11.24. Multi-Station Diagnostic Panels

pycsamt.emtools.plot is the multi-station plotting layer for impedance tensor and tipper response diagnostics. It is the place to go when one station at a time is too narrow, but a map or pseudo-section is too compressed. Its sibling module, pycsamt.emtools.overview, covers the opposite case – one station, every diagnostic at once – and is documented on this page too, since both re-export through the same top-level pycsamt.emtools namespace and cover the same site-level visualisation job.

The page covers six common plotting jobs:

• compact apparent resistivity and phase panels for many stations;

• raw full-tensor station groups;

• response plus tipper quality control panels;

• before/after comparison figures;

• measured-versus-predicted fit grids with per-component RMS misfit labels;

• a single-station “full response” overview combining impedance, induction arrows, and phase tensor ellipses in one figure.

Full callable signatures live in the API reference. This page explains when to use each figure, which arguments matter, and how to build reproducible plotting scripts.

11.24.1. Where This Module Fits

Use pycsamt.emtools.plot after loading and inspecting data. It does not replace the first-look inventory tools in inspect or the specialized tensor/impedance diagnostics. Instead, it gives dense station-group figures for QC, comparison, and reporting.

Function	Best Use	Figure Shape
plot_sites_panels	Quick rho/phase panels for selected stations.	One station per small two-row group.
plot_raw_sites_1d	Full-tensor raw-data review.	One station group, one column per component.
plot_response_tipper	Joint impedance and tipper QC.	Rho/phase rows plus Tx/Ty rows.
plot_sites_compare	Raw vs processed or before vs after.	Paired columns per station.
plot_sites_fit_grid	Observed vs predicted inversion-response review.	Component columns with RMS labels.
plot_response_overview	One station, every diagnostic at once.	Rho/phase columns plus induction-arrow and phase-tensor-ellipse rows.

All functions accept a path, Sites object, collection, or compatible iterable. Internally they normalize inputs with ensure_sites unless a function has a special duplicate-preservation mode.

11.24.2. Load Once, Plot Many

Normalize a survey once and pass the resulting object into every plot that follows; this keeps scripts fast and puts loading failures in one obvious place rather than scattered across each figure call. sites_summary is a cheap way to confirm what actually loaded before spending time on figures – here it doubles as evidence for a claim made throughout this page: L18PLT (an AMT line) carries no tipper, while KAP03 (an MT line, loaded alongside it) does.

>>> from pathlib import Path
>>> from pycsamt.emtools import ensure_sites, sites_summary
>>> survey = ensure_sites(
...     Path("data/AMT/WILLY_DATA/L18PLT"),
...     recursive=True,
...     on_dup="replace",
...     strict=True,
...     verbose=1,
... )
>>> kap = ensure_sites("data/MT/kap03lmt_edis", strict=True)
>>> stations = ["18-001A", "18-007U", "18-016A", "18-018A"]
>>> summary = sites_summary(survey)
>>> print(len(summary), "stations,", "tipper present:", bool(summary["has_tipper"].any()))
28 stations, tipper present: False
>>> summary.head(3)
   station  n_freq  has_tipper  period_min  period_max        lat         lon
0  18-001A      53       False    0.000096    0.992063  32.120300  119.128833
1  18-002U      53       False    0.000096    0.992063  32.121133  119.128900
2  18-003A      53       False    0.000096    0.992063  32.122083  119.128850

Use a short station list for most figures. The functions can plot many stations, but readability falls quickly once every panel carries multiple components and legends.

11.24.3. Quick Station Panels

plot_sites_panels is the simplest overview. Each station gets two stacked panels: apparent resistivity or impedance magnitude on top, phase below. The default components are "xy" and "yx" – the off-diagonal component pair that normally carries the primary TE mode/TM mode information for a 1-D or 2-D earth. Apparent resistivity itself is the standard half-space-equivalent quantity,

\[\rho_a = 0.2\,\frac{|Z|^2}{f},\]

with \(Z\) in field units and \(f\) in hertz; every rho panel on this page plots \(\log_{10}\rho_a\) against period or frequency.

