Radiometer Maps: Calibration & Retrieval Methods

• Radiometer maps are spatial or spatial–temporal representations that transform radiometric measurements into detailed maps of physical state variables in atmospheric, astronomical, and planetary studies.
• They employ diverse calibration and mapmaking methods—such as generalized least squares and destriping—to correct instrument biases and extract actionable data.
• Applications include CMB testing, low-frequency radio imaging, and planetary thermophysics, with retrieval algorithms inferring parameters like PWV, soil moisture, and thermal inertia.

Radiometer maps are spatial or spatial–temporal products derived from radiometric measurements of brightness temperature, antenna temperature, spectral radiance, or flux density, and then organized either as direct maps of the measured quantity or as retrieved maps of a physical state variable. In the cited literature, the term encompasses time–azimuth “atmograms” from scanning 183 GHz water vapor radiometers, nearly full-sky HEALPix maps in Kelvin antenna temperature, 128 pixels-per-degree lunar thermophysical grids, 0.5° × 0.5° Europa brightness-temperature and thermal-property maps, 37 GHz solar scanline reconstructions and butterfly diagrams, diffuse 182 MHz radio maps for short-baseline calibration, and 4D radio-interferometric cubes $D(\theta,\phi,\nu,t)$ instantiated as $S_\nu(\theta,\phi,\nu,t)$ or $T_B(\theta,\phi,\nu,t)$ (Barkats et al., 2018, Anderson et al., 2024, Hayne et al., 2017, Lange et al., 15 Apr 2026, Kivistö et al., 2024, Byrne et al., 2021, Mohan et al., 2017).

1. Scope, forms, and coordinate conventions

Radiometer maps are not restricted to a single geometry or data model. In the CMB site-testing case, maps are made natively in az–el coordinates, and a principal product is the “atmogram,” a 2D image with time on the x-axis and azimuth on the y-axis, built from 183.31 GHz water vapor radiometer timestreams acquired during continuous azimuthal scans at fixed elevation (Barkats et al., 2018). In low-frequency radio astronomy, maps may be full-sky HEALPix products in Galactic coordinates, as in Juno’s six nearly full-sky maps at 0.6–21.9 GHz, or interferometric sky maps of brightness temperature $T(\nu,\hat s)$ across large angular scales, as in MITEoR’s 128–175 MHz northern sky map (Anderson et al., 2024, Zheng et al., 2016).

Other examples are intrinsically swath-based or scanline-based. DRAIN treats the four GMI brightness-temperature fields $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$ as a 2-D image in native GMI swath geometry, with inputs handled as radiometer swath images rather than reprojected lat–lon grids (Viltard et al., 2023). Metsähovi solar maps are constructed from radio intensity sampled along scanlines of the antenna sweep, with subsequent assignment of heliographic coordinates and time–latitude aggregation into a butterfly diagram (Kivistö et al., 2024). Galileo PPR maps of Europa are produced by spherical projection of overlapping scans to a $0.5^\circ \times 0.5^\circ$ grid, while Diviner global maps are gridded Level 2 products at 128 pixels per degree (Lange et al., 15 Apr 2026, Hayne et al., 2017).

A compact summary of representative forms is useful:

Domain	Principal map product	Native quantity
CMB site testing	Atmogram, azimuthal PWV map	$T_b$ or PWV
Low-frequency radio sky	HEALPix diffuse map, full-sky interferometric map	$T_A$, $T_B$, or Jy/sr
Solar radio	Scanline map, butterfly diagram, 4D cube, SPREDS	Intensity, $S_\nu$, or $S_\nu(\theta,\phi,\nu,t)$0
Planetary thermal mapping	Brightness-temperature, albedo, thermal inertia maps	$S_\nu(\theta,\phi,\nu,t)$1, $S_\nu(\theta,\phi,\nu,t)$2, $S_\nu(\theta,\phi,\nu,t)$3
Hydrologic and meteorological retrieval	Soil moisture map, rain-rate map	$S_\nu(\theta,\phi,\nu,t)$4 or retrieved geophysical field

This variety matters because coordinate choice, beam geometry, and revisit pattern determine which modes are measured directly and which must be inferred. A recurrent misconception is that a radiometer map is always a static 2D image of surface brightness. In the cited work, some primary products are explicitly spatial–temporal, including atmograms and 4D cubes, and others are retrieved parameter maps rather than direct radiance maps (Barkats et al., 2018, Mohan et al., 2017).

