GIST Repository
College of Engineering
Department of Environment and Energy Engineering
4. Theses(Ph.D)
Integrating AQUATOX, Ecological Big Data, and Machine Learning for Short-Term Chemical impact Assessment on River Ecosystems
- Abstract
- This thesis develops a methodology for assessing the impact of chemical accidents on aquatic ecosystems using ecological modeling. The study consists of three main parts: First, it generates a quantifiable species-specific ecological parameter database for domestic aquatic ecosystems. This involves estimating representative biomass for 687invertebrate species and 157 fish species in Korean rivers using length-weight relationships and literature data. The study proposes and evaluates a method for applying these parameters to ecological models of domestic river ecosystems. The most accurate method (R2 = 0.5529 for invertebrates, R2 = 0.9349 for fish) used genus-level averages of family-level averages of length- weight constants. Second, the research parameterizes food web interactions between aquatic species in Korean rivers. It employs a Deep Neural Network model, constrained by ecological domain knowledge, to estimate food preferences among species groups. The model is trained on big data from species presence surveys across various regions. The optimal model (R2 = 0.8953, RMSE = 0.2369) was achieved when the maximum trophic level for species consuming producers was set at 3.2, and predation was allowed only when the predator's trophic level exceeded that of the prey. Third, the study applies the AQUATOX model to assess the impact of a phenol spill on the Iksan Stream ecosystem. The model is constructed using water quality, flow, and ecological data from various monitoring points along the stream. A control simulation is established to represent the stream's ecosystem without the phenol spill, followed by a perturbed simulation incorporating observed phenol concentrations (maximum 2.9 mg/L). The model's performance is evaluated using Chlorophyll a concentrations, with a no improvement in the perturbed scenario (zero increase in Nash-Sutcliffe Efficiency). The results demonstrate complex ecosystem responses to the phenol spill, highlighting the importance of considering food web interactions and long-term effects in chemical spill impact assessments. This research contributes to the development of more comprehensive environmental risk assessments and management strategies for aquatic ecosystems in Korea.
Yeom, Jaehoon (염재훈). Integrating AQUATOX, Ecological Big Data, and Machine Learning for Short-Term Chemical impact Assessment on River Ecosystems (AQUATOX 모델, 생태 빅데이터, 머신러닝을 활용한 하천 생태계의 단기 화학물질 노출 영 향 평가). School of Environment and Energy Engineering. 2025. 235p. Prof. Sang Don Kim
- Issued Date
- 2025
- Type
- Thesis
- Alternative Author(s)
- Jaehoon Yeom
- Department
- 대학원 환경에너지공학부
- Advisor
- Kim, Sang Don
- Table Of Contents
- Abstract i
Contents ii
Figure Contents iv
Table Contents viii
Chapter 1 Preface 9
1.1 Overview 9
1.2 Background 12
1.2.1 Chemical accident trends and damage on aquatic environments in the world 12
1.2.2 Definition of Ecological risk assessment (ERA) and community level ERA 14
1.2.3 Modelling approach in high tier ERA 17
1.3 Purpose of this study 18
1.4 Thesis organization 20
Chapter 2 Generating and validating ecological biomass database by using Literature based estimation for
aquatic species 21
2.1 Estimating biomass of aquatic invertebrates in Korean rivers 21
2.1.1 Abstract 21
2.1.2 Introduction 22
2.1.3 Materials and methods 24
2.1.4 Result and Discussions 29
2.1.5 Conclusions 43
2.2 Estimating biomass of fishes in Korean Rivers 45
2.2.1 Abstract 45
2.2.2 Introduction 45
2.2.3 Materials and Methods 47
2.2.4 Result and Discussion 51
2.2.5 Conclusions 69
Chapter 3 Parametrization of food web interaction between species by using machine learning technique 71
3.1 Ecological clustering of aquatic species in Korean rivers 71
3.1.1 Abstract 71
3.1.2 Introduction 71
3.1.3 Materials and Methods 73
3.1.4 Result and Discussions 81
iii
3.1.5 Conclusion 87
3.2 Parametrization of foodweb interaction between aquatic species in Korean rivers 89
