energy systems.
Full text: http://www.itia.ntua.gr/en/getfile/1854/1/documents/EGU_2016_GK_15481.pdf (3409 KB)
A. Sotiriadou, A. Petsiou, E. Feloni, P. Kastis, T. Iliopoulou, Y. Markonis, H. Tyralis, P. Dimitriadis, and D. Koutsoyiannis, Stochastic investigation of precipitation process for climatic variability identification, European Geosciences Union General Assembly 2016, Geophysical Research Abstracts, Vol. 18, Vienna, EGU2016-15137-5, doi:10.13140/RG.2.2.28955.46881, European Geosciences Union, 2016.
The precipitation process is important not only to hydrometeorology but also to renewable energy resources management. We use a dataset consisting of daily and hourly records around the globe to identify statistical variability with emphasis on the last period. Specifically, we investigate the occurrence of mean, maximum and minimum values and we estimate statistical properties such as marginal probability distribution function and the type of decay of the climacogram (i.e. mean process variance vs. scale).
Acknowledgement: This research is conducted within the frame of the undergraduate course "Stochastic Methods in Water Resources" of the National Technical University of Athens (NTUA). The School of Civil Engineering of NTUA provided moral support for the participation of the students in the Assembly.
Full text: http://www.itia.ntua.gr/en/getfile/1658/1/documents/RainP.pdf (3820 KB)
Additional material:
H. Tyralis, N. Mamassis, and Y. Photis, Spatial analysis of electricity demand patterns in Greece: Application of a GIS-based methodological framework, European Geosciences Union General Assembly 2016, Geophysical Research Abstracts, Vol. 18, Vienna, European Geosciences Union, 2016.
We investigate various uses of electricity demand in Greece (agricultural, commercial, domestic, industrial use as well as use for public and municipal authorities and street lighting) and we examine their relation with variables such as population, total area, population density and the Gross Domestic Product. The analysis is performed on data which span from 2008 to 2012 and have annual temporal resolution and spatial resolution down to the level of prefecture. We both visualize the results of the analysis and we perform cluster and outlier analysis using the Anselin local Moran's I statistic as well as hot spot analysis using the Getis-Ord Gi* statistic. The definition of the spatial patterns and relationships of the aforementioned variables in a GIS environment provides meaningful insight and better understanding of the regional development model in Greece and justifies the basis for an energy demand forecasting methodology.
Full text:
H. Tyralis, G. Karakatsanis, K. Tzouka, and N. Mamassis, Analysis of the electricity demand of Greece for optimal planning of a large-scale hybrid renewable energy system, European Geosciences Union General Assembly 2015, Geophysical Research Abstracts, Vol. 17, Vienna, EGU2015-5643, European Geosciences Union, 2015.
The Greek electricity system is examined for the period 2002-2014. The demand load data are analysed at various time scales (hourly, daily, seasonal and annual) and they are related to the mean daily temperature and the gross domestic product (GDP) of Greece for the same time period. The prediction of energy demand, a product of the Greek Independent Power Transmission Operator, is also compared with the demand load. Interesting results about the change of the electricity demand scheme after the year 2010 are derived. This change is related to the decrease of the GDP, during the period 2010-2014. The results of the analysis will be used in the development of an energy forecasting system which will be a part of a framework for optimal planning of a large-scale hybrid renewable energy system in which hydropower plays the dominant role.
Full text:
A. Koukouvinos, D. Nikolopoulos, A. Efstratiadis, A. Tegos, E. Rozos, S.M. Papalexiou, P. Dimitriadis, Y. Markonis, P. Kossieris, H. Tyralis, G. Karakatsanis, K. Tzouka, A. Christofides, G. Karavokiros, A. Siskos, N. Mamassis, and D. Koutsoyiannis, Integrated water and renewable energy management: the Acheloos-Peneios region case study, European Geosciences Union General Assembly 2015, Geophysical Research Abstracts, Vol. 17, Vienna, EGU2015-4912, doi:10.13140/RG.2.2.17726.69440, European Geosciences Union, 2015.
