The document analyzes stochastic models for streamflow data from three rivers in India. It evaluates ten candidate Auto-Regressive Moving Average (ARMA) models to select the best models for forecasting and representing monthly and ten-day streamflow data from the rivers. For the three monthly streamflow series, the models with the maximum likelihood based on criteria are AR(4) for the Cauvery River, ARMA(2,1) for the Hemavathy River, and ARMA(3,1) for the Malaprabha River. For the ten-day Malaprabha River streamflow series, the best model selected is AR(4). The AR(1) model results in the minimum mean
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
Nonlinear State Space Modeling and Control of a Shape Memory Alloy Spring Act...theijes
In this paper, a new state-space model is developed for shape memory alloy (SMA) springs which remarkably captures their nonlinear behavior during phase transformation. The validity of the proposed model is verified by multiple open-loop experimental tests. This state-space model is then used to develop a model-based controller which adjusts the force exerted by the spring using actuating voltage in order to achieve the desired values of input force reference. It will be shown that in spite of nonlinearity of the model and hysteresis phenomenon, the perfect tracking is possible when the frequency of the input force reference is low. However, in high-frequency of input force reference, the perfect tracking is not guaranteed
Optimal operation of single reservoir using artificial neural network safayat...IAEME Publication
This document summarizes a study that developed artificial neural network (ANN) models to determine optimal operating policies for a single reservoir in India. Five different ANN models were created using different combinations of input data. The models either simulated final reservoir storage or optimal water release. The most accurate ANN model was selected by comparing results to a stochastic model. Regression analysis showed the preferred ANN model using current and previous period reservoir storage and inflow as inputs produced results close to the stochastic model, demonstrating ANNs can be an effective alternative for reservoir operation optimization.
OPTIMAL OPERATION OF SINGLE RESERVOIR USING ARTIFICIAL NEURAL NETWORKIAEME Publication
Optimal operating policies of a reservoir have been derived in deterministic and stochastic frame work as well as using Artificial Neural Network (ANN). With different combinations of input data set, five different ANN models have been developed. Out of five models, three models simulate
final storage and two models simulate optimal release. All these models are applied to develop optimal operating policies of Konar, a reservoir in Damodar Valley system in India. Based on data used as input variables, five types of ANN models are developed.
The models, principles and steps of Bayesian time series analysis and forecasting have been established extensively during the past fifty years. In order to estimate parameters of an autoregressive (AR) model we develop Markov chain Monte Carlo (MCMC) schemes for inference of AR model. It is our interest to propose a new prior distribution placed directly on the AR parameters of the model. Thus, we revisit the stationarity conditions to determine a flexible prior for AR model parameters. A MCMC procedure is proposed to estimate coefficients of AR(p) model. In order to set Bayesian steps, we determined prior distribution with the purpose of applying MCMC. We advocate the use of prior distribution placed directly on parameters. We have proposed a set of sufficient stationarity conditions for autoregressive models of any lag order. In this thesis, a set of new stationarity conditions have been proposed for the AR model. We motivated the new methodology by considering the autoregressive model of AR(2) and AR(3). Additionally, through simulation we studied sufficiency and necessity of the proposed conditions of stationarity. The researcher, additionally draw parameter space of AR(3) model for stationary region of Barndorff-Nielsen and Schou (1973) and our new suggested condition. A new prior distribution has been proposed placed directly on the parameters of the AR(p) model. This is motivated by priors proposed for the AR(1), AR(2),..., AR(6), which take advantage of the range of the AR parameters. We then develop a Metropolis step within Gibbs sampling for estimation. This scheme is illustrated using simulated data, for the AR(2), AR(3) and AR(4) models and extended to models with higher lag order. The thesis compared the new proposed prior distribution with the prior distributions obtained from the correspondence relationship between partial autocorrelations and parameters discussed by Barndorff-Nielsen and Schou (1973).
ALPHA LOGARITHM TRANSFORMED SEMI LOGISTIC DISTRIBUTION USING MAXIMUM LIKELIH...BRNSS Publication Hub
The document discusses the alpha logarithm transformed semi-logistic distribution and its maximum likelihood estimation method. It introduces the distribution, provides its probability density function and cumulative distribution function. It then describes generating random numbers from the distribution and outlines the maximum likelihood estimation method to estimate the distribution's unknown parameters. This involves deriving the likelihood function and taking its partial derivatives to obtain equations that are set to zero and solved to find maximum likelihood estimates of the location, scale, and shape parameters.
Assessment of two Methods to study Precipitation PredictionAI Publications
Presipitation analysis plays an important role in hydrological studies. In this study, using 50 years of rainfall data and ARIMA model, critical areas of Iran were determined. For this purpose, annual rainfall data of 112 different synoptic stations in Iran were gathered. To summarize, it could be concluded that: ARIMA model was an appropriate tool to forecast annual rainfall. According to obtained results from relative error, five stations were in critical condition. At 45 stations accrued rainfalls with amounts of less than half of average in the 50-year period. Therefore, in these 45 areas, chance of drought is more than other areas of Iran.
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
Nonlinear State Space Modeling and Control of a Shape Memory Alloy Spring Act...theijes
In this paper, a new state-space model is developed for shape memory alloy (SMA) springs which remarkably captures their nonlinear behavior during phase transformation. The validity of the proposed model is verified by multiple open-loop experimental tests. This state-space model is then used to develop a model-based controller which adjusts the force exerted by the spring using actuating voltage in order to achieve the desired values of input force reference. It will be shown that in spite of nonlinearity of the model and hysteresis phenomenon, the perfect tracking is possible when the frequency of the input force reference is low. However, in high-frequency of input force reference, the perfect tracking is not guaranteed
Optimal operation of single reservoir using artificial neural network safayat...IAEME Publication
This document summarizes a study that developed artificial neural network (ANN) models to determine optimal operating policies for a single reservoir in India. Five different ANN models were created using different combinations of input data. The models either simulated final reservoir storage or optimal water release. The most accurate ANN model was selected by comparing results to a stochastic model. Regression analysis showed the preferred ANN model using current and previous period reservoir storage and inflow as inputs produced results close to the stochastic model, demonstrating ANNs can be an effective alternative for reservoir operation optimization.
OPTIMAL OPERATION OF SINGLE RESERVOIR USING ARTIFICIAL NEURAL NETWORKIAEME Publication
Optimal operating policies of a reservoir have been derived in deterministic and stochastic frame work as well as using Artificial Neural Network (ANN). With different combinations of input data set, five different ANN models have been developed. Out of five models, three models simulate
final storage and two models simulate optimal release. All these models are applied to develop optimal operating policies of Konar, a reservoir in Damodar Valley system in India. Based on data used as input variables, five types of ANN models are developed.
The models, principles and steps of Bayesian time series analysis and forecasting have been established extensively during the past fifty years. In order to estimate parameters of an autoregressive (AR) model we develop Markov chain Monte Carlo (MCMC) schemes for inference of AR model. It is our interest to propose a new prior distribution placed directly on the AR parameters of the model. Thus, we revisit the stationarity conditions to determine a flexible prior for AR model parameters. A MCMC procedure is proposed to estimate coefficients of AR(p) model. In order to set Bayesian steps, we determined prior distribution with the purpose of applying MCMC. We advocate the use of prior distribution placed directly on parameters. We have proposed a set of sufficient stationarity conditions for autoregressive models of any lag order. In this thesis, a set of new stationarity conditions have been proposed for the AR model. We motivated the new methodology by considering the autoregressive model of AR(2) and AR(3). Additionally, through simulation we studied sufficiency and necessity of the proposed conditions of stationarity. The researcher, additionally draw parameter space of AR(3) model for stationary region of Barndorff-Nielsen and Schou (1973) and our new suggested condition. A new prior distribution has been proposed placed directly on the parameters of the AR(p) model. This is motivated by priors proposed for the AR(1), AR(2),..., AR(6), which take advantage of the range of the AR parameters. We then develop a Metropolis step within Gibbs sampling for estimation. This scheme is illustrated using simulated data, for the AR(2), AR(3) and AR(4) models and extended to models with higher lag order. The thesis compared the new proposed prior distribution with the prior distributions obtained from the correspondence relationship between partial autocorrelations and parameters discussed by Barndorff-Nielsen and Schou (1973).
ALPHA LOGARITHM TRANSFORMED SEMI LOGISTIC DISTRIBUTION USING MAXIMUM LIKELIH...BRNSS Publication Hub
The document discusses the alpha logarithm transformed semi-logistic distribution and its maximum likelihood estimation method. It introduces the distribution, provides its probability density function and cumulative distribution function. It then describes generating random numbers from the distribution and outlines the maximum likelihood estimation method to estimate the distribution's unknown parameters. This involves deriving the likelihood function and taking its partial derivatives to obtain equations that are set to zero and solved to find maximum likelihood estimates of the location, scale, and shape parameters.
Assessment of two Methods to study Precipitation PredictionAI Publications
Presipitation analysis plays an important role in hydrological studies. In this study, using 50 years of rainfall data and ARIMA model, critical areas of Iran were determined. For this purpose, annual rainfall data of 112 different synoptic stations in Iran were gathered. To summarize, it could be concluded that: ARIMA model was an appropriate tool to forecast annual rainfall. According to obtained results from relative error, five stations were in critical condition. At 45 stations accrued rainfalls with amounts of less than half of average in the 50-year period. Therefore, in these 45 areas, chance of drought is more than other areas of Iran.
Searching for Dynamical Resemblance Between Time Series: A Method Based on No...Gladstone Alves
This document proposes a method to search for dynamical resemblance between time series based on nonlinear autoregressive models (NARMAX). It involves using NARMAX models to represent time series, estimating models on different data windows, and comparing the eigenvalue variations between models. As a case study, it applies this to monthly river flow data divided into windows. It finds the dynamics of windows 1 and 6 are similar based on linear model forecasts and the proposed NARMAX approach, suggesting a hidden periodic cycle in the data.
Review and Assessment of Turbulence Transition ModelsIJERDJOURNAL
ABSTRACT: Turbulent boundary layer transition can greatly affect flow characteristics such as skin friction, heat transfer, pressure loads, and boundary layer separation point. Accurate computation of such effect is vital to the design of components and vehicles subjected to turbulent transition flows. In this paper we report our review of the existing boundary layer transition models, selection of the boundary layer transition models most appropriate for the existing Reynolds-averaged Navier-Stokes flow solvers, and implementation and evaluation of the selected transition models with some benchmark test cases, ranging from subsonic to hypersonic flows. The objective of this assessment study is not intended to pick the best transition model (in fact, there is no transition model accurately predict the boundary layer transition for all cases tested here). Rather it is to demonstrate that (1) the flow physics of turbulence transition is very complicated and not yet well understood, (2) the applicability of empirical transition models is limited and their use should be cautious due to different transition characteristics for different flow regimes/environments, and (3) further research on prediction of turbulence transition is warranted to improve the accuracy, efficiency and range of applicability.
