the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Aug 2026
Explaining observed surface radiation trends 2001–2023 on Alaska's North Slope
Abstract. Changes in surface radiation are a key contributor to Arctic surface warming. However, the details of Arctic surface radiation change are not fully known due to a lack of multi-decadal observations. Here, we leverage 23 years (2001–2023) of ground-based observations at two neighboring coastal tundra sites on the North Slope of Alaska (NSA) to explain surface radiation change. US Department of Energy Atmospheric Radiation Measurement (ARM) and National Oceanic and Atmospheric Administration (NOAA) facilities in Utgiagvik, Alaska have documented warming (0.9 K/decade since 2001) and surface radiation change. Surface downwelling longwave radiation increases year-round. Surface downwelling and net shortwave radiation decrease during sunlit months (April–September). Decreasing net shortwave radiation is driven by changing cloud properties and increasing surface albedo over exposed tundra in mid-summer. Increasing downwelling longwave radiation during both sunlit and dark months (October–March) is driven by atmospheric warming, increasing water vapor, and changing cloud properties. Atmospheric warming and increasing water vapor explain about half the total observed increase in downwelling longwave radiation, while changing cloud properties explain the rest. Assessed cloud drivers (liquid water path, cloud cover, and low cloud fraction) explain increasing longwave radiation during sunlit months but are insufficient during dark months. Changes in assessed cloud drivers increased longwave radiation in all seasons except SON, when there was no change. These results reveal substantial surface radiative changes on the NSA, where cloud radiative effect changes amplify surface warming during dark months and dampen it during sunlit months.
- Preprint
(922 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 09 Oct 2026)
- RC1: 'Comment on egusphere-2026-4740', Anonymous Referee #1, 20 Sep 2026 reply
-
RC2: 'Comment on egusphere-2026-4740', Anonymous Referee #2, 05 Oct 2026
reply
Review of “Explaining observed surface radiation trends 2001-2023 on Alaska’s North Slope”
Overall:
The Arctic is changing faster than any other region of the planet. A budget of surface energy budget fluxes corresponds to and drive surface warming. Measuring and attributing the energy contributions to the surface energy budget changes is an important scientific endeavor that is undertaken in this study. The authors leverage 23 years of ground station data from the North Slope of Alaska from two sites (ARM and NOAA) to compute the annual cycle and multi-decadal trends in surface radiation budget fluxes and quantify the atmospheric and surface conditions. The authors find significant trends for during both the sunlit and dark seasons; however, the character and drivers of these trends differ. The paper describes these changes in excellent detail. I found the results and descriptions insightful and useful. There are certainly finding is in this work that warrant further investigation including the role of clouds and th eThis work is important and publishable after a minor revision focused on (1) adding a description of the dataset uncertainty and (2) clarity the description of the methodology.
Major Comments:
- The method reads robust and thoughtful. However, after reading it, I’m left with several questions about the methodological choice (see minor comments section below). Also, it seems that it would be challenging based on the description alone for someone to reproduce your results. Overall, the methodology section needs to be clarified including descriptions and justifications for the methodological choices. Many of the choices around the definition of CRE are stated in equations; however, the reasons that these choices need to be made are not explained. Some examples…Why perform the radiative sensitivity study in the form you did using kernels instead of directly computing radiative perturbations using a partial radiative perturbation approach. Why separate into cloud regimes and then never talk about them again in the analysis? What is the relationship between cloud cover and low clouds and how is this accounted for in the analysis.
- Why the lack of closure in LW down during polar night? Could the lack of closure be related to the methodological choice? Consider provide an estimate of the total amount of the change in LWDN that could be explained by a change in the ice water path. Could changes in cloud ice account for the ~3Wm-2 discrepancy? I’m skeptical because the cloud liquid water path changes only account for ~1Wm-2 in Fig. 9 (dark season). Provide more description of the lack of closure in the main text and in the discussion section.
- The work describes a lot of data products, and it is hard to include everything. However, it is critical to the interpretation of the results to include at least a sentence description of the uncertainty. The example I call out below is the pyrgeometer uncertainty and the contribution to trend uncertainty. However, the same is true for cloud variables, surface albedo, etc. Please include an estimate of the uncertainty for each data product these should be available in the dataset description document and/or peer reviewed publications. Description of the measurement uncertainty is needed for radiation and cloud variables.
- Please add references for each datasets, including DOIs to the reference list. This will be required before publication anyway.
Minor Comments:
Line 5: “Utqiagvik” is misspelled
Line 27: grammar comma needed before “which”
Line 100,107: In these sentences, you mention when “underlying data points are assessed to be of good quality.” Please describe the criteria and/or how this was decided.
Line 124: The NOAA flask samples/chemical analysis is missing from Table 1.
Line 136-137: You can “We create profiles…” it is unclear what profile you mean: clouds or temperature/humidity. Also, can you clarify what you mean by “When an individual radiosonde ends…” I think you mean the altitude and individual launch terminates at, but it is unclear. Please reword.
Line 142: Please comment on any concern about not including aerosol in the radiative transfer calculations.