Within the ongoing research project “Combined Renewable Systems for Sustainable Energy Development” (CRESSENDO), we have developed a novel stochastic simulation framework for optimal planning and management of large-scale hybrid renewable energy systems, in which hydropower plays the dominant role. The methodology and associated computer tools are tested in two major adjacent river basins in Greece (Acheloos, Peneios) extending over 15 500 km2 (12% of Greek territory). River Acheloos is characterized by very high runoff and holds ~40% of the installed hydropower capacity of Greece. On the other hand, the Thessaly plain drained by Peneios – a key agricultural region for the national economy – usually suffers from water scarcity and systematic environmental degradation. The two basins are interconnected through diversion projects, existing and planned, thus formulating a unique large-scale hydrosystem whose future has been the subject of a great controversy. The study area is viewed as a hypothetically closed, energy-autonomous, system, in order to evaluate the perspectives for sustainable development of its water and energy resources. In this context we seek an efficient configuration of the necessary hydraulic and renewable energy projects through integrated modelling of the water and energy balance. We investigate several scenarios of energy demand for domestic, industrial and agricultural use, assuming that part of the demand is fulfilled via wind and solar energy, while the excess or deficit of energy is regulated through large hydroelectric works that are equipped with pumping storage facilities. The overall goal is to examine under which conditions a fully renewable energy system can be technically and economically viable for such large spatial scale.
Full text:
Additional material:
See also: http://dx.doi.org/10.13140/RG.2.2.17726.69440
Other works that reference this work (this list might be obsolete):
1. | Stamou, A. T., and P. Rutschmann, Pareto optimization of water resources using the nexus approach, Water Resources Management, 32, 5053-5065, doi:10.1007/s11269-018-2127-x, 2018. |
2. | Stamou, A.-T., and P. Rutschmann, Optimization of water use based on the water-energy-food nexus concept: Application to the long-term development scenario of the Upper Blue Nile River, Water Utility Journal, 25, 1-13, 2020. |
A. M. Filippidou, A. Andrianopoulos, C. Argyrakis, L. E. Chomata, V. Dagalaki, X. Grigoris, T. S. Kokkoris, M. Nasioka, K. A. Papazoglou, S.M. Papalexiou, H. Tyralis, and D. Koutsoyiannis, Comparison of climate time series produced by General Circulation Models and by observed data on a global scale, European Geosciences Union General Assembly 2014, Geophysical Research Abstracts, Vol. 16, Vienna, EGU2014-8529, doi:10.13140/RG.2.2.33887.87200, European Geosciences Union, 2014.
Outputs of General Circulation Models (GCMs) for precipitation are compared with time series produced from observations. Comparison is made on global and hemispheric spatial scale and on annual time scale. Various time periods are examined, distinguishing periods before and after publishing of model outputs. Historical climate time series are compared with the outputs of GCMs for the 20th century and those for the A1B, B1 and A2 emission scenarios for the 21st century. Several indices are examined, i.e. the estimated means, variances, Hurst parameters, cross-correlations etc.
Acknowledgement: This research is conducted within the frame of the undergraduate course "Stochastic Methods in Water Resources" of the National Technical University of Athens (NTUA). The School of Civil Engineering of NTUA provided moral support for the participation of the students in the Assembly.
Full text:
Additional material:
H. Tyralis, and D. Koutsoyiannis, Simultaneous use of observations and deterministic model outputs to forecast persistent stochastic processes, Facets of Uncertainty: 5th EGU Leonardo Conference – Hydrofractals 2013 – STAHY 2013, Kos Island, Greece, doi:10.13140/RG.2.1.3230.4889, European Geosciences Union, International Association of Hydrological Sciences, International Union of Geodesy and Geophysics, 2013.
We combine a time series of a geophysical process with the output of a deterministic model, which simulates the aforementioned process in the past also providing future predictions. The purpose is to convert the single prediction of the deterministic model for the future evolution of the time series into a stochastic prediction. The time series is modelled by a stationary persistent normal stochastic process. The output of the deterministic model comprises the simulation of the historical part of the process and its deterministic future prediction. The complexity of the deterministic model is assumed to be irrelevant to our framework. A multivariate stochastic process, whose first variable is the true (observable) process and the second variable is a process representing the deterministic model, is formed. The covariance matrix function is computed and the distribution of the unobserved part of the stochastic process is calculated conditional on the observations and the output of the deterministic model.
