FLOOD STAGE FORECASTING WITH SUPPORT VECTOR MACHINES1

ABSTRACT: Machine learning techniques are finding more and more applications in the field of forecasting. A novel regression technique, called Support Vector Machine (SVM), based on the statistical learning theory is explored in this study. SVM is based on the principle of Structural Risk Minimization as opposed to the principle of Empirical Risk Minimization espoused by conventional regression techniques. The flood data at Dhaka, Bangladesh, are used in this study to demonstrate the forecasting capabilities of SVM. The result is compared with that of Artificial Neural Network (ANN) based model for one‐lead day to seven‐lead day forecasting. The improvements in maximum predicted water level errors by SVM over ANN for four‐lead day to seven‐lead day are 9.6 cm, 22.6 cm, 4.9 cm and 15.7 cm, respectively. The result shows that the prediction accuracy of SVM is at least as good as and in some cases (particularly at higher lead days) actually better than that of ANN, yet it offers advantages over many of the limitations of ANN, for example in arriving at ANN's optimal network architecture and choosing useful training set. Thus, SVM appears to be a very promising prediction tool.     

An integrated sorption-diffusion model for the calculation of consistent distribution and diffusion coefficients in compacted bentonite

A thermodynamic sorption model and a diffusion model based on electric double layer (EDL) theory are integrated to yield a surface chemical model that treats porewater chemistry, surface reactions, and the influence of charged pore walls on diffusing ions in a consistent fashion. The relative contribution of Stern and diffuse layer to the compensation of the permanent surface charge represents a key parameter; it is optimized for the diffusion of Cs in Kunipia-F bentonite, at a dry density of 400 kg/m3. The model is then directly used to predict apparent diffusivities (Da) of Cs, Sr, Cl-, I- and TcO4- and corresponding distribution coefficients (Kd) of Cs and Sr in different bentonites as a function of dry density, without any further adjustment of surface chemical and EDL parameters. Effective diffusivities (De) for Cs, HTO, and TcO4- are also calculated. All calculated values (Da, De, Kd) are fully consistent with each other. A comparison with published, measured data shows that the present model allows a good prediction and consistent explanation of (i) apparent and effective diffusivities for cations, anions, and neutral species in compacted bentonite, and of (ii) Kd values in batch and compacted systems.     

Sensitivity analysis of radionuclide migration in compacted bentonite: a mechanistic model approach

Mechanistic model calculations for the migration of Cs, Ra, Am and Pb in compacted bentonite have been carried out to evaluate sensitivities with respect to different parameter variations. A surface chemical speciation/electric double layer model is used to calculate: (i) porewater composition and radionuclide speciation in solution and at the bentonite surface, yielding the distribution of mobile and sorbed species and (ii) interaction of diffusing species with negatively charged pore walls to obtain diffusion parameters. The basic scenario considers the interaction of compacted bentonite with a fresh-type groundwater; variations include the presence of bentonite impurities and saline groundwater. It is shown that these scenarios result in significant variations of porewater composition that affect migration via three mechanisms that can partly compensate each other: (1) effects on sorption through radionuclide complexation in solution, and competition of major cations for surface sites; (2) changes in radionuclide solution speciation leading to different diffusing species under different conditions; (3) effects on diffusion through changes in the electric double layer properties of the clay pores as a function of ionic strength.     

GENETIC PROGRAMMING AND ITS APPLICATION IN REAL‐TIME RUNOFF FORECASTING1

ABSTRACT: Genetic programming (GP), a relatively new evolutionary technique, is demonstrated in this study to evolve codes for the solution of problems. First, a simple example in the area of symbolic regression is considered. GP is then applied to real‐time runoff forecasting for the Orgeval catchment in France. In this study, GP functions as an error updating scheme to complement a rainfall‐runoff model, MIKE11/NAM. Hourly runoff forecasts of different updating intervals are performed for forecast horizons of up to nine hours. The results show that the proposed updating scheme is able to predict the runoff quite accurately for all updating intervals considered and particularly for updating intervals not exceeding the time of concentration of the catchment. The results are also compared with those of an earlier study, by the World Meteorological Organization, in which autoregression and Kalman filter were used as the updating methods. Comparisons show that GP is a better updating tool for real‐time flow forecasting. Another important finding from this study is that nondimensionalizing the variables enhances the symbolic regression process significantly.     