>>> import matplotlib.pyplot as plt
>>> from pycsamt.emtools import plot_sites_panels
>>> fig = plot_sites_panels(
...     survey,
...     stations=["18-001A", "18-007U", "18-016A", "18-018A"],
...     components=("xy", "yx"),
...     quantity="rhoa",
...     x_axis="period",
...     ncols=4,
...     show_error_bars=True,
...     show_legend=True,
... )
>>> [fig.axes[i].get_title() for i in range(0, 8, 2)]
['18-001A', '18-007U', '18-016A', '18-018A']
>>> fig.savefig("l18plt_station_panels.png", dpi=200)
>>> plt.close(fig)

Four stations fill the requested four-column grid exactly, one rho/phase pair of axes each, and fig.axes[0::2] (the top, rho, axis of each pair) carries the station name as its title – confirmed above rather than assumed. Use this figure when you want a quick visual sweep of several stations, not a full tensor or tipper review; for those, reach for plot_raw_sites_1d or plot_response_tipper instead.

11.24.4. Choose Rho Or Impedance Magnitude

Most field reports use apparent resistivity, but sometimes you need to inspect the impedance tensor directly. Set quantity="impedance" to plot \(\log_{10}|Z|\) instead of \(\log_{10}\rho_a\) – useful for tensor debugging, instrument checks, or comparison with diagnostics that work in \(Z\)-space rather than resistivity-space.

>>> fig = plot_sites_panels(
...     "data/AMT/WILLY_DATA/L18PLT",
...     stations=["18-001A", "18-016A"],
...     components=("xx", "xy", "yx", "yy"),
...     quantity="impedance",
...     phase_range=(-180.0, 180.0),
...     ncols=2,
...     show_legend=True,
... )
>>> fig.axes[0].get_ylabel()
'$\\log_{10}|Z|$'
>>> fig.savefig("impedance_magnitude_panels.png", dpi=200)

The y-axis label switches automatically with quantity, so a reader can tell which space a saved figure used without checking the call that produced it. Keep quantity="rhoa" for ordinary geophysical response plots and switch to "impedance" only when the tensor itself, not its resistivity-equivalent, is the object under review.

11.24.5. Phase Range And X Axis

The high-level panel function exposes x_axis and phase_range directly rather than requiring you to unwrap phase or convert period to frequency by hand.

>>> fig = plot_sites_panels(
...     "data/AMT/WILLY_DATA/L18PLT",
...     stations=["18-001A", "18-007U"],
...     components=("xy", "yx"),
...     x_axis="frequency",
...     phase_range=(0.0, 360.0),
...     ylim_phase=(0.0, 360.0),
...     ncols=2,
... )
>>> fig.axes[1].get_ylim()
(0.0, 360.0)
>>> fig.savefig("frequency_axis_phase_0_360.png", dpi=200)

The phase axis limits read back exactly as requested, confirming ylim_phase overrides the wrapping range rather than merely suggesting it. Use phase_range=None to show raw, unwrapped phase. Use an explicit range whenever several stations must share one display convention for comparison.

11.24.6. Raw Full-Tensor Panels

plot_raw_sites_1d is built for raw or nearly raw response review. Each station is a group, each selected component is a column, and each component column stacks apparent resistivity above phase.

>>> from pycsamt.emtools import plot_raw_sites_1d
>>> fig = plot_raw_sites_1d(
...     survey,
...     stations=["18-001A", "18-007U", "18-016A"],
...     components=("xx", "xy", "yx", "yy"),
...     raw=True,
...     force_style=False,
...     ncols_groups=3,
...     show_error_bars=True,
...     show_component_legend=True,
... )
>>> len(fig.axes)
24
>>> fig.savefig("raw_full_tensor_panels.png", dpi=200)
>>> plt.close(fig)

Three stations times four components times two rows (rho, phase) is the 24 axes reported above. With raw=True the function uses the package’s plain diagnostic style – black by default – so nothing about the display implies interpretation before you have even looked at it.

11.24.7. Force Component Colours

After the raw diagnostic pass, set force_style=True (with raw=True left on) to restore the usual per-component colours while keeping the identical layout.

>>> fig = plot_raw_sites_1d(
...     "data/AMT/WILLY_DATA/L18PLT",
...     stations=["18-001A", "18-007U", "18-016A"],
...     components=("xy", "yx"),
...     raw=True,
...     force_style=True,
...     ncols_groups=3,
... )
>>> fig.savefig("raw_panels_component_colours.png", dpi=200)

This makes a good second view: look for obvious raw-data problems in plain black first, then re-render with component colours once you want to compare TE mode/TM mode behaviour at a glance.