This document describes a study that compares the performance of threshold autoregressive (TAR) models for fitting nonlinear and non-stationary time series data. Specifically, it examines the smooth transition autoregressive (STAR) model and self-exciting threshold autoregressive (SETAR) model at different orders when fitted to simulated linear, trigonometric, exponential, and polynomial autoregressive data. The performance of the models is assessed using criteria like mean squared error, mean absolute percentage error, and Akaike information criterion. The results show that the SETAR(2,1) model performs best for linear data, while the SETAR(2,2) and STAR(2,3) models perform best for exponential data and the
This document discusses data-driven models and machine learning approaches for water resource systems. It provides an overview of data-driven models and why machine learning is useful for modeling water systems. Specific machine learning algorithms discussed include artificial neural networks, Bayesian multi-layer perceptrons, and relevance vector machines. These algorithms are able to model nonlinear relationships and nonstationary behaviors in water systems. The document also differentiates between supervised and unsupervised learning methods.
Analysis and simulation of strong earthquake ground motions using arma models...Olides Rodriguez
Segmentation of earthquake ground motion time series data and representation of the quasi-stationary segments using ARMA models provides an efficient approach for characterizing strong earthquake ground motions. The document presents a procedure for analyzing and simulating earthquake ground motions that involves segmenting the nonstationary acceleration record into quasi-stationary blocks, fitting ARMA models to each block using maximum likelihood estimation, and using the ARMA models to simulate synthetic accelerograms. The procedure is applied to ground acceleration data from a 1977 Romanian earthquake, with the original and synthetic data found to match well based on statistical comparisons.
Urban strategies to promote resilient cities The case of enhancing Historic C...inventionjournals
This research tackles disaster prevention problems in dense urban areas, concentrating on the urban fire challenge in Historic Cairo district, Egypt, through disaster risk management approach. The study area suffers from the strike of several urban fire outbreaks, that resulted in disfiguring historic monuments and destroying unregulated traditional markets. Therefore, the study investigates the significance of hazard management and how can urban strategies improve the city resilient through reducing the impact of natural and man-made threats. The main findings of the research are the determination of the vulnerability factors in Historic Cairo district, either regarding management deficiency or issues related to the existing urban form. It is found that the absence of the mitigation and preparedness phases is the main problem in the risk management cycle in the case study. Additionally, the coping initiatives adopted by local authorities to address risks are random and insufficient. The study concludes with recommendations which invoke incorporating hazard management stages (pre disaster, during disaster and post disaster) into the process of evolving development planning. Finally, solutions are offered to mitigate, prepare, respond and recover from fire disasters in the case study. The solutions include urban policies, land-use planning, urban design outlines, safety regulation and public awareness and training.
TREATMENT BY ALTERNATIVE METHODS OF REGRESSION GAS CHROMATOGRAPHIC RETENTION ...ijsc
The study treated two closer alternative methods of which the principal characteristic: a non-parametric
method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to
model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other
pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical
molecular descriptors calculated using the software DRAGON. The detection of influential observations for
non-parametric method (LAD) is a problem which has been extensively studied and offers alternative
dicapproaches whose main feature is the robustness.here is presented and compared with the standard
least squares regression .The comparison between methods LAD and OLS is based on the equation of the
hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points
of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling,
shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval
thanks to the concept of robustness to check if the distribution of the errors is really approximate.
Markov-switching multifractal (MSM) is a model of asset returns that incorporates stochastic volatility components of heterogeneous durations.[1][2] MSM captures the outliers, log-memory-like volatility persistence and power variation of financial returns
Treatment by alternative methods of regression gas chromatographic retention ...ijics
The study treated two closer alternative methods of which the principal characteristic: a non-parametric
method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to
model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other
pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical
molecular descriptors calculated using the software DRAGON. The detection of influential observations for
non-parametric method (LAD) is a problem which has been extensively studied and offers alternative
dicapproaches whose main feature is the robustness .here is presented and compared with the standard
least squares regression .The comparison between methods LAD and OLS is based on the equation of the
hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points
of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling,
shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval
thanks to the concept of robustness to check if the distribution of the errors is really approximate.
TREATMENT BY ALTERNATIVE METHODS OF REGRESSION GAS CHROMATOGRAPHIC RETENTION ...ijcisjournal
The study treated two closer alternative methods of which the principal characteristic: a non-parametric method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical molecular descriptors calculated using the software DRAGON. The detection of influential observations for non-parametric method (LAD) is a problem which has been extensively studied and offers alternative dicapproaches whose main feature is the robustness .here is presented and compared with the standard least squares regression .The comparison between methods LAD and OLS is based on the equation of the hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling, shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval thanks to the concept of robustness to check if the distribution of the errors is really approximate.
The Use of ARCH and GARCH Models for Estimating and Forecasting Volatility-ru...Ismet Kale
This document discusses volatility modeling using ARCH and GARCH models. It first provides background on ARCH and GARCH models, noting they were developed to model characteristics of financial time series data like volatility clustering and fat tails. It then describes the specific ARCH and GARCH models that will be used in the study, including the ARCH, GARCH, EGARCH, GJR, APARCH, IGARCH, FIGARCH and FIAPARCH models. The document aims to apply these models to daily stock index data from the IMKB 100 to analyze and forecast volatility, and better understand risk in the Turkish market.
This document describes research using an artificial neural network (ANN) model to predict the length of hydraulic jumps on rough beds. The ANN model takes the Froude number and relative roughness as inputs and predicts the non-dimensional jump length as the output. Experimental data from previous studies was used to train, validate and test the ANN model. The ANN model was found to predict jump length with higher accuracy than an existing empirical equation, with a coefficient of determination of 0.9596, mean absolute percentage error of 6.9231, and root mean square error of 3.2438. A sensitivity analysis showed that the Froude number has a greater influence on jump length prediction than relative roughness.
Treatment by Alternative Methods of Regression Gas Chromatographic Retention ...ijsc
The study treated two closer alternative methods of which the principal characteristic: a non-parametric method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical molecular descriptors calculated using the software DRAGON. The detection of influential observations for non-parametric method (LAD) is a problem which has been extensively studied and offers alternative dicapproaches whose main feature is the robustness.here is presented and compared with the standard least squares regression .The comparison between methods LAD and OLS is based on the equation of the hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling, shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval thanks to the concept of robustness to check if the distribution of the errors is really approximate.
ANN Modeling of Monthly and Weekly Behaviour of the Runoff of Kali River Catc...IOSR Journals
Model is a system, by whose operation; the characteristics of other similar systems can be ascertained. Experimental observation made on a model bear a definite relationship with prototype. So, the model analysis or modeling is actually an experimental method of finding solution of complex flow problems like surface water modeling, sub-surface water modeling etc. Many flow situations are not amenable to theoretical analysis. Modeling is a valuable means of obtaining better understanding of particular situation. Inspired by the functioning of the brain and biological nervous system, Artificial Neural Networks (ANNs) has been applied to various hydrological problems in last two decades. In this study, two ANN models using feed forward – back propagation network are developed to correlate a relationship between rainfall and runoff on monthly and weekly basis for Kali river catchment up to Supa dam in Uttara Kannada District of Karnataka State, India. The developed two models are compared and evaluated using standard statistical parameters to know strength and weaknesses. This performance can be further refined by incorporating more input parameters of catchment properties like soil moisture index; land use and land cover details etc.
Forecasting the commercial Chisawasawa(Lethrinopsspp.) fishery in Lake MalaŵiIJESM JOURNAL
This document summarizes a study that developed time series models to forecast commercial catches of Chisawasawa fish in Lake Malawi. Researchers used data from 1976 to 2010 to identify the best ARIMA model for forecasting annual commercial catches. The ARIMA (1,1) model was selected as it had the lowest normalized Bayesian information criterion value. Forecasts predicted that commercial catches would increase to 1788 tons by 2020, though the confidence intervals included zero, indicating the fishery may have collapsed. The study recommends urgent attention for this fishery and developing similar models for other fisheries in Lake Malawi.
EFFICACY OF NEURAL NETWORK IN RAINFALL-RUNOFF MODELLING OF BAGMATI RIVER BASINIAEME Publication
In this paper, rainfall-runoff model of Bagmati river basin has been developed
using the ANN Technique. Three-layered fced forward network structure with back
propagation algorithm was used to train the ANN model. Different combinations of
rainfall and runoff were considered as input to the network and trained by BP
algorithm with different error tolerance, learning parameter, number of cycles and
number of hidden layers. The sensitivity of the prediction accuracy to the number of
hidden layer neurons in a back error propagation algorithm was used for the study.
The monthly rainfall and runoff data from 2000 to 2009 of Bagmati river basin has
been considered for the development of ANN model. Performance evaluation of the
model has been done by using statistical parameters. Three sets of data have been
used to make several combination of year keeping in view the highest peaks of
hydrographs. First set of data used was from 2000 to 2006 for the calibration and
from 2007 to 2009 for validation. The second set of data was from 2004 to 2009 for
calibration and from 2000 to 2003 for validation. The Third set of data was from 2000
to 2009 for calibration and from 2007 to 2009 for validation. It was found that the
third set of data gave better result than other two sets of data. The study demonstrates
the applicability of ANN approach in developing effective non-linear models of
Rainfall-Runoff process without the need to explicitly representing the internal
hydraulic structure of the watershed
The document describes the development of four forecasting models for monthly precipitation data in Sulaimania governorate, Iraq. Three models are single-variable autoregressive models for the first, second, and third highest daily precipitation values each month. The fourth model is a multi-variable model that incorporates all three precipitation series simultaneously. The models were developed using data from 1992-2006 and verified against data from 2007-2013. Results showed that the single-variable models better preserved statistical properties, though both types of models performed well. The multi-variable model had higher errors, likely due to the high randomness of extreme precipitation values.
