14.6. Frequency Bands#

Every EM method operates over a characteristic frequency range, and knowing that range matters for survey design, quality review, and depth-of-investigation expectations. FrequencyBand captures one method’s operational frequency range together with its period bounds and a rule-of-thumb depth-of-investigation estimate from the skin-depth relationship. MT_BANDS is a ready-made registry covering the principal EM methods used in applied geophysics, and register_band() lets a project add its own named bands without modifying the package. Unlike most classes covered earlier in this section, FrequencyBand is a plain dataclass, not a PyCSAMTObject subclass, and it is general-purpose rather than tied to the EMTF document model – nothing here maps to an EMTF XML element.

14.6.1. Built-In Bands#

MT_BANDS and the mutable REGISTRY (seeded from it) currently hold eight named bands:

>>> from pycsamt.metadata.frequency import MT_BANDS, REGISTRY

>>> print(sorted(REGISTRY.keys()))
['AMT', 'BBMT', 'CSAMT', 'CSEM', 'LAMT', 'LMT', 'MT', 'TEM']

>>> amt = MT_BANDS["AMT"]
>>> print(amt.f_min, amt.f_max)
10.0 100000.0
>>> print(amt.period_range)
(1e-05, 0.1)
>>> print(round(amt.n_decades, 2))
4.0

doi_range_m() turns the band into a rough depth-of-investigation range via the skin-depth formula \(\delta \approx 503.3\sqrt{\rho/f}\), evaluated at f_max (shallow) and f_min (deep). The reference resistivity defaults to 100 \(\Omega \cdot \text{m}\) but can be overridden – a more conductive earth attenuates the same frequency faster, so a lower resistivity always narrows both DOI bounds:

>>> print(amt.doi_range_m())
(15.9, 1591.6)
>>> print(amt.doi_range_m(rho=10.0))
(5.0, 503.3)
>>> print(amt.doi_range_m(rho=1000.0))
(50.3, 5033.0)
>>> print(repr(amt))
FrequencyBand('AMT'  10–1e+05 Hz  DOI≈16–1592 m  [AMT])

14.6.2. Which Band Covers a Frequency#

band_for_frequency() returns every band containing a query frequency, narrowest first. Applied to the real frequency endpoints of the two stations used throughout this section, it shows how a single sounding can straddle several named bands at once:

>>> from pathlib import Path
>>> from pycsamt.seg.edi import EDIFile
>>> from pycsamt.site.base import Site
>>> from pycsamt.metadata.frequency import band_for_frequency

>>> site_willy = Site(EDIFile(Path("data/AMT/WILLY_DATA/L18PLT/18-001A.edi")))
>>> freq_willy = site_willy.freq
>>> print(freq_willy.min(), freq_willy.max(), freq_willy.size)
1.008 10400.0 53

>>> print([b.name for b in band_for_frequency(freq_willy.min())])
['LAMT', 'CSAMT', 'TEM', 'MT', 'BBMT']
>>> print([b.name for b in band_for_frequency(freq_willy.max())])
['AMT']

WILLY’s lowest frequency, 1.008 Hz, falls inside five different named bands at once – it is squarely in the overlap where LAMT, CSAMT, TEM, MT, and BBMT all claim coverage. Its highest frequency, 10.4 kHz, is narrow enough to belong to AMT alone. The Gabbs Valley station used in Processing and Quality spans an even wider range and shows the opposite extreme at its low end:

>>> site_gv = Site(EDIFile(Path("data/gv_data/gv_final_edi/gv100.edi")))
>>> freq_gv = site_gv.freq
>>> print(freq_gv.min(), freq_gv.max(), freq_gv.size)
0.0004882812 767.9902 48

>>> print([b.name for b in band_for_frequency(freq_gv.min())])
['LMT', 'TEM', 'MT', 'BBMT']
>>> print([b.name for b in band_for_frequency(freq_gv.max())])
['AMT', 'LAMT', 'CSAMT', 'MT', 'BBMT']

Note

gv100.edi above is public-domain USGS data; see Provenance and Bibliography for the required citation.

Neither result names the survey’s own acquisition method – both files are conventional broadband MT – band_for_frequency answers “which named bands contain this frequency,” not “which method recorded it.”

