the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Jan 2026
Proposed improvement of the detection and measurements of light precipitation in the Canadian Arctic
Abstract. Snowfall during the extended cold season experienced in Arctic regions is the primary contributor to snowpack evolution, terrestrial components of the water cycle, and many melt-season hydrologic phenomena. Despite this importance, solid precipitation measurements in the Arctic are challenging; frequent periods of light precipitation are often difficult to measure with existing gauge networks, and result in under-estimations of total snowfall during a winter season. This study analyzes the measurement of solid precipitation at the Trail Valley Creek Research Station in the Canadian Northwest Territories, using a weighing precipitation gauge, and micro rain radar. The study period runs from 4 November 2023 to 30 April 2024, with an intensive observation period from 16 March to 2 April 2024, during which detailed manual observations improved our understanding of instrument performance in arctic conditions. The already established weighing gauge was used as the reference for the study and measured a total snowfall (snow water equivalent) during the study period of 68 mm, which increased to 190 mm after corrections for wind and snowfall intensity. Manual observations coupled with radar, however, confirm the difficulty of measuring light precipitation. We present a method of on-site calibration for the reflectivity-snowfall relationship for the micro rain radar, that we use to estimate the low-rate (< 0.2 mm hr-1) snowfall amounts that are commonly missed by weighing gauges. Adding these trace amounts of precipitation, the total snowfall amount increased by another 24 %. While more work is required to confirm these methods in Arctic environments, this study contributes to a better understanding of current measurement systems and can be used to enhance snowfall estimations.
- Preprint
(1544 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- CC1: 'Comment on egusphere-2025-5195', Craig Smith, 05 Feb 2026
- RC1: 'Comment on egusphere-2025-5195', Anonymous Referee #1, 08 Mar 2026
-
RC2: 'Review of “Proposed improvement of the detection and measurements of light precipitation in the Canadian Arctic” by Durat et al.', Maximilian Maahn, 16 Mar 2026
The authors propose a novel method to quantify the amount of snowfall that is missed by traditional precipitation gauges. Measuring snowfall accurately is a long standing, important topic particular in polar regions, the subject of the manuscript is therefore highly relevant and well suited for HESS. In general, the paper is well written, the figures are clear. I do have, however, a couple of comments that I recommend being addressed.
Major Comments
Also light precipitation should eventually accumulate inside the Geonor and eventually trigger the sensitivity threshold. Can the authors comment on why they think this does not allow the Geonor to measure also light precipitation (even though delayed)?
The authors mention they heated the MRR dish. However, we found in https://doi.org/10/gb8r75 that the MRR heating is not always sufficient and half-melted snow can lead to problems due to signal attenuation. In subsequent studies (e.g. https://doi.org/10/w4b ), we therefore deactivated the MRR heating at colder temperatures because dry snow does not attenuate the signal significantly (in contrast to wet snow) and is usually blown away by the wind. For the light snowfall shown in this study, half melted snow is certainly not a problem. But I wonder whether this caused problems when deriving the fit coefficients for the Z-S relation. Did the authors observe any problems with partly melted snow during the IOP when they were at the site?
In particular light snowfall can occur during sublimation of snowfall in the atmosphere before reaching the ground. Fig. 7 shows a nice example and Fig. 6a shows that humidity was typically not 100% during snowfall. Therefore, I am concerned that this could bias the Z-S relation estimates but also overestimate the light snowfall even though the lowest altitude was only 75 m. To correct for this, I would recommend looking at the Ze gradient close to the surface, which is quite often surprisingly linear. Then, Ze can be extrapolated to the surface using the identified gradient form a linear fit. Of course, quality checks need to be employed to make sure the linear fit worked.
I would recommend adding a scatterplot showing the data of that fit. Surface wind speed could be added as a color to assess whether wind undercatch is properly corrected. It might be even an option to include only cases with low surface wind speed to limit the impact of wind undercatch?
As the authors stated, the wind undercatch correction is rather uncertain. How does this uncertainty compare to the uncertainty of the Z-S relation? I wonder what the benefit is of combining gauge and radar data instead of relying only on the radar data after developing a site-specific Z-S relation? Shouldn't this be better because it avoids combining to instruments with different error characteristics?
Minor Comments
L31f: These two references are an odd choice for such a general statement. I would recommend to either replace them or to omit the reference for the first sentence.
L47: Snow -> Snowfall
L84: The authors mention there was also an optical disdrometer and a hotplate deployed, what make and model was used? And at least the optical disdrometer should be able to pick up light precipitation. The authors mention technical problems, but can the optical disdrometer at least be used for a case study?
L86,L261: Fix Reference
Fig1: Does the box in the small map really show the same region as the big one? To me, it looks like the small one includes the coastal region which the big one does not show?
L137: Was the Geonor measuring at 2 m height?
L151: Nice that this finally is tested in the field. When I approached the manufacturer METEK with that idea years ago, they were concerned that the memory (ROM) of the MRR is only specified for a certain number of write accesses, and it could lead to a broken MRR memory within a year or so when switching resolution that often. Can the authors comment on that? Was that overly cautious on METEK’s side or has that been resolved differently?
L242: These are odd values for dBZ, is it possible the authors reported reflectivity in mm6/m3?
Fig.6: I would add a legend with the shading to the Figure
L372: Regarding the data availability, could the data publication be accelerated so it is available when the paper is published?
Citation: https://doi.org/10.5194/egusphere-2025-5195-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 240 | 68 | 25 | 333 | 14 | 16 |
- HTML: 240
- PDF: 68
- XML: 25
- Total: 333
- BibTeX: 14
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This paper highlights some much needed research on the impact of precipitation gauge measurement bias as it relates to small but significant solid precipitation amounts in cold regions and speaks for the need for techniques to augment conventional gauge measurements of solid precipitation in harsh environments.
I do have a comment about the statement made on Page 10 (lines 230-233) about Geonor gauge sensitivity. If interval precipitation is being calculated as the bucket weight differential between time T and time T-1, you are correct to assume that precipitation collected in the bucket that is less than the sensitivity of the gauge will not be reported at time T. However, unless evaporation occurs, that unreported precipitation that has accumulated in the bucket will likely register as an increase in bucket weight in a later interval, usually following more precipitation. The mass is conserved. This means that the timing of the precipitation is impacted, but not the long term total, so those small accumulations are ultimately shown in your accumulated time series in Figure 5. As you point out, many small amounts still go unreported because of wind undercatch (snow that never falls into the bucket in the first place) but this is unrelated to gauge sensitivity.
Also, if timing of precipitation is critical for comparison with the MRR, your raw bucket weight filtering technique may also impact the distribution of measured precipitation between adjacent intervals. Depending on the severity of the noise in your raw data and the effectiveness of your noise filter, this may have a more significant impact on interval totals than gauge sensitivity and should probably be mentioned as a source of uncertainty in your comparisons.