Application of smooth transition autoregressive (star) models for exchange rateAlexander Decker
This document discusses applying Smooth Transition Autoregressive (STAR) models to analyze exchange rate data. Specifically, it evaluates the suitability of Logistic STAR (LSTAR) and Exponential STAR (ESTAR) models. The document outlines the methodology for testing whether exchange rate data exhibits linear or nonlinear properties. It describes testing procedures that involve estimating an auxiliary regression model based on a Taylor series expansion to detect nonlinear behavior. The study applies these LSTAR and ESTAR models to exchange rate data to determine which model provides a better fit and explanation of deviations from the mean exchange rate. The document concludes that an ESTAR model provides the best adjustment for the analyzed exchange rate data series.
This document discusses the use of autoregressive integrated moving average (ARIMA) models in statistical analysis beyond just time series data. It provides examples of using ARIMA models with non-temporal data, where the independent variable is something other than time, such as temperature or longitude. Key points include:
1) ARIMA models only require evenly spaced intervals for the independent variable and do not necessarily need time as the variable. Examples of non-temporal ARIMA models are given for white dwarf star populations and the distribution of attorneys.
2) Temperature can act as a "time proxy" for white dwarf stars since temperature and time are monotonically related as the stars cool.
3) ARIM
ARIMA is a statistical model used to forecast future values in a time series by examining patterns in past observations. It stands for autoregressive integrated moving average. The ARIMA model incorporates autoregressive (AR) terms to model dependence on past values, differences (I) to make the data stationary, and moving average (MA) terms to model lagged forecast errors. The ARIMA(p,d,q) parameters include p for AR order, d for differencing, and q for MA order. ARIMA is useful when data exhibits trends or seasonality and can be applied after differencing to make the time series stationary.
Sen, Z. (1974) Propiedades de muestras pequeñas de modelos estocásticos estac...SandroSnchezZamora
This document provides an abstract for a thesis submitted for the degree of Doctor of Philosophy in the University of London in November 1974.
The thesis examines the small sample properties of parameters for stationary stochastic hydrological models. It aims to improve techniques for simulating observed streamflow sequences and substitute computationally expensive Monte Carlo techniques with exact or approximate analytical expressions.
Specifically, it analytically derives the small sample expectations of parameters like variance, standard deviation, and serial correlation coefficients for normal stationary processes. It also derives expectations for variables related to Hurst's phenomenon, including the range and resealed range of cumulative departures from the sample mean. A new model called the white Markov process is proposed for modeling Hurst's law
Molz, FJ, Rajaram, H. y Lu, S. (2004). Modelos de heterogeneidad basados en...SandroSnchezZamora
This document summarizes research on using fractal-based stochastic models to characterize heterogeneity in subsurface hydrology. It discusses how modern measurement techniques have revealed highly irregular and nonstationary property distributions in porous media. Researchers have explored using nonstationary stochastic processes with stationary increments to model these, which relates to the mathematical theory of stochastic fractals. The document reviews applications of Gaussian and Levy stochastic fractals and multifractals to model properties like porosity and hydraulic conductivity. It also discusses limitations like models based on increments being less predictive than stationary models, and issues around measuring media that exhibit variations on all scales. The conclusion is that additional data is needed to advance fractal-based theories, especially for 3D fracture networks.
More Related Content
Similar to Mujumdar, PP y Kumar, DN Modelos estocásticos de caudal algunos estudios de caso..pdf
Searching for Dynamical Resemblance Between Time Series: A Method Based on No...Gladstone Alves
This document proposes a method to search for dynamical resemblance between time series based on nonlinear autoregressive models (NARMAX). It involves using NARMAX models to represent time series, estimating models on different data windows, and comparing the eigenvalue variations between models. As a case study, it applies this to monthly river flow data divided into windows. It finds the dynamics of windows 1 and 6 are similar based on linear model forecasts and the proposed NARMAX approach, suggesting a hidden periodic cycle in the data.
Review and Assessment of Turbulence Transition ModelsIJERDJOURNAL
ABSTRACT: Turbulent boundary layer transition can greatly affect flow characteristics such as skin friction, heat transfer, pressure loads, and boundary layer separation point. Accurate computation of such effect is vital to the design of components and vehicles subjected to turbulent transition flows. In this paper we report our review of the existing boundary layer transition models, selection of the boundary layer transition models most appropriate for the existing Reynolds-averaged Navier-Stokes flow solvers, and implementation and evaluation of the selected transition models with some benchmark test cases, ranging from subsonic to hypersonic flows. The objective of this assessment study is not intended to pick the best transition model (in fact, there is no transition model accurately predict the boundary layer transition for all cases tested here). Rather it is to demonstrate that (1) the flow physics of turbulence transition is very complicated and not yet well understood, (2) the applicability of empirical transition models is limited and their use should be cautious due to different transition characteristics for different flow regimes/environments, and (3) further research on prediction of turbulence transition is warranted to improve the accuracy, efficiency and range of applicability.
This document describes a study that compares the performance of threshold autoregressive (TAR) models for fitting nonlinear and non-stationary time series data. Specifically, it examines the smooth transition autoregressive (STAR) model and self-exciting threshold autoregressive (SETAR) model at different orders when fitted to simulated linear, trigonometric, exponential, and polynomial autoregressive data. The performance of the models is assessed using criteria like mean squared error, mean absolute percentage error, and Akaike information criterion. The results show that the SETAR(2,1) model performs best for linear data, while the SETAR(2,2) and STAR(2,3) models perform best for exponential data and the
This document discusses data-driven models and machine learning approaches for water resource systems. It provides an overview of data-driven models and why machine learning is useful for modeling water systems. Specific machine learning algorithms discussed include artificial neural networks, Bayesian multi-layer perceptrons, and relevance vector machines. These algorithms are able to model nonlinear relationships and nonstationary behaviors in water systems. The document also differentiates between supervised and unsupervised learning methods.
Analysis and simulation of strong earthquake ground motions using arma models...Olides Rodriguez
Segmentation of earthquake ground motion time series data and representation of the quasi-stationary segments using ARMA models provides an efficient approach for characterizing strong earthquake ground motions. The document presents a procedure for analyzing and simulating earthquake ground motions that involves segmenting the nonstationary acceleration record into quasi-stationary blocks, fitting ARMA models to each block using maximum likelihood estimation, and using the ARMA models to simulate synthetic accelerograms. The procedure is applied to ground acceleration data from a 1977 Romanian earthquake, with the original and synthetic data found to match well based on statistical comparisons.
Urban strategies to promote resilient cities The case of enhancing Historic C...inventionjournals
This research tackles disaster prevention problems in dense urban areas, concentrating on the urban fire challenge in Historic Cairo district, Egypt, through disaster risk management approach. The study area suffers from the strike of several urban fire outbreaks, that resulted in disfiguring historic monuments and destroying unregulated traditional markets. Therefore, the study investigates the significance of hazard management and how can urban strategies improve the city resilient through reducing the impact of natural and man-made threats. The main findings of the research are the determination of the vulnerability factors in Historic Cairo district, either regarding management deficiency or issues related to the existing urban form. It is found that the absence of the mitigation and preparedness phases is the main problem in the risk management cycle in the case study. Additionally, the coping initiatives adopted by local authorities to address risks are random and insufficient. The study concludes with recommendations which invoke incorporating hazard management stages (pre disaster, during disaster and post disaster) into the process of evolving development planning. Finally, solutions are offered to mitigate, prepare, respond and recover from fire disasters in the case study. The solutions include urban policies, land-use planning, urban design outlines, safety regulation and public awareness and training.
TREATMENT BY ALTERNATIVE METHODS OF REGRESSION GAS CHROMATOGRAPHIC RETENTION ...ijsc
The study treated two closer alternative methods of which the principal characteristic: a non-parametric
method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to
model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other
pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical
molecular descriptors calculated using the software DRAGON. The detection of influential observations for
non-parametric method (LAD) is a problem which has been extensively studied and offers alternative
dicapproaches whose main feature is the robustness.here is presented and compared with the standard
least squares regression .The comparison between methods LAD and OLS is based on the equation of the
hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points
of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling,
shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval
thanks to the concept of robustness to check if the distribution of the errors is really approximate.
Markov-switching multifractal (MSM) is a model of asset returns that incorporates stochastic volatility components of heterogeneous durations.[1][2] MSM captures the outliers, log-memory-like volatility persistence and power variation of financial returns
Treatment by alternative methods of regression gas chromatographic retention ...ijics
The study treated two closer alternative methods of which the principal characteristic: a non-parametric
method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to
model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other
pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical
molecular descriptors calculated using the software DRAGON. The detection of influential observations for
non-parametric method (LAD) is a problem which has been extensively studied and offers alternative
dicapproaches whose main feature is the robustness .here is presented and compared with the standard
least squares regression .The comparison between methods LAD and OLS is based on the equation of the
hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points
of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling,
shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval
thanks to the concept of robustness to check if the distribution of the errors is really approximate.
TREATMENT BY ALTERNATIVE METHODS OF REGRESSION GAS CHROMATOGRAPHIC RETENTION ...ijcisjournal
The study treated two closer alternative methods of which the principal characteristic: a non-parametric method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical molecular descriptors calculated using the software DRAGON. The detection of influential observations for non-parametric method (LAD) is a problem which has been extensively studied and offers alternative dicapproaches whose main feature is the robustness .here is presented and compared with the standard least squares regression .The comparison between methods LAD and OLS is based on the equation of the hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling, shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval thanks to the concept of robustness to check if the distribution of the errors is really approximate.
The Use of ARCH and GARCH Models for Estimating and Forecasting Volatility-ru...Ismet Kale
This document discusses volatility modeling using ARCH and GARCH models. It first provides background on ARCH and GARCH models, noting they were developed to model characteristics of financial time series data like volatility clustering and fat tails. It then describes the specific ARCH and GARCH models that will be used in the study, including the ARCH, GARCH, EGARCH, GJR, APARCH, IGARCH, FIGARCH and FIAPARCH models. The document aims to apply these models to daily stock index data from the IMKB 100 to analyze and forecast volatility, and better understand risk in the Turkish market.
This document describes research using an artificial neural network (ANN) model to predict the length of hydraulic jumps on rough beds. The ANN model takes the Froude number and relative roughness as inputs and predicts the non-dimensional jump length as the output. Experimental data from previous studies was used to train, validate and test the ANN model. The ANN model was found to predict jump length with higher accuracy than an existing empirical equation, with a coefficient of determination of 0.9596, mean absolute percentage error of 6.9231, and root mean square error of 3.2438. A sensitivity analysis showed that the Froude number has a greater influence on jump length prediction than relative roughness.
