16. Forward Modelling

pycsamt.forward contains the in-process forward modelling layer of pyCSAMT v2. It provides resistivity model containers, 1-D/2-D/quasi-3-D electromagnetic solvers, synthetic dataset generators, noise models, and diagnostic plotting utilities.

Forward modelling answers a controlled physical question:

If this earth model were true, what response would the survey record?

That question sits underneath every serious inversion workflow. Inversion searches for a model that explains observed data; forward modelling computes the data predicted by a known model. A good inversion setup therefore depends on forward modelling for sensitivity tests, synthetic recovery experiments, training data, uncertainty design, and quality control.

• 16.1. Overview
  • 16.1.1. Forward And Model Integrations
  • 16.1.2. Package Map
  • 16.1.3. Core Workflow
  • 16.1.4. Choosing A Path
  • 16.1.5. Relationship To Theory
• 16.2. Forward Modelling Concepts
  • 16.2.1. The Forward Problem
  • 16.2.2. Physical Inputs
  • 16.2.3. Dimensionality
  • 16.2.4. 1-D Forward Models
  • 16.2.5. MT, CSAMT, And TEM Assumptions
  • 16.2.6. 2-D Forward Modelling
  • 16.2.7. Quasi-3-D Forward Modelling
  • 16.2.8. Frequency, Period, And Depth Sensitivity
  • 16.2.9. Response Quantities
  • 16.2.10. Synthetic Priors
  • 16.2.11. Noise And Realism
  • 16.2.12. Synthetic Recovery
  • 16.2.13. Common Failure Modes
  • 16.2.14. Good Practice Checklist
  • 16.2.15. Next Steps
• 16.3. Forward Configuration
  • 16.3.1. Configuration Classes
  • 16.3.2. Template Files
  • 16.3.3. 1-D Configuration
  • 16.3.4. Geological Priors
  • 16.3.5. TEM Configuration
  • 16.3.6. 2-D Configuration
  • 16.3.7. 2-D Grid Tuning
  • 16.3.8. 3-D Configuration
  • 16.3.9. 3-D Configuration Notes
  • 16.3.10. Validation
  • 16.3.11. Summaries And Provenance
  • 16.3.12. What To Record
  • 16.3.13. Recommended Run Layout
  • 16.3.14. Common Mistakes
  • 16.3.15. Next Steps
• 16.4. Solvers And Grids
  • 16.4.1. Solver Map
  • 16.4.2. 1-D Layered Models
  • 16.4.3. 1-D MT And CSAMT Solvers
  • 16.4.4. 1-D TDEM Solver
  • 16.4.5. 2-D Profile Grid Concepts
  • 16.4.6. 2-D Profile Grid Constructors
  • 16.4.7. 2-D MT Solver
  • 16.4.8. 3-D Volume Grid Concepts
  • 16.4.9. 3-D Volume Grid Constructors
  • 16.4.10. Quasi-3-D Solver
  • 16.4.11. Grid Design Rules
  • 16.4.12. Response Containers And Feature Arrays
  • 16.4.13. Debugging Solver Experiments
  • 16.4.14. Next Pages
• 16.5. Maxwell Adapter Layer
  • 16.5.1. Legacy And Validated Engines
  • 16.5.2. Package Map
  • 16.5.3. Core Workflow
  • 16.5.4. Choosing A Path
  • 16.5.5. Relationship To Theory
• 16.6. Maxwell Problem, Mesh, And Result Contracts
  • 16.6.1. Two Canonical Axis Orders
  • 16.6.2. Receivers And Immutability
  • 16.6.3. Problem Identity And Provenance
  • 16.6.4. Diagnostics And Result Validity
  • 16.6.5. Persistence Without Pickle
  • 16.6.6. The Triangular Mesh Contract
  • 16.6.7. Common Mistakes
  • 16.6.8. Next Pages
• 16.7. Maxwell Solver Meshes
  • 16.7.1. Choosing A Mesh Builder
