UNESCO–Al Fozan International Prize
Precision Irrigation ML
Auditable Computational-Feasibility Study for Sustainable Farming
Auditable computational-feasibility study (AI-Powered Precision Irrigation for Sustainable Vegetable Farming) prepared for the UNESCO–Al Fozan International Prize, benchmarking Logistic Regression, Random Forest, and XGBoost against persistence, threshold, and FAO-56 baselines across 1,800 counterfactual water-balance scenarios with a 3D twin dashboard.
System ID
04
precision-irrigation-ml
Counterfactuals
M-011,800
Water-Balance Runs
Eval Split
M-0260/20/20
Chronological + 24h Embargo
Benchmarks
M-033 ML vs 3 Base
LR/RF/XGB vs FAO-56
Telemetry Source
M-04NASA POWER
3D Twin + Pytest
Section 01
Problem Statement
Agricultural ML proposals often lack temporal leakage guards, physical water-balance baselines, and reproducible audits for scientific review.
Section 02
System Constraints
- >Prevent temporal leakage using chronological 60/20/20 splits and a 24h embargo
- >Benchmark ML classifiers against persistence, threshold, and FAO-56 baselines
- >Deliver reproducible Pytest verification alongside an interactive 3D twin UI
Section 03
System Architecture & Dataflow
NASA POWER Climate Data
Chronological 60/20/20 + 24h Embargo
LR / RF / XGBoost vs FAO-56
1,800 Scenarios & 3D Twin UI
Section 04
Key Engineering Decisions
- 01Enforced chronological 60/20/20 splits with a 24-hour embargo and Pytest checks to eliminate look-ahead leakage.
- 02Evaluated Logistic Regression, Random Forest, and XGBoost against persistence, threshold, and FAO-56 baselines across 1,800 counterfactual water-balance scenarios.
- 03Built an interactive Next.js 3D twin dashboard backed by NASA POWER API inputs and a full Audit Report and Data Pack.