Treatment by Alternative Methods of Regression Gas Chromatographic Retention ...ijsc
The study treated two closer alternative methods of which the principal characteristic: a non-parametric method (the least absolute deviation (LAD)) and a traditional method of diagnosis OLS.This was applied to model, separately, the indices of retention of the same whole of 35 pyrazines (27 pyrazines with 8 other pyrazines in the same unit) eluted to the columns OV-101 and Carbowax-20M, by using theoretical molecular descriptors calculated using the software DRAGON. The detection of influential observations for non-parametric method (LAD) is a problem which has been extensively studied and offers alternative dicapproaches whose main feature is the robustness.here is presented and compared with the standard least squares regression .The comparison between methods LAD and OLS is based on the equation of the hyperplane, in order to confirm the robustness thus to detect by the meaningless statements and the points of lever and validated results in the state approached by the tests statistics: Test of Anderson-Darling, shapiro-wilk, Agostino, Jarque-Bera, graphic test (histogram of frequency) and the confidence interval thanks to the concept of robustness to check if the distribution of the errors is really approximate.
ANN Modeling of Monthly and Weekly Behaviour of the Runoff of Kali River Catc...IOSR Journals
Model is a system, by whose operation; the characteristics of other similar systems can be ascertained. Experimental observation made on a model bear a definite relationship with prototype. So, the model analysis or modeling is actually an experimental method of finding solution of complex flow problems like surface water modeling, sub-surface water modeling etc. Many flow situations are not amenable to theoretical analysis. Modeling is a valuable means of obtaining better understanding of particular situation. Inspired by the functioning of the brain and biological nervous system, Artificial Neural Networks (ANNs) has been applied to various hydrological problems in last two decades. In this study, two ANN models using feed forward – back propagation network are developed to correlate a relationship between rainfall and runoff on monthly and weekly basis for Kali river catchment up to Supa dam in Uttara Kannada District of Karnataka State, India. The developed two models are compared and evaluated using standard statistical parameters to know strength and weaknesses. This performance can be further refined by incorporating more input parameters of catchment properties like soil moisture index; land use and land cover details etc.
Forecasting the commercial Chisawasawa(Lethrinopsspp.) fishery in Lake MalaŵiIJESM JOURNAL
This document summarizes a study that developed time series models to forecast commercial catches of Chisawasawa fish in Lake Malawi. Researchers used data from 1976 to 2010 to identify the best ARIMA model for forecasting annual commercial catches. The ARIMA (1,1) model was selected as it had the lowest normalized Bayesian information criterion value. Forecasts predicted that commercial catches would increase to 1788 tons by 2020, though the confidence intervals included zero, indicating the fishery may have collapsed. The study recommends urgent attention for this fishery and developing similar models for other fisheries in Lake Malawi.
EFFICACY OF NEURAL NETWORK IN RAINFALL-RUNOFF MODELLING OF BAGMATI RIVER BASINIAEME Publication
In this paper, rainfall-runoff model of Bagmati river basin has been developed
using the ANN Technique. Three-layered fced forward network structure with back
propagation algorithm was used to train the ANN model. Different combinations of
rainfall and runoff were considered as input to the network and trained by BP
algorithm with different error tolerance, learning parameter, number of cycles and
number of hidden layers. The sensitivity of the prediction accuracy to the number of
hidden layer neurons in a back error propagation algorithm was used for the study.
The monthly rainfall and runoff data from 2000 to 2009 of Bagmati river basin has
been considered for the development of ANN model. Performance evaluation of the
model has been done by using statistical parameters. Three sets of data have been
used to make several combination of year keeping in view the highest peaks of
hydrographs. First set of data used was from 2000 to 2006 for the calibration and
from 2007 to 2009 for validation. The second set of data was from 2004 to 2009 for
calibration and from 2000 to 2003 for validation. The Third set of data was from 2000
to 2009 for calibration and from 2007 to 2009 for validation. It was found that the
third set of data gave better result than other two sets of data. The study demonstrates
the applicability of ANN approach in developing effective non-linear models of
Rainfall-Runoff process without the need to explicitly representing the internal
hydraulic structure of the watershed
The document describes the development of four forecasting models for monthly precipitation data in Sulaimania governorate, Iraq. Three models are single-variable autoregressive models for the first, second, and third highest daily precipitation values each month. The fourth model is a multi-variable model that incorporates all three precipitation series simultaneously. The models were developed using data from 1992-2006 and verified against data from 2007-2013. Results showed that the single-variable models better preserved statistical properties, though both types of models performed well. The multi-variable model had higher errors, likely due to the high randomness of extreme precipitation values.
Application of smooth transition autoregressive (star) models for exchange rateAlexander Decker
This document discusses applying Smooth Transition Autoregressive (STAR) models to analyze exchange rate data. Specifically, it evaluates the suitability of Logistic STAR (LSTAR) and Exponential STAR (ESTAR) models. The document outlines the methodology for testing whether exchange rate data exhibits linear or nonlinear properties. It describes testing procedures that involve estimating an auxiliary regression model based on a Taylor series expansion to detect nonlinear behavior. The study applies these LSTAR and ESTAR models to exchange rate data to determine which model provides a better fit and explanation of deviations from the mean exchange rate. The document concludes that an ESTAR model provides the best adjustment for the analyzed exchange rate data series.
This document discusses the use of autoregressive integrated moving average (ARIMA) models in statistical analysis beyond just time series data. It provides examples of using ARIMA models with non-temporal data, where the independent variable is something other than time, such as temperature or longitude. Key points include:
1) ARIMA models only require evenly spaced intervals for the independent variable and do not necessarily need time as the variable. Examples of non-temporal ARIMA models are given for white dwarf star populations and the distribution of attorneys.
2) Temperature can act as a "time proxy" for white dwarf stars since temperature and time are monotonically related as the stars cool.
3) ARIM
ARIMA is a statistical model used to forecast future values in a time series by examining patterns in past observations. It stands for autoregressive integrated moving average. The ARIMA model incorporates autoregressive (AR) terms to model dependence on past values, differences (I) to make the data stationary, and moving average (MA) terms to model lagged forecast errors. The ARIMA(p,d,q) parameters include p for AR order, d for differencing, and q for MA order. ARIMA is useful when data exhibits trends or seasonality and can be applied after differencing to make the time series stationary.
Similar to Mujumdar, PP y Kumar, DN Modelos estocásticos de caudal algunos estudios de caso..pdf (20)
Sen, Z. (1974) Propiedades de muestras pequeñas de modelos estocásticos estac...SandroSnchezZamora
This document provides an abstract for a thesis submitted for the degree of Doctor of Philosophy in the University of London in November 1974.
The thesis examines the small sample properties of parameters for stationary stochastic hydrological models. It aims to improve techniques for simulating observed streamflow sequences and substitute computationally expensive Monte Carlo techniques with exact or approximate analytical expressions.
Specifically, it analytically derives the small sample expectations of parameters like variance, standard deviation, and serial correlation coefficients for normal stationary processes. It also derives expectations for variables related to Hurst's phenomenon, including the range and resealed range of cumulative departures from the sample mean. A new model called the white Markov process is proposed for modeling Hurst's law
Molz, FJ, Rajaram, H. y Lu, S. (2004). Modelos de heterogeneidad basados en...SandroSnchezZamora
This document summarizes research on using fractal-based stochastic models to characterize heterogeneity in subsurface hydrology. It discusses how modern measurement techniques have revealed highly irregular and nonstationary property distributions in porous media. Researchers have explored using nonstationary stochastic processes with stationary increments to model these, which relates to the mathematical theory of stochastic fractals. The document reviews applications of Gaussian and Levy stochastic fractals and multifractals to model properties like porosity and hydraulic conductivity. It also discusses limitations like models based on increments being less predictive than stationary models, and issues around measuring media that exhibit variations on all scales. The conclusion is that additional data is needed to advance fractal-based theories, especially for 3D fracture networks.
Lawrance, AJ y Kottegoda, NT (1977). Modelado estocástico de series temporale...SandroSnchezZamora
The document discusses statistical modeling of riverflow time series in hydrology. It aims to produce stochastic simulations that resemble important properties of historical riverflow data, such as patterns of low and high flows. These simulations are then used as inputs to model water resource systems. The paper describes the non-Gaussian and periodic features of typical riverflow time series, including seasonality, distributions, dependence, and crossing properties. It discusses both traditional time series models and hydrologically motivated long memory and broken line models that have been adapted for use in hydrology. The paper aims to bring these models to the attention of statisticians while also making their theory more accessible to hydrologists. It identifies areas where current methodology could be consolidated and where further
Todorovic, P. (1978). Modelos estocásticos de inundaciones. Investigación de ...SandroSnchezZamora
This document discusses stochastic models for analyzing floods based on partial duration series of streamflow data. It presents three stochastic models that depend on assumptions about exceedances above a base level in the streamflow data. The models improve upon each other by having less restrictive assumptions. The distribution of the largest flood volume in a time interval is determined. Comparison of theoretical and observed distributions shows the assumptions are not unduly restrictive.
Paschalis, A., Molnar, P., Fatichi, S. y Burlando, P. (2013). Un modelo estoc...SandroSnchezZamora
This document presents a new stochastic space-time model called STREAP for simulating high-resolution precipitation fields. STREAP is a three-stage hierarchical model that mimics the precipitation formation process. The first stage simulates storm arrival as an alternating renewal process. The second stage models the temporal evolution of mean areal precipitation intensity and wet area using a bivariate Gaussian process. The third stage simulates the two-dimensional storm structure over time as a random field. STREAP was applied to weather radar data in Switzerland and was able to reproduce important statistical features of precipitation across spatial and temporal scales. It performed better than an existing space-time point process model in describing spatial precipitation patterns.
Yevjevich, V. (1987). Modelos estocásticos en hidrología. Hidrología e Hidráu...SandroSnchezZamora
This document discusses stochastic models in hydrology. It defines stochasticity in hydrologic processes and compares it to determinism. Hydrologic time series are influenced by astronomical, environmental, and anthropogenic factors and exhibit tendencies like trends and periodicities as well as randomness. Effective stochastic models decompose time series into trend, periodic, and stochastic components. Key aspects of stochastic hydrology include extracting information from data, transferring information between variables, simulating processes, and forecasting random variables. Non-stationarities from periodic patterns, trends, and rare disruptive events pose challenges for modeling.