Experimental and modeling studies on sorption and diffusion of radium in bentonite

The sorption and desorption behavior of radium on bentonite and purified smectite was investigated as a function of pH, ionic strength and liquid to solid ratio by batch experiments. The distribution coefficients (Kd) were in the range of 10(2) to > 10(4) ml g-1 and depended on ionic strength and pH. Most of sorbed Ra was desorbed by 1 M KCl. The results for purified smectite indicated that Ra sorption is dominated by ion exchange at layer sites of smectite, and surface complexation at edge sites may increase Ra sorption at higher pH region. Reaction parameters between Ra and smectite were determined based on an interaction model between smectite and groundwater. The reaction parameters were then used to explain the results of bentonite by considering dissolution and precipitation of minerals and soluble impurities. The dependencies of experimental Kd values on pH, ionic strength and liquid to solid ratio were qualitatively explained by the model. The modeling result for bentonite indicated that sorption of Ra on bentonite is dominated by ion exchange with smectite. The observed pH dependency was caused by changes of Ca concentration arising from dissolution and precipitation of calcite. Diffusion behavior of Ra in bentonite was also investigated as a function of dry density and ionic strength. The apparent diffusion coefficients (Da) obtained in compacted bentonite were in the range of 1.1 x 10(-11) to 2.2 x 10(-12) m2 s-1 and decreased with increasing in dry density and ionic strength. The Kd values obtained by measured effective diffusion coefficient (De) and modeled De were consistent with those by the sorption model in a deviation within one order of magnitude.     

Polyfluorinated telomers in indoor air of Japanese houses

The fluorotelomer alcohols (FTOHs) have been detected in various environmental compartments, including indoor and outdoor air, in North America and Europe. In our previous studies, FTOHs were detected at a relative higher concentration in outdoor air in the Keihan (Kyoto–Osaka, one of the major industrial zones) area, Japan compared to reported data. The exposure level of FTOHs in indoor air in the Keihan area remains unclear. In the present study, indoor air FTOH concentrations were investigated using a passive air sampler containing activated carbon felts. The indoor air sampling was conducted in 49 households of the Keihan area, during winter and summer 2008. Most samples contained 6:2 FTOH, 8:2 FTOH, 10:2 FTOH and 8:2 FTOAc. The median concentration of 8:2 FTOH (5.84 ng m^?3) was highest among fluorotelomers, followed by those of 10:2 FTOH (1.12 ng m^?3), 6:2 FTOH (0.29 ng m^?3), and others. Significant correlations among fluorotelomers were observed in collected samples. The association between housing conditions and 8:2 FTOH concentrations showed that samples collected from bed rooms have higher 8:2 FTOH concentrations than those collected from other locations. In addition, samples collected in winter showed lower levels of 8:2 FTOH than those collected in summer. These findings suggest that 8:2 FTOH is the predominant component among fluorotelomers in indoor air, and that there are emission sources of fluorotelomers in indoor environments of the Keihan area. Further investigations into the origins of fluorotelomers are needed to evaluate indoor contamination with fluorotelomers.     

Nest and maternal origin can influence morphology and locomotor performance of hatchling green turtles (Chelonia mydas) incubated in field nests

In numerous laboratory experiments involving the incubation of reptile eggs, both the maternal origin of eggs and the incubating environment (nest effect) have been demonstrated to influence hatchling phenotype. Although different hatchling phenotypes have been reported from natural nests, the separate effects of maternal origin and nest on hatchling phenotype in natural nests have not been demonstrated because in natural nests the two effects are confounded with each other. Here, we use a split clutch design to experimentally separate nest effects from maternal origin effects in field nests of green turtles (Chelonia mydas). We found both maternal origin and nest to influence hatchling morphology and locomotor performance in some but not all field nests. By using egg mass (maternal origin effect) and nest temperature (nest effect) in multiple regression analysis, we found maternal origin had a greater influence than nest temperature on the morphological attributes of hatchling mass and carapace size, but nest temperature had a greater influence than maternal origin on the performance attributes of self-righting time, self-righting propensity, swim thrust during the first 30 min of swimming, and power stroke rate during the first 30 min of swimming.