2. Measurement principles and calibration chains

Radiometer maps inherit their semantics from the radiometric observable. For the 183 GHz WVRs, the instrument reports sky brightness directly in Rayleigh–Jeans units, with

$S_\nu(\theta,\phi,\nu,t)$5

and Dicke-switching between internal hot and ambient loads at 5 Hz provides continuous calibration and stability; both loads are regulated to better than 1 mK (Barkats et al., 2018). Juno’s cruise maps instead begin from calibrated antenna temperature time streams, using one-second cycles that include a noise-diode-on sky sample and an internal load sample, followed by multi-minute running averages to remove receiver temperature/DC offsets and set gain (Anderson et al., 2024).

In thermal planetary applications, the calibration target is commonly brightness temperature in a Planck framework rather than a Rayleigh–Jeans one. Europa PPR conversion assumes graybody emission with emissivity $S_\nu(\theta,\phi,\nu,t)$6 and solves

$S_\nu(\theta,\phi,\nu,t)$7

while Diviner reports a brightness temperature that is the blackbody temperature consistent with the measured spectral radiance in that band (Lange et al., 15 Apr 2026, Hayne et al., 2017). In ASTER TIR geological mapping, Level 1B digital numbers are converted to at-sensor radiance through

$S_\nu(\theta,\phi,\nu,t)$8

and subsequent temperature normalization is applied before mineralogical index construction (Corrie et al., 2011).

Absolute scaling can also be external to the mapmaker. In MWA solar imaging, interferometric snapshot maps are placed on an absolute flux scale by matching the disc-integrated image flux to an independently estimated solar flux density $S_\nu(\theta,\phi,\nu,t)$9, with

$T_B(\theta,\phi,\nu,t)$0

and calibrated maps given by $T_B(\theta,\phi,\nu,t)$1 (Mohan et al., 2017). PoLRa uses cold-sky calibration to convert raw voltages to brightness temperatures, followed by filtering rules such as $T_B(\theta,\phi,\nu,t)$2 and a maximum threshold of 320 K (Zhang et al., 2024).

The calibration architecture therefore depends on the platform and observable: Dicke-switching, internal loads, noise diodes, cold-sky references, external calibrators such as Cygnus A and Cassiopeia A, and independent total-power spectra all appear in the cited literature (Barkats et al., 2018, Anderson et al., 2024, Zhang et al., 2024, Zheng et al., 2016, Mohan et al., 2017).

3. Mapmaking formalisms

Once calibrated, the data are projected into map space through platform-specific pointing and inversion operators. The most explicit formulation in the cited work is generalized least squares. For az–el PWV maps over one hour, the WVR time-ordered data $T_B(\theta,\phi,\nu,t)$3 are projected to a pixelized map $T_B(\theta,\phi,\nu,t)$4 via a pointing matrix $T_B(\theta,\phi,\nu,t)$5 and noise covariance $T_B(\theta,\phi,\nu,t)$6: $T_B(\theta,\phi,\nu,t)$7 The paper emphasizes that filtering transfer functions $T_B(\theta,\phi,\nu,t)$8 suppress large angular scales and must be deconvolved when interpreting low-frequency structure (Barkats et al., 2018).

MITEoR adopts an analogous linear system for full-sky interferometric imaging,

$T_B(\theta,\phi,\nu,t)$9

with minimum-variance estimator

$T(\nu,\hat s)$0

regularized in practice because $T(\nu,\hat s)$1 is ill-conditioned for partially sampled sky modes (Zheng et al., 2016). Juno’s cruise maps use destriping with MADAM, representing long-time-scale offsets as $T(\nu,\hat s)$2 in

$T(\nu,\hat s)$3

and extending the model for R1 and R2 to include multiplicative gain drifts in an NPIPE-like self-calibration iteration (Anderson et al., 2024).

Swath and scanline products are built differently. The 182 MHz diffuse MWA map uses Fast Holographic Deconvolution, horizon-to-horizon snapshots, deconvolution-free reimaging on baselines $T(\nu,\hat s)$4, and position-dependent Tukey weighting before HEALPix mosaicking (Byrne et al., 2021). Metsähovi solar maps fit a circular disk to scanline samples, correct for limb brightening and beamwidth convolution through a model $T(\nu,\hat s)$5, and then transform fitted image-plane coordinates into heliographic latitude and longitude (Kivistö et al., 2024). Europa’s PPR brightness temperatures are mosaicked after geometry filtering, and Diviner’s global grids are assembled from a near-polar orbital sampling pattern that completes a global mapping cycle in about one month (Lange et al., 15 Apr 2026, Hayne et al., 2017).