This document discusses stochastic models that can be used to represent growth and steady state phenomena in hydrology. It reviews two types of growth models: the cyclic growth model and the random configuration model. For steady state phenomena like time series, models are generally restricted to a Gaussian type with or without autocorrelation. Self-similarity models can lead to physically absurd conditions if extrapolated to high frequencies. The document focuses on cyclic growth models, where growth occurs through successive similar stages over time intervals, producing a self-similar system that can be viewed as composed of similar repeating "cells".
Machiwal, D. y Jha, MK (2012). Modelado estocástico de series de tiempo. En A...SandroSnchezZamora
This document provides an overview of stochastic modelling and different stochastic processes that are commonly used, including:
- Purely random (white noise) processes where data points are independent and identically distributed
- Autoregressive (AR) processes where each data point is modeled as a linear combination of previous data points plus noise
- Moving average (MA) processes where each data point is modeled as a linear combination of previous noise terms plus a constant
- Autoregressive moving average (ARMA) processes which combine AR and MA processes
- Autoregressive integrated moving average (ARIMA) processes which explicitly include differencing to make time series stationary
Comparative analysis between traditional aquaponics and reconstructed aquapon...bijceesjournal
The aquaponic system of planting is a method that does not require soil usage. It is a method that only needs water, fish, lava rocks (a substitute for soil), and plants. Aquaponic systems are sustainable and environmentally friendly. Its use not only helps to plant in small spaces but also helps reduce artificial chemical use and minimizes excess water use, as aquaponics consumes 90% less water than soil-based gardening. The study applied a descriptive and experimental design to assess and compare conventional and reconstructed aquaponic methods for reproducing tomatoes. The researchers created an observation checklist to determine the significant factors of the study. The study aims to determine the significant difference between traditional aquaponics and reconstructed aquaponics systems propagating tomatoes in terms of height, weight, girth, and number of fruits. The reconstructed aquaponics system’s higher growth yield results in a much more nourished crop than the traditional aquaponics system. It is superior in its number of fruits, height, weight, and girth measurement. Moreover, the reconstructed aquaponics system is proven to eliminate all the hindrances present in the traditional aquaponics system, which are overcrowding of fish, algae growth, pest problems, contaminated water, and dead fish.
Software Engineering and Project Management - Introduction, Modeling Concepts...Prakhyath Rai
Introduction, Modeling Concepts and Class Modeling: What is Object orientation? What is OO development? OO Themes; Evidence for usefulness of OO development; OO modeling history. Modeling
as Design technique: Modeling, abstraction, The Three models. Class Modeling: Object and Class Concept, Link and associations concepts, Generalization and Inheritance, A sample class model, Navigation of class models, and UML diagrams
Building the Analysis Models: Requirement Analysis, Analysis Model Approaches, Data modeling Concepts, Object Oriented Analysis, Scenario-Based Modeling, Flow-Oriented Modeling, class Based Modeling, Creating a Behavioral Model.
AI for Legal Research with applications, toolsmahaffeycheryld
AI applications in legal research include rapid document analysis, case law review, and statute interpretation. AI-powered tools can sift through vast legal databases to find relevant precedents and citations, enhancing research accuracy and speed. They assist in legal writing by drafting and proofreading documents. Predictive analytics help foresee case outcomes based on historical data, aiding in strategic decision-making. AI also automates routine tasks like contract review and due diligence, freeing up lawyers to focus on complex legal issues. These applications make legal research more efficient, cost-effective, and accessible.
Redefining brain tumor segmentation: a cutting-edge convolutional neural netw...IJECEIAES
Medical image analysis has witnessed significant advancements with deep learning techniques. In the domain of brain tumor segmentation, the ability to
precisely delineate tumor boundaries from magnetic resonance imaging (MRI)
scans holds profound implications for diagnosis. This study presents an ensemble convolutional neural network (CNN) with transfer learning, integrating
the state-of-the-art Deeplabv3+ architecture with the ResNet18 backbone. The
model is rigorously trained and evaluated, exhibiting remarkable performance
metrics, including an impressive global accuracy of 99.286%, a high-class accuracy of 82.191%, a mean intersection over union (IoU) of 79.900%, a weighted
IoU of 98.620%, and a Boundary F1 (BF) score of 83.303%. Notably, a detailed comparative analysis with existing methods showcases the superiority of
our proposed model. These findings underscore the model’s competence in precise brain tumor localization, underscoring its potential to revolutionize medical
image analysis and enhance healthcare outcomes. This research paves the way
for future exploration and optimization of advanced CNN models in medical
imaging, emphasizing addressing false positives and resource efficiency.
Advanced control scheme of doubly fed induction generator for wind turbine us...IJECEIAES
This paper describes a speed control device for generating electrical energy on an electricity network based on the doubly fed induction generator (DFIG) used for wind power conversion systems. At first, a double-fed induction generator model was constructed. A control law is formulated to govern the flow of energy between the stator of a DFIG and the energy network using three types of controllers: proportional integral (PI), sliding mode controller (SMC) and second order sliding mode controller (SOSMC). Their different results in terms of power reference tracking, reaction to unexpected speed fluctuations, sensitivity to perturbations, and resilience against machine parameter alterations are compared. MATLAB/Simulink was used to conduct the simulations for the preceding study. Multiple simulations have shown very satisfying results, and the investigations demonstrate the efficacy and power-enhancing capabilities of the suggested control system.
Design and optimization of ion propulsion dronebjmsejournal
Electric propulsion technology is widely used in many kinds of vehicles in recent years, and aircrafts are no exception. Technically, UAVs are electrically propelled but tend to produce a significant amount of noise and vibrations. Ion propulsion technology for drones is a potential solution to this problem. Ion propulsion technology is proven to be feasible in the earth’s atmosphere. The study presented in this article shows the design of EHD thrusters and power supply for ion propulsion drones along with performance optimization of high-voltage power supply for endurance in earth’s atmosphere.
Mujumdar, PP y Kumar, DN Modelos estocásticos de caudal algunos estudios de caso..pdf
1. Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=thsj20
Download by: [188.72.127.226] Date: 13 June 2016, At: 00:28
Hydrological Sciences Journal
ISSN: 0262-6667 (Print) 2150-3435 (Online) Journal homepage: http://www.tandfonline.com/loi/thsj20
Stochastic models of streamflow: some case
studies
P. P. MUJUMDAR & D. NAGESH KUMAR
To cite this article: P. P. MUJUMDAR & D. NAGESH KUMAR (1990) Stochastic models
of streamflow: some case studies, Hydrological Sciences Journal, 35:4, 395-410, DOI:
10.1080/02626669009492442
To link to this article: http://dx.doi.org/10.1080/02626669009492442
Published online: 29 Dec 2009.
Submit your article to this journal
Article views: 112
View related articles
Citing articles: 9 View citing articles
2. Ilydrological Sciences - Journal - des Sciences Hydrologiques, 35,4, 8/1990
Stochastic models of streamflow: some case
studies
P. P. MUJUMDAR* & D. NAGESH KUMAR
Department of Civil Engineering, Indian Institute ofScience,
Bangalore, 560 012, India.
Abstract Ten candidate models of the Auto-Regressive Moving
Average (ARMA) family are investigated for representing and
forecasting monthly and ten-day streamflow in three Indian rivers.
The best models for forecasting and representation of data are
selected by using the criteria of Minimum Mean Square Error
(MMSE) and Maximum Likelihood (ML) respectively. The
selected models are validated for significance of the residual
mean, significance of the periodicities in the residuals and
significance of the correlation in the residuals. The models
selected, based on the ML criterion for the synthetic generation
of the three monthly series of the Rivers Cauvery, Hemavathy and
Malaprabha, are respectively AR(4), ARMA(2,1) and ARMA(3,1).
For the ten-day series of the Malaprabha River, the AR(4) model
is selected. The AR(1) model resulted in the minimum mean
square error in all the cases studied and is recommended for use
in forecasting flows one time step ahead.
Modèles stochastiques de l'écoulement - quelques études de cas
Résumé Dix modèles test de la famille ARMA ont été examinés
pour représenter et prévoir les débits mensuels et décadaires de
trois rivières Indiennes. Les meilleurs modèles pour prévoir et
représenter les données ont été choisis en utilisant respectivement
le critère du Moindre Carré (MMCE) et le Maximum de
Probabilité (MP). Les modèles choisis ont été validés pour la
signification de la moyenne résiduelle, la signification des
périodicités dans les résidus et la signification de la corrélation
dans les résidus. Les modèles choisis basés sur le critère du MP
pour la production synthétique de séries de débits de trois mois
des rivières Kaveri, Hemavathi et Malaprabha sont
respectivement AR(4), ARMA(2,1) et ARMA(3,1). Pour les
séries décadaires de la rivière Malaprabha, le modèle AR(4) a
été choisi. Le modèle AR(1) a conduit à la valeur minimale de
moyenne quadratique dans tous les cas étudiés et on le
recommande en prévoyant les débits avec un pas de temps en
avance.
*now with the Department of Civil Engineering, Indian Institute of Technology, Bombay 400 076,
India.
Open for discussion until 1 February 1991 395
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
3. P. P. Mujumdar & D. Nagesh Kumar 396
INTRODUCTION
The development and use of stochastic models of hydrological phenomena
play an important role in water resources engineering, including their use to
forecast river flows. The choice of the right model for a given hydrological
series is an important aspect of the modelling process. The statistical models
that are best suited for three South Indian rivers, viz. the Cauvery,
Malaprabha and Hemavathy, are investigated herein. Many models of the
ARMA (Auto-Regressive Moving Average) family are considered in this
study, and for each river a model is selected for the representation of data
and for one step ahead forecasting. As demonstrated below, the best models
for these two needs are often not the same. The ARIMA (Auto-Regressive
Integrated Moving Average) models are deliberately excluded from the study,
as differencing the series (which is an essential feature of such models) causes
the variance to increase continuously and hence such models cannot be used
for the simulation of data (Kashyap & Rao, 1976). They may, however, be
used for one step ahead forecasting, where they may perform as well as the
ARMA models.
Box & Jenkins (1970) give a method to estimate the orders of the AR
and MA terms of a model based on autocorrelations and partial auto-
correlations. Procedures for estimating these orders from the given data based
on testing residuals, given in Kashyap & Rao (1976), are used in the present
study.