Line 151: Please describe the empirical approach for low cloud fraction or point to where this is described in the text.
Line 164-165: It is stated that CRE_LWP distributions are estimated from…radiative transfer…with cloud cover set to zero. I’m don’t follow. If cloud cover is zero in the calculation, then there would be no radiative effect of LWP. Do you mean cloud cover “anomaly”?
Line 173: I follow that you are saying here about the “weights” however this is the first time that weights are brought up and you point back to (2) like they were mentioned. Adding a sentence when you introduce (2) to state how the weights (f_i) is used to create a monthly mean would help clarify the methodology.
Line 211-217: This is an interesting comparison. In addition, a contribution to the difference that not mentioned is the uncertainty in the surface radiation measurement. Error bars should be included and/or described. I would speculate that about half of the ARM-NOAA difference could be instrument uncertainty. Please comment.
Line 236-239: Here you start using oC instead of K. I don’t follow why you would switch units. If there isn’t a strong reason, I recommend using K throughout.
Line 215-217: I know this is out of order…I just read lines 244-250. These two sections do not seem to jive. You say that “the difference results from the earlier melt-out date at the NOAA site” when discussing the decadal differences. However, lines 244-250 state that there is no change in snow season timing. Please clarify.
Line 289-290: Can you make a comment on what this means or what the source of the discrepancy is?
Citation: https://doi.org/10.5194/egusphere-2026-4740-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 257 | 121 | 47 | 425 | 47 | 45 |
- HTML: 257
- PDF: 121
- XML: 47
- Total: 425
- BibTeX: 47
- EndNote: 45
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript investigates changes in the surface radiation budget at the North Slope of Alaska over the period 2001–2023, using long-term observations from the ARM and neighboring NOAA sites at Utqiagvik. The authors document increasing downwelling longwave radiation throughout the year and decreasing net shortwave radiation during the sunlit months. They then attribute these observed changes to changes in atmospheric temperature, water vapor, surface albedo, and several cloud properties. To do so, they combine long-term observational records with radiative transfer calculations and simplified radiative kernels. The main result is a strong seasonal contrast: during the sunlit season, changes in clouds and surface albedo increase shortwave cooling and largely offset the longwave effect of atmospheric warming and moistening, whereas during the dark season, cloud changes enhance the increase in downwelling longwave radiation.
Overall, this is a very interesting and valuable piece of work. A particular strength is the successful effort in bringing together the many different observational datasets needed to address this question. The combination of two independent long-term surface radiation records with observations of temperature, water vapor, surface albedo, cloud cover, liquid water path, and cloud-base height provides an unusually comprehensive observational basis for investigating changes in the Arctic surface radiation budget. The study goes beyond simply documenting trends and attempts to understand which observed changes in the atmosphere and at the surface can account for them.
My major concern is with the presentation of the methodology. After several readings of Sections 2.2 and 2.3, I still have only a somewhat vague understanding of how exactly the radiative anomalies are reconstructed from the observations and radiative-transfer calculations. I think the underlying approach is powerful, but the current description makes it difficult for the reader to understand and, consequently, to evaluate it.
The authors already walk the reader through the complete reconstruction procedure, but I miss some details on exactly how the calculations are performed and how you go from the individual radiosonde profiles and radiative transfer calculations to the monthly values in the end. A concrete example for one relatively simple driver, such as temperature, would also make the method more accessible.
Addressing these questions in the revised manuscript would, in my view, substantially improve the transparency and reproducibility of the analysis. I would then even say that this study is an excellent piece of work.
Specific comments:
lines 119-120: Cloud cover here refers to a temporal cloud occurrence, right? What is the temporal resolution? This actually holds for all data sets. What is the original temporal resolution, and how did you process/sample the data during the process?
How homogeneous are the data products over time? E.g., what about changes in biases in the MWR products due to different instruments operated, changing biases over time due to calibrations,…? How does this affect the identified trends? Have there been any changes in the operation of the instruments over time?
Does rain pose a problem for the MWR data? Since a shift from snow to rain is likely, would that also introduce uncertainties in IWV and LWP trend estimates (because wetting of the MWR radome might yield unreliable data more frequently)?
Section 2.2:
“We then perturb a single driver for every valid profile in RRTM”…
What is meant by a valid profile? What is the baseline state that you perturb? The original radiosonde profile? Or do you calculate a mean state from all radiosonde profiles?
How is it perturbed? And what is meant by anomaly? In reference to what?
Do you get a K for each radiosonde profile? How do you derive monthly values/sensitivities in the end?
Are these Ks only calculated for the first 11 years? How do you know which sensitivity values (K) has to be used in the other years?
Can you explicitly state the process: starting from one radiosonde and using one variable as an example?
Section 2.3
I am also somehow lost in the formalism here. This whole section is so confusing, even after reading it several times. I would simply ask you to rewrite that part and explain it in simple terms. E.g., I stumble over “driver bin k”, “normalized histogram of driver x by bin”
“Cloud properties are treated differently….” and the following part. I am sorry, but you lost me completely from here on.