Observational Temperature: Concepts & Applications

• Observational temperature is a measurement-dependent construct defined through calibrated instrument readings, fixed points, and ordered hotness levels.
• It is applied across domains including atmospheric muon corrections, radiative transfer, and cosmological observations by employing effective, brightness, and transport-weighted methods.
• Practical applications span calibrated forecasting, bias correction in lake temperature modeling, and instrument-response analyses, linking observed signals with theoretical temperature.

Searching arXiv for the cited works to ground the article in current records. Observational temperature denotes temperature as it is accessed through observation, calibration, and inference rather than as a directly seen microscopic attribute. In Mareš’s phenomenological analysis, temperature is constructed from thermoscopes, fixed points, and ordered “hotness levels”; in applied work, it appears as an effective atmospheric temperature for muon corrections, a brightness temperature in radiative transfer, a surface or scalar temperature derived from model fields and in situ sensors, or a thermal-bath parameter constrained indirectly by cosmological observables (Mares, 2016). This suggests that observational temperature is not a single object but a family of measurement-dependent constructs linked by reproducible procedures, equilibrium assumptions, and instrument-specific response functions.

1. Phenomenological foundation and scale construction

Mareš’s starting point is that temperature is not observed directly in microscopic motions. What is observed is a reproducible state of a body revealed by a thermoscope. In his formulation, a thermoscope is a homogeneous two-parameter system $(X,Y)$ with one conjugate variable fixed, $Y=Y_0$, and the other, $X$, serving as the thermoscopic variable under diathermic contact. The reading $X_k(P)$ belongs to the instrument, while the thermoscopic state $P\in H_k$ is attributed to the body. Thermal equilibrium is operationalized by diathermic partitions and constant conjugate variables under fixed external conditions, yielding an observational form of the zeroth-law idea (Mares, 2016).

This framework becomes quantitative by imposing order. Thermoscopic states in a given instrument range $H_k$ are ordered by the instrument reading, and the principle of indifference requires that different thermoscopes distinguish the same ordering in overlapping ranges. Fixed points $F$ then allow different thermoscopes to be “sewn together” into a single ordered hotness series,

$H=\bigcup_k H_k .$

An empirical temperature scale $\theta$ is an order-preserving one-to-one mapping of $H$ onto a segment of the rational numbers. At this stage, observational temperature is only an order scale: it encodes hotter and colder, but not yet ratios or physically meaningful differences (Mares, 2016).

To obtain a physical temperature, Mareš requires a ratio scale. Temperature scales that define a physical quantity must satisfy

$Y=Y_0$0

with a constant similarity factor $Y=Y_0$1. The two classical auxiliary constructions are the Carnot–Kelvin scale and the ideal-gas scale. Kelvin’s prescription $Y=Y_0$2 gives an absolute scale through reversible heat engines, while the ideal-gas thermometer gives

$Y=Y_0$3

Mareš emphasizes that, with suitable units, $Y=Y_0$4, so ideal-gas thermometry and the ideal heat-engine construction are equivalent routes to physical temperature. The modern Kelvin scale is then fixed by assigning the triple point of water the exact value $Y=Y_0$5 (Mares, 2016).

A central implication is that phenomenological and statistical temperature are not identical by definition. Statistical mechanics gives

$Y=Y_0$6

but this theoretical temperature coincides with phenomenological temperature only under specific assumptions: macroscopic equilibrium, valid thermodynamic description, and an appropriate thermometer-system coupling. Mareš treats that identification as empirically contingent rather than automatic (Mares, 2016).

2. Effective, environmental, and transport-weighted temperatures

In atmospheric muon observations, temperature is not the thermometer reading of the air column but a weighted quantity determined by how different atmospheric layers modulate the muon flux. For the Nagoya multidirectional muon telescope, the temperature effect is written in the integral method as

$Y=Y_0$7

where $Y=Y_0$8 are differential temperature coefficients. This is recast through an integral coefficient

$Y=Y_0$9

and an effective temperature $X$0, so that the observational correction becomes

$X$1

For Nagoya, the net ground-level temperature effect is overall negative, the vertical coefficient is about $X$2, the effective-temperature and mass-average methods differ by $X$3, and the annual temperature-driven modulation disappears after correction. The coefficient becomes more negative up to about $X$4 zenith angle and then less negative at $X$5, reflecting the increasing positive temperature component at higher energies (Berkova et al., 2017).