A popular decision rule for comparing models in the time series
literature is the Akaike Information Criterion (AIC) (Akaike, 1974). However,
investigations, both theoretical (Kashyap, 1980) as well as numerical, have
indicated flaws in the AIC rule. Firstly, the AIC has no optimal property, i.e.
it does not minimize the average value of any criterion function. Secondly, the
AIC rule is not consistent, i.e. the probability that the decision rule will
choose a wrong model does not go to zero even when the number of
observations tends to infinity (Shibata, 1976).
Rao et al. (1982) have given a rule that is consistent. They have
generalized the technique used by Kashyap (1977) in which a minimum
probability of error rule was developed for comparing generalized AR models
of different orders.
In some time series applications, the given data are often transformed by
a non-linear transformation such as a Box-Cox transformation (Box & Cox,
1964). By employing the methods of Granger & Newbold (1976) it is now
possible to obtain Minimum Mean Square Error (MMSE) forecasts of the
original series when the data have been changed by a non-linear trans-
formation.
Several investigators have used Bayes decision theory for choosing the
model type and order (Valdes et al, 1979; Schwartz, 1978). Akaike (1979)
interprets the AIC criterion as a Bayes rule.
In constructing an appropriate model for a given streamflow series, the
following procedure is usually followed: (a) the selection of the appropriate
type of model among AR, MA, ARMA, ARIMA and seasonal ARIMA
models; (b) the choice of orders for the selected model; (c) the estimation of
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
4. 397 Stochastic models of streamflow - some case studies
the parameters in the model using the given streamflow series; and (d)
validation of the model by residual testing and by simulation. This procedure
is applied to identify models for forecasting and synthetic generation of four
streamflow series from three rivers in Karnataka State, India. The three rivers
considered for the study are the Cauvery, Hemavathy & Malaprabha. In the
case of the Malaprabha, both monthly and ten-daily series are studied,
whereas only monthly series are considered for the other two rivers. Table 1
gives the summary details of the data used for the four series.
In the following paragraphs, the selection of models of the ARMA
family with the two criteria, MLE and MMSE, is discussed.
Table 1 Data used for the study
Stream & site name Period for which
data are available
Type of data
Cauvery at Krishna Raja
Sagara Reservoir
Hemavathy at AkJahebbal
Malaprabha at Manoli
Malaprabha at Manoli
June 1934 - May 1974
(40 years)
June 1916 - May 1974
(58 years)
June 1950
(35 years)
June 1950
(35 years)
May 1985
May 1985
Monthly
Monthly
Monthly
Ten-daily
MODEL DESCRIPTION
The models belonging to the ARMA family may be written as:
y(t) 2 $ 4>jy(t •j) l%ejW(t j) + C + w(t) (1)
*1
8
where {y(t), t-1,2, } is the series being modelled;
mn is the number of AR parameters;
is the ;'th AR parameter;
is the number of MA parameters;
is the j'th MA parameter;
CJ
is a constant; and
{w(t), t=l,2, } is the residual series.
The important assumptions involved in such models are that [w(t)] has
zero mean with terms which are uncorrelated and form an independently
identically distributed random variable.
By choosing different values of m1 and m2 different models of the
ARMA family can be generated. The simplest model belonging to this family
would be the AR(1) model, for which m1 = 1 and m2 = 0, i.e.:
y(t) = 0j y(t - 1) + C + w(t) (2)
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
5. P. P. Mujumdar & D. Nagesh Kumar 398
ARMA models may be used with different transformations of the
original (observed) series (Granger & Newbold, 1976; Granger & Anderson,
1978). Commonly used transformations are the logarithm transform (Box &
Jenkins, 1970) and the square root transform (McLeod et al., 1977). These
transformations decide the class to which the model belongs. The observed
series and the standardized series also constitute important classes. A
standardized series {x), in this context, is defined as the series {y):
xt-x.
'•-IT
in which jT. is the estimate of the mean streamflow of the period / (month or
ten days) to which / belongs and St is the estimate of the standard deviation
of the streamflows of the period /'. Standardization ensures the removal of
periodicities inherent in the process. In the present work, only the
standardized series are considered for the selection of models. It may be
more useful to study different classes of models and select the best model for
each of the classes and compare the performances of these models before a
model is finally selected.
Both contiguous and non-contiguous models are studied. The non-
contiguous models account for the most significant periodicities without
considering the intermediate terms which may be insignificant. For example, a
non-contiguous AR(3) model with significant periodicities at first, fourth and
twelfth lags would be,
y(t) = ^ y(t - 1) + % y(t - 4) + 012 y{t - 12) + C + wit) (3)
The moving average terms are similarly considered. The obvious
advantage of non-contiguous models is the reduction in the number of
parameters to be estimated while accounting for the significant periodicities.
Exactly which terms to include in the non-contiguous models would have to
be decided based on the spectral analysis of the series under consideration. In
the present case, the term corresponding to the twelfth lag was included for
all the monthly series and that at 36th lag for the ten day series of the
Malaprabha river was considered. In the latter case, a year consists of exactly
36 periods. Thus a non-contiguous ARMA(3,3) model for the monthly series
will be:
y(t) = 0j y(t - 1) + 02 y(t - 2) + <t>n y(t - 12) +
B1 w(t - 1) + 92 w(t -2) + 912w(r - 12) + C + w(f) (4)
In the following paragraphs, the model selection based on the Maximum
Likelihood Estimate (MLE) and the Mean Square Error (MSE) criteria is
discussed and the results are presented for the three rivers chosen for the
study. The validation tests carried out on the selected models are
subsequently presented.
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
6. 399 Stochastic models of streamflow - some case studies
MODEL SELECTION
The problem of model selection is an important one in time series analysis as
there are infinitely many possible models and the choice of a wrong model
may result in a costly decision. Out of these possible models however, only a
few need to be considered for modelling a given streamflow sequence. AR
parameters of up to order 6 and MA parameters of up to order 2 would, in
general, serve the purpose. In this study, therefore, only the following models
are investigated: AR(1), AR(2), ..., AR(6), ARMA(1,1), ARMA(2,1),
ARMA(3,1), ARMA(1,2) and ARMA(2,2).
A model may be selected as the best among those investigated by using
the following two criteria: Maximum Likelihood rule (ML) and Mean Square
Error (MSE). Many other criteria are available and these two are
representative of those available. Both these methods are used for the
selection of the best model for each of the three rivers considered. The two
criteria used for the model selection are discussed below.
Maximum likelihood rule
Selection of a model by this criterion involves evaluating a likelihood value
for each of the candidate models and choosing the model which gives the
highest value. The general form of the log-likelihood function for the /th
model for a Gaussian process is (Kashyap & Rao, 1976):
L.= In p{z,<t>i)]-ni (5)
where L. is the likelihood value;
p is the probability density function;
z is the vector of the historical series;
$. is the vector of the parameters and residual variance,
(e1( e2,...., 4>v 4>2 , P);
p;. is the residual variance; and
n. is the number of parameters.
An immediate observation of the likelihood function (equation (5)) is
that, in general, as the number of parameters, n., increases, the likelihood
value decreases. Thus it is to be expected that the ML rule selects models
with a small number of parameters. This is the principle of parsimony
propounded by Box & Jenkins (1970). A particular likelihood function within
this general framework is (Kashyap & Rao, 1976):
-N
In pi - ni
'" „2,
Vl ffl.
Ur(0
-t=
-N m
l
— - In 2n + —
2 2
r
P "
»
in
J y J
2
(6)
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
7. P. P. Mujumdar&D. Nagesh Kumar 400
where p is the variance of iy(t)}.
Usually ml « N, in which case equation (6) may be written as (Kashyap
& Rao, 1976):
L.
N
[in p.) - n. (V)
Equation (7) is the likelihood function that is used to select the model
for a given site. The twelve models mentioned above are the candidate
models. Table 2 gives the likelihood values for these twelve models for each
of the three streams for both contiguous and non-contiguous models. It is
apparent that, for a given model, the value of the likelihood function differs
significantly from one site to another. This large difference may be due to the
difference in the lengths of the data available as the variance of the residuals
(or the logarithm of it) is not likely to cause the observed magnitude of
change in the values of the likelihood function. The relative values of the
likelihood function for different models when applied to a given site, rather
than those for a given model for different sites, is of interest. In Table 2, the
values shown with an asterisk superscript are the maximum values in their
respective rows. For the Cauvery river at the KRS reservoir site, the model
corresponding to the maximum likelihood value is AR(4). This is in
Table 2 Maximum likelihood values
Site Contiguous ARMA
(1,0) (2,0) (3,0) (4,0) (5,0) (6,0) (1,1) (1,2) (2,1) (2,2) (3,1)
KRS Reservoir
(Monthly)
Hemavathy
(Monthly)
Malaprabha
(Monthly)
Malaprabha
(Ten-daily)
29.33 28.91 28.96 31.63 30.71 29.90 30.58 29.83 29.83 28.80 29.45
22.53 22.55 22.64 22.94 22.47 21.51 23.38 24.96* 24.48 23.94 22.37
0.588 0.830 -0.16 -0.86 -0.68 -0.63 0.660 -0.07 -0.74 -1.12 -1.19*
59.71 60.26 59.87 61.97* 60.97 60.80 60.66 61.52 58.94 58.64 60.91
Site Non-contiguous ARMA
(2,0) (3,0) (4,0) (5,0) (6,0) (7,0) (2,2) (2,3) (3,2) (3,3) (4,2)
KRS Reservoir
(Monthly)
Hemavathya
(Monthly)
Malaprabha
(Monthly)
Malaprabha
(Ten-daily)
28.52 28.12 28.21 30.85 29.94 29.12 29.81 28.82 28.48 28.06 28.65
22.52 22.49 22.57 23.09 22.76 21.77 23.37 24.81* 24.33 22.87 22.85
-0.37 -0.17 -1.16 -1.86 -1.66 -1.63 -0.33 -1.07 -1.74 -2.11 -2.34*
58.80 59.33 58.94 61.13* 60.14 59.98 59.73 60.64 58.03 57.67 60.06
a - AR & MA parameters at 12th lag (ref. equation (4))
b - AR & MA parameters at 36th lag (ref. equation (4))
* indicates the selected model.
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
8. 401 Stochastic models ofstreamflow - some case studies
accordance with observation of the spectral analysis of the series. For the
Hemavathy river at Akkihebbal, the model selected is ARMA (1,2). For the
monthly series of the Malaprabha river at Manoli, ARMA(3,1) is chosen, whereas
for the ten-day series of the same river the AR(4) model is selected.
