Comparison of different automatic methods for estimating snow water equivalent

<strong class="journal-contentHeaderColor">Abstract.</strong> An airborne gamma-ray remote sensing technique provides a strong potential to estimate reliable snow water equivalent (SWE) in forested environments where typical remote sensing techniques have large uncertainties. This study explores the utility of assimilating the temporally (up to four measurements during a winter period) and spatially sparse airborne gamma SWE observations into a land surface model to improve SWE estimates in forested areas in the northeastern U.S. Here, we demonstrate that the airborne gamma SWE observations add value to the SWE estimates from the Noah land surface model with multiple parameterization options (Noah-MP) via assimilation despite the limited number of the measurements. Improvements are witnessed during the snow accumulation period while reduced skills are seen during the snow melting period. The efficacy of the gamma data is greater for areas with lower vegetation cover fraction and topographic heterogeneity ranges, and it is still effective in reducing the SWE estimation errors for areas with higher topographic heterogeneity. The gamma SWE data assimilation (DA) also shows a potential of extending the impact of flight line-based measurements to adjacent areas without observations by employing a localization approach. The localized DA reduces the modeled SWE estimation errors for adjacent grid cells up to 32-km distances from the flight lines. The enhanced performance of the gamma SWE DA is evident when the results are compared to those from assimilating the existing satellite-based SWE retrievals from the Advanced Microwave Scanning Radiometer 2 (AMSR2) for the same locations and time periods. Although there is still room for improvement, particularly for the melting period, this study shows that the gamma SWE DA is a promising method to improve the SWE estimates in forested areas.

An airborne gamma-ray remote sensing technique provides a strong potential to estimate reliable snow water equivalent (SWE) in forested environments where typical remote sensing techniques have large uncertainties. This study explores the utility of assimilating the temporally (up to four measurements during a winter period) and spatially sparse airborne gamma SWE observations into a land surface model to improve SWE estimates in forested areas in the northeastern U.S. Here, we demonstrate that the airborne gamma SWE observations add value to the SWE estimates from the Noah land surface model with multiple parameterization options (Noah-MP) via assimilation despite the limited number of the measurements. Improvements are witnessed during the snow accumulation period while reduced skills are seen during the snow melting period. The efficacy of the gamma data is greater for areas with lower vegetation cover fraction and topographic heterogeneity ranges, and it is still effective in reducing the SWE estimation errors for areas with higher topographic heterogeneity. The gamma SWE data assimilation (DA) also shows a potential of extending the impact of flight line-based measurements to adjacent areas without observations by employing a localization approach. The localized DA reduces the modeled SWE estimation errors for adjacent grid cells up to 32-km distances from the flight lines. The enhanced performance of the gamma SWE DA is evident when the results are compared to those from assimilating the existing satellite-based SWE retrievals from the Advanced Microwave Scanning Radiometer 2 (AMSR2) for the same locations and time periods. Although there is still room for improvement, particularly for the melting period, this study shows that the gamma SWE DA is a promising method to improve the SWE estimates in forested areas.

Area-based measurements of snow water equivalent (SWE) are important for understanding earth system processes such as glacier mass balance, winter hydrological storage in drainage basins and ground thermal regimes. Remote sensing techniques are ideally suited for wide-scale area-based mapping with the most commonly used technique to measure SWE being passive-microwave, which is limited to coarse spatial resolutions of 25 km or greater, and to areas without significant topographic variation. Passive-microwave also has a negative bias for large SWE. Repeat-pass synthetic aperture radar interferometry (InSAR) as an alternate technique allows measurement of SWE change at much higher spatial resolution. However, it has not been widely adopted because: (1) the phase unwrapping problem has not been robustly addressed, especially for interferograms with poor coherence and; (2) SWE change maps scaled directly from repeat-pass interferograms are not an absolute measurement but contain unknown offsets for each contiguous coherent area. We develop and test a novel method for repeat-pass InSAR based dry-snow SWE estimation that exploits the sensitivity of the dry-snow refraction-induced InSAR phase to topographic variations. The method robustly estimates absolute SWE change at spatial resolutions of < 1 km, without the need for phase unwrapping. We derive a quantitative signal model for this new SWE change estimator and identify the relevant sources of bias. The method is demonstrated using both simulated SWE distributions and a 9-year RADARSAT-2 spotlight-mode dataset near Inuvik, NWT, Canada. SWE results are compared to in situ snow survey measurements and estimates from ERA5 reanalysis. Our method performs well in high-relief areas and in areas with high SWE (> 150 mm), thus providing complementary coverage to other passive- and active-microwave based SWE estimation methods. Further, our method has the advantage of requiring only a single wavelength band and thus can utilize existing spaceborne synthetic aperture radar systems. In application, a first order analysis of SWE trends within three drainage basins suggests that differences between basin-level accumulations are a function of major landcover types, and that re-vegetation following a forest-tundra fire that occurred over 50 years ago continues to affect the spatial distribution of SWE accumulation in the study area.