Full text:
H. Tyralis, and D. Koutsoyiannis, A Bayesian approach to hydroclimatic prognosis using the Hurst-Kolmogorov stochastic process, European Geosciences Union General Assembly 2012, Geophysical Research Abstracts, Vol. 14, Vienna, doi:10.13140/RG.2.2.24273.74089, European Geosciences Union, 2012.
It has now been well recognized that hydrological processes exhibit a scaling behaviour, also known as the Hurst phenomenon. An appropriate way to model this behaviour is to use the Hurst-Kolmogorov stochastic process. This process is associated with large scale fluctuations and also enhanced uncertainty in the parameter estimation. When we have to make a prognosis for the future evolution of the process, the total uncertainty must be evaluated. The proper technique to do this is provided by Bayesian methods. We develop a Bayesian framework with Monte Carlo implementation for the uncertainty estimation of future prognoses assuming a Hurst-Kolmogorov stochastic process with a non-informative prior distribution of parameters. We derive the posterior distribution of the parameters and use it to make inference for future hydroclimatic variables.
Full text:
P. Kossieris, D. Koutsoyiannis, C. Onof, H. Tyralis, and A. Efstratiadis, HyetosR: An R package for temporal stochastic simulation of rainfall at fine time scales, European Geosciences Union General Assembly 2012, Geophysical Research Abstracts, Vol. 14, Vienna, 11718, European Geosciences Union, 2012.
A complete software package for the temporal stochastic simulation of rainfall process at fine time scales is developed in the R programming environment. This includes several functions for sequential simulation or disaggregation. Specifically, it uses the Bartlett-Lewis rectangular pulses rainfall model for rainfall generation and proven disaggregation techniques which adjust the finer scale (hourly) values in order to obtain the required coarser scale (daily) value, without affecting the stochastic structure implied by the model. Additionally, a repetition scheme is incorporated in order to improve the Bartlett-Lewis model performance without significant increase of computational time. Finally, the package includes an enhanced version of the evolutionary annealing-simplex optimization method for the estimation of Bartlett-Lewis parameters. Multiple calibration criteria are introduced, in order to reproduce the statistical characteristics of rainfall at various time scales. This upgraded version of the original HYETOS program (Koutsoyiannis, D., and Onof C., A computer program for temporal stochastic disaggregation using adjusting procedures, European Geophysical Society, 2000) operates on several modes and combinations thereof (depending on data availability), with many options and graphical capabilities. The package, under the name HyetosR, is available free in the CRAN package repository.
Remarks:
Software page: http://itia.ntua.gr/en/softinfo/3/
Full text:
Other works that reference this work (this list might be obsolete):
1. | #Montesarchio, V., F. Napolitano, E. Ridolfi and L. Ubertini, A comparison of two rainfall disaggregation models, In Numerical Analysis and Applied Mathematics ICNAAM 2012: International Conference of Numerical Analysis and Applied Mathematics, AIP Conference Proceedings, Vol. 1479, 1796-1799, 2012. |
2. | #Villani, V., L. Cattaneo, A. L. Zollo, and P. Mercogliano, Climate data processing with GIS support: Description of bias correction and temporal downscaling tools implemented in Clime software, Euro-Mediterranean Center on Climate Change (RMCC) Research Papers, RP0262, 2015. |
3. | Förster, K., F. Hanzer, B. Winter, T. Marke, and U. Strasser, An open-source MEteoroLOgical observation time series DISaggregation Tool (MELODIST v0.1.1), Geoscientific Model Development, 9, 2315-2333, doi:10.5194/gmd-9-2315-2016, 2016. |
4. | Devkota, S., N. M. Shakya, K. Sudmeier-Rieux, M. Jaboyedoff, C. J. Van Westen, B. G. Mcadoo, and A. Adhikari, Development of monsoonal rainfall intensity-duration-frequency (IDF) relationship and empirical model for data-scarce situations: The case of the Central-Western Hills (Panchase Region) of Nepal, Hydrology, 5(2), 27, doi:10.3390/hydrology5020027, 2018. |