For a given series, the choice of a contiguous model or a non-
contiguous model is decided by the relative likelihood values for the two
models. Thus for the monthly series of the Cauvery and Hemavathy,
contiguous models are adequate whereas for the Malaprabha a non-
contiguous ARMA(4,2) model is the best one among those considered. Table
3 gives the parameters estimated for the models selected to represent the
four series. The values indicated within the brackets are the standard error
values associated with the parameters. For a parameter to be significant, its
absolute value must be larger than the standard error.
Table 3 Parameter values and their standard error for models selected on ML
rule basis
Site Model Parameters with their standard error
selected in brackets
0.2137 (0.0644) ,
0.0540 (0.0661)
-0.0157(0.1070)
0.9015 (0.0605) :
0.1737 (0.0659) ,
0.9402 (0.0650) ,
-0.0897(0.0560) .
0.1425 (0.1124) '
0.4852 (0.0410) ,
-0.0051(0.0457) ,
0.1243 (0.0974) '
f * ? -
•>*4-
• • 9
/ =
,
- c =
;4>? =
,-9i =
;<t>?-
;<S>4-
0.0398 (0.0659)
0.1762 (0.0652)
0.5404 (0.0838)
-0.1706(0.1418)
-0.0830(0.0612)
0.6059 (0.1756)
0.0441 (0.0456)
0.1034 (0.0416)
Models such as those in Table 3 are often used for the synthetic
generation of data. Sequences generated by such models are used for the
design of reservoirs. Such simulated sequences would obviously be different
from one model to another. Designs based on such sequences would thus
depend on the right choice of model. The maximum likelihood estimate
criterion is suited for the selection of a model for simuhtion purpose. For
short-term forecasting, such as one step ahead forecasting, the mean square
error (MSE) criterion may be more useful (Kashyap & Rao, 1976). Selection
of a model based on an MSE criterion is known as the prediction approach,
and is discussed below.
Prediction approach (MSE criterion)
The procedure involved in this approach is quite simple and can be
KRS Reservoir ARMA(4,0) <t>1
(Monthly) A
C3
Hemavathy ARMA(1,2) $1
(Monthly) g
Malaprabha ARMA(3J) 4>j
(Monthly) A
C3
Malaprabha ARMA(4,0) tp.
(Ten-daily) A J
C3
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
9. P. P. Mujumdar & D. Nagesh Kumar 402
summarized as follows: (a) estimate the parameters of different models using
a portion, usually half, of the available data; (b) forecast the second half of
the series one step ahead by using the candidate models; (c) estimate the
MSE corresponding to each model; and (d) select the modeMhat results in
the least value of the MSE. The one step ahead forecast, y[(t + l)/t], for
ARMA {mv m2) is given by:
Ç[(t + l)lt] = ijl} 4>j y(t - j) + Z™J <t>j w(t -j)*C (8)
y[(t + l)/f] represents the forecast streamflow for the time, t + 1, given the
streamflow up to and including the time, t. The one step ahead forecast error
is given by:
e(t + 1) = y(t + 1) - y[(t + l)/t] (9)
When the series consists of N observations, the first N/2 observations are
used for the parameter estimation of the candidate models. The streamflows
from N/2 + 1 to N are forecast by using these models and their errors
calculated. The MSE for a model is then given by:
TN
e(i)2
MSE . — (10)
Table 4 gives MSE values for contiguous as well as non-contiguous models
for all the series considered. For all the cases the simplest model, AR(1),
results in the least value of the MSE, underlining the fact that for one step
ahead forecasting quite often the simplest model is sufficient. Also, in the
present case, as the number of parameters increases, the MSE increases,
which is an interesting result contrary to the common belief that models with
larger numbers of parameters give better forecasts. For all the four series of
the streamflows considered, the AR(1) model is strongly recommended for
use in forecasting the series one step ahead. However it should be noted that
this is not a general conclusion and for other time series a similar analysis
has to be carried out separately to decide the model that suits the particular
sequence the best. Table 5 gives the estimated parameters, with their standard
error in brackets, for the models selected on the MSE criteria.
The exercise so far has been to identify a model both for simulation and
for forecasting. Before the model is used, however, it has to be validated. The
major assumptions that have gone into the construction of the model must
be checked for their validity in the selected model. The following section
discusses the validation tests carried out in this study and the results.
VALIDATION TESTS
The following tests are carried out to examine whether the following
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
10. 4 03 Stochastic models ofstreamflow - some case studies
Table 4 Mean square error values
Site Contiguous ARMA
(1,0) (2,0) (3,0) (4,0) (5,0) (6,0) (1,1) (1,2) (2,1) (2,2) (3,1)
KRS Reservoir
(Monthly)
Hemavathy
(Monthly)
Malaprabha
(Monthly)
Malaprabha
(Ten-daily)
KRS Reservoir"
(Monthly)
Hemavathy®
(Monthly)
Malaprabha
(Monthly)
Malaprabha
(Ten-daily)
0.97 1.92 2.87 3.82 4.78 5.74 2.49 2.17 3.44 4.29 1.89
0.78* 1.54 2.31 3.08 3.85 4.62 0.98 0.81 0.80 1.35 2.41
0.77* 1.54 2.31 3.08 3.85 4.62 0.98 0.81 0.80 1.35 2.41
0.62* 1.24 1.85 2.47 3.09 3.72 0.77 1.02 1.07 2.44 1.62
Site Non-contiguous ARMA
(2,0) (3,0) (4,0) (5,0) (6,0) (7,0) (2,2) (2,3) (3,2) (3,3) (4,2)
0.96 1.89 2.84 3.79 4.74 5.70 2.42 1.99 2.52 1.15 1.71
0.76* 1.51 2.25 3.00 3.75 4.49 1.04 1.52 8.68 0.83 1.39
0.77* 1.54 2.31 3.07 3.85 4.62 0.98 0.81 0.80 1.24 3.76
0.62* 1.24 1.85 2.47 3.09 3.72 0.68 0.88 1.13 2.04 1.65
a - AR & MA parameters at 12th lag (ref. equation (4))
b - AR & MA parameters at 36th lag (ref. equation (4))
* indicates the selected model.
Table 5 Parameter values and their standard error for models selected on MSE
basis
Site
KRS Reservoir
(Monthly)
Hemavathy
(Monthly)
Malaprabha
(Monthly)
Malaprabha
(Ten-daily)
Model
selected
ARMA(1,0)
ARMA(l.O)
ARMA(1,0)
ARMA (1,0)
Parameters with their standard error
in brackets
<pj = 0.2557 (0.0627) ; C = -0.009 (0.0765)
<pj = 0.4204 (0.0502) ; C = -0.1816 (0.0864)
4>j = 0.3742 (0.0654) ; C = 0.1404 (0.1110)
ipj = 0.5311 (0.0348) ; C = -0.1189 (0.0780)
assumptions used in building the model are in fact valid for the model
selected:
(a) the residual series {w(t)} has zero mean;
(b) no significant periodicity is present in the residual series; and
(c) the residual series is uncorrelated.
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
11. P. P. Mujumdar & D. Nagesh Kumar 404
The residual series is constructed from equation (1) as follows:
m^ ""2
w(t) = y(t) - Ij,] fyd - ;) - I^f 9. w(t - /) + C (11)
Fig. 1 shows the histogram of residuals for the inflows into the KRS
reservoir resulting from an AR(4) model. The histogram is skewed to the
right, suggesting that a skewed distribution such as the log-Pearson type III
should be used in a simulation. A similar skewness is also seen in the
histograms of the residuals from the other models selected.
VJ
Q
/,
I i"7i rr m rzl
1 2 3 4 5 6 7 S 9 1 0 1 1 12 13 14 15 16 17 18 19 20
Fig. 1 Histogram of residuals.
All the validation tests are carried out on the residual series only. The
tests are summarized briefly in the following paragraphs.
Test 1 Significance of the residual mean
The purpose of this test is to examine the validity of the assumption that the
series {w(t)} has zero mean. For this purpose a statistic, n(w), is defined as:
n(w) = N* wl pYi
(12)
where w is the estimate of the residual mean; and
A
p is the estimate of the residual variance.
The statistic, *)(w), is approximately distributed as t(a, N - 1), where a
is the significance level at which the test is being carried out. If the value of
i){w) 5 t(a, N - 1), then the mean of the residual series is not significantly
different from zero and hence the series passes this test. Table 6 gives the
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
12. 405 Stochastic models ofstreamflow - some case studies
values of the statistic, n(w), and f(a, N - 1) for all the models selected for
the different sites. At the 95% significance level, it is observed that the
residual series passes the test in all the cases. This must be true when the
models are fitted to the standardized series.
Table 6
Model
ARMA (1,0)
ARMA (2,0)
ARMA (3,0)
ARMA (4,0)
ARMA (5,0)
ARMA (6,0)
ARMA (1,1)
ARMA (1,2)
ARMA (2,1)
ARMA (2,2)
Results of tests 1 and 2 for KRS data
Test 1
ri
0.002
0.006
0.008
0.025
0.023
0.018
0.033
0.104
0.106
0.028
'0.95 <239
>
1.645
1.645
1.645
1.645
1.645
1.645
1.645
1.645
1.645
1.645
Test 2
V) value
One
0.527
1.027
1.705
3.228
3.769
4.190
4.737
6.786
7.704
6.857
for the periodicity
Two
1.092
2.458
4.319
6.078
7.805
10.130
10.090
10.670
12.120
13.220
Three
0.364
0.813
1.096
0.948
1.149
1.262
2.668
2.621
2.976
3.718
Four
0.065
0.129
0.160
0.277
0.345
0.441
0.392
0.372
0.422
0.597
F095(2, 238)
3.00
3.00
3.00
3.00
3.00
3.00
3.00
3.00
3.00
3.00
Test 2 Significance of the periodicities
For the model to be applicable, the residual series, {w(f)}, must not have any
significant periodicity in it. The following test is carried out to ensure that
this is, in fact, true. This test is conducted for different periodicities, and the
significance of each of the periodicities is tested. A statistic, n(w), is defined
as:
n(w)
Jl
(N - 2)
4 p
(13)
where y2
= or + 3 ;
px = VN fe [w(t) - « cos {Wxt) - B sin ( ^ f ) ] 2
] ;
â = 2/N l£i w(t) cos (Wxt)
B = 2IN 1 ^ w(t) sin (Wf) and
2fllWl is the periodicity for which the test is being carried out.
