| 08 May 2026
The missing drifts: Widespread and systematic underestimation of heterogeneity in mountain snow density and SWE across scales
Abstract. Near-real-time estimates of mountain snow water equivalent (SWE) are essential for streamflow forecasting and other applications, and snowpack heterogeneity controls summer water supply volumes, peak flow magnitudes, and other snowmelt runoff dynamics. Spatially resolved SWE quantification often depends on extrapolation from sparse regional measurements. When these measurements are not representative, the resulting spatial datasets are misleading. Here, we leverage complete density profiles from 36 snow pits (including eight at least 2 m deep) and repeat airborne lidar surveys spanning the Wind River Range (Wyoming, USA) to quantify missing heterogeneity in widely available snow density and SWE datasets at the point scale, grid scale (3 m to 1 km), and watershed scale. Our field measurements strategically constrain expected endmembers of snowpack variability, including rarely sampled snowpack zones such as deep drifts, steep slopes, avalanche runouts, and high elevations. Complete vertical profiling of a 5.9 m drift in a nivation hollow reveals a bulk density of 585 kg/m3, rivaling the density of avalanche debris and exceeding the predictions of prior empirical snow density models by 14 % to 35 %. In contrast, we contemporaneously observe bulk densities as low as 339 kg/m3 in adjacent forested areas. All 88 of the in-situ daily snow monitoring sites (SNOTEL) in the State of Wyoming are located in forested areas, which cannot account for alpine snowpack heterogeneity. This missing density and depth heterogeneity propagates into watershed-scale assessments. Across three large-domain near-real-time gridded SWE datasets, the standard deviation of SWE at the 500 m grid-scale is underestimated by 33 % to 75 % compared to our lidar-based data aggregated to the same resolution. These extrapolated SWE datasets underestimate the snowpack storage in a heavily drifted 1 km2 glacial cirque basin by 65 % to 73 % despite only 2 % to 17 % underestimation of watershed-average SWE, underscoring the problem of missing heterogeneity. Spatially resolved SWE quantification in mountain regions can benefit from lidar and strategic field density sampling to better account for dense wind drifts and other drivers of heterogeneity in near-real-time.
Competing interests: Author ENB is the owner of Mountain Hydrology LLC, which contracted for data acquisition and partially funded ENB. Author JWB has financial interests in Airborne Snow Observatories, Inc., which acquired the lidar data used here. Authors CAJ, SDS, JAW, and AAH received funding through a Mountain Hydrology LLC subaward to the University of Nevada, Reno.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
This study directly addresses what I would call the measurement and estimation problem: using 36 snow and repeat airborne lidar across the Wind River Range to show that widely used snow density and SWE datasets systematically underrepresent heterogeneity. A key finding is that complete profiling of a 5.9 m drift in a nivation hollow yields a bulk density of 585 kg/m³ — exceeding prior empirical bulk snow density model predictions by 14–35%. At the watershed scale, three near-real-time gridded SWE datasets underestimate total SWE storage in a heavily drifted 1 km² glacial cirque basin by 65–73%, despite underestimating watershed-average SWE by only 2–17%.
The same research group recently published two other papers dealing with snow heterogeneity in the Wind River Range, Wyoming. Compared to these papers, this study shows:
After reading the large preprint (42 pages without references) I was disappointed as there is unfortunately only little new scientific insight in this study. Based on the above three points, only the first is new, i.e. showing the density distribution in such a deep snow drift and providing an explanation for the vertical and lateral differences within the nivation hollow. Unfortunately, it is only one profile from one winter.
However, already the finding that empirical models of snow density underestimate the measured bulk density is not surprising, because, as the authors themselves write: “Many of the models tested are being used beyond their intended scope.”
Similarly, regarding the second of the above points. It is not surprising that most of current available products fail to reflect the spatial heterogeneity, because, as the authors themselves write “Gridded SWE products that do not explicitly consider snow transport processes are unlikely to capture the spatial variability”. However, they were trained to reflect watershed-average, where they perform mostly quite well.
