A Non-Parametric Bayesian Network Hydrologic Model a Case Study of a Lowland Catchment

A Non-Parametric Bayesian Network Hydrologic Model a Case Study of a Lowland Catchment

A Non-Parametric Bayesian Network Hydrologic Model A Case Study of a Lowland Catchment Sjoerd Gnodde ANON-PARAMETRIC BAYESIAN NETWORK HYDROLOGIC MODEL ACASE STUDY OF A LOWLAND CATCHMENT ANON-PARAMETRIC BAYESIAN NETWORK HYDROLOGIC MODEL ACASE STUDY OF A LOWLAND CATCHMENT MSc Thesis to obtain the degree of Master of Science at Delft University of Technology, to be defended online on June 29, 2020 at 14:001. by Sjoerd GNODDE 1Please contact the author if you want to attend the defence. Thesis committee: dr.ir. O. Morales-Nápoles, TU Delft (Chairman) dr. M. Hrachowitz, TU Delft dr. E. Ragno, TU Delft ir. B. Dekens, Witteveen+Bos dr.ir. J. Hoch, Universiteit Utrecht Keywords: Lowland catchments, Hydrologic modeling, Non-parametric Bayesian Net- works, Data-driven models, Model optimization Cover design Anniek Keijer & Sjoerd Gnodde Photo back cover The river Vledder Aa during the Water-Op-Maat project middenloop Vled- der Aa. Photo by Corné Joziasse. An electronic version of this thesis is available at http://repository.tudelft.nl/. I believe that we do not know anything for certain, but everything probably. Christiaan Huygens CONTENTS Summary xi Preface xv 1 Introduction 1 1.1 Problem statement and objective . 1 1.2 Previous research into Bayesian networks and similar methods in hydrology . 3 1.3 Overview research goals and design. 4 2 Copulas and non-parametric Bayesian networks 7 2.1 Bayesian network . 7 2.2 Sklar’s theorem . 8 2.3 Types of copula . 8 2.3.1 Gaussian . 9 2.3.2 Archimedean copulas . 10 2.4 Fit a copula to data . 13 2.4.1 Maximum Likelihood Estimation . 13 2.4.2 Directly from correlation. 13 2.5 Application of copulas in non-parametric Bayesian networks. 14 2.6 Multivariate normal method . 15 2.6.1 Process. 15 2.6.2 Correlation matrix . 16 2.6.3 Conditional MVN parameters . 16 2.7 Random sampling . 18 2.8 Calculating conditional distributions . 18 2.8.1 By sampling regularly . 19 2.8.2 By sampling randomly . 19 3 Case Study 21 3.1 Selecting case study . 21 3.2 Vledder, Wapserveense and Steenwijker Aa . 21 3.2.1 History . 22 3.2.2 Recent works in the catchment . 22 3.2.3 Catchment delineation . 23 3.2.4 Waterways . 24 3.2.5 Geology . 25 3.2.6 Land use . 25 3.2.7 Climate . 26 3.2.8 Artificial structures and water fluxes . 26 vii viii CONTENTS 3.3 Data. 27 3.3.1 Overview. 27 3.3.2 Discharge . 27 3.3.3 KNMI data . 29 3.3.4 Precipitation . 30 3.3.5 Groundwater level . 32 3.3.6 Surface water level . 33 3.3.7 Soil moisture. 33 3.3.8 NDVI. 34 3.3.9 Processing and filtering . 35 3.4 Water balance. 35 3.4.1 Monthly scale . 36 3.4.2 Complete timeframe - Budyko framework . 37 4 Initial model and performance 39 4.1 Initial model . 39 4.2 Determining performance . 40 4.2.1 NSE . 40 4.2.2 Kling-Gupta efficiency . 41 4.3 k-fold cross-validation . 41 4.4 Performance per observation . 42 5 Testing the copula assumption 43 5.1 Introduction . 43 5.2 Autocorrelation test . 44 5.2.1 Unconditional autocorrelation. 44 5.2.2 Partial Autocorrelation . 44 5.3 Multidimensional Cramér-von Mises test . 45 5.3.1 Theory . 45 5.3.2 Results . 47 5.4 Absolute differences . 47 5.5 Quadrant Pearson correlation. 48 5.5.1 Tail dependence in data . 48 5.5.2 Tail dependence for copula types . 49 6 Selection and implementation of the marginal distribution 51 6.1 Empirical cumulative density function . 52 6.2 Altered logistic function. 52 6.2.1 Initial values for fitting the CDF . 53 6.2.2 Optimal number of fit parameters . 53 6.3 Gaussian mixture model . 53 6.3.1 Initial values for fitting the CDF . 55 6.3.2 Optimal number of normal distributions . 55 6.4 Inverse cumulative function . 56 6.4.1 Fitting the CDFs to the data . 56 CONTENTS ix 6.5 Preferred CDF. 56 6.5.1 Optimal fit . 56 6.5.2 Extrapolation . 57 6.5.3 Computational duration . 57 6.5.4 Conclusion. 57 6.6 Shifting the CDFs to increase extrapolation . 57 6.6.1 Shift samples outwards . 58 6.6.2 Limit shifted range . 58 6.6.3 Differentiate in extrapolation sides . 59 6.6.4 Implementation . 59 7 Model parameters and error sensitivity 63 7.1 Most likely value from sampled values . 63 7.2 Number of samples to use . 64 7.2.1 Good fit of the target variable . 64 7.2.2 Confidence intervals . 65 7.2.3 Computational time . 65 7.3 Subset period data . 66 7.4 Sensitivity input error . 66 7.4.1 Systematic errors or bias . 67 7.4.2 Random errors . 67 7.4.3 Systematic errors in the discharge measurements . 70 8 Bayesian network layout 73 8.1 Criteria . 73 8.2 Strategy . 74 8.3 Connections in the model based on influences on the variables . 74 8.4 Selected implementations . 77 8.5 Final model . 78 9 Results Bayesian network and benchmark models 81 9.1 General results Bayesian network . 81 9.1.1 Complete k-fold test . 81 9.1.2 Factors KGE . 81 9.1.3 Error per observed values . 83 9.1.4 Not fixing all variables . 83 9.1.5 Predicting all variables . 84 9.1.6 Predicting all days . 85 9.2 Benchmark models . 86 9.2.1 Saturated Bayesian network . ..

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    156 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us