3.2.1 Abstract 89
3.2.2 Introduction 89
3.2.3 Materials and Methods 90
3.2.4 Result and Discussions 99
3.2.5 Conclusions 106
Chapter 4 Assessment of AQUATOX model in phenol exposure accdent on Iksan river 108
4.1 Ecological models and AQUATOX model for ecological risk assessments 108
4.1.1 Comparison of ecological models 108
4.1.2 Concept of AQUATOX model 109
113
4.1.3 Methodology to assess chemical impact with AQUATOX model 115
4.2 AQUATOX model for phenol exposure in Iksan River 116
4.2.1 Abstract 116
4.2.2 Introduction 117
4.2.3 Materials and Methods 119
4.2.4 Result and Discussions 128
4.2.5 Conclusions 156
Chapter 5 Conclusions 158
References 161
Appendix 171
Curriculum Vitae 227
Acknowledgement 232
iv
Figure Contents
Figure 1. Pros and cons of various tiers in ERA (Ecological Risk Assessment) [5] 15
Figure 2. Goals of the study 19
Figure 3. Chapter organization 20
Figure 4. Scheme of methodology to interpolate LWRs coefficients 25
Figure 5. Scheme of methodology to evaluate LWRs coefficient 28
Figure 6. Mean length distribution of invertebrate species linving in Korean rivers 29
Figure 7. Predicted dry weight versus Measured dry weight of invertebrate species with no exception
on outliers (Genus level) 33
Figure 8. Predicted dry weight versus Measured dry weight of invertebrate species with no exception
on outliers (Family level) 34
Figure 9. Predicted dry weight versus Measured dry weight of invertebrates species with no exception
on outliers (Order level) 35
Figure 10. Predicted dry weight versus Measured dry weight of invertebrate species with exception on
outliers (Genus level) 36
Figure 11. Predicted dry weight versus Measured dry weight of invertebrate species with exception on
outliers (Family level) 37
Figure 12. Predicted dry weight versus Measured dry weight of invertebrate species with exception on
outliers (Order level) 38
Figure 13. LWRs coefficient matching process for mollusk species 40
Figure 14. Scheme of methodology to evaluate Length-Weight coefficient for fish 48
Figure 15. Number of LWRs Coefficients by Order level 51
Figure 16. Length distribution of target fish species living in Korean rivers. 52
Figure 17. Predicted wet weight versus Measured wet weight of fish species with no exception on
outliers (Genus level, CG ~ G4G Groups) 55
Figure 18. Predicted wet weight versus Measured wet weight of fish species with no exception on
outliers (Genus level, G5G ~ G9G Groups). 56
Figure 19. Predicted wet weight versus Measured wet weight of fish species with no exception on
outliers (Family level, CF ~ G4F Groups) 57
Figure 20. Predicted wet weight versus Measured wet weight of fish species with no exception on
outliers (Family level, G5F ~ G9F Groups) 58
Figure 21. Predicted wet weight versus Measured wet weight of fish species with no exception on
outliers (Order level, CO ~ G4O Groups) 59
Figure 22. Predicted wet weight versus Measured wet weight of fish species with no exception on
outliers (Order level, G5O ~ G9O Groups) 60
Figure 23. Predicted wet weight versus Measured wet weight of fish species with exception on outliers
(Genus level, CG ~ G4G) 63
Figure 24. Predicted wet weight versus Measured wet weight of fish species with exception on outliers
(Genus level, G5G ~ G9G) 64
v
Figure 25. Predicted wet weight versus Measured wet weight of fish species with exception on outliers
(Family level, CF ~ G4F Groups) 65
Figure 26. Predicted wet weight versus Measured wet weight of fish species with exception on outliers
(Family level, G5F ~ G9F Groups) 66
Figure 27. Predicted wet weight versus Measured wet weight of fish species with exception on outliers
(Order level, CO ~ G4O Groups) 67
Figure 28. Predicted wet weight versus Measured wet weight of fish species with exception on outliers
(Order level, C5O ~ G9O Groups) 68
Figure 29. Ecological clustering process of invertebrate species groups 79
Figure 30. Ecological clustering process of fish species groups 80
Figure 31. Result of correlation analysis of invertebrates group with two parameters (Locomotion and
Biomass) 81
