• 054903 - NATURAL RESOURCES MANAGEMENT [1ST MOD]

The course develops knowledge and skills for the advanced modelling, management, and control of natural resources systems. The emphasis is on the operational and real time control aspects of natural resources modeling and management, with a focus on water resources systems and extensive reference to real world case studies. Topics include: natural resources systems modeling for management (from stochastic processes, through linear models, to non-linear model and machine learning), predictive modelling, uncertainty and sensitivity analysis, model diagnostics, stochastic and robust optimal control, approximate control methods (e.g. simulation-based approaches, approximate dynamic programming, reinforcement learning), real time control, and complexity reduction methods. The course is organized into two modules: the 1^st module develops topics from data to well validated models, the 2^nd covers algorithms and methods to design optimal decisions and negotiation support systems. The course is aimed at preparing graduate students to work in environmental and water resources engineering field.

Students are expected to acquire knowledge and understanding of the basics of natural resources systems modelling, including how to identify the right model for a given problem and run model diagnostics (uncertainty analysis, sensitivity analysis). And also build knowledge and understanding of the basics of natural resources systems planning, management, and real time control, including how to formulate optimization and optimal control problems, select the most suitable solving approach, and use effectively and efficiently the information available.

The theoretical knowledge built through class lectures and individual studying will be complemented by computer based tutorials, where the students will learn how to turn theory into practice and solve real world cases. The extensive use of real world examples of applications of the theoretical material presented should also help the students in developing strong practical skills.

Students are supposed to be able to autonomously address real world decision-making problems involving water resources planning and management, by selecting and building the most appropriate tools for the problem at hand, running diagnostic analysis tools, ultimately designing robust and reliable solutions.

The course also includes seminars from national and international experts and the presentation, in a two days final workshop, of two scientific papers on selected topics by group of students. Students will learn how to extract key information and messages from scientific literature and reorganize them in a logically, graphically and scientifically sound way for the benefit of the public.

The mix of learning experiences (class lectures, computer tutorials, group working) should facilitate a deep understanding of the course topics and consolidated ability in approaching real world problems.

The course is delivered via lectures and seminar sessions on selected topics.  There will be weekly tutorials on computer-based analysis and modelling works that will contribute to the course assessment. The course is organized into two modules. Detailed topics covered include:

Module 1:

1. Introduction to the course - natural resources: what they are, why they are under stress and major challenges to sustainable management.
2. Modelling for management - why do we need models and how we use them; prediction vs simulation: performance metrics; calibration methods, cross-validation.
3. From linear to weakly non-linear modelling - recap of stochastic processes, linear models, weakly-non linear (TVP, SDP models);
4. Machine learning techniques: non linear data-driven models, e.g. Artificial Neural Network, Tree based regressors, deep learning.
5. Uncertainty and model disgnostics – stochasticity vs deep uncertainty; uncertainty and sensitivity analysis; multimodelling approaches; multiobjective model calibration.

Module 2:

1. Optimal planning - the optimal planning problem; overview of solving approaches; examples from the real world; multi-objective vs many-objective; beyond linear programming: from global optimization algorithms to metaheuristics; rival problem framework.
2. Optimal control - the optimal control problem; Direct Policy Search, Stochastic Dynamic Programming, Approximate Dynamic Programming and Reinforcement Learning, data-driven and partially data-driven control: how to select and use exogenous information.
3. Real time control - deterministic and stochastic Model Predictive Control; deterministic simulation; MonteCarlo simulation.
4. Dimensionality and complexity - non dynamic emulation modeling (iterative Response Surface); model reduction/dynamic emulation modeling; objective reduction; from multi to many objective; multidimensional visualization tools.
5. Introduction to behavioural modelling – coupled human natural systems modelling; data-driven behavioural modelling; introduction to agent based systems.

• SDG6 - CLEAN WATER AND SANITATION
• SDG7 - AFFORDABLE AND CLEAN ENERGY
• SDG13 - CLIMATE ACTION

A good knowledge of math and statistics, and basic knowledge of mathematical modelling will facilitate understanding and elaboration.

The final exam comprises a written part (68%), a computer based assigment (16%), and a computer based contest at the end of the semester (16%).