Probabilistic Prediction for Monthly Streamflow through Coupling Stepwise Cluster Analysis and Quantile Regression Methods

Investigating the dynamic characteristic of hydrological processes is of vital significance for environmental protection. In this study, the stepwise cluster analysis (SCA) method was used for monthly streamflow simulation in a hypothetical case. According to SCA, a cluster tree was formulated through training the data of monthly temperature, precipitation and streamflow from 2004 to 2010. Then, the generated tree was used to reproduce monthly streamflow in calibration period (i.e., 2004-2010) and validation period (i.e., 2011-2013). A comparison of SCA and multiple linear regression (MLR) was conducted to reflect the complex relationship of meteorological parameters (e.g., precipitation) and hydrological parameters. Model performance was assessed using Nash-Sutcliffe efficiencies (NSE), the determination coefficient (R 2), the root-mean-square error (RMSE) and the mean absolute error (MAE). NSE and R2 obtained from SCA are higher than that obtained from MLR. RMSE and MAE obtained from SCA are smaller than that obtained from MLR, indicating a better coincidence between simulated streamflow and the observed values in SCA. Results indicated that SCA has advantage in revealing the nonlinear relationship among precipitation, temperature and streamflow.

The effects of a discrete wavelet-transformation data-preprocessing method on neural-network-based monthly streamflow prediction models in producing energy from small hydro power plants in the Pungwe River basin in Mozambique were investigated. Data from a Vanduzi gauging station in Pungwe River basin were collected. Eight different single-step-ahead monthly stream flow neural prediction models were developed. Coupled simulation involving use of MATLAB and of a Wavelet-Neural Network was employed. Different models were tested using the same sample in each case, an Artificial Neural Network (ANN) being found to performance best. The major objective of the research project was to analyze the monthly stream flow predictions in the Pungwe River, to be able to make as appropriate decisions as possible during dry or wet spells, and also to resolve as effectively as possible conflicts regarding water resourses.

This study investigates the effect of discrete wavelet transform data pre-processing method on neural network-based successive-station monthly streamflow prediction models. For this aim, using data from two successive gauging stations on Coruh River, Turkey, we initially developed eight different single-step-ahead neural monthly streamflow prediction models. Typical three-layer feed-forward (FFNN) topology, trained with Levenberg-Marquardt (LM) algorithm, has been employed to develop the best structure of each model. Then, the input time series of each model were decomposed into subseries at different resolution modes using Daubechies (db4) wavelet function. At the next step, eight hybrid neuro-wavelet (NW) models were generated using the subseries of each model. Ultimately, root mean square error and Nash-Sutcliffe efficiency measures have been used to compare the performance of both FFNN and NW models. The results indicated that the successive-station prediction strategy using a pair of upstream-downstream records tends to decrease the lagged prediction effect of single-station runoff-runoff models. Higher performances of NW models compared to those of FFNN in all combinations demonstrated that the db4 wavelet transform function is a powerful tool to capture the non-stationary feature of the successive-station streamflow process. The comparative performance analysis among different combinations showed that the highest improvement for FFNN occurs when simultaneous lag-time is considered for both stations.

A Comparative Study on Prediction of Monthly Streamflow Using Hybrid ANFIS-PSO Approaches

This study aims to construct the model for the length of hospital stay for patients with COVID-19 using quantile regression and Bayesian quantile approaches. The quantile regression models the relationship at any point of the conditional distribution of the dependent variable on several independent variables. The Bayesian quantile regression combines the concept of quantile analysis into the Bayesian approach. In the Bayesian approach, the Asymmetric Laplace Distribution (ALD) distribution is used to form the likelihood function as the basis for formulating the posterior distribution. All 688 patients with COVID-19 treated in M. Djamil Hospital and Universitas Andalas Hospital in Padang City between March-July 2020 were used in this study. This study found that the Bayesian quantile regression method results in a smaller 95% confidence interval and higher value than the quantile regression method. It is concluded that the Bayesian quantile regression method tends to yield a better model than the quantile method. Based on the Bayesian quantile regression method, it investigates that the length of hospital stay for patients with COVID-19 in West Sumatra is significantly influenced by Age, Diagnoses status, and Discharge status.