Area-based measurements of snow water equivalent (SWE) are important for understanding earth system processes such as glacier mass balance, winter hydrological storage in drainage basins and ground thermal regimes. Remote sensing techniques are ideally suited for wide-scale area-based mapping with the most commonly used technique to measure SWE being passive-microwave, which is limited to coarse spatial resolutions of 25 km or greater, and to areas without significant topographic variation. Passive-microwave also has a negative bias for large SWE. Repeat-pass synthetic aperture radar interferometry (InSAR) as an alternate technique allows measurement of SWE change at much higher spatial resolution. However, it has not been widely adopted because: (1) the phase unwrapping problem has not been robustly addressed, especially for interferograms with poor coherence and; (2) SWE change maps scaled directly from repeat-pass interferograms are not an absolute measurement but contain unknown offsets for each contiguous coherent area. We develop and test a novel method for repeat-pass InSAR based dry-snow SWE estimation that exploits the sensitivity of the dry-snow refraction-induced InSAR phase to topographic variations. The method robustly estimates absolute SWE change at spatial resolutions of < 1 km, without the need for phase unwrapping. We derive a quantitative signal model for this new SWE change estimator and identify the relevant sources of bias. The method is demonstrated using both simulated SWE distributions and a 9-year RADARSAT-2 spotlight-mode dataset near Inuvik, NWT, Canada. SWE results are compared to in situ snow survey measurements and estimates from ERA5 reanalysis. Our method performs well in high-relief areas and in areas with high SWE (> 150 mm), thus providing complementary coverage to other passive- and active-microwave based SWE estimation methods. Further, our method has the advantage of requiring only a single wavelength band and thus can utilize existing spaceborne synthetic aperture radar systems. In application, a first order analysis of SWE trends within three drainage basins suggests that differences between basin-level accumulations are a function of major landcover types, and that re-vegetation following a forest-tundra fire that occurred over 50 years ago continues to affect the spatial distribution of SWE accumulation in the study area.

The assessment of snow water equivalent (SWE) is crucial for hydrological studies in glaciated catchments to quantify accumulation and ablation of a seasonal snow cover for both, the glaciated and non-glaciated terrain. Presently, the majority of the SWE assessment on glaciated terrain relies on manual measurements once or a few times per year, given the limited techniques for continuous SWE monitoring and the challenging conditions in a high mountain environment. Cosmic Ray Neutron Sensors (CRNS) offer to overcome these limitations providing sub-daily SWE estimates derived from neutron counts.This study employs a CRNS installed on an Alpine glacier (Hintereisferner, Austria) over three years, complemented with an additional CRNS for one winter roughly 300 m lower in elevation along the glacier&amp;#8217;s central flow line. Comparing CRNS outputs with frequent manual SWE measurements, the results demonstrate close agreement in SWE and snow density. The CRNS were found to be remarkably resilient in harsh conditions, providing nearly continuous hourly data over the measurement period. Additionally, the study evaluates at the CRNS locations the performance of two snow models, which might be considered as end members of model complexity &amp;#8211; SNOWPACK and &amp;#916;SNOW in its latest version. While SNOWPACK, with its physically-based approach, yields the best results, &amp;#916;SNOW stands out for its simplicity, requiring only daily snow depth observations as input and performs almost as well as SNOWPACK in terms of mean absolute SWE error.&amp;#160;Comparing the SWE measurements with winter precipitation from weighing gauges distributed in the catchment gives interesting details of precipitation gradients with elevation. Precipitation gradients interpolated from non-glaciated to glaciated terrain are considerably higher than on non-glaciated terrain only. On the latter, empirical bulk correction factors are frequently applied, which might fail on glaciers due to their different topographic setting. This highlights the need of separate treatment of snow on glaciers in hydrological models for correct SWE representation across the catchment.&amp;#160;