Figure 32. Result of correlation analysis of fish group with two parameters (Feeding habit and
Biomass) 82
Figure 33. Result of K-means clustering for invertebrate groups 82
Figure 35. Mean trophic level result of K-means clustering for invertebrate groups (I 42 ~ I 82) 83
Figure 34. Mean trophic level result of K-means clustering for invertebrate groups (I 1 ~ I 41) 83
Figure 36. Outlier result of K-means clustering for invertebrate groups (I 1 ~ I 41) 84
Figure 37. Outlier result of K-means clustering for invertebrate groups (I 42 ~ I 82) 84
Figure 38. Result of K-means clustering for fish groups 85
Figure 39. Mean trophic level result of K-means clustering for fish groups 86
Figure 40. Outlier result of K-means clustering for fish groups 87
Figure 41. Scheme of food preference table for 113 species groups from clustering 93
Figure 42. Scheme of Existence database for aquatic species monitored in Korean river during 2010 to
2020 94
Figure 43. Scheme of Mean trophic level database and distribution assumption 94
Figure 44. X input database for DDN (Deep Neural Network) 95
Figure 45. Y input database for DDN (Deep Neural Network) 95
Figure 46. Scheme for food preference model training methodology by using DNN 96
Figure 47. Model efficiency of expecting food preference table 101
Figure 48. Food preference of each fish group on whole fish group 102
Figure 49. Food preference of each fish group on whole Invertebrates group 103
Figure 50. Food preference of each fish group on whole producer group 103
Figure 51. Food preference of each invertebrates group (I1 ~ I41) on whole invertebrate group 104
Figure 52. Food preference of each invertebrates group (I42 ~ I82) on whole invertebrate group 104
Figure 53. Food preference of each invertebrates group (I1 ~ I41) on whole producer group 105
Figure 54. Food preference of each invertebrates group (I42 ~ I82) on whole producer group 105
vi
Figure 55. Schematic figure of sub-model interaction in the AQUATOX model [33] 110
Figure 56. Window for the parameter of food web interaction in AQUATOX 113
Figure 57. Weibull distribution for the time varying affected fraction of organism from exposure [33]
. 114
Figure 58. Process of simulating AQUATOX model for ecological assessment [146] 115
Figure 59. Site explanation of Iksancheon 119
Figure 60. Monitoring data of pheonol concentration (mg/L) for Iksancheon (Iksancheon3 monitoring
site) 120
Figure 61. Monitoring data of water quality for Iksancheon (Iksancheon3 monitoring site) 121
Figure 62. Monitoring data of flow rate (m3/day) for Iksancheon (Insu monitoring site) 121
Figure 63. Partial correlation analysis of waterquality parameters in Iksancheon 122
Figure 64. Interpolated result for water quality parameters of Iksancheon2 by using LSTM model 129
Figure 65. Interpolated result of flow rate (m3/day) of Iksancheon (Insu monitoring site) by using
LSTM and GRU models 129
Figure 66. Scheme of food web in Iksancheon (Using boundary condition TL 1 predation TL limit -
3.2/ Predator TL 0.3 > Prey TL) 132
Figure 67. Scheme of food web in Iksancheon (Using boundary condition TL 1 predation TL limit -
3.5/ Predator TL 0.3 > Prey TL) 133
Figure 68. Scheme of food web in Iksancheon (Using boundary condition TL 1 predation TL limit -
3.8/ Predator TL 0.3 > Prey TL) 133
Figure 69. Scheme of food web in Iksancheon (Using boundary condition TL 1 predation TL limit –
4.1/ Predator TL 0.3 > Prey TL) 134
Figure 70. SSD curve for Cypriniformes (LC50 for 2 days exposure of phenol) 135
Figure 71. SSD curve for Cypriniformes (LC50 for 1 day exposure of phenol) 136
Figure 72. SSD curve for Cypriniformes (LC50 for 4 days exposure of phenol) 136
Figure 73. SSD curve for Diptera (LC50 for 2 days exposure of phenol) 137
Figure 74. SSD curve for Coleoptera (LC50 for 2 days exposure of phenol) 137
Figure 75. SSD curve for Hemiptera (LC50 for 2 days exposure of phenol) 138
Figure 76. SSD curve for Snail (LC50 for 2 days exposure of phenol) 139
Figure 77. SSD curve for species living in Iksancheon (NOEC values for exposure of phenol) 140
Figure 78. Comparison of monitored and modeled inflow flow rate (m3/day) for target segment during
