Evaluating FEW Nexus in the Coupled Natural-Human System with Agent- Based Modeling Y. C. Ethan Yang, Ph.D. GISP P. C. Rossin Assistant Professor 2018/10/04 @ CUAHSI Cyberseminar Department of Civil and Environmental Engineering
Outline Introduction Water-Food-Energy-Environment Nexus Methodology Agent-based Modeling ABM Study 1 Human-ecosystem interaction ABM Study 2 Accommodating natural variability ABM Study 3 Addressing human behavior uncertainty Remark and the way forward 2/36
Introduction Overarching Science Question: How do the complex interactions between humans and nature affect the Earth system at different spatial and temporal scales? A systems approach with nexus thinking 3/36
Introduction How to get married? Systems approach We need to break down the big question and understand all different parts of the it Nexus thinking We also need to understand the connections between different parts Pick a style 4/36
Water-Food-Energy-Environment Nexus Water-Food-Energy-Environment (WFEE) Nexus Why is it important? 1.1 billion have no access to clean water 1.3 billion live without electricity 1.02 billion are hungry Environmental sustainability Water Need more in the future! Food and energy production 5/36
Water-Food-Energy-Environment Nexus Conventional solutions treat these security issues separately, but they are connected to each other. Without nexus thinking, solving a problem in one sector can cause negative effects in another one. 6/36
Water-Food-Energy-Environment Nexus Science Questions for understanding the WFEE Nexus using the systems approach with nexus thinking: What is the first step to quantify human-ecosystem interaction in the WFEE Nexus? Human-ecosystem interaction in the WFEE Nexus - the Yellow River (Yang et al. 2009; Yang et al. 2012) How can we accommodate the dynamics of natural variability in the WFEE Nexus? Coupling a human model with a process-based model - the Mekong River (Khan et al. 2017; Yang et al. 2018) WFEE Nexus How can we address the human behavioral uncertainty in the WFEE Nexus? Developing the human model using Bayesian Inference Map the San Juan River (Hyun et al. 2018) 7/36
Methodology Agent-based Modeling What do you mean by a human model? Not just a traditional socioeconomic model, but a more complex and realistic modeling approach: Agent-Based Modeling (ABM) 8/36
Methodology Agent-based Modeling What is ABM? From Computer Science: Distributed Artificial Intelligence An agent is an object that is driven by its own utility function follows behavioral rules interacts with other agents and the environment autonomously Why ABM? Most top-down model regarding human decision in WFEE Nexus study suffer from various limitations such as policy implementation, complete information exchange, and large computational costs. 9/36
ABM Study 1 Human-ecosystem interaction What is the first step to quantify humanecosystem interaction in the WFEE Nexus? Humanecosystem interaction in the WFEE Nexus - the Yellow River (Yang et al. 2009; Yang et al. 2012) 10/36
ABM Study 1 Human-ecosystem interaction The second largest river in China flows through nine provinces Arid to semi-arid climate Overuse of water resources results in streamflow cutoff since 1972 Zero-flow days Zero-flow distance (km) Zero-flow days Zero-flow distance (km) 11/36
ABM Study 1 Human-ecosystem interaction Traditional management approach: Unified Water Flow Regulation (UWFR, top-down control) Water market scenario (bottom-up management) A water trading mechanism is used to account for water rights violations, i.e., water users (defined as agents) are allowed to buy or sell water 12/36
ABM Study 1 Human-ecosystem interaction 20 21 Agent system map 52 water use agents 5 reservoir agents 3 ecosystem agents 1 6 7 8 9 R 1 12 R 2 13 5 11 4 10 2 3 15 14 34 35 18 16 36 19 17 31 32 33 37 29 24 22 26 25 R 3 23 27 30 40 39 38 R 4 28 41 42 46 R 5 44 49 43 48 47 51 E 1 49 50 E 2 52 E 3 A i A i Water use agents Water use agents with source flow Mainstream inflow source Tributary inflow source R i Reservoir agents Mainstream E i Ecosystem agents Tributary 13/36
