Processing Pages#

The processing pages are where the web app does scientific work: quality control, corrections, advanced diagnostics, TDEM, the processing pipeline, forward and inversion modelling, interpretation, the results viewer, and the agents. Every page is an interactive surface over the same pycsamt.app.desktop.controllers used by the desktop app — the callbacks delegate to the package, so results are identical to the Python API and interchangeable with the desktop GUI.

Most pages share a common shape: a left-hand control column with a primary action button (Generate, Run, Plot, Apply…), tabs or a group selector across the plotting area, and per-page Lines and Stations pickers so an operation can be scoped to a subset of the survey.

Shared Controls#

Before the individual pages, three patterns recur:

Lines / Stations pickers — most pages let you restrict the operation to chosen lines and stations. These are independent of, but seeded by, the global Lines drawer.
The primary action — pages do not recompute continuously. A single button (Plot, Generate, Run …, Apply) triggers the work, which keeps large surveys responsive.
Result tabs — modelling and results pages separate the plot, the numerical summary, and the run log into tabs so you can inspect each without losing the others.

Quality Control#

The Quality Control page groups every QC diagnostic behind one control column. Use it to decide whether data are healthy before correcting or modelling.

Filters — restrict to Lines, Stations, and a Component (XY, YX, or Both).
Groups — Overview, Coverage, Noise / SNR, Skew / Dim, Static Shift, and Distortion. Each group exposes a set of plots.
Plot list — for example QC Quicklook (multi-panel), Coverage quality heatmap, Frequency confidence section, Confidence band summary, and Confidence profile.
Parameters — plot-specific inputs such as the metric (for example coherence), the outlier method (for example IQR), and a threshold.
Refresh / Auto — redraw on demand, or auto-redraw as settings change.

Read QC before deciding on corrections: sparse frequency coverage, erratic error bars, or isolated low-SNR stations are the kind of findings that justify frequency editing or static-shift work on the next page.

Correction#

The Correction page applies pyCSAMT’s catalogue of corrections as a non-destructive chain: 25 methods across 6 categories, previewed before they are committed, with per-step undo.

Category and Correction method — choose from the catalogue (for example Static Shift → AMA (spatial average)). A short description explains what the method does.
Parameters — method-specific inputs (for example sort key, half-window, and kernel for a spatial-average static-shift correction).
View modes — Before / After, ρ/φ View, Overlay, and Difference compare the raw and corrected responses.
Show and Comp — display Raw, Corrected, or Both, for the XY, YX, or both components, across up to Max stn stations.
Actions — Preview shows the effect without committing; Apply adds the step to the chain; Undo removes the last step; Reset All returns to raw data.

The chain is the important idea: corrections are staged and reversible, so you can preview a static-shift method, apply it, preview a noise method on top, and still undo back to raw at any point. Corrected EDIs are exported from this page; see Exports And Reproducibility.

Advanced Plots#

The Advanced Plots page produces the survey-scale diagnostics used for strike, dimensionality, and depth reasoning.

Tabs — Strike, Phase Tensor, Induction, Impedance/Z, Depth, and Survey Tools.
Active Lines and Stations — scope the diagnostic to chosen lines and stations.
Parameters — plot-specific inputs and Figure size presets.
Generate — build the selected diagnostic; a short Description explains what the current plot shows.

These are whole-survey versions of the per-line diagnostics in Profile View — use them for strike consistency and dimensionality across all lines at once.

TDEM#

The TDEM page handles time-domain electromagnetic data with its own folder browser and a fixed tab bar of plot categories.

Tabs — Decay / Rho, Survey Section, Map & Overview, and Dashboard.
Each tab exposes a set of plots, a figure-size preset, and a colour map.
Point the folder browser at a TDEM dataset, choose a plot, and generate it.

TDEM is grouped under Analysis in the rail alongside Advanced Plots.

Processing Pipeline#

The Pipeline page runs the standard survey workflow as an ordered, inspectable sequence. It is the eight-step load → QC → edit → correct → strike → export chain, run step by step or all at once.

The steps are:

#	Step	What it does
1	Load Data	Use the loaded survey, or browse to a fresh EDI folder.
2	QC Screening	Drop low-confidence frequencies by threshold.
3	Frequency Edit	Edit frequencies by confidence (recover or trim).
4	Static Shift Correction	Apply a static-shift correction (for example AMA).
5	Noise Removal	Denoise the impedance response.
6	Strike Analysis	Estimate geoelectric strike.
7	Strike Rotation	Rotate to the strike frame.
8	Export	Write processed EDIs and products to an export folder.

Controls

Run Step runs the current step; Run All runs the remaining steps in order; Skip advances without running; Reset returns to the start.
For each step you choose a Method and see a Description; the Steps list and the numbered progress track show where you are.
The output pane has Log, Preview, and Status tabs. The log reports each step (— Step 2 done —) and surfaces failures (for example ERROR: SVD did not converge if a static-shift solve fails) so you can adjust parameters and re-run.
The Export step writes to a folder you choose with Select Export Folder.

The pipeline is the repeatable counterpart to the interactive pages: the same QC, correction, and strike operations, run as a recorded sequence you can re-run on new data.

Forward Modelling#

The Forward Model page computes model responses in 1-D, 2-D, and 3-D.

The Forward Model page showing 3-D MT model slices for a halfspace — Forward Model on the 3-D MT tab: model slices for a halfspace, with grid, background, and frequency-range controls on the left.#

Tabs — 1-D, 2-D MT, and 3-D MT (quasi-3D).
Model Type — for example Halfspace or a layered/preset model.
Background & Grid — background resistivity, station layout, grid dimensions (Nx × Ny × Nz), and extents.
Frequency Range — log10(Hz) min/max and the number of points.
Run Forward computes the response; the plotting area shows the model and its response for the selected type, component, and frequency.