A closely related but different usage appears in observational site characterization. For Timau National Observatory, “surface temperature” is the ERA5-derived horizontal surface temperature at the site elevation, obtained by cubic interpolation in longitude, latitude, and height. Over 2002–2021, the median surface temperature is $X$6, the typical daily maximum–minimum difference is about $X$7, the median absolute hourly temperature change is about $X$8, and the annual mean trend is about $X$9. Here observational temperature is an environmental state variable for observatory design and operation, particularly for seeing, dome equilibration, and enclosure thermal control (Priyatikanto et al., 2024).

In heliospheric plasma measurements, the observed ion temperature is inferred from thermal speed rather than direct calorimetry. For Solar Orbiter O$X_k(P)$0,

$X_k(P)$1

and the measured scalar temperature from $X_k(P)$2 to $X_k(P)$3 AU is well fit by an adiabatic profile

$X_k(P)$4

with no significant heating over that range. In this setting, observational temperature is a transport-weighted kinetic parameter inferred from moments of an ion distribution function rather than from local thermometric equilibrium (Rivera et al., 12 Aug 2025).

3. Radiative, spectroscopic, and instrument-response temperatures

In radiative observations, the primary observable is intensity rather than temperature. For protoplanetary disks, the brightness temperature $X_k(P)$5 is defined by

$X_k(P)$6

with exact inversion

$X_k(P)$7

For an optically thick line in LTE, $X_k(P)$8 approximates the physical temperature of the $X_k(P)$9 layer. The paper on empirical temperature measurement in disks shows that the best estimator is the line peak emission map without subtracting continuum emission. Continuum subtraction systematically underestimates the gas temperature because the line can absorb the continuum; once beam dilution and noise are included, the peak line brightness temperature is within $P\in H_k$0–$P\in H_k$1 of the physical temperature of the emitting region, assuming optically thick emission (Weaver et al., 2018).

In the corona, the relevant observational temperature is an emission-weighted quantity filtered through a broad instrumental response. Synthetic EUV intensity is computed as

$P\in H_k$2

so the image samples a density-squared-weighted temperature distribution rather than the true maximum plasma temperature. In simulations of the coronal kink instability, parallel thermal conduction reduces the local peak temperature from about $P\in H_k$3 to about $P\in H_k$4, yet broad-band TRACE/AIA $P\in H_k$5 images change only modestly because the response functions span a broad temperature range. The principal observational effect is a blurring of internal structure, not a direct display of the true peak temperature (Botha et al., 2011).

Nebular spectroscopy provides another case in which different temperature diagnostics are not observationally equivalent. In the giant H II region H 1013, optical collisionally excited line ratios, Balmer-jump estimates, and Spitzer mid-infrared fine-structure lines probe different temperature weightings. Detailed Cloudy/pyCloudy modeling constrained by optical and mid-IR data yields weak temperature fluctuations with $P\in H_k$6 and a best-model oxygen abundance $P\in H_k$7, arguing against previously claimed $P\in H_k$8–$P\in H_k$9. At the same time, the model does not reproduce the oxygen recombination lines, implying that the RL–CEL discrepancy requires an explanation other than temperature fluctuations (Stasinska et al., 2013).

4. Temperature as a cosmological observable

The cosmic microwave background turns temperature into a directly constrained cosmological observable. In a variable dark-energy model with $H_k$0, radiation no longer obeys the standard adiabatic law trivially. The modified temperature–redshift relation derived in the paper is

$H_k$1

which reduces to $H_k$2 for $H_k$3 and $H_k$4. Using Sunyaev–Zel’dovich and quasar-absorption measurements up to $H_k$5, the temperature data alone give $H_k$6, corresponding to $H_k$7; combining with supernovae, BAO, CMB anisotropy, and $H_k$8 yields $H_k$9 and $F$0. Within uncertainties, the model is indistinguishable from a cosmological constant (Jetzer et al., 2011).