Water consumption Water price ABM Study 1 Human-ecosystem interaction Water consumption p i * Water availability Equilibrium state numerical loop Equilibrium state Water right x i * Water right p i * water selling water right>x i * water buying water right <x i * Water price numerical loop 14/36
Streamflow Storage Streamflow Water Con. Streamflow ABM Study 1 Human-ecosystem interaction ABM calibration 7 6 5 4 3 2 1 0 Observed UWFR 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Obs-A21 UWFR-A21 1 2 3 4 5 6 7 8 9 101112 Month Agent 18 16 14 12 10 8 6 4 2 0 Obs-R1 Obs-R2 Obs-R3 Obs-R4 Obs-R5 UWFR-R1 UWFR-R2 UWFR-R3 UWFR-R4 UWFR-R5 1 2 3 4 5 6 7 8 9 10 11 12 Month 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Obs-A5 UWFR-A5 1 2 3 4 5 6 7 8 9 101112 Month 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Obs-E3 UWFR-E3 1 2 3 4 5 6 7 8 9 101112 Month 15/36
Water Con. GDP. ABM Study 1 Human-ecosystem interaction Basin-wide water consumption and GDP 10.0 8.0 6.0 Price Water Market (ABM) UWFR Conventional (UWFR) 140 130 120 110 Price Water Market (ABM) UWFR Conventional (UWFR) 4.0 100 2.0 90 0.0 1 2 3 4 5 6 7 8 9 10 11 12 Month Conventional 34.52 billion m 3 Water Market 33.80 billion m 3 80 1 2 3 4 5 6 7 8 9 10 11 12 Month Conventional 1246.68 billion RMB Water Market 1270.01 billion RMB Difference: 23.33 billion RMB 16/36
Streamflow. Streamflow. Streamflow. ABM Study 1 Human-ecosystem interaction Effect on ecosystem agents Total water buying by agents 4.25 BCM/year 44 52 E 3 3.5 3.0 2.5 2.0 1.5 Ecosystem Agent 3 Total water selling by agents 7.51 BCM/year Difference 3.26 BCM/year 46 43 47 49 48 1.0 51 49 E 1 E 2 Ecosystem Agent 1 50 1.0 0.5 0.0 1.0 1 2 3 4 5 6 7 8 9 10 11 12 Month Ecosystem Agent 2 Cost for governments 0.8 0.6 0.8 0.6 2.95 billion RMB < 23.33 billion RMB 0.4 0.4 Cost < Benefit self-sustaining system! 0.2 0.0 1 2 3 4 5 6 7 8 9 10 11 12 Month 0.2 0.0 1 2 3 4 5 6 7 8 9 10 11 12 Month 17/36
Streamflow (cms) So, what s next? Quantify the natural-human interaction was the easiest part of this WFEE nexus (that s why I did it) A bigger challenge of this is how to simulate human s decision making under year-by-year natural variability? 7000 6000 5000 4000 3000 2000 1000 0 1991-2007 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Crop area decisions Hydropower decisions 18/36
ABM Study 2 Accommodating natural variability How can we accommodate the dynamics of natural variability in the WFEE Nexus? Coupling a human model with a processbased model - the Mekong River (Khan et al. 2017; Yang et al. 2018) 19/36
ABM Study 2 Accommodating natural variability The Mekong River Basin is the sixth largest river basin in the world. Over 50 million people depend upon the River for food, water, and economic sustenance, and the Basin is home to several diverse ecosystems. The primary energy sources in the Mekong River Basin is hydropower (China-72%, Myanmar- 75%, Laos- 91%, Cambodia- 57%, Vietnam- 47% and Thailand- 30%). Similar water use agent definition as the Yellow River study, 32 dams, 23 eco hotpots 20/36
ABM Study 2 Accommodating natural variability Soil and Water Assessment Tool (SWAT) A river basin scale model developed to quantify the impact of land management practices in large, complex watersheds. Open-source Physical process-based daily simulation Fully or semi - distributed - HRU Calibrated: 1983-2007 21/36
ABM Study 2 Accommodating natural variability Streamflow variability will affect human decision Streamflow Initial Reservoir and Agriculture settings User-defined preferences and cooperation setting Crop Yield Day 1, Year 1 Day 1, Year 2 SWAT Reservoir Storage and Release Day 365, Year 1 Day 365, Year 2 ABM Other agents decisions Reservoir Characteristics Agriculture, Hydropower and Ecological Targets Reservoir Operations Irrigated Area Coverage Day 365, Last Year Agriculture, Hydropower and Ecologic performance summary 22/36