5. | Cordeiro, M. R. C., J. A. Vanrobaeys, and H. F. Wilson, Long-term weather, streamflow, and water chemistry datasets for hydrological modelling applications at the upper La Salle River watershed in Manitoba, Canada, 6(1), 41-57, Geoscience Data Journal, doi:10.1002/gdj3.67, 2019. |
6. | #Thomson, H., and L. Chandler, Tailings storage facility landform evolution modelling, Proceedings of the 13th International Conference on Mine Closure, A. B. Fourie & M. Tibbett (eds.), Australian Centre for Geomechanics, Perth, 385-396, 2019. |
7. | Sun, Y., D. Wendi, D. E., Kim, and S.-Y. Liong, Deriving intensity–duration–frequency (IDF) curves using downscaled in situ rainfall assimilated with remote sensing data, Geoscience Letters, 6(17), doi:10.1186/s40562-019-0147-x, 2019. |
8. | Oruc, S., I. Yücel, and A. Yılmaz, Investigation of the effect of climate change on extreme precipitation: Capital Ankara case, Teknik Dergi, 33(2), doi:10.18400/tekderg.714980, 2021. |
9. | Hayder, A. M., and M. Al-Mukhtar, Modelling the IDF curves using the temporal stochastic disaggregation BLRP model for precipitation data in Najaf City, Arabian Journal of Geosciences, 14, 1957, doi:10.1007/s12517-021-08314-6, 2021. |
10. | Diez-Sierra, J., S. Navas, and M. del Jesus, Neoprene: An open-source Python library for spatial rainfall generation based on the Neyman-Scott process, doi:10.2139/ssrn.4092195, 2022. |
11. | Cordeiro, M. R. C., K. Liang, H. F. Wilson, J. Vanrobaeys, D. A. Lobb, X. Fang, and J. W. Pomeroy, Simulating the hydrological impacts of land use conversion from annual crop to perennial forage in the Canadian Prairies using the Cold Regions Hydrological Modelling platform, Hydrology and Earth System Sciences, 26, 5917-5931, doi:10.5194/hess-26-5917-2022, 2022. |
D. Koutsoyiannis, S. Kozanis, and H. Tyralis, A general Monte Carlo method for the construction of confidence intervals for a function of probability distribution parameters, European Geosciences Union General Assembly 2011, Geophysical Research Abstracts, Vol. 13, Vienna, EGU2011-1489, doi:10.13140/RG.2.2.33147.31527, European Geosciences Union, 2011.
We derive an algorithm which calculates an exact confidence interval for a distributional parameter of location or scale family, based on a two-sided hypothesis test on the parameter of interest, using some pivotal quantities. We use this algorithm to calculate approximate confidence intervals for the parameter or a function of the parameter of one-parameter distributions. We show that these approximate intervals are asymptotically exact. We modify the algorithm and use it to obtain approximate confidence intervals for a parameter or a function of parameters for multi-parameter distributions. We compare the results of the method with those obtained by known methods of the literature for the normal, the gamma and the Weibull distribution and find them satisfactory. We conclude that the proposed method can yield approximate confidence intervals, based on Monte Carlo simulations, in a generic way, irrespectively of the distribution function, as well as of the type of the parameters or the function of parameters.
Full text:
H. Tyralis, and D. Koutsoyiannis, Performance evaluation and interdependence of parameter estimators of the Hurst-Kolmogorov stochastic process, European Geosciences Union General Assembly 2010, Geophysical Research Abstracts, Vol. 12, Vienna, EGU2010-10476, doi:10.13140/RG.2.2.27118.00322, European Geosciences Union, 2010.
We investigate three methods for simultaneous estimation of the Hurst parameter (H) and the standard deviation (σ) for a Hurst-Kolmogorov stochastic process, namely the maximum likelihood method and two methods based on the variation of the standard deviation or of the variance with time scale. We show that the simultaneous estimation of the two parameters is important, albeit not given appropriate attention in the literature, because of the interdependence of the two parameter estimators. In addition, we test the performance of the three methods for a range of sample sizes and H values, through a simulation study and we compare it with other known results for other estimators of the literature.