Warm inflation constrains temperature differently: the thermal bath is not observed directly but through its imprint on $F$1, $F$2, and $F$3. In the two-field model with

$F$4

Planck data restrict the allowed thermal regime. For effectively single-field behavior, only weak dissipation is allowed, $F$5; for mass ratios $F$6 and $F$7, the bounds relax only slightly; for $F$8, strong warm inflation with $F$9 remains viable. Observationally, this means that high $H=\bigcup_k H_k .$0 is strongly constrained in the single-field limit but can be compatible with Planck once multi-field dynamics and isocurvature transfer broaden the allowed $H=\bigcup_k H_k .$1 domain (Wang et al., 2019).

CMB temperature measurements are also vulnerable to observational systematics. In scan-based anisotropy mapping, the measured temperature after dipole subtraction can acquire a residual

$H=\bigcup_k H_k .$2

where $H=\bigcup_k H_k .$3 is an effective line-of-sight error encompassing pointing, sidelobe, dipole-direction, and timing errors. After full-sky scanning, these structured residuals can generate artificial large-scale anisotropies, including quadrupole and octopole patterns correlated with the scan strategy (Liu et al., 2011).

5. Observational temperature as target, truth, and state variable in forecasting

In forecasting problems, observational temperature is often the target variable against which models are calibrated. In Falling Creek Reservoir, in situ water temperature is observed at 10 depths from $H=\bigcup_k H_k .$4 to $H=\bigcup_k H_k .$5, treated as the field truth $H=\bigcup_k H_k .$6, and used to define a bias signal relative to a Gaussian-process surrogate of the General Lake Model: $H=\bigcup_k H_k .$7 A second GP models this discrepancy, and the resulting bias-corrected GP surrogate combines the process-based model with the observational record (Holthuijzen et al., 2024).

In the forecasting configuration, recent observational temperature also enters as a state descriptor: $H=\bigcup_k H_k .$8 This five-day average of observed $H=\bigcup_k H_k .$9 and $\theta$0 temperatures helps the forecast adapt to the current thermal state rather than relying only on climatology. The paper reports that the bias-corrected GP surrogate outperforms raw GLM, a climatological model, and other baselines in forecast accuracy and uncertainty quantification up to two weeks into the future (Holthuijzen et al., 2024).

This use case broadens the meaning of observational temperature. It is neither merely a thermoscopic reading nor solely a radiative proxy. It is a reference field, a calibration target, and an assimilated state variable whose uncertainty propagates through the forecast system.

6. Extended regimes, limits, and recurring controversies

A recurring theme is that observational temperature depends on admissible couplings and on the structure of the measuring device. Mechanical thermometer models make this explicit. For FPU-like chains, long mode-thermalization times do not prevent a thermometer from reading the temperature when it is coupled to microscopic degrees of freedom rather than to normal modes. For long-range interacting systems, the thermometer reads the microcanonical temperature rather than the canonical one in ensemble-inequivalent regions. For systems with absolute negative temperatures, a thermometer must have bounded total energy; an unbounded quadratic thermometer strongly perturbs the system and changes the sign of its temperature, whereas a bounded thermometer with kinetic term $\theta$1 can read $\theta$2 without destroying the state (Baldovin et al., 2017).

Another recurring issue is that observational temperature can be systematically biased by simplifying reductions. The Nagoya study finds that the empirical Duperier method gives noticeably worse corrections than effective-temperature or mass-average methods because the single-level assumption is too crude (Berkova et al., 2017). In protoplanetary disks, continuum subtraction removes genuine line flux and biases temperatures low when the line is optically thick (Weaver et al., 2018). In broad-band EUV imaging, brightness patterns cannot reveal the true peak plasma temperature because the response is broad and the intensity is weighted by $\theta$3 (Botha et al., 2011). In CMB mapping, the dipole-cleaning stage can imprint artificial anisotropy if the observational geometry is slightly wrong (Liu et al., 2011).

At the conceptual level, Mareš’s analysis keeps the strongest limitation in view: phenomenological temperature and statistical temperature coincide only under additional assumptions, and the identification remains incomplete and sometimes controversial outside ordinary macroscopic equilibrium (Mares, 2016). Observational temperature is therefore best understood not as a single universally given quantity, but as a rigorously constructed bridge between observable signals and thermal interpretation. The bridge may be direct, as in fixed-point thermometry; weighted, as in $\theta$4; radiative, as in $\theta$5; or inferential, as in data-assimilative forecasting. Across these cases, the central scientific task is the same: to specify the observable, the weighting, the equilibrium hypothesis, and the calibration that make a temperature claim meaningful.