11.24.8. Shared Labels Or Axis Labels

plot_raw_sites_1d can lay out its station and axis labels two ways. label_mode="shared" draws the station name and axis labels once per group, as fig.text calls that float above and beside the whole group; label_mode="axis" instead sets each panel’s own Axes.set_ylabel and leaves no group-level figure text at all. The difference is visible directly in the object graph, not just the rendered figure:

>>> fig_shared = plot_raw_sites_1d(
...     "data/AMT/WILLY_DATA/L18PLT",
...     stations=["18-001A"],
...     components=("xx", "xy", "yx", "yy"),
...     label_mode="shared",
...     show_component_legend=False,
...     ncols_groups=1,
...     figsize_scale=(8.0, 4.0),
... )
>>> fig_shared.subplots_adjust(left=0.20)
>>> sorted(t.get_text() for t in fig_shared.texts)
['$\\log_{10}T$ (s)', '$\\log_{10}\\rho_a$ ($\\Omega\\,\\mathrm{m}$)', '18-001A', 'Phase ($^\\circ$)']
>>> fig_shared.savefig("raw_shared_labels.png", dpi=200)

>>> fig_axis = plot_raw_sites_1d(
...     "data/AMT/WILLY_DATA/L18PLT",
...     stations=["18-001A"],
...     components=("xx", "xy"),
...     label_mode="axis",
...     show_component_legend=False,
...     ncols_groups=1,
...     figsize_scale=(6.0, 4.0),
... )
>>> [t.get_text() for t in fig_axis.texts]
[]
>>> [ax.get_ylabel() for ax in fig_axis.axes if ax.get_ylabel()]
['$\\log_{10}\\rho_a$ ($\\Omega\\,\\mathrm{m}$)', 'Phase ($^\\circ$)']
>>> fig_axis.savefig("raw_axis_labels.png", dpi=200)

fig_shared.texts carries four entries – the station name plus rho, phase, and period-axis labels – while fig_axis.texts is empty because the same information lives on the axes themselves instead. label_mode="shared" is the more compact choice for dense reports; label_mode="axis" is sometimes easier while debugging a single group, since every label belongs to a concrete Axes you can query or move independently.

11.24.9. Use Display Control

plot_raw_sites_1d and plot_response_tipper read display policy from PYCSAMT_CONTROL unless you pass a specific control object. This is how one script changes x-axis convention, phase wrapping, and rho display consistently across every figure it produces.

>>> from pycsamt.api.control import PYCSAMT_CONTROL
>>> with PYCSAMT_CONTROL.context(
...     x__view="frequency",
...     rho__view="linear",
...     phase__range=(0.0, 360.0),
... ):
...     fig = plot_raw_sites_1d(
...         "data/AMT/WILLY_DATA/L18PLT",
...         stations=["18-001A"],
...         components=("xy", "yx"),
...         label_mode="axis",
...         force_style=True,
...         show_component_legend=False,
...         ncols_groups=1,
...         figsize_scale=(6.0, 4.0),
...     )
...
>>> fig.axes[0].get_ylabel()
'$\\rho_a$ ($\\Omega\\,\\mathrm{m}$)'
>>> fig.axes[2].get_ylim()
(0.0, 360.0)
>>> fig.savefig("raw_panels_frequency_linear_rho.png", dpi=200)
>>> plt.close(fig)

The rho label switches from $\log_{10}\rho_a$ to a plain linear $\rho_a$, and the phase axis limits read back exactly (0.0, 360.0) – both driven entirely by the active control, with no other argument touched. Use this pattern whenever one report needs a consistent, non-default display style; it is more reliable than editing labels or axis limits after the fact on each figure separately.

11.24.10. Response Plus Tipper

plot_response_tipper adds tipper rows to the impedance response layout. It needs a survey that actually carries vertical-field data, so this and the next example switch to KAP03 (data/MT/kap03lmt_edis) – confirmed below to carry tipper on every station, unlike L18PLT.

