| # Photosynthesis Prediction Model – Purpose & Background |
|
|
| This project builds a **two-stage photosynthesis prediction pipeline for grapevines**: |
|
|
| - **Stage 1 (Mechanistic)**: Use on-site crop sensor data from the Seymour plot to compute leaf photosynthesis rate \(A\) using the **Farquhar et al. (1980)** model with **Greer & Weedon (2012)** parameterization for *Vitis vinifera*. This includes temperature-dependent Vcmax/Jmax, electron transport, and a Crop Water Stress Index (CWSI) based on leaf–air temperature differences and humidity. |
| - **Stage 2 (ML)**: Train simple machine‑learning models to predict \(A\) **using only external IMS weather station data** (station 43, Sde Boker). This lets us estimate grapevine photosynthesis from standard meteorological inputs without needing the full sensor stack. |
|
|
| ### Domain context |
|
|
| - Site: solar wine‑farm experimental vineyard near Yeruham, Israel (Seymour plot). |
| - Crop: grapevine (*Vitis vinifera*). |
| - On‑site sensors: PAR, leaf temperature, air temperature, CO₂, VPD, humidity, and crop spectral indices (PRI, NDVI variants). |
| - External weather: IMS station 43 (Sde Boker) providing temperature, solar radiation, humidity, rain, wind, and pressure. |
|
|
| ### Design principles |
|
|
| - **Mechanistic first, then ML**: Ground the target \(A\) in a physiological model before fitting data‑driven models. |
| - **Strict separation of data sources**: Stage 1 uses only on‑site sensors; Stage 2 uses only IMS weather features to avoid leakage. |
| - **Reproducible, well‑documented pipeline**: Clear directory layout, config in `config/settings.py`, and this `context/` folder as the long‑term project memory. |
|
|
|
|