  • 16.7.2. Mesh Design Parameters
  • 16.7.3. A Padded 2-D Solver Mesh
  • 16.7.4. Skin Depth And Mesh Quality
  • 16.7.5. Near-Surface Resolution Convergence
  • 16.7.6. Receiver Placement And Terrain
  • 16.7.7. A 3-D Solver Mesh
  • 16.7.8. Persisting A Mesh Model
  • 16.7.9. Graded Triangular Meshes
  • 16.7.10. Common Mistakes
  • 16.7.11. Next Pages
• 16.8. Maxwell Backend Registry
  • 16.8.1. Capability Assessment
  • 16.8.2. The Backend Protocol
  • 16.8.3. Registering Backends
  • 16.8.4. Private Registries
  • 16.8.5. Common Mistakes
  • 16.8.6. Next Pages
• 16.9. Maxwell Adapters
  • 16.9.1. Choosing An Adapter
  • 16.9.2. In-Repository 2-D Solve
  • 16.9.3. Research-Only 3-D Solve
  • 16.9.4. Topography-Aware Triangular FEM
  • 16.9.5. External Solver Adapters
  • 16.9.6. Common Mistakes
  • 16.9.7. Next Pages
• 16.10. Maxwell Analytic Benchmarks
  • 16.10.1. Acceptance Thresholds
  • 16.10.2. Building And Running A Case
  • 16.10.3. Reusing A Solved Result
  • 16.10.4. Case Identity And Provenance
  • 16.10.5. Validation Guardrails
  • 16.10.6. Common Mistakes
  • 16.10.7. Next Pages
• 16.11. Maxwell Caching And Batch Solving
  • 16.11.1. Content-Addressed Caching
  • 16.11.2. Corruption And Quarantine
  • 16.11.3. Pruning A Cache
  • 16.11.4. Solving Many Problems At Once
  • 16.11.5. Retries And Terminal Failures
  • 16.11.6. Stopping Early
  • 16.11.7. Concurrency And Backend Identity
  • 16.11.8. Common Mistakes
  • 16.11.9. Next Pages
• 16.12. Synthetic Datasets And Noise
  • 16.12.1. 1-D Dataset Generation
  • 16.12.2. ForwardDataset Contract
  • 16.12.3. Saving, Loading, And Splitting
  • 16.12.4. Configuration-Driven Generation
  • 16.12.5. TDEM Datasets
  • 16.12.6. Geological Priors
  • 16.12.7. Noise Models
  • 16.12.8. Clean And Noisy Dataset Pairs
  • 16.12.9. Pseudo-3-D Survey Datasets
  • 16.12.10. 2-D And Quasi-3-D Solver Datasets
  • 16.12.11. Dataset QA
  • 16.12.12. Recommended Archive Layout
  • 16.12.13. Common Mistakes
  • 16.12.14. Next Pages
• 16.13. Forward Plotting
  • 16.13.1. Plot Selection Guide
  • 16.13.2. Saving Figures
  • 16.13.3. 1-D Response Plots
  • 16.13.4. 1-D Model Plots
  • 16.13.5. 1-D Composite View
  • 16.13.6. 2-D Grid Model Plots
  • 16.13.7. 2-D Pseudo-Sections
  • 16.13.8. 2-D Lateral Response Profiles
  • 16.13.9. 3-D Model Slice Plots
  • 16.13.10. 3-D Response Maps
  • 16.13.11. 3-D Response Sections
  • 16.13.12. 3-D Tensor Component Panels
  • 16.13.13. Plotting Noisy Responses
  • 16.13.14. Plotting Dataset Samples
  • 16.13.15. Plotting Checklist
  • 16.13.16. Common Mistakes
  • 16.13.17. Next Pages
• 16.14. From Forward Modelling To Inversion
  • 16.14.1. The Handoff Contract
  • 16.14.2. Why Synthetic Recovery Comes First
  • 16.14.3. 1-D MT Recovery
  • 16.14.4. 1-D TDEM Recovery
  • 16.14.5. Stitched 2-D Profile Recovery
  • 16.14.6. True 2-D Forward Response Handoff
  • 16.14.7. Backend Choice After Forward Tests
  • 16.14.8. Error Model Handoff
  • 16.14.9. What To Compare
  • 16.14.10. Common Failure Modes
  • 16.14.11. Recommended Workflow