>>> from pycsamt.emtools import plot_response_tipper
>>> kap_summary = sites_summary(kap)
>>> bool(kap_summary["has_tipper"].all())
True
>>> fig = plot_response_tipper(
...     kap,
...     stations=["kap103", "kap142", "kap151"],
...     components=("xy", "yx"),
...     tipper_components=("tx", "ty"),
...     tipper_span_group=True,
...     ylim_tipper=(-2.5, 2.5),
...     ncols_groups=3,
... )
>>> fig.savefig("kap03_response_tipper.png", dpi=200)
>>> plt.close(fig)

With tipper_span_group=True each tipper row spans the whole station group, which usually reads best when there are only two impedance components and tipper should register as a station-level property rather than a per-component one. Running this same function against an AMT/CSAMT line with no tipper would still render – it would just have nothing meaningful to draw in the \(T_x\)/\(T_y\) rows, which is why the has_tipper check above matters before you invest in the figure.

11.24.11. Compact Tipper Rows

Set tipper_span_group=False for a compact, repeated layout where the tipper rows sit directly under each component column instead of spanning the group.

>>> fig = plot_response_tipper(
...     "data/MT/kap03lmt_edis",
...     stations=["kap151"],
...     components=("xx", "xy", "yx", "yy"),
...     tipper_components=("tx", "ty"),
...     tipper_span_group=False,
...     show_tipper_error_bars=False,
...     show_component_legend=False,
...     ncols_groups=1,
...     figsize_scale=(7.2, 5.6),
...     shared_x_label_pad=0.11,
... )
>>> fig.savefig("kap151_response_tipper_compact.png", dpi=200, bbox_inches="tight")

Use the compact layout for one or two stations where every tensor component matters. Use the spanning layout from the previous section when comparing several stations and fewer, larger axes read better than many small ones.

11.24.12. Before And After Comparison

plot_sites_compare pairs stations from two datasets column by column. The usual use is raw versus processed, before versus after a static shift correction, or original versus smoothed. There is no real post-processing run bundled with these docs, so a light frequency-domain moving average stands in honestly for “after”: real numbers from a real, if simple, transform rather than anything fabricated.

>>> from pycsamt.emtools import plot_sites_compare, smooth_mavg
>>> raw = survey
>>> smoothed = smooth_mavg(raw, k=5)
>>> fig = plot_sites_compare(
...     raw,
...     smoothed,
...     stations=["18-001A", "18-007U", "18-016A"],
...     components=("xy", "yx"),
...     labels=("raw", "smoothed k=5"),
...     quantity="rhoa",
...     x_axis="period",
...     ncols_groups=3,
...     show_legend=True,
... )
>>> [ax.get_title() for ax in fig.axes if ax.get_title()]
['18-001A', '18-007U', '18-016A']
>>> fig.savefig("raw_vs_smoothed_compare.png", dpi=200)
>>> plt.close(fig)

The title lives once per station group, spanning the raw/after column pair, which is exactly the three stations requested above. The function pairs stations by name; if the second dataset were missing a station, the corresponding after-column would simply come back blank rather than raising an error.

11.24.13. Compare Impedance Instead Of Rho

The comparison view also accepts quantity="impedance". This matters whenever a processing step changes the complex impedance tensor directly and you want to see that change without it being partly hidden behind the apparent-resistivity conversion.

>>> processed = smooth_mavg(raw, k=3)
>>> fig = plot_sites_compare(
...     raw,
...     processed,
...     stations=["18-001A", "18-016A"],
...     components=("xx", "xy", "yx", "yy"),
...     quantity="impedance",
...     phase_range=(-180.0, 180.0),
...     labels=("raw", "processed"),
...     ncols_groups=2,
... )
>>> fig.savefig("impedance_before_after.png", dpi=200)

Use fixed ylim_rhoa/ylim_phase whenever a before/after comparison needs the same visual scale across several separately saved figures – otherwise auto-scaling can make two genuinely similar curves look artificially different, or vice versa.

