Remote Sensing Modelling

Lessons and portfolio activities

1. 3.1
   Prediction, Inference and ExplanationPrediction Framing Studio · 120–160 minutes
2. 3.2
   Define the Target and Prediction UnitTarget Contract Laboratory · 140–190 minutes
3. 3.3
   Design Predictors and Modelling HypothesesPredictor Hypothesis Workshop · 150–210 minutes
4. 3.4
   Build the Modelling Dataset and Pre-register the ExperimentExperiment Design and Data Integrity Laboratory · 180–240 minutes
5. 3.5
   What Does a Useful Model Need to Beat?Baseline and Skill Laboratory · 150–210 minutes
6. 3.6
   Trees, Ensembles and BoostingTree and Ensemble Mechanism Laboratory · 160–220 minutes
7. 3.7
   XGBoost from First PrinciplesXGBoost Mechanism Studio · 190–260 minutes
8. 3.8
   Train the First Defensible XGBoost ModelFirst Model Reproducibility Laboratory · 210–280 minutes
9. 3.9
   Validation Is Part of the ModelValidation Claim and Evidence Laboratory · 170–230 minutes
10. 3.10
    Spatial, Grouped and Leave-Location-Out ValidationSpatial Validation Design Laboratory · 210–290 minutes
11. 3.11
    Temporal and Spatiotemporal ValidationTemporal Transfer and Drift Laboratory · 190–260 minutes
12. 3.12
    Nested Model Selection and Leakage PreventionNested Evidence and Leakage Audit Laboratory · 220–300 minutes
13. 3.13
    Hyperparameter OptimisationControlled Search Design Laboratory · 210–290 minutes
14. 3.14
    Early Stopping, Regularisation and Learning DynamicsLearning Dynamics and Early-stopping Laboratory · 220–300 minutes
15. 3.15
    Feature Selection, Redundancy and StabilityFeature Relevance and Stability Laboratory · 210–290 minutes
16. 3.16
    Imbalanced Classification and Decision ThresholdsRare-habitat Decision Threshold Laboratory · 220–300 minutes
17. 3.17
    Regression EvaluationRegression Evidence and Diagnostic Laboratory · 220–300 minutes
18. 3.18
    Classification Evaluation and Probability QualityClassification and Probability-quality Laboratory · 230–310 minutes
19. 3.19
    Residual Geography and Structured FailureStructured Failure and Residual Geography Laboratory · 230–310 minutes
20. 3.20
    Model Interpretation Without Causal OverclaimingPredictive Interpretation and Claim-boundary Laboratory · 240–320 minutes
21. 3.21
    Domain of Applicability and ExtrapolationDomain of Applicability Signature Laboratory · 250–340 minutes
22. 3.22
    What Uncertainty Means in Predictive EOPredictive Uncertainty Reasoning Laboratory · 210–290 minutes
23. 3.23
    Prediction Intervals and Quantile ApproachesQuantile Prediction Interval Laboratory · 240–330 minutes
24. 3.24
    Conformal Prediction and Empirical CoverageStructured Split-conformal Coverage Laboratory · 250–340 minutes
25. 3.25
    Uncertainty and Applicability MapsPrediction Evidence Mapping Signature Laboratory · 250–350 minutes
26. 3.26
    Raster Inference at ScaleOperational Raster Inference Laboratory · 250–350 minutes
27. 3.27
    Google Earth Engine for Modelling WorkflowsCloud EO Modelling Component Laboratory · 250–340 minutes
28. 3.28
    Local ML versus Earth Engine MLModelling Architecture Decision Studio · 230–320 minutes
29. 3.29
    Monitoring Through Repeated PredictionsRepeated Prediction and Drift Laboratory · 250–350 minutes
30. 3.30
    Reproducibility, Model Cards and Operational QAOperational Model Package Signature Laboratory · 270–370 minutes
31. Capstone
    Environmental Monitoring ProjectIndependent Module Capstone · 40–60 hours