ABM Study 2 Accommodating natural variability Rule-based simulation ABM setup ABM Supporting Data files and Web-User Input 23/36
ABM Study 2 Accommodating natural variability How streamflow variation + agent s preference affect results? Agent s water use preference on AG AG rank =1 Agent s water use preference on HP HP rank =1 AG rank =3 HP rank =3 24/36
ABM Study 2 Accommodating natural variability Level of Cooperation Drought in Vietnam (Mekong Delta) causes it to request for help from upstream countries Tham Theun 2 Dam in Central Laos releases more water as response Third party impacts Negative - Reduced food production in Central Laos Positive - Increased food production in Cambodia Normal runs Tham Theun 2 helps Vietnam More water released from the dam during the response year Less water released from the dam during the following year 25/36
Ok, and then? So if you think natural variability is hard to model, modeling human behavior for WFEE nexus is even harder. Human behavior is affected by so called common sense. Which route will this guy take to go home with shortest time? (c) (a) (b) 26/36
ABM Study 3 Addressing human behavior uncertainty How can we address the human behavioral uncertainty in the WFEE Nexus? Developing the human model using Bayesian Inference Map the San Juan River (Hyun et al. 2018) 27/36
ABM Study 3 Addressing human behavior uncertainty The San Juan River Basin is located at the borders of four States: Utah, Colorado, Arizona, and New Mexico It is the largest tributary of the Colorado River Basin A typical Western US basin that involve multiple water users (farmers, cities and power plants) and need to consider not fully utilized Indian water right. 28/36
ABM Study 3 Addressing human behavior uncertainty River-routing and reservoir management modeling: RiverWare RiverWare ABM PLEXOS ABM (farmers are defined as agents, how much to irrigated is the decision) Looking backward but acting forward 29/36
ABM Study 3 Addressing human behavior uncertainty Bayesian Inference (BI) map As the internal thinking process in farmer s brain to decide the probability of expanding the irrigated area Cost-Loss model As the external economic factors that affect the probability of expanding the irrigated area action/probability 1-p p Increase C C Decrease 0 L 30/36
ABM Study 3 Addressing human behavior uncertainty How do we know the coupled ABM-RiverWare model can provide reasonable results? Check historical irrigated area change 31/36
ABM Study 3 Addressing human behavior uncertainty How can we use this model to address risk perception? Basin-wide results Individual farmer results 32/36
ABM Study 3 Addressing human behavior uncertainty What will happen if the expanding irrigated area becomes expensive? 33/36
Remark and the way forward Coupled human-natural complex system WFEE Nexus Quantify Human-nature interaction in the WFEE Nexus Evaluate natural variability in the WFEE Nexus Address human behavior uncertainty in the WFEE Nexus 34/36
Remark and the way forward The ongoing DoE project to dig deeper into the spatial and temporal scale issue in WFEE Nexus. How can we use the ABM approach to upscale human decision uncertainty? Compare the San Juan-ABM model vs. Colorado River Simulation System (CRSS)-ABM A new NSF project to develop a coupled ABM framework (a food module, an energy module and a water module) for the Columbia River Basin s WFEE reliability, resilience, and vulnerability. How can we adopt structured public participation in the ABM framework? Can we use the ABM framework to facilitate the treaty negotiation? 35/36
Acknowledgement: Dr. Ximing Cai (UIUC), Dr. Jian-Shi Zhao (Tsinghua), Dr. Claudia Ringler (IFPRI), Dr. Hua Xie (IFPRI), Dr. Hassaan Khan (Stanford), Dr. Shih-Yu Huang (Lehigh), Dr. Jin-Young Hyun (UMass) and Dr. Vincent Tidwell (Sandia) Ph.D. student and postdoc positions available in CAWS! Please contact me if you are interested yey217@lehigh.edu Department of Civil and Environmental Engineering