Full text:
H. Tyralis, Integrated management of surface water resources of Acheloos and Peneios river basins, May 2015.
Full text:
H. Tyralis, An introduction to R programming language, 39 pages, November 2011.
Full text: http://www.itia.ntua.gr/en/getfile/1230/1/documents/TYralisNaresuanRlang2011.pdf (547 KB)
H. Tyralis, A brief introduction to Bayesian statistics, 29 pages, November 2011.
Full text: http://www.itia.ntua.gr/en/getfile/1229/1/documents/TYralisNaresuan2011.pdf (657 KB)
H. Tyralis, Spatial and temporal analysis of electricity demandin Greece, MSc thesis, 95 pages, July 2016.
In this study we analyse the electricity demand in Greece for the time period 2002-2014 and we simulate the electricity demand in Greece and Thessaly. We visualize the electricity demand in Greece for the time period 2002-2014. We search the relationship between the energy demand and the Gross Domestic Product (GDP) and the temperature. The analysis shows that in general the conclusions of Psiloglou et al. (2009) are still valid. However, since 2010, when the GDP began to decline, the pattern of the energy demand has changed in the winter. Moreover, we investigate the effectiveness of the energy demand forecasting system of the Independent Transmission System Operator (IPTO). The results of the analysis are useful for forecasting the energy demand and for generating synthetic time series. The data and the code of this work are available as supporting material. We analyse the spatial patterns of the energy demand. We present patterns of energy demand for various uses and combinations of energy demand variables for various uses with variables such as the GDP, the population, the area and the population density for the time period 2008-2012. We perform clusters and outliers analysis, hot spot analysis and grouping analysis. The most important results of these analyses are presented in the main body of work, while the total of 1 125 Figures produced during the work, are provided as supporting material. We believe that the results are useful to understand issues related to the spatial distribution of the energy demand in Greece and the developmental orientation of the state, to improve previous works, which use socioeconomic variables. We simulate the total energy demand in Greece and Thessaly, after the removal of the energy demand for agricultural use. Furthermore, we simulate the energy demand for agricultural use in Thessaly. The simulation concerns a long time period. It is performed to assess the long-term properties of an energy project. Therefore, it neglects features such as the dependence on initial conditions. The simulation of energy demand is performed after the removal of the annual averages and the observed periodicities. The synthetic series is produced using a bootstrap method with blocks of random length derived from a geometric distribution. The removed elements are added to the synthetic time series. Summarizing the results of the analysis in the time domain, we observe during the day two local maxima at approximately 12:00 and 20:00, and two local minima at about 04:00 and 16:00. We observe daily maxima of the energy demand on Wednesdays or Thursdays and daily minima on Sundays. The shape of the daily energy demand is almost independent of the month. We observe local maxima of the monthly energy demand in January and July and local minima in October and April. The December monthly energy demand increased significantly after the hydrological year 2011. We observe less variation in energy demand in the period between December and April. A linear model for the relationship between the energy demand and the GDP seems reasonable. Moreover, we observe that the GDP increase results in increased energy demand (or vice versa). Nevertheless, the GDP increase in the hydrological year 2008 was followed by a decrease in energy demand. We observe a global minimum of the energy demand for temperatures at about 17-18° C, and local maxima of temperatures at about 3° C and 32° C. A regression line for the energy demand (the y axis) and temperature (the x axis) would be a convex curve. The IPTO’s daily predictions usually overestimate the energy demand. The results of the investigation will be useful for the construction of a short and a medium-term energy demand forecasting model. Regarding the spatial analysis, we show some Figures from the support material which present significant results. It seems that in Greece’s middle there is an area which is classified as industrial but also agricultural. The Greek islands are mainly characterized as commercial, while the area around Athens is characterized by high values of household energy use. Regarding its development model Greece is divided to the main continental part, which includes areas in the middle and North, to Central Greece and to island regions. The island regions are similar to the Athens area and Peloponnese in specific cases. These results may be useful for the management of the electricity grid and for finding optimal policies for the development model of the state. In regard to the simulation, we did not remove the industrial energy demand use from the data. The industrial use is characterized by different periodicities compared e.g. with the household use. The result of interest was the production of a synthetic time series to simulate a power system and assess its long-term performance. For this specific application it was sufficient to examine only the time series electricity demand. However, if we are interested in forecasting we must take into account the forecasts of temperature, GDP etc. In this case it is necessary to correlate the electricity demand with other variables.