Abstract. During the World Meteorological Organization (WMO) Solid Precipitation Intercomparison Experiment (SPICE), automated measurements of snow water equivalent (SWE) were made at the Sodankylä (Finland), Weissfluhjoch (Switzerland) and Caribou Creek (Canada) SPICE sites during the northern hemispheric winters of 2013/14 and 2014/15. Supplementary intercomparison measurements were made at Fortress Mountain (Kananaskis, Canada) during the 2013/14 winter. The objectives of this analysis are to compare automated SWE measurements with a reference, comment on their performance and, where possible, to make recommendations on how to best use the instruments and interpret their measurements. Sodankylä, Caribou Creek and Fortress Mountain hosted a Campbell Scientific CS725 passive gamma radiation SWE sensor. Sodankylä and Weissfluhjoch hosted a Sommer Messtechnik SSG1000 snow scale. The CS725 operating principle is based on measuring the attenuation of soil emitted gamma radiation by the snowpack and relating the attenuation to SWE. The SSG1000 measures the mass of the overlying snowpack directly by using a weighing platform and load cell. Manual SWE measurements were obtained at the intercomparison sites on a bi-weekly basis over the accumulation–ablation periods using bulk density samplers. These manual measurements are considered to be the reference for the intercomparison. Results from Sodankylä and Caribou Creek showed that the CS725 generally overestimates SWE as compared to manual measurements by roughly 30–35 % with correlations (r2) as high as 0.99 for Sodankylä and 0.90 for Caribou Creek. The RMSE varied from 30 to 43 mm water equivalent (mm w.e.) and from 18 to 25 mm w.e. at Sodankylä and Caribou Creek, which had respective SWE maximums of approximately 200 and 120 mm w.e. The correlation at Fortress Mountain was 0.94 (RMSE of 48 mm w.e. with a maximum SWE of approximately 650 mm w.e.) with no systematic overestimation. The SSG1000 snow scale, having a different measurement principle, agreed quite closely with the manual measurements at Sodankylä and Weissfluhjoch throughout the periods when data were available (r2 as high as 0.99 and RMSE from 8 to 24 mm w.e. at Sodankylä and from 56 to 59 mm w.e. at Weissfluhjoch, where maximum SWE was approximately 850 mm w.e.). When the SSG1000 was compared to the CS725 at Sodankylä, the agreement was linear until the start of ablation when the positive bias in the CS725 increases substantially relative to the SSG1000. Since both Caribou Creek and Sodankylä have sandy soil, water from the snowpack readily infiltrates into the soil during melt, even if the soil is frozen. However, the CS725 does not differentiate this water from the unmelted snow. This issue can be identified, at least during the late spring ablation period, with soil moisture and temperature observations like those measured at Caribou Creek. With a less permeable soil and surface runoff, the increase in the instrument bias during ablation is not as significant, as shown by the Fortress Mountain intercomparison.

Seasonal snow occurs in high latitude and altitude regions of the globe, and throughout the winter and post-winter period can rapidly alter the makeup of these regions. Commonly studied snow features include snow depth, snow water equivalent (SWE), and snow density. These features can be measured on the ground while also capable of being remotely sensed with airborne and spaceborne instruments. Individually both approaches can be utilized to assess these snow features. However, field-based techniques tend to be limited to smaller spatial scales while remotely sensed methods tend to excel at larger spatial scales. At local scales (10 km2) a hybrid technique can be employed to better estimate such snow features. This can be realized by utilizing machine learning modeling to upscale field measurements with high-resolution remote sensing imagery. We performed this over a 10 km2 area in Sodankyl&amp;#228;, Finland by combining repeat field snow depth and SWE data with 2-meter resolution WorldView-2 (WV-2) and Light Detection and Ranging (LiDAR) data over a winter period between the middle of December 2022 to the end of April 2023. Snow depth field measurements were upscaled to a local spatial scale with an object-based machine learning approach before harnessing the estimated snow depth products to permit an enhanced estimation of SWE to the same local scale from more limited field measurements. Snow density was then determined from the predicted snow depth and SWE. A weighted ensemble approach of multiple machine learning models proved to be most effective compared to the chosen base models of Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Multiple Linear Regression (MLR). The fluctuating outputs from these features over the winter period were found to strongly connect to dry and wet peatbogs and with forests containing carbon and mineral surface soils.

