Welcome to CEVE 543!!
Lecture
2023-08-21
Introductions
Today
Introductions
Climate Hazards
What makes environmental data unique?
Course Organziaton
Should you take this class?
About me
- Dr. James Doss-Gollin
- Assistant professor in CEVE
- Interested in bridging Earth science, data science, and decision science to improve climate risk management and long-term infrastructure planning
- Hometown: New Haven, CT (❤️ for Houston, NYC, and Luque, Paraguay)
- Doss-Gollin lab
TA
- Yuchen Lu
- Third year Ph.D. student in CEVE
- Currently researching statistical methods to estimate the probability of extreme precipitation
- Hometown: Wuhan, China (via Pittsburgh)
Your turn!
- Your name
- Your field and program of study
- Your hometown(s)
What is one thing you hope to learn this semester?
Take a moment to think, write it down, and then we’ll share.
Climate Hazards
Today
Introductions
Climate Hazards
What makes environmental data unique?
Course Organziaton
Should you take this class?
Floods in Paraguay, 2015
Figure 1: Municipalidad de Asunción
TX Cold Snap, 2021
What other climate hazards do you know about?
Take a moment to think, write it down, and then we’ll share.
How do we manage climate risks?
- Reduce emissions to prevent future climate change (“mitigation”)
- Real-time monitoring and forecasting
- Building codes and design standards
- Insurance
- and much more!
Bayesian Decision Theory
Expected reward \(R\) (equivalently utility, loss, etc.) for taking some decision \(a \in \mathcal{A}\): \[ \mathbb{E}(R(a)) = \int_{\mathcal{S}} R(a, {\bf{s}}) p({\bf{s}}) d{\bf{s}} \] Crucial insight: \[ \mathbb{E}(R(a)) \neq R(a, \mathbb{E}(\bf{s})) \]
Implications
- We often care about extremes
- Uncertainty (especially of extremes) matters
- What makes a “good” estimates of \(p(\bf{s})\)?
- Physically accurate / realistic
- High spatial and temporal resolution to quantify impacts on people and infrastructure
- Large ensemble sizes to quantify uncertainty
- Multiple scenarios (of “deep” uncertainties)
What makes environmental data unique?
Today
Introductions
Climate Hazards
What makes environmental data unique?
Course Organziaton
Should you take this class?
Fat tails
Bonnafous et al. (2017)
Quasi-periodic oscillations
Doss-Gollin et al. (2019)
Nonstationarity
Fagnant et al. (2020)
Vary on multiple temporal scales
Doss-Gollin et al. (2019)
Spatial structure
Farnham et al. (2018)
Emphasis on extremes
Doss-Gollin & Keller (2023)
Deep Uncertainty
Walker et al. (2013)
Course Organziaton
Today
Introductions
Climate Hazards
What makes environmental data unique?
Course Organziaton
Should you take this class?
Syllabus
Lectures
- Tests will cover material from lectures and labs
- Slides will be posted ahead of time on course website (see instructions for printing to PDF)
- Occasional readings (assigned ahead of time on Canvas)
- I am not a mind reader! Ask questions.
Labs (10%)
- Build your hands-on computational skills
- Most weeks, generally Friday
- Apply conecpts from lectures to simple problems
- Graded on a 3 point scale
- Due one week after the in-class lab; solutions will be posted and discussed so no late submissions (turn in what you have)
Tests (40%)
- Material from lecture, assigned readings, and labs
- Always a review session
Projects (40%)
- Apply concepts from class and lab to a more complex and open-ended problem
- Each module (except intro) centers on a project
Three rainfall-focused projects planned:
- Downscaling
- Frequency analysis
- Weather typing
Participation (10%)
Some ways to participate include:
- Attending every class
- Asking questions in class
- Answering questions on Canvas
- Coming to office hours
We will co-grade your participation for every module
Should you take this class?
Today
Introductions
Climate Hazards
What makes environmental data unique?
Course Organziaton
Should you take this class?
Job opportunities
Growing climate analytics opportunities in:
- Insurance
- Finance
- Agriculture
- Engineering
And more! These require an understanding of climate, probability, statistics, coding, and communication.
Pre-requistes: Linear Algebra
You need basic matrix notation and multiplication, but not much more. Let \[ A = \left[ \begin{matrix} a & b \\ c & d \end{matrix} \right], \quad B = \left[ \begin{matrix} e & f \\ g & h \end{matrix} \right], \quad x = \left[ \begin{matrix} k \\ \ell \end{matrix} \right], \quad \]
You should be able to (with note-checking as needed!) figure out:
- \(A_{2,1}\)
- \(A + B\)
- \(AB\)
- \(A x\)
- \(x^T x\)
- \(x x^T\)
Pre-requisites: Probability and Statistics
You should have a course in applied statistics. You should be able to:
- Compute summary statistics of a sample
- Define joint, marginal, and conditional distributions
- Understand probability density functions, quantiles, and cumulative distribution functions
- Explain a few probability distributions and where they are appropriate
- Perform and interpret linear regressions
Pre-requisites: Coding
We will use the Julia programming language. I think you’ll find it easy and fun to learn!
- No experience in Julia is expected
- Some prior experience coding (R, Python, Matlab, C, etc.) is suggested
- If you have no prior coding experience, you will need to put in extra effort to learn the basics
Questions?
- Wednesday: “What drives uncertain climate hazard?”
- Friday: “Lab 01: Setting up Julia, GitHub, and Quarto”