These differences are not merely implementation details. They determine whether the map is mean-zero, whether a monopole is externally imposed, how beam smearing enters, and which angular or temporal modes are attenuated by preprocessing.

4. From radiometric maps to retrieved geophysical fields

A large fraction of radiometer maps are retrieved products built on forward models. In the WVR case, each channel measures a band-averaged brightness temperature modeled as

$T(\nu,\hat s)$6

and PWV is obtained by minimizing

$T(\nu,\hat s)$7

This converts multi-channel brightness maps into PWV maps and associated spatial and temporal statistics (Barkats et al., 2018).

For soil moisture, PoLRa uses the $T(\nu,\hat s)$8–$T(\nu,\hat s)$9 model,

$$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$0

with $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$1, together with SMAP Single Channel Algorithm, Regularized Dual Channel Algorithm, or Dual Channel Algorithm variants to map soil moisture from measured $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$2 (Zhang et al., 2024). DRAIN uses a different inversion strategy: a U-net ingests full GMI swath images and retrieves 99 rain-rate quantiles per pixel, with the median used as the default estimator (Viltard et al., 2023).

In planetary thermophysics, the retrieved variables are often albedo, thermal inertia, and structure parameters. Diviner fits rock-removed nighttime regolith brightness temperatures with a 1-D conduction model whose single free parameter per pixel is the $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$3-parameter, then derives a fixed-temperature thermal inertia map $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$4 (Hayne et al., 2017). Europa PPR uses the KRC thermal model to invert day–night temperature pairs for Bond albedo $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$5 and thermal inertia $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$6, and then interprets $$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$7 with porous-ice conductivity models to infer porosity and grain-size bounds (Lange et al., 15 Apr 2026). ASTER TIR geological mapping transforms normalized radiance into the Carbonate Index,

$$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$8

the Quartz Index,

$$(37\mathrm{H},37\mathrm{V},89\mathrm{H},89\mathrm{V})$$9

and the Mafic Index,

$0.5^\circ \times 0.5^\circ$0

thereby producing mineralogical radiometer maps rather than direct thermal maps (Corrie et al., 2011).

Jupiter’s north polar MWR maps illustrate a further class of retrieval: multi-angle, multi-frequency atmospheric inversion. Polar-mean nadir brightness temperatures and limb-darkening spectra from 22, 10, 5.2, 2.6, and 1.25 GHz jointly constrain vertical temperature, ammonia, and water, with Markov chain Monte Carlo retrievals supporting two equally plausible scenarios: a dry-adiabatic profile with slight NH$0.5^\circ \times 0.5^\circ$1 depletion at a few bars, or a moist-adiabatic profile with uniform ammonia (Hu et al., 14 May 2026).

5. Statistical interpretation and scientific uses

Radiometer maps are often used less as pictures than as statistical fields. For CMB site testing, azimuthal PWV maps yield a correlation function

$0.5^\circ \times 0.5^\circ$2

a second-order structure function

$0.5^\circ \times 0.5^\circ$3

power spectra, knee frequencies, autocorrelation times, and the one-number site metric

$0.5^\circ \times 0.5^\circ$4

computed over the 0.02–0.2 Hz band relevant for degree-scale CMB polarization scanning (Barkats et al., 2018).

In 21 cm cosmology, diffuse radio maps close the modeling gap on short baselines. The 182 MHz MWA map provides 1.1–9.4° structure in Stokes $0.5^\circ \times 0.5^\circ$5, and validation is performed with the “fraction of signal modeled” metric comparing simulated and measured visibilities; adding the diffuse map improves short-baseline agreement so that accuracy now approaches that of longer baselines typically used for EoR calibration (Byrne et al., 2021). MITEoR’s full-sky map similarly serves as a foreground model and calibration target for compact arrays, with direct access to a position-dependent PSF and map covariance (Zheng et al., 2016).

Solar applications make especially heavy use of temporal dimensionality. The 4D cube $0.5^\circ \times 0.5^\circ$6 preserves spatial, spectral, and temporal information simultaneously, while SPREDS extracts a local dynamic spectrum from an arbitrary region (Mohan et al., 2017). Metsähovi’s long time series instead bins scanline-derived measurements by time and heliographic latitude to construct a 37 GHz butterfly diagram spanning solar cycles 21 to 24 and extending near to the poles (Kivistö et al., 2024).