Snow water equivalent (SWE) can be measured using low-cost Global Navigation Satellite System (GNSS) sensors with one antenna placed below the snowpack and another one serving as a reference above the snow. The underlying GNSS signal-based algorithm for SWE determination for dry- and wet-snow conditions processes the carrier phases and signal strengths and derives additionally liquid water content (LWC) and snow depth (HS). So far, the algorithm was tested intensively for high-alpine conditions with distinct seasonal accumulation and ablation phases. In general, snow occurrence, snow amount, snow density and LWC can vary considerably with climatic conditions and elevation. Regarding alpine regions, lower elevations mean generally earlier and faster melting, more rain-on-snow events and shallower snowpack. Therefore, we assessed the applicability of the GNSS-based SWE measurement at four stations along a steep elevation gradient (820, 1185, 1510 and 2540 m a.s.l.) in the eastern Swiss Alps during two winter seasons (2018–2020). Reference data of SWE, LWC and HS were collected manually and with additional automated sensors at all locations. The GNSS-derived SWE estimates agreed very well with manual reference measurements along the elevation gradient and the accuracy (RMSE = 34 mm, RMSRE = 11 %) was similar under wet- and dry-snow conditions, although significant differences in snow density and meteorological conditions existed between the locations. The GNSS-derived SWE was more accurate than measured with other automated SWE sensors. However, with the current version of the GNSS algorithm, the determination of daily changes of SWE was found to be less suitable compared to manual measurements or pluviometer recordings and needs further refinement. The values of the GNSS-derived LWC were robust and within the precision of the manual and radar measurements. The additionally derived HS correlated well with the validation data. We conclude that SWE can reliably be determined using low-cost GNSS-sensors under a broad range of climatic conditions and LWC and HS are valuable add-ons.

Abstract. Despite decades of effort, there remains an inability to measure snow water equivalent (SWE) at high spatial resolutions using remote sensing. Passive gamma ray spectrometry is one of the only well-established methods to reliably remotely sense SWE, but airborne applications to date have been limited to observing kilometre-scale areal averages. Noting the increasing capabilities of unoccupied aerial vehicles (UAVs) and miniaturization of passive gamma ray spectrometers, this study tested the ability of a UAV-borne gamma spectrometer and concomitant UAV-borne lidar to quantify the spatial variability of SWE at high spatial resolutions. Gamma and lidar observations from a UAV (UAV-gamma and UAV-lidar) were collected over two seasons from shallow, wind-blown, prairie snowpacks in Saskatchewan, Canada, with validation data collected from manual snow depth and density observations. A fine-resolution (0.25 m) reference dataset of SWE, to test UAV-gamma methods, was developed from UAV-lidar snow depth and snow survey snow density observations. The ability of UAV-gamma to resolve the areal average and spatial variability of SWE was promising with appropriate flight characteristics. Survey flights flown at a velocity of 5 m s−1, altitude of 15 m, and line spacing of 15 m were unable to capture the average or spatial variability of SWE within the uncertainty of the reference dataset. Slower, lower, and denser flight lines at a velocity of 4 m s−1, altitude of 8 m, and line spacing of 8 m were able to successfully observe areal average SWE and its variability at spatial resolutions greater than 22.5 m. Using a combination of UAV-based gamma SWE and UAV-based lidar snow depth improved the spatial representation of SWE substantially and permitted estimation of SWE at a spatial resolution 0.25 m with a ± 14.3 mm error relative to the reference SWE dataset. UAV-borne gamma spectrometry to estimate SWE is a promising and novel technique that has the potential to improve the measurement of shallow prairie snowpacks, and when combined with UAV-borne lidar snow depths, can provide fine-resolution, high-accuracy estimates of prairie SWE. Research on optimal hardware, data processing, and interpolation techniques is called for to further improve this remote sensing product and explore its application in other environments.