Streamflow prediction is crucial for planning future developments and safety measures along river basins, especially in the face of changing climate patterns. In this study, we utilized monthly streamflow data from the United States Bureau of Reclamation and meteorological data (snow water equivalent, temperature, and precipitation) from the various weather monitoring stations of the Snow Telemetry Network within the Upper Colorado River Basin to forecast monthly streamflow at Lees Ferry, a specific location along the Colorado River in the basin. Four machine learning models—Random Forest Regression, Long short-term memory, Gated Recurrent Unit, and Seasonal AutoRegresive Integrated Moving Average—were trained using 30 years of monthly data (1991–2020), split into 80% for training (1991–2014) and 20% for testing (2015–2020). Initially, only historical streamflow data were used for predictions, followed by including meteorological factors to assess their impact on streamflow. Subsequently, sequence analysis was conducted to explore various input-output sequence window combinations. We then evaluated the influence of each factor on streamflow by testing all possible combinations to identify the optimal feature combination for prediction. Our results indicate that the Random Forest Regression model consistently outperformed others, especially after integrating all meteorological factors with historical streamflow data. The best performance was achieved with a 24-month look-back period to predict 12 months of streamflow, yielding a Root Mean Square Error of 2.25 and R-squared (R2) of 0.80. Finally, to assess model generalizability, we tested the best model at other locations—Greenwood Springs (Colorado River), Maybell (Yampa River), and Archuleta (San Juan) in the basin.

The SWAT model was used to estimate the runoff and sediment yield of Lolab watershed. The model was calibrated, validated, and assessed for evaluation to model ambiguity using Nash–Sutcliffe coefficient (NSE) and coefficient of determination (R2). Ten highly sensitive parameters were recognized for stream flow simulation of which CN2 (Initial SCS CN II value) factor was the most sensitive one and four highly sensitive parameters were recognized for sediment yield simulation of which SPCON (Linear parameters for sediment re-entrainment) was most sensitive one. The model was calibrated for a time period between 1993 to 2000 and validated from 2001 to 2004 for flow and sediment yield. The predicted and observed stream flow and sediment yields generally matched well. The results of the model calibration and validation showed reliable estimates of monthly stream flow (R2=0.74 and ENS=0.68) and yearly stream flow (R2=0.90 and ENS=0.68) during the calibration period and monthly stream-flow (R2=0.85 and ENS=0.83) and yearly stream-flow (R2=0.99 and ENS=0.91) during the validation period. For sediment yield, this study shows antremendous model efficiency of monthly sediment yield (R2=0.80 and ENS=0.79) and yearly sediment yield (R2=0.86 and ENS=0.78) during the calibration period and monthly sediment yield (R2=0.88 and ENS=0.86) and yearly sediment yield (R2=0.83 and ENS=0.58) during the validation period. This study showed that the SWAT model is competent of predicting sediment yields and hence can be used as a tool for water resources planning and management in the study watershed.

Accurate streamflow prediction plays a pivotal role in hydraulic project design, nonpoint source pollution estimation, and water resources planning and management. However, the highly non-linear relationship between rainfall and runoff makes prediction difficult with desirable accuracy. To improve the accuracy of monthly streamflow prediction, a seasonal Support Vector Regression (SVR) model coupled to the Soil and Water Assessment Tool (SWAT) model was developed for 13 subwatersheds in the Illinois River watershed (IRW), U.S. Terrain, precipitation, soil, land use and land cover, and monthly streamflow data were used to build the SWAT model. SWAT Streamflow output and the upstream drainage area were used as two input variables into SVR to build the hybrid SWAT-SVR model. The Calibration Uncertainty Procedure (SWAT-CUP) and Sequential Uncertainty Fitting-2 (SUFI-2) algorithms were applied to compare the model performance against SWAT-SVR. The spatial calibration and leave-one-out sampling methods were used to calibrate and validate the hybrid SWAT-SVR model. The results showed that the SWAT-SVR model had less deviation and better performance than SWAT-CUP simulations. SWAT-SVR predicted streamflow more accurately during the wet season than the dry season. The model worked well when it was applied to simulate medium flows with discharge between 5 m3 s-1 and 30 m3 s-1, and its applicable spatial scale fell between 500 to 3000 km2. The overall performance of the model on yearly time series is “Satisfactory”. This new SWAT-SVR model has not only the ability to capture intrinsic non-linear behaviors between rainfall and runoff while considering the mechanism of runoff generation but also can serve as a reliable regional tool for an ungauged or limited data watershed that has similar hydrologic characteristics with the IRW.