The statistic, n(w), is distributed approximately as Fa(2, N - 2), a being the
significance level. The periodicity corresponding to W1 is not significant if:
n(w) « FJ2, N-2)
This test was carried out on the residual series resulting from each of
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
13. P. P. Mujumdar&D. Nagesh Kumar 406
the four streamflow series considered. Table 6 presents the values of *7(w)
for the different periodicities tested. All the periodicities tested were found
to be insignificant for the models selected and thus the models passed
the test.
Another test carried out for the significance of periodicities is the
cumulative periodogram test, also known as Bartlett's test (Bartlett, 1946).
Unlike the previous test which has to be carried out for one periodicity
at a time, this test is conducted to detect the first significant periodicity
in the series. If a significant periodicity is observed, the next significant
periodicity will be detected by carrying out the test on the series from
which the first periodicity is removed, and so on. The test is briefly
explained below:
Define y= [2/N 1 ^ w(t) cos (Wkt)]2
+ [2/N 1 ^ w(t) sin (Wkt)]2
k = l,2,....,/v72 (14)
yk y2
C
°mPUteg
*=
if (15)
It is noted that 0 S gk S 1. The plot of gk versus fc is known as
the cumulative periodogram. On the cumulative periodogram two
confidence limits are drawn. These are given by ±/(N/2)y
The value of
X prescribed (Kashyap & Rao, 1976) is 1.35 for 95% confidence and 1.65
for 99% confidence. If all the values of gk lie within the significance
band, then there is no significant periodicity present in the series. When
one of the gk values lies outside the significance band (the subsequent
values will also lie outside the band), the periodicity corresponding to that
value of gk is significant.
Fig. 2 shows the cumulative periodogram of the residual series resulting
from the AR(4) model applied to the KRS inflows. All the values of gk lie
within the significance band, thus confirming the result of the earlier test that
no significant periodicity is present in the residual series. The same result is
also observed in the case of the models selected for each of the other three
streamflow series considered. To contrast the cumulative periodogram, shown
in Fig. 2, the cumulative periodogram of the original series (without stan-
dardizing) is shown in Fig. 3. It is seen from Fig. 3 that, corresponding to k
= 40, the periodicity is significant. This value of k corresponds to a periodicity
of 12 months.
Of the two tests mentioned in this section, the latter (Bartlett's test) is
more convenient computationally. These two tests are carried out both on the
original series and on the residual series. In the cases studied, the conclusions
drawn from the two tests do not differ from each other at any time.
However, the cumulative periodogram test is preferred because of its ability
to test all the periodicities at a time.
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
14. 407 Stochastic models of streamflow - some case studies
Fig. 2 Cumulative periodogram for residuals.
Fig. 3 Cumulative periodogram for monthly data.
Test 3 White noise test
An important assumption in the models studied is that the residual series,
{w(t)}, is a white noise sequence (or that the series is uncorrelated). In this
section the residuals are tested for absence of correlation. Two tests are
carried out for this purpose and the results are compared.
Whittle's test This test (Whittle, 1952) involves the construction of the
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
15. P. P. Mujumdar & D. Nagesh Kumar 408
covariance matrix. The covariance Rk at lag k of the series, {w(t)}, is estimated by:
Rk = 1I(N - k) l £ M w(j) w(j - k) (16)
k =
°'1
'2
Anax
The value of &max is normally chosen as 15% of the sample size., i.e.
kmgx = 0.15N. The covariance matrix, TnV is then constructed as:
«l
R
0 * 1 ^ 2 R
km
ax
R
* i?1
^0
* w 2
^nax 'taax4
"" °
(17)
This is a square symmetric matrix of size «1 = /cmax.
A statistic, n(w), is defined as:
n(w) = (N/nl - 1) (p/Pl - 1)
where pQ is the lag zero correlation coefficient (= 1) and
det I".
(18)
Pi
nl
det r nl-l
The matrix Tnll is constructed by eliminating the last row and the last
column from the matrix TnV The statistic, n(yv), defined by equation (18) is
distributed approximately as Fa(nl, N - nl). If n(w) « F (ni, N - nl) then
the residual series is uncorrelated. This test was carried out on the residual
series resulting from the different models considered for all the four
streamflow series. Table 7 gives the values of the statistic, n(w), for the
models applied to the KRS inflows. From Table 7 it is seen that when
nl = 25, ARMA(1,2) and ARMA(2,1) do not pass the test. In all other
cases the test is successful. The result from the other sites indicates that the
models selected for each series also passed the test.
Portmanteau test This test also uses the covariance, Rk, defined earlier.
The statistic, ri(w), is defined as:
n(w) = (N-nl) Z& (RJR0)2
(19)
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
16. 4 09 Stochastic models of streamflow - some case studies
Table 7 Whittle's testfor KRS data (N = 240)
F
0.95 <nl
- N
~ nl)
Model
ARMA (1,0)
ARMA (2,0)
ARMA (3,0)
ARMA (4,0)
ARMA (5,0)
ARMA (6,0)
ARMA (1,1)
ARMA (1,2)
ARMA (2,1)
ARMA (2,2)
ni = 73
1.29
n
0.642
0.628
0.606
0.528
0.526
0.522
0.595
0.851
0.851
0.589
ni = 49
1.39
T)
0.917
0.898
0.868
0.743
0.739
0.728
0.854
1.256
1.256
0.845
ni = 25
1.52
T)
0.891
0.861
0.791
0.516
0.516
0.493
0.755*
1.581*
1.581
0.737
indicates that the model does not pass the test,
This is distributed approximately as x^inl). If ri(w) « xa(nl), then the
series is uncorrelated at the significance level, a. The value of nl is normally
chosen as Q.15N. However the test was carried out for different values of nl.
Table 8 gives the results of this test. It is seen that the residuals of all
models except ARMA(1,2) and ARMA(2,1) pass the test.
Kashyap & Rao (1976) have proved that the portmanteau test is
uniformly inferior to Whittle's test and recommended the latter for
application.
Table 8 Portmanteau testfor KRS data (N = 240)
^•0.95^k
maJ
Model
ARMA (1,0)
ARMA (2,0)
ARMA (3,0)
ARMA (4,0)
ARMA (5,0)
ARMA (6,0)
ARMA (1,1)
ARMA (1,2)
ARMA (2,1)
ARMA (2,2)
k n = 48
max
65.0
31.44
32.03
30.17
20.22
19.84
19.64
29.89
55.88
55.88
28.62
k =36
max
50.8
33.41
34.03
32.05
21.49
21.08
20.87
31.76*
59.38*
59.38
30.41
k =24
max
36.4
23.02
24.47
21.61
11.85
11.75
11.48
22.24*
48.37*
48.37
20.39
k =12
max
21.0
14.80
15.17
13.12
4.31
4.14
3.79
12.76*
39.85*
38.85
11.25
indicates that the model does not pass the test.
CONCLUSIONS
Streamflow sequences for three south Indian rivers, viz. the Cauvery,
Hemavathy and Malaprabha, are modelled. Ten candidate models of the
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016
17. P. P. Mujumdar & D. Nagesh Kumar 410
ARMA family are studied and the best model for each of the four
streamflow series is selected (three monthly and one ten-day series). The best
models resulting from the maximum likelihood criterion for the three monthly
series of the Cauvery, Hemavathy and Malaprabha are respectively AR(4),
ARMA(2,1) and ARMA(3,1). For the ten-day series of the Malaprabha, the
model selected is AR(4). Selection of models based on a minimum mean
square error criterion results in an AR(1) model for all the four streamflow
series considered. The selected models are validated by tests on residuals for
the significance of residual mean, the significance of periodicities (Bartlett's
test) and the significance of correlations (Whittle's test and Portmanteau test).
These tests revealed that the models selected by the two criteria pass all the
tests and hence these models are recommended for use in practice for the
three rivers.
Acknowledgements This work originated from a series of lectures on Time
Series Analysis delivered by Dr. A. Ramchandra Rao, Professor, School of
Civil Engineering, Purdue University, currently Visiting Professor, Indian
Institute of Science, Bangalore. The authors wish to thank him for his
encouragement and guidance.
REFERENCES
Akaike, H. (1979) Bayesian extension of the minimum AIC procedure. Biometrika. 66, 237-242.
Akaike, H. (1974) A new look at statistical model identification. IEEE Trans. Automatic Control
AC-19, 716-722.
Bartlett, M. S. (1946) On the theoretical specification of sampling properties of autocorrelated
time series. /. Roy. Statist. Soc. B8, 27.
Box, G. E. P. & Cox, D.R. (1964) An analysis of transformations. /. Roy. Statist. Soc. B26, 211.
Box, G. E. P. & Jenkins, G. M. (1970) Time Series Analysis. Holden Day, San Francisco,
California, USA.
Granger, C. W. J. & Anderson, A. G. (1978) Non-linear time series modeling. In: Applied Time
Series Analysis (ed. D. F. Findely), Academic Press, New York, USA.
Granger, C. W. J. & Newbold, P. (1976) Forecasting transformed series. /. Roy. Statist. Soc.
B38 (2), 189-203.
Kashyap, R. L. (1977) Bayesian comparison of dynamic models. IEEE Trans. Automatic Control
AC-22, 715-727.
Kashyap, R. L. (1980) Inconsistency of AIC rule for estimating the order of auto regressive
models. IEEE Trans. Automatic Control. AC-25, 996-998.
Kashyap, R. L. & Rao, A. R. (1976) Dynamic Stochastic ModelsfromEmpirical Data. Academic
Press, New York, USA.
McLeod, A. I., Hipel, K. W. & Lennox, W. C. (1977) Advances in Box-Jenkins modeling. 2.
Applications. Wat. Resour. Res. 13 (3), 577-586.
Rao, A. R., Kashyap, R. L. & Liang-Tsi Mao (1982) Optimal choice of type and order of river
flow time series models. Wat. Resour. Res. 18 (4), 1097-1109.
Shibata, R. (1976) Selection of the order of an autoregressive model by Akaike's information
criterion. Biometrika 63, 117-126.
Schwartz, G. (1978) Estimating the dimensions of a model. Ann. Statist. 6 (2), 461-464.
Whittle, P. (1952) " Tests of fit in time series. Biometrika 39, 309-318.
Valdes, J. B., Vicens, G. J. & Rodriguez-Iturbe, I. (1979) Choosing among alternative
hydrologie regression models. Wat. Resour. Res. 15 (2), 347-358.
Received 23 August 1989; accepted 20 December 1989
Downloaded
by
[188.72.127.226]
at
00:28
13
June
2016