Index

Tutorial: AgReFed-ML

AgReFed-ML is a machine learning framework specifically designed for agricultural soil modeling. It transforms sparse soil measurements into detailed spatial-temporal prediction maps of soil properties and their uncertainties using advanced Gaussian Process regression. The system combines traditional machine learning models (like Random Forest and Bayesian Linear Regression) as mean functions with sophisticated spatial correlation modeling to make accurate predictions even with limited data. It’s particularly valuable for precision agriculture applications like carbon accounting, soil moisture mapping, and property change detection over time.

Source Repository: https://github.com/Sydney-Informatics-Hub/AgReFed-ML

flowchart TD A0["Gaussian Process Models "] A1["Mean Function Models "] A2["Prediction Workflows "] A3["Data Preprocessing Pipeline "] A4["Model Evaluation and Cross-Validation "] A5["Spatial-Temporal Modeling Framework "] A6["Uncertainty Quantification System "] A7["Notebook-Based Workflows "] A8["Synthetic Data Generation "] A2 -- "Uses for prediction" --> A0 A2 -- "Uses as baseline" --> A1 A0 -- "Incorporates" --> A6 A2 -- "Uses for preparation" --> A3 A4 -- "Evaluates" --> A0 A4 -- "Tests performance" --> A1 A5 -- "Extends with 4D kernels" --> A0 A7 -- "Demonstrates usage" --> A2 A8 -- "Provides test data" --> A4

Chapters

  1. Notebook-Based Workflows
  2. Prediction Workflows
  3. Data Preprocessing Pipeline
  4. Mean Function Models
  5. Gaussian Process Models
  6. Uncertainty Quantification System
  7. Spatial-Temporal Modeling Framework
  8. Synthetic Data Generation
  9. Model Evaluation and Cross-Validation