TY - CHAP A1 - Bazartseren, Bunchingiv A1 - Holz, Klaus-Peter T1 - Data Preprocessing for Hydrological and Hydromorphological Studies using Neural Networks T2 - Simulation in Umwelt- und Geowissenschaften, Workshop Cottbus 2002 N2 - The use of artificial neural networks for various hydrologic and hydromorphological studies is being investigated. Application of the Artificial Intelligence (AI) methods, especially the neural networks for water related studies is expanding lately, due to its advantages such as being less subjected to he constraints of the physical considerations and quick delivery of feasible and cost-effective responses. Primary stage results of the research are presented in two separate examples.The first example is the runoff modelling for the case of Yellowstone River, USA at the outlet of a high-altitude lake using neural networks, in which the influence of melting ice and snow cover is carefully considered. A wavelet transform is used for smoothening the input signal for the neural network model in order to improve the accuracy and prevent the inconsistency of neural network solutions. A morphological evolution study along cross-shore profiles at the Kiel Bay, Baltic Sea coast using ANN provides the second case. Data preprocessing or in this particular case downsampling of bathymetry measurements through a number of cross shore profiles was done by a wavelet transform.The data oriented approaches, such as neural networks often have to deal with the abundant data or noisy observations, which normally require a thorough analysis, preprocessing or downsampling, to enable a satisfactory performance of the models. The above case studies emphasize the importance and necessity of the data analysis and preprocessing. KW - artificial neural networks KW - coastal morphological study KW - hydrological studies KW - hydromorphological studies Y1 - 2002 SN - 3-8322-0733-3 SP - 53 EP - 64 PB - Shaker CY - Aachen ER - TY - CHAP A1 - Bazartseren, Bunchingiv A1 - Holz, Klaus-Peter T1 - Data Reduction for Coastal Morphological Study Using a Wavelet Transform T2 - Advances in hydro-science and -engineering, volume V, proceedings of the 5th International Conference on Hydro-Science and -Engineering, Warsaw University of Technology, Faculty of Environmental Engineering, Warsaw, Poland, September 18 -20, 2002 N2 - A possibility of compressing coastal morphological data by a wavelet transform was investigated. This was motivated by our intention to use Artificial Neural Networks (ANN) to simulate or predict the evolution of coastal morphology. ANN models for the above purpose should use the bathymetry measurements through a number of cross shore profiles for simulation or prediction, which consequently would cause the dimensionality dilemma in the input space. Therefore, a wavelet transform was chosen as an alternative method for a data compression.Wavelet analysis is a signal processing technique, which captures the time varying frequencies. A profile data is transformed into coefficients of high or low frequency components for different level of wavelet decomposition tree, from which the original profile can be restored back. The data reduction is realized, when the low frequency component coefficients after some level of wavelet decomposition would represent the profile. Thus, depending on the complexity of the profile shape and level of the decompositions, it is possible to reduce the data points of the profile several times.The reconstruction of the original data from up to the fourth level of approximation coefficients was reasonably accurate and at the same time the profile data was reduced to a permissible number as input to the ANN model for prediction. The reconstruction error will further be compared to those of Principal Component Analysis. The study has proven the wavelet transform to be a possible alternative of data reduction for the considered case. KW - wavelet transform KW - coastal morphology KW - data reduction Y1 - 2002 SN - 83-90617-77-3 CY - Warschau ER - TY - CHAP A1 - Bazartseren, Bunchingiv A1 - Holz, Klaus-Peter T1 - Improving the Performance of a Neural Network by Smoothening its Input Signal T2 - Hydroinformatics 2002, proceedings of the Fifth International Conference on Hydroinformatics, Vol. 1, Cardiff, UK, July 1st - 5th 2002 N2 - The article reports an attempt to obtain a better generalization of neural network model by smoothening its input signal. The considered case is a river runoff modelling at the outlet of a high-altitude lake. For the test location, the dominant peak of the annual hydrograph is induced by snow melt in the late spring. The summer rainfall has a less significance to the streamflow and the variability of daily runoff is very low, due to a regulating effect of the lake. The ANN model derived the daily average runoff quite accurately, with small inconsistencies in the output, using only a few input variables. To improve the model performance, the time series of air temperature was smoothened by a wavelet transform and Savitzky-Golay algorithm to compare with the normal procedure of moving window averaging. After using the smoothened data by different techniques, a noticeable improvement has been attained in the quality of the output of the individual ANN models, while the accuracy of the solutions were comparable. KW - neural networks KW - wavelet transform KW - Savitzky-Golay filter KW - runoff modelling Y1 - 2002 SN - 1-8433-9022-1 SP - 743 EP - 748 PB - IWA CY - London ER - TY - CHAP A1 - Bazartseren, Bunchingiv A1 - Hildebrandt, Gerald A1 - Holz, Klaus-Peter T1 - Hydroinformatic Web application and Web service for real-time water level presentation and short-term prediction T2 - Environmental communication in the information society, proceedings of the 16th International Conference Informatics for Environmental Protection, Vol. 2, September 25 - 27, 2002, University of Technology, Vienna, Austria N2 - Modern urbanization tends to cause a fast response time between a heavy precipitation and consequent runoff processes in a river basin. Therefore, it is even more essential nowadays to enable citizens to have a rapid and flexible access to information on the prevention or restoration measures in cases of flood, on the basis of the cutting edge advances of the Information and Communication Technology (ICT). The paper contains a description of a real-time Web application and service for water level observation, processing, presentation and a short term prediction of a river water level in the area of interest. The Artificial Neural Networks (ANN) are implemented for cost-effective water level prediction for a short horizon. On the whole, the Web applications and services should form a part of a publicly accessible Web based flood crisis management system. KW - Web application KW - Web service KW - real-time water level presentation KW - short-term prediction Y1 - 2002 SN - 3-9500036-7-3 SP - 604 EP - 610 PB - ISEP, International Society for Environmental Protection CY - Vienna ER -