Planetary radiometer maps support process inference on geologic timescales. Diviner’s $0.5^\circ \times 0.5^\circ$7 maps show remarkable global uniformity in lunar fines, superposed with high-thermal-inertia young crater interiors and ejecta, low-inertia cold spots, and a high-thermal-inertia deposit near the antipode of Tycho crater (Hayne et al., 2017). Europa’s radiometer-derived thermal inertia and porosity maps reveal a low-inertia equatorial band on the leading hemisphere, higher inertia at mid-latitudes and on the trailing equator, little correlation with most geologic units except the Pwyll ejecta, and agreement with modeled sputtering rates that supports sputtering-driven sintering (Lange et al., 15 Apr 2026). Jupiter polar maps show circumpolar cyclone anomalies detectable down to about 8 bar and a north-pole brightness structure consistent with either reduced ammonia opacity or moist-adiabatic warming (Hu et al., 14 May 2026).

6. Limitations, ambiguities, and interpretive cautions

Several limitations recur across otherwise disparate radiometer-map applications. First, absolute zero level is not always observed. The 182 MHz interferometric diffuse map is explicitly mean-zero in all Stokes because there is no absolute zero-spacing, so polarization fraction and angle in the image domain cannot be directly inferred (Byrne et al., 2021). Juno’s sky maps adopt the ARCADE 2 model plus the CMB as the monopole assignment,

$0.5^\circ \times 0.5^\circ$8

and the paper states that this is a calibration choice (Anderson et al., 2024). Solar interferometric maps likewise require external disc-integrated flux information because snapshot images do not measure the total power (Mohan et al., 2017).

Second, preprocessing can erase or bias scientifically relevant modes. High-pass filtering and polynomial detrending in WVR data attenuate the largest angular scales, making transfer-function deconvolution necessary for low-$0.5^\circ \times 0.5^\circ$9 interpretation (Barkats et al., 2018). Juno’s low-frequency maps exhibit pixelization noise near the Galactic plane at $T_b$0, requiring either $T_b$1 products or pixelization-correction maps (Anderson et al., 2024). Beam size also imposes hard limits: the WVR beam is about $T_b$2 FWHM, Diviner’s native footprint is roughly $T_b$3, and ASTER TIR’s 90 m resolution is too coarse to resolve internal lithologic sequences within ophiolites (Barkats et al., 2018, Hayne et al., 2017, Corrie et al., 2011).

Third, model dependence is intrinsic to retrieved radiometer maps. Europa retrievals assume $T_b$4, constant $T_b$5 with depth, and no explicit volumetric solar absorption; the inferred properties are therefore “apparent” values at the PPR footprint scale (Lange et al., 15 Apr 2026). PoLRa soil-moisture retrieval depends on surface roughness, soil effective temperature, NDVI-derived $T_b$6, and dielectric model choice, with vertically polarized brightness temperatures generally more stable than horizontally polarized ones (Zhang et al., 2024). DRAIN’s rain-rate maps exhibit systematic underestimation above about $T_b$7 and larger errors over mountains and in winter (Viltard et al., 2023).

Finally, radiometer maps may support more than one physical interpretation. Jupiter’s polar MWR spectra are fit equally well by a dry adiabat with slight NH$T_b$8 depletion starting near $T_b$9 bar and by a moist adiabat with uniform NH$T_A$0 and latent heating near the water cloud, while the inferred 6–7 K polar warmth at 1 bar is close to the 1-sigma uncertainty level (Hu et al., 14 May 2026). Such cases illustrate that a radiometer map is not synonymous with a unique physical solution; angular dependence, frequency leverage, and prior assumptions control the degeneracy structure.

Across these applications, radiometer maps function as calibrated, geometry-aware, and often model-dependent representations of radiative fields. Their central strength is that they link instrument-native observables to spatial or spatial–temporal structure at scientifically relevant scales, whether the target is atmospheric water vapor, diffuse synchrotron emission, solar radio variability, lunar regolith thermophysics, Europan porosity, Jovian polar composition, soil moisture, mineralogical contrast, or precipitation morphology (Barkats et al., 2018, Byrne et al., 2021, Hayne et al., 2017, Lange et al., 15 Apr 2026, Anderson et al., 2024).