Data Handoff Call - Raw Dataset Received + MVP Scope Confirmed

Follow-up call after the Sep 21 kickoff. The real dataset has now arrived (8 towers, 30-min + daily files, May 2023 through Sep 2026) and Shell confirmed exactly what to build first and how they want forecasting validated later.

• Confirmed two-phase order: (1) visualization dashboard first - "like a weather app but with history" - because it's fast to demo, then (2) predictive forecasting on the same data as phase 2, not a separate effort.
• Data is here: 8 CSV pairs (30-min + daily) covering 2023-05-10 through 2026-09-15, one pair per flux tower. Files currently live in the team's Senior Design/output folder - see the task about moving this to shared storage (~109MB total across 16 files, too big to keep emailing around).
• Tower-to-plot mapping (from the farm map in Resources): Tower 013 = Sorghum Control, Tower 003 = Sorghum-Treatment-1, Tower 006 = Sorghum-Treatment-2, Tower 002 = Pasture-Treatment-1, Tower 014 = Pasture Control.
• NEEDS CONFIRMATION WITH SHELL: Towers 001, 004 and 005 appear on the farm map but without a named plot label - we need Shell to confirm what treatment those three represent before we label any charts with them.
• MVP variable priority, in the sponsor's own words: don't try to visualize everything at once - lead with weather (temperature/rainfall/humidity), soil moisture, and the carbon/gas exchange numbers (NEE/GPP/RECO). Those are the "primary" ones per the sponsor.
• How they want forecasting validated (phase 2, later): train on part of the historical record, hold back the rest as a test set, predict forward (they gave a 3-month forecast example), then compare the prediction to what actually happened once that period passes. Split by time, not randomly.
• This data also feeds Shell's carbon-credit / carbon-removal reporting, not just our class project - useful context for "why this matters" in presentations.
• The farm also has agricultural practice history for these plots (fertilizer, planting/harvest dates) that isn't in our CSVs yet - worth asking Shell whether that's available to layer in as context.
• Timeline: sponsor would like to see first real results by end of October/November. They explicitly said our current ~1-2 week delay is fine and not to stress about it - better to start once we're fully set up than to rush.
• Sponsor wants a second call scheduled soon with colleague Hemshikha to see Shell's own internal live-visualization tool and finalize data-transfer logistics - Monday/Thursday afternoons generally work for us.
• Sponsor mentioned openness to the team presenting this work at a conference, plus possible summer internship/research opportunities tied to this project specifically.