control simulation of AQUATOX model 141
Figure 79. Comparison of monitored and modeled water quality of target segment during control
simulation of AQUATOX model (Calibration of water quality in control simualtion) 142
Figure 80. Result of modeled biomass density (g/ m2 dry) for each species group of target segment
during control simulation of AQUATOX model 144
Figure 81. result of modeled phenol concentration (ug/L) of target segment during perturbed
simulation of AQUATOX model 146
Figure 82. Comparison of Chlorophyll a concentration (ug/L) of target segment during control and
vii
perturbed simulation of AQUATOX model 146
Figure 83. Difference in Chlorophyll a concentration (ug/L) of target segment between control and
perturbed simulation of AQUATOX model (%) 147
Figure 84. Difference in Biomass denstiy (g/m2 dry) of each species group of target segment between
control and perturbed simulation of AQUATOX model (%) 148
Figure 85. Percent perturbation (%) in Biomass denstiy (g/m2 dry) of each species group of target
segment between control and perturbed simulation of AQUATOX model 150
Figure 86. MAD analysis result of phytoplankton groups in Iksancheon 151
Figure 89. MAD analysis result of Invertebrates groups in Iksancheon 152
Figure 88. MAD analysis result of fish group in Iksancheon 153
Figure 89. Chlorophyll a concentration (ug/L) from control, perturbed simualtion and monitored data
. 155
viii
Table Contents
Table 1. Data sources of parameters used for LWRs 27
Table 2. Availability of LWRs coefficients for generating group coefficients 30
Table 3. Data point and outlier information for invertebrate species (C means Control, G1~ G4 mean
Group 1 ~ Group 4, G, F, and O mean Genus, Family, and Order level each – CG group means
Control group and Genus level approximation) 31
Table 4. Target group of mollusk species for generating mean weight of species 39
Table 5. Generated mean weight for mollusk species 41
Table 6. Analysis on database for monitored weight of fish species (C means Control, G1~ G9 mean
Group 1 ~ Group 9, G, F, and O mean Genus, Family, and Order level each – CG group means
Control group and Genus level approximation) 53
Table 7. Result of data analysis for each group and ourlier ratio (Purple color means R2 >0.5, red color
means R2 > 0.7, bold red color means R2 >0.9) 62
Table 8. Data sources for investigation of ecological indicators 74
Table 9. Characteristics of each ecological group of fish species 75
Table 10. Characteristics of each feeding habit group of fish species 75
Table 11. Characteristics of each ecological group of invertebrate species 76
Table 12. Characteristics of each locomotion group of invertebrate species 76
Table 13. Locomotion score 78
Table 14. Feeding habit score 78
Table 15. Hyperparameters for DNN learning of food preference 97
Table 16. Boundary condition with Domain knowledge for generating Food preference table
(Maximum trophic level to eat P group and Trophic Level Condition between Predator and Prey
groups) 98
Table 17. Model performance of generated food preference table for each Domain knowledge
condition 100
Table 18. Comparison of ecological models for aquatic environments 108
Table 19. Model process of plant compartment in AQUATOX [33] 111
Table 20. Model process of the animal compartment in AQUATOX 112
Table 21. Hyperparameter condition of LSTM model for training water quatlity regression between
upstream and downstream of Iksancheon 123
Table 22. Hyperparameter condition of LSTM and GRU models for interpolation of flow rate in
Iksancheon (Insu monitoring site) 124
Table 23. Model efficiency of spatial interpolation using LSTM on water quality parameters of
Iksancheon2 128
Table 24. Occurence of species groups in Iksancheon 130
Table 25. Food preference table for Iksancheon environemnt - relative preference (%, Using boundary
condition TL 1 predation TL limit 3.5/ Predator TL 0.3 > Prey TL) 131
- Degree
- Doctor
- Appears in Collections:
- Department of Environment and Energy Engineering > 4. Theses(Ph.D)
- 파일 목록
-
-
관련 파일이 존재하지 않습니다.
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.