| 11 Nov 2025
Fabrication, Calibration, and Deployment of a Custom-Built Radiometer
Abstract. A custom-built spectroradiometric system was developed to measure spectral actinic flux for atmospheric photochemistry. The radiometer incorporates a UV-enhanced fibre optic cable, a compact CCD spectrograph, and an interchangeable polytetrafluoroethylene (PTFE) receiver head. Five receiver designs (domes, cones, and flat tops) were fabricated and evaluated for angular response and wavelength-dependent efficiency using collimated UV and visible LEDs. The variable cone design showed the most uniform angular sensitivity, minimizing the need for post-processing corrections. Wavelength calibration was conducted using emission LEDs from 265 to 523 nm, and the full system response was characterized from 196 to 888 nm. The system was evaluated in a controlled atmospheric chamber and in outdoor field conditions. Under clear skies, outdoor testing validated the radiometer’s effectiveness in measuring photon fluxes with an 11.2 % uncertainty. These photon fluxes were used to calculate photolysis rate constants (J-values) with an uncertainty of 15.3 % for both NO2 and O3. Although not fully cosine-corrected, the radiometer demonstrated sufficient stability and repeatability for deployment in research settings, particularly where commercial instruments are unavailable. This study establishes a low-cost, customizable tool for actinic flux measurements in support of atmospheric photochemistry.
This work tested multiple actinic flux optical designs based on commercially accessible materials for field use. Actinic flux optics are generally in short supply. Thus, the efforts in this work to produce low-cost actinic flux optics would greatly benefit the scientific community. Unfortunately, the paper lacks sufficient detail to assess the designs and quality of the resulting measurements. Most importantly:
The work requires major corrections and additional analysis to be considered for publication. However, I do encourage these additional efforts to examine and further improve the actinic flux optical design.
Major comments
Line 25: Here is one more formative, peer-reviewed reference for photochemical rates:
Line 33: Add Bohn and Lohse, 2023.
Lines 38-41: These references do not apply to “instrumentation for actinic flux”. They are radiance and irradiances measurements and not directly relevant to photochemistry. The irradiance measurements have a cosine response which falls to zero at the horizon. This does not represent the full 4 pi steradian (actinic flux) response of molecules to the light environment. Also, those referenced instruments are not generally optimized for the UV and visible light that drives photochemistry. Perhaps a careful read of Madronich, 1987 (not the same 1987 article referenced in this article) could help with the apparent disconnect between irradiance and photochemistry.
Here are suggested references for double-monochromators with PMTs for actinic flux
For diode arrays:
For CCDs
Line 44: Bohn and Lohse 2023 do not provide “cosine error corrections” because they are not using cosine weighted (irradiance) optics. The introduction in this work correctly notes the complex optical corrections to achieve the “isotropic angular response across all zenith and azimuth angles” for actinic flux.
Line 54: Again, these references do not describe photochemically relevant instruments. The instrument papers noted above in this review include more appropriate angular calibration techniques (e.g. Fig 2 of Shetter and Mueller. 1999)
Line 83 and in Figures 2 and 3: The angle range of 0 – 180 deg is traditionally aligned to be be -90 to 90. That is, 0 deg is the zenith position, directly above the optic. This relates back to “cosine” optics. These irradiance optics collect overhead light fully (i.e. cos(0 deg) = 1) and collect no light at the horizon (i.e., cos(90 deg)=0). Actinic flux optics follow the same tradition but ideally collect light fully at all angles from -90 to 90 degrees. See instrument references noted above (e.g. Jakel, Hofzumahaus, Shetter, etc).
Line 89 More details about the power meter are required, including model number, calibration details, etc.
Line 100: Humidity and breeze do not directly affect radiometric measurements in the UV/VIS range. They do indirectly affect cloud formation and dust aerosols, respectively.
Line 107: Was the albedo really set at 0.9999 or is that a typo. That is brighter than fresh snow. Perhaps 0.09999 was used?
Line 109-10 Perhaps I missed it but I do not see the model outputs in the paper or data output at the BYU Scholar archive.
Line 116-7: I would not expect the 265 and 523 nm LEDs to show the same response because the scattering efficiency of the material will have a wavelength dependence. Perhaps a plot of the response as a function of wavelength for one of the optics would be helpful.
Line 117: There are seven led lamps used in the calibration. Where only six used for the angular response calibrations?