Full text: http://www.itia.ntua.gr/en/getfile/1626/1/documents/Spatio-temporal_energy_analysis_Greece_Msc_thesis.pdf (7210 KB)
Additional material:
H. Tyralis, Use of Bayesian techniques in hydroclimatic prognosis, PhD thesis, 166 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, November 2015.
Additional material:
A. Koukouvinos, A. Efstratiadis, D. Nikolopoulos, H. Tyralis, A. Tegos, N. Mamassis, and D. Koutsoyiannis, Case study in the Acheloos-Thessaly system, Combined REnewable Systems for Sustainable ENergy DevelOpment (CRESSENDO), 98 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, October 2015.
This report describes the validation of methodologies and computer tools that have been developed in the context of the research project, in the interconnected river basin system of Acheloos and Peneios. The study area is modelled as a hypothetically closed and autonomous (in terms of energy balance) system, in order to investigate the perspectives of sustainable development at the peripheral scale, merely based on renewable energy.
Related project: Combined REnewable Systems for Sustainable ENergy DevelOpment (CRESSENDO)
Full text: http://www.itia.ntua.gr/en/getfile/1613/1/documents/Report_EE4a.pdf (8010 KB)
A. Efstratiadis, N. Mamassis, Y. Markonis, P. Kossieris, and H. Tyralis, Methodological framework for optimal planning and management of water and renewable energy resources, Combined REnewable Systems for Sustainable ENergy DevelOpment (CRESSENDO), 154 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, April 2015.
We describe a stochastic simulation and optimization framework for hybrid renewable energy systems, based on effective coupling of different models. Initially, we explain the problem of combined management of water and energy resources, we introduce the main concepts and highlight the peculiarities of the problem, by means of methodology and computational implementation. Next is presented the general context, which is based on the combined use of an hourly simulation model for the renewables of a specific study area (wind and solar units), and a daily simulation model for the water resource system and the associated energy components. The models are fed by synthetic time series of hydrological inflows, wind velocity, solar radiation and electricity demand over the study area, for the generation of which are used appropriate stochastic schemes. The theoretical background of all models and related software systems is based on original methodologies or existing approaches that have been improved or generalized in the context of the research project.
Related project: Combined REnewable Systems for Sustainable ENergy DevelOpment (CRESSENDO)
Full text: http://www.itia.ntua.gr/en/getfile/1599/1/documents/Report_EE2.pdf (3766 KB)
H. Tyralis, and A. Efstratiadis, "National Programme for the Management and Protection of Water Resources" and "Impacts of climate change to surface and groundwater resources of Greece": Comparative presentation, September 2012.
Two documents are compared: (1) the report of the National Programme for the Management and Protection of Water Resources, elaborated by NTUA within a research project, and (2) the report entitled "Impacts of climate change to surface and groundwater resources of Greece", elaborated by a research team of Athens University (EKPA) in June 2011, for the Bank of Greece. A large part (~40%) of the two documents are identical.
Remarks:
The report of the National Programme for the Management and Protection of Water Resources: http://itia.ntua.gr/el/docinfo/782/
Web site of the Bank of Greece which contains, among other things, the report of the Stournaras team: http://www.bankofgreece.gr/Pages/el/klima/relevant.aspx (accessed 2012/09/07)
Full text: http://www.itia.ntua.gr/en/getfile/1285/1/documents/MasterPlanComparison_3.pdf (8176 KB)
Additional material:
H. Tyralis, A brief introduction to Bayesian statistics, 24 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, April 2011.
Full text: http://www.itia.ntua.gr/en/getfile/1148/1/documents/2011Tyralis_IntroBayesianStats.pdf (314 KB)