Objective: To assess various marginal effects of nutrient intakes, health behaviours and nutrition knowledge on the entire distribution of body mass index (BMI) across individuals. Design: Quantitative and distributional study. Setting: Taiwan Methods: This study applies Becker’s (1965) model of health production to construct an individual’s BMI determining function. Quantile regression method is used to analyse the entire distribution of BMI across a nationwide survey sample in Taiwan. Endogeneity associated with choice variables of nutrient intakes, health behaviours and nutrition knowledge in determining BMI is corrected using a two-stage least-squares method. Results: Empirical results indicate that energy, oleic acid and cholesterol increase BMI, while fibre, calcium and sodium reveal a reverse effect. For overweight and obese females, protein lowers BMI. For underweight and slightly to severely obese males, vitamin C reduces BMI. Jogging lowers BMI, but drinking raises non-obese individuals’ BMI. For males whose BMI is in the ideal weight to slightly overweight ranges, nutrition knowledge reduces BMI, but this effect is insignificant for females. Conclusions: Several important conclusions can be drawn. (1) Having a BMI which is too high or too low is a risk factor to health. (2) Quantile regression is a useful empirical method to evaluate crucial factors determining the entire distribution (high, medium and low) of BMI. (3) The econometric problem of endogeneity in regressors has to be considered and corrected when using a quantile regression approach. (4) BMI-relevant health education and policy can be evaluated and derived using quantile regression effects of an individual’s choices of nutrient intakes, health behaviours and nutrition knowledge.

This thesis consists of three empirical studies on what drives stock market dynamics. The first empirical study explores the effect of crude oil price changes on the stock market returns of oil-exporting countries and oil-importing countries as well as those of a number of global stock indices. Using the Ordinary Least Squares (OLS) approach as well as the more robust Quantile Regression (QR) approach to explore the relationship between crude oil and stock market dynamics. The empirical findings suggest that the QR approach provides further insights compared to the OLS approach. For instance, the QR approach is able to identify specific quantiles where a significant relation exists. In particular crude oil price increases tend to have a negative impact on the stock market returns for some oil-exporting countries (such as Mexico, Iraq, Ecuador, and Venezuela) and a positive effect for other oil-exporting countries (such as Brazil and Algeria). However, the OLS approach suggests that these relationships are insignificant at the level of the mean. Overall, the empirical findings confirm that the QR approach can reveal more information about the relationship between crude oil price changes and stock market return across different quantiles of their distribution.The second study explores the extent to which implied volatility extracted from commodity markets and developed stock markets can predict the implied volatility of stock markets in BRICS countries. Using daily data from 2011 to 2016 and employing the newly developed Bayesian Graphical Vector Autoregressive (BGVAR) model of Ahelegbey et al. (2016) which does not suffer from over-parameterization and the identification problems associated with traditional VAR frameworks, this study finds that implied volatilities extracted from global and regional stock markets have a significant predictive power over the implied volatilities in BRICS stock markets. However, the predictive power of implied volatility from commodity markets are significant only in the case of South Africa.The third empirical study analyses the relationship between illiquidity and stock market returns in the G7 and BRICS countries. More specifically, this study explore the extent to which the Amihud (2002) illiquidity measure can improve the explanatory power of three commonly used asset pricing models, namely the Capital Asset Pricing Model (CAPM), the Fama-French three-factor model and the Carhart four-factor model. The empirical analysis is based on 15 years of monthly data on the returns of seven stock portfolios: 100 largest companies (Largest100), small value (S/V), small neutral (S/N), small growth (S/G) stocks, big value (B/V) stocks, big neutral (B/N) stocks, and big growth (B/G) stocks. The findings suggest that incorporating illiquidity as an additional factor results in a significant improvement in the explanatory power of these asset pricing models across several of the sample countries (8 countries in the case of the CAPM and Carhart four-factor model, and 6 countries in the case of the Fama-French three-factor model). For example, in the US adding illiquidity to the CAPM leads to an increase of the goodness of fit by 2.6% in the B/V portfolio, and for the Fama-French three-factor model the goodness of fit increases by up to 3% in all portfolios Moreover, the goodness of fit increases in all portfolios in the US by adding illiquidity to the Carhart four-factor model, with an up to 36% increase in the B/N portfolio.