14.6.3. Overlap and Clipping#

overlaps() and intersection() compare two bands directly, and clip_frequencies() filters a real frequency array down to one band’s range:

>>> csamt = MT_BANDS["CSAMT"]
>>> mt = MT_BANDS["MT"]
>>> print(amt.overlaps(csamt), amt.intersection(csamt))
True (10.0, 10000.0)
>>> print(amt.overlaps(mt), amt.intersection(mt))
True (10.0, 1000.0)

>>> clipped = amt.clip_frequencies(freq_willy)
>>> print(clipped.size, freq_willy.size)
40 53

Thirteen of WILLY’s 53 frequencies – everything below 10 Hz – fall outside the AMT band and are dropped by clip_frequencies(). That is expected: WILLY is a broadband AMT-class survey extending below the nominal AMT floor, not a sign that a quarter of the sounding is somehow invalid. Use clipping to restrict a plot or an inversion sounding to one named band’s nominal range, not as a data-quality filter – Processing and Quality’s DataQuality is the right tool for that.

14.6.4. Extending the Registry#

register_band() adds a custom FrequencyBand to the global REGISTRY without touching the package. The committed pycsamt.airborne package (see the top-level user guide) has no frequency-band registry of its own, so a project working with a MobileMT-style system might register its two frequency stacks directly:

>>> from pycsamt.metadata.frequency import FrequencyBand, register_band, frequency_range

>>> low = FrequencyBand(
...     name="MOBILEMT_LOW", label="MobileMT low stack",
...     f_min=25.0, f_max=6000.0, method="AIRBORNE",
... )
>>> high = FrequencyBand(
...     name="MOBILEMT_HIGH", label="MobileMT high stack",
...     f_min=6000.0, f_max=115000.0, method="AIRBORNE",
... )
>>> register_band(low)
>>> register_band(high)
>>> print(frequency_range("AIRBORNE"))
(25.0, 115000.0)

That result is the method-level union frequency_range()’s docstring promises: two bands sharing method="AIRBORNE", neither named exactly "AIRBORNE", so the query falls through to the union branch. That branch is easy to miss in the built-in registry, because it never actually fires there – every MT_BANDS band’s method string is also the name of a band in its own right (AMT is both a band and a method, and so is every other entry), so frequency_range()’s direct name lookup always wins first and the union path is dead code for the shipped registry. It only becomes observable once a project registers multiple bands under one shared method that is not itself a band name, as above.

Registering with an uppercase name also matters more than it looks. frequency_range() upper-cases its query before checking REGISTRY for a matching key, which works seamlessly for the built-in bands because they are all uppercase already. A band registered with a lowercase or mixed-case name is invisible to name-based lookup through frequency_range(), under either casing of the query – even though it is genuinely present in REGISTRY and fully reachable through band_for_frequency() or direct dict access:

>>> lower = FrequencyBand(
...     name="mobilemt_mid", label="a lowercase-keyed band",
...     f_min=1000.0, f_max=2000.0, method="AIRBORNE",
... )
>>> register_band(lower)
>>> print("mobilemt_mid" in REGISTRY)
True
>>> try:
...     frequency_range("mobilemt_mid")
... except KeyError as exc:
...     print(type(exc).__name__)
KeyError

Follow the built-in convention – register custom bands with an uppercase name – unless the only access pattern needed is REGISTRY[name] or band_for_frequency(), neither of which uppercases anything.

14.6.5. Choosing the Right Function#

Need	Function/method	Notes
Look up a named band’s range	`REGISTRY[name]` or `frequency_range()`	`frequency_range` needs an uppercase name to match by name.
Which bands contain a frequency	`band_for_frequency()`	Sorted narrowest to widest; scans by range, no name-key lookup.
Rough depth of investigation	`doi_range_m()` / `doi_estimate()`	Skin-depth estimate only; not a substitute for real forward modelling.
Filter an array to one band	`clip_frequencies()`	A range filter, not a data-quality tool – see Processing and Quality.
Add a project-specific band	`register_band()`	Use an uppercase `name` to match the built-in convention.

14.6.6. Next Steps#

This closes the current pass through Metadata. The remaining pycsamt.metadata modules – geology (resistivity priors for synthetic layered-earth generation) and rocks – are forward- modelling scaffolding rather than EMTF-adjacent metadata, and are deliberately not covered here; see the module docstrings directly if needed. From here:

Site and Survey Metadata through Processing and Quality cover the format-neutral metadata that populates an EMTF document;
the next step in the broader documentation roadmap is a deep pass through user_guide/emtf/, which can now cross-reference this section instead of re-explaining these classes.