Teaching

CE 4110/6250 Environmental Systems Modeling and Management

This course emphasizes the formulation of environmental management issues as optimization problems. Simulation models will be presented and then combined with optimization algorithms. Environmental issues to be addressed may include power capacity expansion, air quality, water quality, water supply, and reservoir operations. Optimization techniques presented include linear programming, dynamic programming, nonlinear programming and genetic algorithms. Learning objectives include:

• Simulate a variety of simple environmental systems.
• Formulate and solve linear, nonlinear, dynamic, and stochastic programming problems.
• Use optimization software to analyze environmental systems and interpret the output to guide management decisions.
• Understand the importance of accounting for uncertainty in environmental decision-making problems.
• Learn to present a structured, organized systems analysis.

Python coding examples from the course can be found here: https://github.com/EnvSystemsUVA/CodingExamples

CE 6280 Stochastic Hydrology

The goal of this course is to illustrate the importance of uncertainty analysis in hydrology. Whether data-driven or mechanistically-driven, all hydrologic models have errors, and the data used to build such models represent a finite sample of a stochastic process. Quantifying the uncertainty in hydrologic model predictions due to limited data availability and model error is important for informing the design and management of civil infrastructure systems.

In this course, students learn stochastic methods used in hydrology for this purpose. Specific statistical concepts covered include applications of extreme value theory to estimate flood and drought statistics, regionalization methods for predictions in ungauged basins, and trend analysis of historical time series. Students should leave the course with an understanding of how to apply these methods in practice to design civil infrastructure systems that are robust to hydrologic uncertainty.

SYS 4021 Linear Statistical Models

What are the contributing factors to the severity of train accidents? How do you predict if an e-mail is spam? How can you translate goal-directed problems such as these into actionable decisions and meaningful recommendations that can have vast societal implications? How can you harness multi-dimensional, heterogeneous data to analyze the problem? In this course, we explore Evidence Informed Systems Engineering (EISE) practices and how they can be applied to difficult, open-ended problems.

The primary tools for EISE come from linear statistical models and this course demonstrates the use of these models for problem understanding, prediction, and control. We learn how to formulate hypotheses, build statistical models to test them, and make recommendations based on our findings. These steps can be laden with biases, for example in the data available to test these hypotheses, and in the metrics used to assess success. We learn how to identify and prevent these biases to ensure equitable outcomes.

The specific modeling tools we cover include principal components analysis, multivariate linear regression, logistic regression, time series analysis, and simulation and bootstrapping. In class, we concentrate on the theory and practice of model construction, while weekly labs assess your understanding of the theory and ability to apply it in practice. Projects provide open-ended problem solving situations that illustrate the broad applicability of the methods in a setting similar to what you will encounter in the real world. We hope these projects illustrate the value of statistical modeling and that the course provides a foundation for future learning.

Interactive Reservoir Simulation Game

Slides explaining the reservoir simulation game (link) and a Jupyter notebook on Google Colab to play it (link).

Weather Generation Methods

Slides explaining parametric and non-parametric methods for synthetic weather generation, as well as approaches to condition them on climate change projections and seasonal climate forecasts. (link)

Introduction to Information Theory

Slides and coding exercise introducing information theory and metrics of synergy, uniqueness and redundancy as defined by Goodwell et al. (2017). (link)