11.24.14. Measured Versus Predicted Fit Grid

plot_sites_fit_grid is built for inversion QC: it pairs measured and predicted stations by name, aligns predicted values onto the measured frequency grid, plots both curves, and annotates each component panel with its own RMS misfit. Writing the per-datum log-space residual as \(r_i = \log_{10}\rho_a^{\mathrm{meas}} - \log_{10}\rho_a^{\mathrm{pred}}\) (or the equivalent in \(\log_{10}|Z|\) when quantity="impedance") and its display-space error as \(\sigma_i\), the annotated value is

\[\mathrm{RMS} = \sqrt{\frac{1}{N}\sum_{i=1}^{N} \left(\frac{r_i}{\sigma_i}\right)^2}\]

when errors are available, falling back to the unweighted root-mean square of \(r_i\) when they are not. This is the same error-normalized misfit used throughout pyCSAMT’s inversion tooling: a value near 1 means the fit is consistent with the assigned uncertainty, while a much larger value means the “prediction” departs from the data by more than its own error bars would allow.

>>> from pycsamt.emtools import plot_sites_fit_grid
>>> measured = survey
>>> predicted = smooth_mavg(measured, k=5)
>>> fig = plot_sites_fit_grid(
...     measured,
...     predicted,
...     stations=["18-001A", "18-016A"],
...     components=("xy", "yx"),
...     quantity="rhoa",
...     phase_range=(-180.0, 180.0),
...     ncols_groups=2,
...     show_mode_legend=True,
... )
>>> sorted(t.get_text() for ax in fig.axes for t in ax.texts if "rms" in t.get_text())
['ZXY  rms=3.65', 'ZXY  rms=5.49', 'ZYX  rms=3.48', 'ZYX  rms=4.49']
>>> fig.savefig("measured_vs_predicted_fit_grid.png", dpi=200)
>>> plt.close(fig)

Every panel comes back with an RMS well above 1, from 3.48 to 5.49 across the two stations and two components requested. That is expected, not a bug in the demonstration: a five-point moving average was never fit to this survey’s error bars in the first place, so it lands several times outside them even though the raw curve shapes agree closely by eye. A real inversion response, fit to those same uncertainties, would be expected to land close to \(\mathrm{RMS} \approx 1\) instead. Treat this section’s numbers as a demonstration of the metric, not a QC result to reproduce.

11.24.15. Understand TE And TM Fit Colours

The fit grid uses separate fit colours for TE mode-like and TM mode-like components:

• xx and xy use the TE fit colour;

• yx and yy use the TM fit colour.

>>> fig = plot_sites_fit_grid(
...     measured,
...     predicted,
...     stations=["18-001A"],
...     components=("xx", "xy", "yx", "yy"),
...     color_fit_te="#2ca02c",
...     color_fit_tm="#d62728",
...     lw_fit=2.2,
...     ls_fit="-",
...     ncols_groups=1,
...     figsize_scale=(8.0, 4.0),
...     show_mode_legend=False,
... )
>>> fig.savefig("fit_grid_custom_fit_colours.png", dpi=200, bbox_inches="tight")

Use custom fit colours when assembling figures for publication or when the package defaults conflict with another report’s existing colour convention.

11.24.16. Full-Response Station Overview

Every figure so far puts several stations side by side. Sometimes the opposite view is what a report needs: one station, but the complete picture – apparent resistivity and phase for all four impedance components, induction arrows, and phase tensor ellipses, all sharing one period axis. plot_response_overview (in the sibling pycsamt.emtools.overview module, re-exported from pycsamt.emtools) builds exactly that: the classic MT “full response” quicklook, reworked to read display convention from PYCSAMT_CONTROL/PYCSAMT_STYLE like every other function on this page rather than hard-coding colours or axis choices.

The example switches datasets one more time, to the Gabbs Valley USGS MT survey bundled at data/gv_data – three real stations (gv100, gv130, gv163) that, unlike L18PLT, carry a full \(2\times 2\) impedance tensor (\(Z_{xx}\) and \(Z_{yy}\) included, not just the off-diagonal pair) as well as tipper, which this figure needs to fill every row.