Fig 2: The angular response of the optics is not addressed properly. I presume the “phase angle” is the angular response of the optics to a point source. If so, none of the optics have a flat actinic flux response across the hemisphere (0-180 deg). The response shown is neither a square wave (for actinic flux), or a cosine response (for irradiance).
The authors mention they could perform a “cosine” correction this is not the correct terminology. A cosine correction accounts for deviations from a cosine weighted angular response, as used for irradiance. Actinic flux optical corrections aim to approximate a flat hemispherical response. As shown, these optics have significant diurnal biases that could strongly impact the daily measurement cycles and introduce biases into subsequent chemical analyses.
Finally, more discussion would be helpful to understand the differences between the optical designs. What could be done to further improve the design? What is the response beyond the 180 degree range of this graph?
Lines 121-122: The discussion of the <10% asymmetry of the 90 deg azimuthal rotation would seem to be insufficient as each of the receivers clearly show significant azimuthal asymmetries between 0 and 180 deg. For example, the variable cone design has a relative response of ~58% at 0 degrees and 35% at 180 deg. Additionally, the peak of the Uni Dome appears to be ~10 degrees offset from the zenith position at 90 deg. This needs to be explained and I recommend showing the results at 90 deg (and perhaps additional angles) for at least one optic.
Figs 2 and 3: Grid lines at 0, 90 and 180 deg should be included as key angles in this analysis. (As noted above, I would first rotate the coordinates to -90, 0, 90 deg).
Figs 2-4: Define and discuss the marker uncertainties
Line 122: Actinic flux does not have a cosine response. Please review the angular response sections in Hofzumahaus, 1999, Jakel, 2005 and Shetter and Mueller 1999. They all show angular response curves with relatively flat responses out to ~85 deg, as needed for actinic flux. A cosine response is limiting for actinic flux because it underrepresents the area close to the horizon. If aiming for a cosine response, how do you convert to actinic flux? If aiming for a flat actinic flux response, more discussion is needed to assess what could be done to create a more uniform actinic flux response across the hemisphere (0-180 deg).
Fig 4: Define and describe the red markers and error bars for the data collected.
Lines 136-40: The stray light assumptions are not correct and the claim of insignificant stray light from the low-cost spectrometer in this experiment is highly improbable. Stray light correction is vital for atmospheric single monochromators measurements in the UV-B. The issue lies in the 4-5 orders of magnitude difference in light intensity between 300 and 400 nm. A fraction of a percent of the visible light scattered inside the spectrometer (stray light) and hitting the 300 nm pixel easily overwhelms the signal. Thus, careful UV-B stray light characterization is necessary to measure actinic flux and calculate UV-B sensitive photolysis frequencies, particularly jO3.
Jakel et al., 2007 provides an analysis technique using a series of long pass filters that demonstrate the influence of broad spectral regions on UV-B signals. Unfortunately, this work relied on thin, isolated mercury lines that are insufficient to demonstrate the effect. I am unclear if the current analysis even includes a dark correction to remove the electronic offset inherent in CCD spectrometers. The only clue is in the data files that log the “Electric dark correction enabled” is labeled “true”.
Unless convincingly proven otherwise, this small spectrometer should be presumed to have a significant or overwhelming stray light effect in the UV-B. I would strongly recommend foregoing UV-B and jO3 analysis with this spectrometer without a full consideration of stray light and dark current. Unfortunately, the TUV comparison for jO3 does not alleviate the concerns (see comments below). Again, direct comparison with established instrumentation would definitively determine the accuracy and appropriateness of this system for jO3 measurements.
Fig 4 and line 143: What fitting was used to generate this line and what is the reasoning behind a 10% uncertainty?
Section 3.2: The uncertainty analysis neglects the uncertainties inherent stray light or angular response (“cosine”) corrections. Admittedly, many of the past analyses in the literature have not fully considered these factors, but they should be noted here.
Also, the cross-section and quantum yield uncertainties are better described as biases and not correctly added in quadrature. Fortunately, both measurement and model are using the same cross-sections and quantum yields to calculate photolysis frequencies (this should be noted here and then the references do not need to be repeated in sections 3.2 and 3.4). Thus, these biases do not affect the comparison to TUV.
Section 3.3: I am not certain of the scientific or technical value of this section. Please further explain the purpose of the graph which only demonstrates the measured mercury lamp spectrum through an FEP chamber film. The chamber may indeed be “sufficiently strong for photochemical studies” but I am not sure of the contribution to this work.