Varying coefficient model (VCM) is extensively used in various scientific fields due to its capability of capturing the changing structure of predictors. Classical mean regression analysis is often complicated in the existence of skewed, heterogeneous and heavy-tailed data. For this purpose, this work employs the idea of model averaging and introduces a novel comprehensive approach by incorporating quantile-adaptive weights across different quantile levels to further improve both least square (LS) and quantile regression (QR) methods. The proposed procedure that adaptively takes advantage of the heterogeneous and sparse nature of input data can gain more efficiency and be well adapted to extreme event case and high-dimensional setting. Motivated by its nice properties, we develop several robust methods to reveal the dynamic close-to-truth structure for VCM and consistently uncover the zero and nonzero patterns in high-dimensional scientific discoveries. We provide a new iterative algorithm that is proven to be asymptotic consistent and can attain the optimal nonparametric convergence rate given regular conditions. These introduced procedures are highlighted with extensive simulation examples and several real data analyses to further show their stronger predictive power compared with LS, composite quantile regression (CQR) and QR methods.

The present study was conducted in Baitarani basin up to Anandapur gauging station of Odisha covering an area of 8603.7 km2. Pre-processing of basin from digital elevation model (DEM) was done using HEC-Geo-HMS extension and spatial analyst tool in ArcGIS. These pre-processed files were then imported to HEC-HMS for simulating runoff. In this study, runoff simulation was done using two methods, viz., composite and distributed curve number (CN) approaches. SCS curve number method was used for computation of runoff volume, SCS UH method for direct runoff, constant- monthly varying base flow method for base flow and Muskingum method for flow routing. The model was calibrated and validated using both composite and distributed CN approaches. Data from 1st January, 2007 to 31st December, 2013 were used for calibration and 1st January, 2014 to 31st December, 2016 were used for validation. During the calibration period of composite CN approach, the statistical parameters like Nash-Sutcliffe efficiency (NSE), Coefficient of determination (R2), Percent bias (PBIAS) and RMSE-observations standard deviation ratio (RSR) were found to be 0.51, 0.63, 12.82 and 0.7, respectively and during the validation period they were found to be 0.53, 0.54, -19.73 and 0.7, respectively. In case of distributed CN approach, the statistical parameters like NSE, R2, PBIAS and RSR were found to be 0.62, 0.63, -8.64 and 0.6, respectively during the calibration period and 0.67, 0.66, -2.25 and 0.6, respectively during the validation period. The study indicated that distributed CN approach is more accurate than composite CN approach in simulation of runoff using HEC-HMS model.

Statiscal Downscaling (SD) was a method to model the relationship between local scale data as response variable and global scale data as predictor variable. The response variable was the rainfall and the predictor variable was the global circulation model (GCM) ouput. In general, GCM output had a large dimension and multicollinearity so the principal component analysis was used to solve this problem. This study modeled the rainfall with the selected principal component using quantile regression. This functional relationship could be parametric or nonparametric relationship. In this case, nonparametric functional relationship used spline to accommodate the extreme value with the quantile regression. The result showed that the quantile spline model was better than the quantile polynomial regression model especially in predicting the extreme values in the 90 th and 95 th quantile with correlation values of 0.95 and 0.93.

Hydrological models have emerged as essential tools for examining the hydrological processes of complex watersheds addressing various environmental and water resource issues. This study focuses on modeling the hydrology of the Merguellil watershed in central Tunisia using the Soil and Water Assessment Tool (SWAT). The SWAT model is a physical modeling tool developed to forecast hydrological processes and is well-documented as an effective tool for resource water management. The primary objective of this research is to evaluate the SWAT model's performance in simulating monthly hydrological processes within the Merguellil watershed in central semi-arid Tunisia. Calibration of the model was conducted from 2002 to 2011, followed by a validation from 2012 to 2017. Sensitivity analysis identified key parameters, including Curve Number, Slope Length, and effective hydraulic conductivity, as the most sensitive. The findings demonstrate that the model exhibits satisfactory performance according to goodness-of-fit criteria during both the calibration and validation phases. The Nash–Sutcliffe Efficiency (NSE) were 0.65 and 0.41, respectively, for calibration and validation periods. The coefficient of determination (R²) and Kling-Gupta efficiency (KGE) are both equal to 0.7, the RMSE-observations standard deviation ratio (RSR) is less than or equal to 0.6, for calibration. The Percent Bias (PBIAS) indicates that the model overestimates the discharges by +23.5% during the calibration period. In addition, the runoff in Merguellil watershed demonstrated a notable spatio-temporal variability, significantly influenced by the complexity and heterogeneity of its environment.

Heterogeneity in German Residential Electricity Consumption: A quantile regression approach