SM Notes- Fall 2024

[sofia-mujica]

Weekly Notebook Entry — Week 4

Overview

• Week Span: 9/9/2024 to 9/13/2024

Tasks for This Week

• [x] Task 1: Attend GNI Symposium and present work from Microclimate Spring 2024!

Tasks

Task 1: GNI Symposium- Complete

• Description: Presentation of research and intake of related research.
• Outcome: Strong presentation and panel session! We learned about Lag Llama; a potential course of action or comparison for future work.

Personal Reflection

• It was fulfilling to see a glimpse of the work being done worldwide for the simulation of improved urban environments. I enjoyed the opportunity to expand my perception and awareness of the research being done in the field. I made personal connections with individuals that might benefit our work as we progress. Being able to present research in a professional environment as an undergraduate was incredibly special to me, and I look forward to the possibility of being able to do so again!

Additional Notes & Resources

• GNI Abstracts - Abstract lists from paper sessions at GNI Symposium & Expo 2024. Use password: gnisymposium2024

---

Plans for Next Week

• [x] Revisit and share methods observed from GNI for potential review.

[sofia-mujica]

Weekly Notebook Entry — Week 5

Overview

• Week Span: 9/16/2024 to 9/20/2024

Tasks for This Week

• [x] Task 1: Report GNI takeaways.

Tasks

Task 1: Share GNI learnings- Complete

• Description: Continue to review papers covering the UWG model and note its shortcomings.
• Outcome & Reflection: UWG is validated for use in major cities over neighborhoods. This design supports the setting of our sensitivity analysis from the previous semester. This semester's focus will shift to validating our simulated data with the recorded weather data provided by Dr. Stone.

General Takeaway, Skills & Knowledge Acquired

• Decided against learning more about Lag Llama.

---

Plans for Next Week

• [x] Look into improvements for UWG implementation.

[sofia-mujica]

Weekly Notebook Entry — Week 6

Overview

• Week Span: 9/23/2024 to 9/27/2024

Tasks for This Week

• [x] Task 1: Review new UWG papers/ progress.

Tasks

Task 1: UWG Review- Complete

• Description:
• Link to Work: "Improvements, extensions, and validation of the Urban Weather Generator (UWG) for performance-oriented neighborhood planning"
• Notes: Improvements made to calculate mean radiant temperature to consider outdoor human thermal comfort and blue infrastructure effects. Controlling of site-layout, building form, and green-blue infrastructure. Parameters site coverage ratio, average building height, and facade to site ratio are used to describe 2D infinitely long urban canyon-- additional parameters are added to describe homogenous rectangular 3D building array: individual buildings over interlaced structures. Scattered solar radiation collected from sky view factors. MRT used for short and longwave radiation fluxes from human body exposed to environment: direct MRT, diffuse MRT, longwave MRT. GeoEye-1 remote sensing image maps? Significant improvement to UWG radiation process. Current UWG fails to consider reduced solar radiation on road with increased vegetation-- new version better evaluates efficacy of thermal mitigation measures resultant of vegetation coverage.

General Takeaway, Skills & Knowledge Acquired

• The additions to the UWG might relate to work done with the mobility team- add-ons to further evaluate human comfort resultant of thermal conditions.
• Impacts of planning elements affect neighborhoods differently based on unique thermal environments- seasonal and geographical factors should also be considered.

Team Meetings

• Meeting Highlights: Will stick with UWG and perform second validation this semester while also looking into machine learning alternatives. Data extraction for UWG will prove useful for producing surrogate models, like with random forest! Patrick will help breakdown the semester goal into realistic stepping stones for the team.

Personal Reflection

• Glad to be furthering work with UWG and to have a larger team to push our work forward!

---

Plans for Next Week

• [x] Task 1: Review other material/ repositories of interest.

[sofia-mujica]

Weekly Notebook Entry — Week 7

Overview

• Week Span: 9/30/2024 to 10/04/2024

Tasks for This Week

• [x] Task 1: Explore different directions.

Tasks

Task 1: Exploration- Complete

• Description:
• Link to Work: Comparison of Prithvi WxC and Deep Learning models.
• Notes on Prithvi WxC Model: Foundational model for weather and climate. Zer-shot applications in masked reconstruction and forecasting. Uses AI to model atmospheric dynamics. Trained primarily on NASA's MERRA-2 dataset. Leverages historical data and AI-models to improve cost-efficiency, speed, and accuracy in forecasting. Downscales large-scale climate models for increased resolution. Splits map into individual windows and performs modulo masking, alternation between local and global attention.
• Notes on Deep Learning Model: Uses Geo-layer, LSTM layer, and Kriging layer. Comparison with traditional microclimate data accessing methods and classical machine learning models to reach baseline. Data provided for temperature, humidity, solar radiation, etc from weather stations. Downscales historical data from original weather stations to finer grid. Predictions in high-spatial and high-temporal directions aid in establishment of urban/ district digital twins.

General Takeaway, Skills & Knowledge Acquired

• Weekly presentation narrowed efforts to focus on the Deep Learning model repository.
• Achieves spatial resolution of 1-meter grid!
• Importance of incorporating spatial, temporal, and geographical knowledge (LULC) in the model shown-- look into UWG for validation of thorough LULC data usage.

Personal Reflection

• Should download code to try running on my own- look into how this can be implemented further for our use.

---

Plans for Next Week

• [x] More exploration of deep learning model repository.

[sofia-mujica]

Weekly Notebook Entry — Week 8

Overview

• Week Span: 10/07/2024 to 10/11/2024

Tasks for This Week

• [x] Task 1: Further research on deep learning model.

Tasks

Task 1: Break down model- Complete

• Description: The repository for our model of interest for this semester is super clunky with little commentary or explanations of the authors' work. Before jumping into this model, we want to go through and add comments to each section to make sense of the model.

General Takeaway, Skills & Knowledge Acquired

• Fixed bugs with VSCode from last semester to be able to use the same editor and GitHub desktop app for seamless committing.

Team Meetings

• Meeting Highlights: Shift to working with deep learning model, reproduce the work done with this model for GT campus and use this information to make comments about the UWG, begin work by forking their repository into ours and attempting to understand their work.

Personal Reflection

• A bit of a slower week with exams.. spent more time looking into the model before breaking it down.

---

Plans for Next Week

• [x] Start cleaning up code.

[sofia-mujica]

Weekly Notebook Entry — Week 9

Overview

• Week Span: 10/14/2024 to 10/18/2024

Tasks for This Week

• [x] Clean code and add comments.

Tasks

Task 1: Clean up code- Complete

• Description: deep learning model is super confusing to read with minimal comments and nondescript variables. for us to be able to understand this model, we want to start by breaking down segments of code to make it legible for our team before beginning to use it.
• Outcome & Reflection:

General Takeaway, Skills & Knowledge Acquired

Team Meetings

• Meeting Highlights: xx no team meetings this week.

Personal Reflection

---

Plans for Next Week

[sofia-mujica]

Weekly Notebook Entry — Week 10

Overview

• Week Span: 10/21/2024 to 10/25/2024

Tasks for This Week

• [ ] Task 1:

Tasks

Task 1: - Complete

• Description:
• Outcome & Reflection:

General Takeaway, Skills & Knowledge Acquired

Team Meetings

• Meeting Highlights:

Personal Reflection

---

Plans for Next Week