>>> from pycsamt.emtools import ensure_sites, sites_summary, plot_response_overview
>>> gv = ensure_sites("data/gv_data/gv_final_edi", strict=True)
>>> gv_summary = sites_summary(gv)
>>> print(len(gv_summary), "stations,", "tipper present:", bool(gv_summary["has_tipper"].all()))
3 stations, tipper present: True
>>> fig = plot_response_overview(gv, station="gv100")
>>> len(fig.axes)
7
>>> fig.axes[0].get_yscale(), fig.axes[0].get_xscale()
('log', 'log')
>>> fig.axes[2].get_yscale()
'linear'
>>> fig.savefig("gv100_response_overview.png", dpi=200, bbox_inches="tight")

Seven axes come back for the default two-column layout: apparent resistivity and phase for the off-diagonal (\(Z_{xy}\), \(Z_{yx}\)) pair on the left, the diagonal (\(Z_{xx}\), \(Z_{yy}\)) pair on the right, one induction-arrow row, one phase-tensor-ellipse row, and the ellipse row’s own colourbar axis. The apparent-resistivity row confirms as genuinely log-log (get_yscale() == "log" over a log x-axis, not \(\log_{10}\rho_a\) values plotted on a linear axis) and the phase row as linear-over-log, i.e. semilog – both draw Matplotlib’s own per-decade grid rather than sparse gridlines over pre-logged numbers, matching the reference quicklook convention this figure reproduces.

Reading the panels from top to bottom: the resistivity and phase curves behave exactly as in the earlier sections, just with all four components present at once – \(Z_{xx}\) and \(Z_{yy}\) stay small relative to the off-diagonal pair through the middle of the period range, as expected for a survey without strong 3-D distortion there, then grow noticeably at the longest periods where the error bars also widen – a data-quality effect documented in the survey’s own USGS data release, Peacock et al. (2021), not a plotting artefact. The induction-arrow row draws one arrow per period for the real and imaginary tipper parts, in the same real/imag colour pairing used by plot_response_tipper’s \(T_x\)/\(T_y\) rows earlier on this page; each arrow’s vertical extent is the physically meaningful part, while its slight horizontal fan is a schematic separation between neighbouring periods, not a period shift. The bottom row draws one phase-tensor ellipse per period, coloured by Skew by default: the pale, near-white ellipses through the middle of the period range indicate low skew (consistent with the comparatively regular, near-1-D-looking sounding curves above them), while the more saturated red/blue ellipses at the shortest and longest periods flag the same noisier bands already visible in the resistivity row.

11.24.17. A Full-Range Ellipse Colouring

The default skew colouring is the more diagnostic choice – it flags departures from 1-D/2-D structure directly – but some reports expect the other common convention: colouring by \(\phi_{\min}\) (the phase tensor’s minor-axis angle) over its natural \(0^\circ\)-\(90^\circ\) range. Pass c_by="phimin_deg" with an explicit clim to get it; note this is deliberately not the same as c_by="phi_min", which colours by the raw phase-tensor singular value (tan units) instead of its arctan in degrees and saturates near one end of a 0-90 scale instead of using it fully.

>>> fig = plot_response_overview(
...     gv,
...     station="gv130",
...     c_by="phimin_deg",
...     clim=(0.0, 90.0),
...     cmap="turbo",
... )
>>> cbar_ax = fig.axes[-1]
>>> cbar_ax.get_xlabel()
'$\\phi_{min}$ (°)'
>>> cbar_ax.get_xlim()
(0.0, 90.0)
>>> fig.savefig("gv130_response_overview_phimin.png", dpi=200, bbox_inches="tight")

The colourbar now spans the full requested \(0^\circ\)-\(90^\circ\) range and every ellipse gets a distinct hue rather than clustering near one end – exactly the “very saturated, easy to scan” look a \(\phi_{\min}\)-coloured strip is expected to have, and the reason this convention, not skew, is the one to reach for when the ellipse row needs to be readable at a glance rather than used as a 3-D-structure flag. The colourbar itself lives in its own reserved row below the ellipse strip by default (cbar_orientation="horizontal"); pass cbar_orientation="vertical" to restore a right-hand gutter column instead. Either way the space is reserved in the layout up front, so the arrow row above and the ellipse row stay pixel-aligned on the shared period axis – a colourbar attached the ordinary Figure.colorbar-via-divider way would otherwise narrow just the ellipse row after the fact, leaving a given period at a different pixel column in each row.

Two more knobs are worth knowing about before reaching for this function on a new survey. show_diag=False drops the \(Z_{xx}\)/\(Z_{yy}\) column entirely, narrowing the figure to a single column when a survey’s diagonal terms are not of interest or are too noisy to be useful; the arrow and ellipse rows then span that one column instead of two. And x_view and log_log_rho control the axis conventions independently of PYCSAMT_CONTROL’s shared global state – x_view="log10_period" switches to the \(\log_{10}T\,(\mathrm{s})\) linear-axis convention used by plot_raw_sites_1d earlier on this page, and log_log_rho=False does the same for apparent resistivity, without needing a PYCSAMT_CONTROL.context(...) block for either.

