Start Page Menu Items Interactive Features Hotkeys Tutorial

DC2DInvRes - Tutorial

Standard data inversion procedure

("Load Data" and "Full inversion" works but learn by playing)
  1. Load a data file (File>>Open Data File)
  2. Estimate data errors by configuration factor and electrode variations
  3. Add other files to the data (and save all as separate data file)
  4. Display apparent resistivities, IP, voltages and eliminate apparently bad data by clicking
  5. Display model and edit model parameters (type in resistivity and click on cells)
  6. If a-priori information (e.g. layering) is known, enter resistivity values or draw model
    (save starting model by Model>>Export Model)
  7. If not and layering is to be seen in the data, choose Inversion>>Layered Model
  8. Create Sensitivities (Inversion>>Sensitivities) and explore them (Display>>Sensitivities)
    (if necessary, change model parameterization and check sensitivities again)
  9. Inversion options: Gauss-Newton method, Smoothness Constraints, Combine Cells, global&line search on
    (1st order Smoothness for broad structures, 2nd for bounded bodies)
  10. Choose fixed lambda value (10-30) or manual choosing
  11. Click Inversion>>One Inversion step
  12. experiment with different lambdas by hand or with slider (go back every time)
    (if resistivities get too low, try "lower resistivity bound" in Options>>Inversion
  13. Full inversion, watch forward accuracy and chi^2 (below 10 is acceptable)
    (if forward accuracy becomes too large, increase z-refining
    (if chi^2 stays large, increase error estimates or decrease lambda and start over)
  14. Watch data misfit during inversion, should ideally represent randoms at the end
    (practical using the "Show>>Compare Data" option once)
  15. check other inversion options, if necessary
  16. Watch resolution measures, esp. resolution radius and Model Cell resolution
  17. Save model (Model>>Export) or workspace(File>>Save Workspace)
  18. Visualize model, data, ... and export into graphics files

Synthetic data modelling / experimental design

  1. Make new data using Data>>Create Data Set and watch error estimates
  2. Edit model by Model>>Parameter and draw model
  3. Calculate forward response using Data>>Forward, add noise with Data>>Add noise
  4. Data>>Set response as data
  5. Inversion>>Homogeneous model or starting model
  6. Inversion and resolution (see 9.-16. above)