Use forward modelling to build intuition for what a target looks like in the data, and to generate training or test responses for the inversion page.

Inversion#

The Inversion page runs traditional, AI-neural, PINN, and hybrid inversions in 1-D, 2-D, and 3-D.

Problem dimension — for example 2-D Profile (U-Net).
Architecture — the network or solver (for example UNet2D (encoder–decoder)).
Forward solver (training data) — the physics used to generate training responses (for example MT 1-D).
Network Config and Training — collapsible panels for network and optimisation settings.
Frequency Range — the band used for the inversion.
Result tabs — Result, Convergence, Statistics, Log, and Data Fit separate the model, the loss/RMS curve, the numbers, the run log, and the observed-vs-predicted fit.
Run Inversion starts the run.

The header names the available families — Traditional · AI Neural · PINN · Hybrid — and the supported dimensionalities. To browse the output of an external solver run instead, use the Results View page below.

Results View#

The Results View page (Inversion Results Viewer) browses, inspects, and exports inversion outputs from external solvers — ModEM, Occam2D, and MARE2DEM.

Results Folder — point at a solver output folder and Load it.
Solver — Auto-detect, or force ModEM (3-D), Occam2D, or MARE2DEM. The panel confirms what was loaded (for example MODEM loaded · 74 iter · RMS=3.057).
Tabs — Convergence, Section, Depth Map, All Profiles, Covariance, Response, and Pseudo.
Display — resistivity range and maximum depth.
Generate Plot redraws with the current settings; Export PNG saves the figure.

The Results View Convergence tab showing RMS misfit versus iteration — The Convergence tab: RMS misfit versus iteration, with the best RMS and the RMS=1 target marked.#

The Results View Response tab showing observed versus predicted data — The Response tab: observed-versus-predicted apparent resistivity and phase per station, with per-station RMS — the direct check of how well the model fits the data.#

Interpretation#

The Interpretation page turns processed data and models into geological and hydrological interpretation products — 42 workflows across 9 categories.

Categories — Setup, Geology, Hydrology, Field, EM Diag, Uncert., Advanced, Fusion, and Export.
Plot / Analysis — the specific workflow (for example Constraint misfit — observed vs modelled field data).
Data source — Raw (loaded) or Corrected.
Display options — figure size, colour map, and workflow-specific controls.
Generate builds the product; Export (the Export category) saves interpretation outputs.

AI Agents#

The AI Agents page exposes pyCSAMT’s agents — 36 agents across 10 categories, spanning LLM, processing, and inversion — in two modes.

Agent Runner lists agents you can configure and run directly.

The AI Agents page in Agent Runner mode running the Dimensionality agent — Agent Runner. Search and pick an agent, scope it by line and station, set parameters, and **Run Agent**; the run history and last result (log, figure, summary) appear on the right.#

The agent list is grouped — LLM agents (Interpretation, Report, Code Generation, EDI Export), Workflow agents (Workflow Orchestrator, Pipeline, Batch Survey), and Processing agents (QC Quicklook, Dimensionality, Static Shift, and more).
Selecting an agent shows its description, Filter by line and Station filter pickers, and its Parameters.
Run Agent runs it; Run History logs each run, and Last Result shows the Log, Figure, and Summary.

Chat is a conversational assistant over the loaded survey.

The AI Agents Chat tab with a proposed plan and quick actions — The Chat tab. Describe what you want in plain language; the assistant proposes a plan it can run, and **Quick Actions** cover common tasks.#

Type a request such as “Run quality control on all stations and identify bad SNR” and the assistant proposes a runnable plan. Quick Actions — Run full QC, Correct static shift, Dimensionality, Prep inversion, Interpret results — trigger common workflows directly.

The agents need an LLM provider and API key for the LLM-backed features; configure these in the Settings drawer (see Navigation And Layout). For a dedicated, full-screen conversational surface, use the standalone Agent Master application.

Tools#

The Tools menu on the command bar collects standalone utilities that open in their own drawer, independent of the current page.

Group	Tools
Analysis	Strike Analyzer, EDI Validator
Conversion & Export	Format Converter, Batch Export Plots
Geospatial	Coordinate Transformer, Elevation Enrichment
Visualisation	Station Response Inspector, Strike Profile Viewer, Phase Tensor Map
Classification & Editing	Dimensionality Classifier, Frequency Editor
Modelling	(modelling utilities)

Selecting a tool opens it in a drawer with an icon rail of the available tools and the selected tool’s controls and output.

The Strike Analyzer tool drawer showing a consensus strike rose diagram — The **Strike Analyzer** tool: a consensus-strike rose diagram and a per-line strike distribution, computed from the currently loaded survey.#

The Batch Export Plots tool is the fastest way to export many figures at once; see Exports And Reproducibility.

How Pages Delegate To The Package#

None of these pages implement science of their own. A control change updates a Dash store, and the primary-action callback calls the matching pycsamt.app.desktop.controllers function — the same one the desktop GUI calls and the same computation the Python API performs. This is why a survey processed in the web app can be picked up unchanged in the desktop app or in a script, and why the results are reproducible outside the browser.

Next Steps#

Exports And Reproducibility – save figures, corrected EDIs, and inversion products.
Maps And Profiles – the spatial views that precede processing.
Troubleshooting – when a page is blank, a run fails, or an agent errors.