Snow depth observations can be leveraged with data assimilation (DA) to improve estimation of snow density and snow water equivalent (SWE). A key consideration for mission and campaign design is how snow depth retrieval characteristics (including observation timing/frequency and sampling error) influence SWE accuracy and uncertainty in a DA framework. To quantify these effects, we implement a particle filter (PF) assimilation technique to assimilate depth and validate this approach against observed snow density and SWE at 49 snow telemetry sites across 9 years. We sample from continuous in situ snow depth records to test a range of measurement timing and sampling error scenarios representative of remote sensing capabilities. Assimilation reduces density bias by over 40% and SWE bias by over 70% across climate zones and in both wet and dry years. There is little incremental benefit to SWE accuracy when assimilating more than one depth observation near peak accumulation. SWE estimates are less sensitive to observation timing than sampling error. Alternatively, more frequent depth observations improve melt‐out date timing and reduce SWE uncertainty, a key consideration when evaluating the operational utility of DA. In matching depth observations, the PF mostly acts to increase model precipitation inputs, while not systematically shifting other parameter values or forcings across the climate zones represented with the study sites. This demonstrates that precipitation is the largest source of model error. With DA, density errors are still nontrivial (above 10%), illuminating the need for further improvements to modeled density to estimate SWE within specified error limits.

Direct measurement of snow water equivalent (SWE) in snow-dominated mountainous areas is difficult, thus its prediction is essential for water resources management in such areas. In addition, because of nonlinear trend of snow spatial distribution and the multiple influencing factors concerning the SWE spatial distribution, statistical models are not usually able to present acceptable results. Therefore, applicable methods that are able to predict nonlinear trends are necessary. In this research, to predict SWE, the Sohrevard Watershed located in northwest of Iran was selected as the case study. Database was collected, and the required maps were derived. Snow depth (SD) at 150 points with two sampling patterns including systematic random sampling and Latin hypercube sampling (LHS), and snow density at 18 points were randomly measured, and then SWE was calculated. SWE was predicted using artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS) and regression methods. The results showed that the performance of ANN and ANFIS models with two sampling patterns were observed better than the regression method. Moreover, based on most of the efficiency criteria, the efficiency of ANN, ANFIS and regression methods under LHS pattern were observed higher than the systematic random sampling pattern. However, there were no significant differences between the two methods of ANN and ANFIS in SWE prediction. Data of both two sampling patterns had the highest sensitivity to the elevation. In addition, the LHS and the systematic random sampling patterns had the least sensitivity to the profile curvature and plan curvature, respectively.

Snow depth observations and modeled snow density can be combined to calculate snow water equivalent (SWE). In this approach, SWE uncertainty is dominated by snow density uncertainty, which depends on meteorological data quality and process representation (e.g., compaction) in models. We test whether assimilating snow depth observations with the particle filter can improve modeled snow density, thus improving SWE estimated from intermittent depth observations. We model snowpack at Mammoth Mountain (California) over water years 2013–2016, assuming monthly snow depth data (e.g., sampling intervals relevant to lidar or manual surveys) for assimilation, and validate against observed SWE and density. The particle filter reduced density and SWE root‐mean‐square error by 27% and 28% relative to open loop simulations when using high‐quality, point location forcing. Assimilation gains were greater (35% and 51% reduction in density and SWE root‐mean‐square error) when using coarse‐resolution North American Land Data Assimilation System phase 2 meteorology. Ensembles created with both meteorological and compaction perturbations led to the greatest model improvements. Because modeled depth and density were both generally lower than observations, assimilation favored particles with higher precipitation and thus more overburden compaction. This moved depth and density (therefore SWE) closer to observations. In contrast, ensemble generation that varied only compaction parameters degraded performance. These results were supported by synthetic experiments with prescribed error sources. Thus, assimilation of snow depth data from lidar or other techniques can likely improve snow density and SWE derived at the basin scale. However, supplementary in situ observations are valuable to identify primary error sources in simulated snow depth and density.