Line 148: The atmospheric UV-B signal is significantly lower than UV-A and thus I would expect a higher uncertainty for jO3 (even neglecting stray light). This is not accounted for when using strong UV-B LED lamps for calibration.
Section 3.4 The duration of the field study is insufficient. To properly assess the optics requires multiple full sampling days under varying atmospheric conditions. However, the field data only covers ~6 hours near mid-day under relatively clear sky conditions. While the data is shown to be in rough agreement with the clear-sky model, this comparison is too simplistic. For one, the angular response is relatively flat at the low solar zenith angles during the measurements. The dataset should include low sun angles where optical design is most critical. For another, cloud and aerosol conditions change the direct solar beam and diffuse fraction of light and the optic needs to represent all conditions. Finally, if these new optics are to be considered as a cheap alternative to established optics, they should be directly intercompared (though that could be deferred to a later study).
Fig 6a: The spectra do not resemble surface atmospheric spectra above ~500 nm. The strong peak centered around 550 nm and a very low value at 900nm is not expected in either surface actinic flux or irradiance spectra. To demonstrate, I compared this data with TUV model (out to 700 nm). The model shows approximately 10% drop in signal from 500 to 700 nm for actinic flux and ~25% drop for irradiance. The corresponding drop in this figure is much larger at ~50%.
The spectral comparison to the extraterrestrial flux (ASTM G-173 AM 1.5 G) is noted in the text for the unit change wavelength offset but does not account for the unusual surface atmospheric shape. Perhaps this is a calibration issue at higher wavelengths. A calibration with NIST-traceable sources or intercomparison with an established actinic flux instrument would help determine if the LED light calibration is sufficient across the spectral range.
I recommend replacing this plot with a single spectrum in comparison with TUV.
Line 166: What is the value of the total summed flux from the spectrometer at one moment in time. This is only a function of the instrument wavelength range and not of atmospheric or photochemical relevance.
Line 180: AtChem is mentioned but why are no AtChem results shown? This should be expanded or removed.
Lines 180-1: Where is the factor of 3? A factor of 3 is quite large if the uncertainty is only 15.3% for the measurements. What is the reason for this discrepancy?
Lines 183-4: Perhaps the morning measurement is too low because of the uncorrected angular response.
Lines 186-7: The cross-sections and quantum yields in the model and measurement are described as identical in the paper. Perhaps the authors mean that the implantation of the cross-sections and quantum yields in the data and in the model may have slight discrepancies.
Fig 6B: This 311 nm actinic flux plot would benefit from a TUV model comparison to help address the jO3 questions below. Perhaps add an additional wavelength to represent jNO2 (e.g. 380 nm).
Fig 6C and 6D. I attempted to reproduce the TUV model results using the parameters specified in Section 2.4. However, this was complicated by two factors.
I was able to approximately calculate the same jNO2 in Fig 6a if I assumed a total actinic flux output. The downwelling-only would reduce the values by ~10%.
I cannot, with any combination of factors, reproduce the jO3 shown in Fig 6b. Applying the same parameters used to match total jNO2, the corresponding total modeled jO3 is a factor of ~3 higher than shown in the figure. I suggest the authors revisit this analysis and clearly define all parameters used in the calculations.
In addition, the figure includes the uncertainty from the cross-sections and quantum yield uncertainties in the measurements. Identical references were noted in both measurement and model so they are not relevant to the measurement/model comparison. The more fundamental comparison is between the measured and modeled actinic flux where the uncertainty is not tied to these molecular parameters. As mentioned above, a spectral comparison and a plot of individual wavelengths vs time would be revealing.
Line 212: What is the relevance of “wavelength range” and “low saturation levels” to stray light?
Line 214: This paper describes spectrally-resolved spectrometer measurements. They are not “broadband” measurements that average over a spectral range. Perhaps this refers only to the optics. They could indeed be used in broadband instruments and stability would be crucial.
Technical corrections
Line 35: Typo on references, “Bohn et al.(Jäkel et al., 2007)” . Note, the two referenced Bohn papers and Jakel et al., 2007 do not mention cosine corrections as each study used 2 pi steradian optics that do not require such corrections.
Line 49: What does “Quality, 2021, 2023” refer to in the references?