The third of the above points is connected to the second point. The finding that SNOTEL sites are not representative on the watershed scale is not only valid for Wyoming. It is valid also for other regions and other in-situ networks and has been demonstrated already many times (e.g. doi.org/10.1002/hyp.6128, doi.org/10.1002/hyp.9355)
Given this fact (just one single new scientific finding), I believe that the study, in its current form, is not worthy of publication in TC. If at all, I strongly suggest changing the focus and the title so that the weight is on the density profile in the nivation hollow. I do not like the current title for two reasons: First, not only deep snow drifts are missing in current approaches, but probably also the opposite like ridges with bare ground. Second, it is not all new that most current approaches underestimate the heterogeneity of snow in mountainous terrain (e.g. doi.org/10.5194/hess-19-1339-2015)
Finally, the author totally misses mentioning the current literature, which tries to tackle the missing heterogeneity problem. Here is just a small set of examples…
doi.org/10.5194/tc-15-743-2021
doi.org/10.5194/tc-18-3533-2024
doi.org/10.3389/feart.2024.1393260
doi.org/10.3389/feart.2023.1308269
doi.org/10.1002/2016WR019872
doi.org/10.5194/tc-18-4315-2024
Here, some input to individual lines in the manuscript:
L21: Accumulation areas of typical mountain glaciers are often at locations which profit from snow drift. Mass balance measurements before the melt season frequently show densities around 600 kg/m3 (e.g. doi.org/10.5194/tc-13-3413-2019)
L40: Pfohl et al. in review: Could not be checked.
L291: There are more newer, more enhanced models available, which were developed based on the same data set (doi.org/10.1016/j.coldregions.2025.104435, doi.org/10.5194/hess-25-1165-2021).
L376-389: Belongs to introduction
L455: Fig.4: snow stratigraphy and snowpack temperatures would help a lot to explain density trends.
L501: “we there”?
L503: “lower density”. I strongly recommend to consistently use “bulk density” and reserve “mean density” for the average of bulk densities from different profiles.
L543: Fig: 6: snow stratigraphy would help a lot to explain density trends.
L545: “the other tree snow pits”. I strongly recommend introducing a table providing an indicator, snow depth, elevation, aspect, density, density-trend for each snow pit and provide the indicator in the figures.
L557: Snowpack temperatures: I wonder if it was always measured in the shade when shoveling took several hours?
L562: “The coldest temperatures in the Downs Mountain summit drift are near the ground surface”. That might have been due to permafrost?
L573: “the temperature profile is reversed compared to what we would expect for depth hoar formation”. Not necessarily, as the depth hoar formation typically happens during a period of several days where the temperature measurement is only valid at the time of the measurement.
L600: That also depends on the aspect of the forest (doi.org/10.5194/hess-30-1691-2026)
L764: “avalanche debris is the densest snow within the high-elevation Downs Mountain vicinity”. Just one sample?
L768: “Since snow near the surface of the drift exceeds the vertically integrated mean density”. This can easily be caused by wind- or melt-crust.
L791: “Tabler (2003) emphasizes the pressure of overlying snow”. Tabler developed is method for snow drifts along road fences and was never thinking about special cases like nivation hollows.
L799: “For the first time, our study quantifies the watershed-scale impact of ignoring wind-packing effects on snow density.” Snow density heterogeneity?
L814: The best model shows only an underestimation of 3%! It is in sect. 3.5 and not 3.4!
L882: Current progress in forest-snow is missing: doi.org/10.5194/hess-24-2545-2020, doi.org/10.5194/hess-27-2099-2023 or doi.org/10.1029/2020WR029064
L890: “overrepresentation of in-situ stations in forested regions”. This might be true for the SNOTEL network, but it is not true for other mountainous snow networks. Also, the opposite has been observed (doi.org/10.1002/hyp.10295).
L915: “colocation issues between the snow pillow (which measures SWE) and the snow depth sensor (which measures snow depth) could contribute to inconsistent density calculations”. Please provide any literature for this claim.
L938: “the two deepest snow pits (4.0 and 5.9 m) have the third- and fourth-highest representativeness scores.” Not shown - please provide table/figure.
L1022: “Airborne lidar surveys provide the key to accurate quantification of snow depth heterogeneity”. Right, but please also mention that they are expensive and therefore rarely available in other mountain regions.