Seasonal snow is an essential component of regional and global water and energy cycles, particularly in snow-dominant regions that rely on snowmelt for water resources. Land surface models (LSMs) are a common approach for developing spatially and temporally complete estimates of snow water equivalent (SWE) and hydrologic variables at a large scale. However, the accuracy of the LSM-based SWE outputs is limited and unclear by mixed factors such as uncertainties in the meteorological boundary conditions and the model physics. In this study, we assess the SWE, snowfall, precipitation, and air temperature products from a twelve-member ensemble – with four LSMs and three meteorological forcings – using automated SWE, precipitation, and temperature observations from 809 Snowpack Telemetry stations over the western U.S. Results show that the mean annual maximum LSM SWE is underestimated by 268 mm. The timing of peak SWE from the LSMs is on average 36 days earlier than that of the observations. By the date of peak SWE, winter accumulated precipitation is underestimated (forcings mean: 485 mm vs. stations: 690 mm). In addition, the precipitation partitioning physics generates different snowfall estimates by an average of 113 mm with the same forcing data. Even though there are widespread cold biases (up to 3 °C) in the temperature forcings, larger ablations and lower ratios of SWE to total precipitation are found even in the accumulation period, indicating that melting physics in LSMs drives some SWE uncertainties. Based on the principal component analysis, we find that precipitation bias has the largest contribution to the first principal component, which accounts for more than half of the total variance. The results provide insights into prioritizing strategies to improve SWE estimates from LSMs for hydrologic applications.

Comparisons between snow water equivalent (SWE) and river discharge estimates are important in evaluating the SWE fields and to our understanding of linkages in the freshwater cycle. In this study, we compared SWE drawn from land surface models and remote sensing observations with measured river discharge (Q) across 179 Arctic river basins. Over the period 1988‐2000, basin‐averaged SWE prior to snowmelt explains a relatively small (yet statistically significant) fraction of interannual variability in spring (April–June) Q, as assessed using the coefficient of determination (R2). Averaged across all basins, mean R2s vary from 0·20 to 0·28, with the best agreement noted for SWE drawn from a simulation with the Pan‐Arctic Water Balance Model (PWBM) forced with data from the European Centre for Medium‐Range Weather‐Forecasts (ECMWF) Re‐analysis (ERA‐40). Variability and magnitude in SWE derived from Special Sensor Microwave Imager (SSM/I) data are considerably lower than the variability and magnitude in SWE drawn from the land surface models, and generally poor agreement is noted between SSM/I SWE and spring Q. We find that the SWE versus Q comparisons are no better when alternate temporal integrations–using an estimate of the timing in basin thaw–are used to define pre‐melt SWE and spring Q. Thus, a majority of the variability in spring discharge must arise from factors other than basin snowpack water storage. This study demonstrates how SWE estimated from remote sensing observations, or general circulation models (GCMs), can be evaluated effectively using monthly discharge data or SWE from a hydrological model. Copyright © 2007 John Wiley &amp; Sons, Ltd.

Seasonal snow cover of the Northern Hemisphere (NH) is a major factor in the global climate system, which makes snow cover an important variable in climate models. Monitoring snow water equivalent (SWE) at continental scale is only possible from satellites, yet substantial uncertainties have been reported in NH SWE estimates. A recent bias-correction method significantly reduces the uncertainty of NH SWE estimation, which enables a more reliable analysis of the climate models' ability to describe the snow cover. We have intercompared the CMIP6 (Coupled Model Intercomparison Project Phase 6) and satellite-based NH SWE estimates north of 40° N for the period 1982–2014, and analyzed with a regression approach whether temperature (T) and precipitation (P) could explain the differences in SWE. We analyzed separately SWE in winter and SWE change rate in spring. The SnowCCI SWE data are based on satellite passive microwave radiometer data and in situ data. The analysis shows that CMIP6 models tend to overestimate SWE, however, large variability exists between models. In winter, P is the dominant factor causing SWE discrepancies especially in the northern and coastal regions. This is in line with the expectation that even too cold temperatures cannot cause too high SWE without precipitation. T contributes to SWE biases mainly in regions, where T is close to 0 °C in winter. In spring, the importance of T in explaining the snowmelt rate discrepancies increases. This is to be expected, because the increase in T is the main factor that causes snow to melt as spring progresses. Furthermore, it is obvious from the results that biases in T or P can not explain all model biases either in SWE in winter or in the snowmelt rate in spring. Other factors, such as deficiencies in model parameterizations and possibly biases in the observational datasets, also contribute to SWE discrepancies. In particular, linear regression suggests that when the biases in T and P are eliminated, the models generally overestimate the snowmelt rate in spring.