How davidsamah/jordan-agricultural-water-stress-v0 works
This is a research prototype for forecasting next-month shallow-soil-moisture stress in the Northern Jordan Valley, not an operational alerting service. Its main workflow turns processed climate observations into a leakage-safe target and lagged feature table, trains and evaluates a Random Forest against persistence and seasonal-climatology baselines, then produces uncertainty, sensitivity, visualization, and reporting outputs. The README also describes literature review, data acquisition, and verification, but those implementations were not sampled; their relationships are therefore not asserted.
Main entry points: train_v0.py Train and compare · plot_results.py Result plots · model_dataset_lagged.csv Lagged model dataset · test_predictions_2024.csv 2024 predictions
Components 23
Environmental data
- CHIRPS precipitation
data/chirps/processed - ERA5-Land variables
data/era5land/processed
Target and features
- Climate feature table
data/processed_v0/climate_features_2018_2024_monthly.csv - Complete features
data/processed_v0/features_complete.csv - Stress target builder
scripts/build_target.py - Target dataset
data/processed_v0/model_dataset.csv - Lag and season features
scripts/build_features.py - Lagged model dataset
data/processed_v0/model_dataset_lagged.csv - Leakage checks
scripts/leakage_check.py
Forecast evaluation
- Train and compare
scripts/train_v0.py - Random Forest model
models/random_forest_v0.joblib - Evaluation metrics
results/metrics_v0.json - 2024 predictions
results/test_predictions_2024.csv - Feature importance
results/feature_importance.csv - Bootstrap uncertainty
scripts/bootstrap_test.py - Threshold sensitivity
scripts/threshold_sensitivity.py - Sensitivity results
results/threshold_sensitivity.csv
Research outputs
- Result plots
scripts/plot_results.py - Research figures
figures - Report generator
scripts/make_report.py - Experiment report
reports/v0_experiment_report.md - Research evidence
evidence - Study notes
notes
Connections 21
- Complete features → Stress target builder: provides input
scripts/build_target.py - Stress target builder → Target dataset: writes target
scripts/build_target.py - Target dataset → Lag and season features: provides input
scripts/build_features.py - Lag and season features → Lagged model dataset: writes features
scripts/build_features.py - Lagged model dataset → Train and compare: provides input
scripts/train_v0.py - Lagged model dataset → Leakage checks: checks ordering
scripts/leakage_check.py - Train and compare → Random Forest model: saves model
scripts/train_v0.py - Train and compare → Evaluation metrics: writes metrics
scripts/train_v0.py - Train and compare → 2024 predictions: writes predictions
scripts/train_v0.py - Train and compare → Feature importance: writes importance
scripts/train_v0.py - 2024 predictions → Threshold sensitivity: reads predictions
scripts/threshold_sensitivity.py - Threshold sensitivity → Sensitivity results: writes results
scripts/threshold_sensitivity.py - Complete features → Result plots: reads features
scripts/plot_results.py - 2024 predictions → Result plots: reads predictions
scripts/plot_results.py - Feature importance → Result plots: reads importance
scripts/plot_results.py - Result plots → Research figures: writes figures
scripts/plot_results.py - Evaluation metrics → Report generator: reads metrics
scripts/make_report.py - Feature importance → Report generator: reads importance
scripts/make_report.py - Report generator → Experiment report: writes report
scripts/make_report.py - CHIRPS precipitation → Climate feature table: contributes precipitation
README.md - ERA5-Land variables → Climate feature table: contributes variables
README.md