| 27 Aug 2026
Radiative Constraints on VIIRS Nighttime Detectability of Lower-Tropospheric Methane
Abstract. This study investigates the feasibility of detecting nighttime methane (CH₄) plumes in the lower troposphere using operational thermal infrared (TIR) imagery from the VIIRS sensor M12 midwave IR (MWIR) channel (3.7 µm), through a comprehensive radiative transfer sensitivity analysis using MODTRAN6 anchored to a real atmospheric profile. Simulations indicate that the atmosphere in the M12 band is optically thick (τ_eff ≈ 6.81) at background methane concentrations, and the detectable signal is therefore driven by the plume's thermal emission (path radiance) rather than surface absorption. We find that while plumes confined to the shallow nocturnal boundary layer produce a negligible signal regardless of concentration, plumes extending into the deeper residual layer yield substantially larger brightness temperature perturbations. At the native pixel scale, theoretical simulations indicate that detection (SNR ≥ 2) is probable only under an extreme super-emitter scenario at a 9× baseline concentration (~800 % enhancement). Incorporating observationally constrained surface emissivity retrievals from the Combined ASTER and MODIS Emissivity over Land (CAMEL) dataset at 3.6 µm into the uncertainty budget propagated a surface-derived brightness temperature uncertainty of σemis = 0.217 K, exceeding the instrument noise-equivalent temperature difference (NEΔT) and establishing an irreducible noise floor independent of window size. Thus, surface emissivity heterogeneity, rather than instrument noise, is the dominant limitation on detectability, and spatial averaging provides only limited benefit in mixed-covered scenes.
1. General Comments
The remote detection of methane leaks or releases during the nighttime is an important topic, as the majority of airborne and space-based sensors operate in the SWIR and therefore require solar illumination. Hence we have very limited ability for nighttime methane measurement. This is an unfortunate gap not just because methane emissions are episodic and we are not looking more than half the time, but also because the lower boundary layer and lower turbulent mixing significantly changes dispersal dynamics. The MWIR (~3.3-3.4um) methane absorption band for remote sensing is not widely studied, and the competing effects of strong absorption but low ground thermal radiance (relative to LWIR) make its utility unclear. So the authors' investigation of whether an existing space asset (VIIRS), specifically the M12 filter band which touches only the low-line-strength edge of the band, could provide utility, even though the result is essentially a negative one, is a useful contribution to the community. Incorporating measured CH4 profiles and satellite-derived ground scene variability adds important realism to the modeling.
In addition to references already included, the article comparing MWIR and LWIR methane detection from an airborne hyperspectral system (SEBASS), Scafutto and Filho, Remote Sensing 10, p.1237 (2018) could be added for background.
2. Specific Comments
2.1: The authors' approach of using the vetted and established MODTRAN platform, which solves the multiple-scattering problem, to do the radiative transfer is reasonable and provides a trustworthy result that can be repeated/verified by others in the community. However, it also draws attention to one limitation of the paper, which is the failure to connect the methane "enhancement factor" relative to background levels to a broadly understood release rate (kg/hr, etc.). While the authors caveat (lines 164-165) that "the perturbations are intended to approximate super-emitter-like conditions, rather than to convert the EPA threshold directly into a column scaling factor," the submission would be strengthened if a stronger connection could be made between the enhancement factors modeled and the range of release rates that might yield the same absorption.
The authors label an enhancement factor of 2 (that is, twice the background level up to boundary layer of either 350m or 1,850m) as representing the bottom of the "super-emitter" regime, which is generally defined as 100 kg/hr. This assertion should be supported or explained further. Further explanation is requested because the corresponding release rates may be much higher than what is implied. First, the reviewer accepts that in the small-signal limit, a large quantity of methane, such as that caused by a release plume, that fills a volume much smaller than the sensor pixel IFOV, will result in the same signal that the same quantity of gas spread out uniformly throughout the pixel volume would. Because of this, the approach of using MODTRAN, which treats every layer as a uniform concentration, is valid even in the case where complete mixing has not had time to occur, such as from a point release. However, it may take a lot of release to get to the enhancement factors the authors model. As an example, consider a 100 kg/hr release (the lower limit of the "super-emitter" definition) in a 1 meter/s wind. In this low-wind case, even with some mixing, much of the release would clear the 750-meter pixel in at most 750 seconds, depositing just 21 kg of methane into the pixel volume in steady state. 21 extra kg of methane in the 350-meter high pixel (NBL case) represents a change in average concentration across the entire pixel of just 162 ppbv, or just 31 ppbv across the taller 1,850-meter pixel volume (RL case). The authors' lowest labeled super-emitter enhancement value of 100% represents a concentration increases around 1,800 ppbv, more than an order of magnitude higher than this 100 kg/hr case. In the limiting case of near-zero wind, and very strong turbulent mixing and advection, the same 100 kg/hr of release would of course stay in the single pixel volume much longer, increasing the change in average concentration. How realistic is this case? (The reviewer is not an expert here.)
A detailed modeling of steady-state point-release plume dispersal under realistic atmospheric conditions, identifying what is typical along the spectrum of low-wind/strong mixing to strong wind/lower mixing, is likely outside of the scope of the article. However, because a back of the envelope calculation suggests that even a relatively low-wind case may result in concentration enhancements much lower than those considered, some effort to connect the regime modeled to release rate is warranted, ideally by introducing relevant references and presenting some discussion of atmospheric dynamics that would allow connecting the release rate to accumulation of gas in the pixel volume. As the overall conclusions are essentially negative (that the VIIRS sensor cannot detect or can only just barely detect the lower tropospheric concentration enhancements modeled), this addition would only strengthen that conclusion.
2.2: The introduction to the discussion of spatial averaging from lines 280 – 308 is confusing, and could be clearer if tightened up. Essentially, the authors highlight that averaging multiple pixel measurements together will reduce the sensor noise by the square root of the number of pixels, but that this will add in quadrature with the ground emissivity variability. Then, they show from satellite measurements that it the ground emissivity variability is indeed the limiting variance, an important result. Table 4 and the caption are mostly sufficient to explain this. Absent from the discussion of spatial averaging is that the single pixel IFOV of 750 meter is already a large pixel relative to a typical point emission plume, and that averaging multiple pixels will reduce the measured concentration if the plume does not fill the entire larger effective pixel.
3. Technical Corrections
Line 66 "confounding" typo.