11.24.18. Choosing The Right Figure

Here is a practical decision path:

I only need a quick rho/phase overview

Use plot_sites_panels with components=("xy", "yx") and a short station list.

I want a raw tensor QC view

Use plot_raw_sites_1d with all four components and raw=True.

The survey has tipper and I need to inspect it with response curves

Use plot_response_tipper. Confirm tipper exists with sites_summary or list_missing_sections first, as done above.

I processed the data and need before/after panels

Use plot_sites_compare with the original and processed Sites objects.

I have model predictions or inversion responses

Use plot_sites_fit_grid and read the RMS labels component by component.

I need the complete picture for one station -- full tensor, arrows, and ellipses

Use plot_response_overview. Confirm the station carries a full \(2\times 2\) tensor and tipper first, the same way sites_summary confirmed it for gv100 above.

11.24.19. Common Pitfalls

Too many stations in one figure

Start with four to six stations. If you need every station, create multiple figures by station group.

Comparing figures with auto-scaled axes

Use fixed ylim_rhoa and ylim_phase when the visual comparison matters.

Using tipper plots on AMT lines with no tipper

The figure can still render, but it will not add useful interpretation. Check sites_summary first, as this page does.

Treating a smoothed dataset as a real prediction

A smoothed copy is useful for demonstrating the plotting API, but it is not a forward-model response – the inflated RMS values above are the direct consequence of that gap.

Ignoring display control

If a report switches between period and frequency axes, or between linear and log rho displays, use PYCSAMT_CONTROL.context so all plots share the same convention, rather than editing axes after the fact.

11.24.20. Save A Multi-Figure Plot Bundle

The following script writes the core report figures for one line in a single pass: a station-panel grid, raw full-tensor panels, a raw-vs-smoothed comparison, and a measured-vs-predicted fit grid, all from the same survey and station subset used throughout this page.

</>View the executed report-bundle source codeClick to inspect and copy the complete code

def make_plot_report_bundle() -> None:
    """Write the four report figures used by the "Save A Multi-Figure
    Plot Bundle" section: a station-panel grid, raw full-tensor panels,
    a raw-vs-smoothed comparison, and a measured-vs-predicted fit grid,
    all built from the same L18PLT survey and station subset.
    """
    survey = ensure_sites(L18PLT, strict=True)
    stations = ["18-001A", "18-007U", "18-016A", "18-018A"]
    processed = smooth_mavg(survey, k=5)

    fig = plot_sites_panels(
        survey,
        stations=stations,
        components=("xy", "yx"),
        ncols=4,
        show_legend=True,
    )
    fig.savefig(IMAGES / "user-guide-emtools-plot-15-01.png", dpi=200)
    plt.close(fig)

    fig = plot_raw_sites_1d(
        survey,
        stations=stations[:3],
        components=("xx", "xy", "yx", "yy"),
        ncols_groups=3,
    )
    fig.savefig(IMAGES / "user-guide-emtools-plot-15-02.png", dpi=200)
    plt.close(fig)

    fig = plot_sites_compare(
        survey,
        processed,
        stations=stations[:3],
        components=("xy", "yx"),
        labels=("raw", "smoothed"),
        ncols_groups=3,
    )
    fig.savefig(IMAGES / "user-guide-emtools-plot-15-03.png", dpi=200)
    plt.close(fig)

    fig = plot_sites_fit_grid(
        survey,
        processed,
        stations=stations[:2],
        components=("xy", "yx"),
        ncols_groups=2,
    )
    fig.savefig(IMAGES / "user-guide-emtools-plot-15-04.png", dpi=200)
    plt.close(fig)

Each of the four panels above comes from one of the functions covered earlier on this page, run back to back against the same four-station subset – a convenient single script to keep on hand once a survey has a settled station list and you just want the standard report figures regenerated.

11.24.21. Worked Example

The gallery example demonstrates the same plotting layer on bundled AMT and MT surveys, including a tipper-capable MT line and display-control overrides.

Open the rendered gallery page here: Multi-station diagnostic panels (pycsamt.emtools.plot).