Water Equivalent Research Articles

LSTM-Based Data Integration to Improve Snow Water Equivalent Prediction and Diagnose Error Sources

Abstract Accurate prediction of snow water equivalent (SWE) can be valuable for water resource managers. Recently, deep learning methods such as long short-term memory (LSTM) have exhibited high accuracy in simulating hydrologic variables and can integrate lagged observations to improve prediction, but their benefits were not clear for SWE simulations. Here we tested an LSTM network with data integration (DI) for SWE in the western United States to integrate 30-day-lagged or 7-day-lagged observations of either SWE or satellite-observed snow cover fraction (SCF) to improve future predictions. SCF proved beneficial only for shallow-snow sites during snowmelt, while lagged SWE integration significantly improved prediction accuracy for both shallow- and deep-snow sites. The median Nash–Sutcliffe model efficiency coefficient (NSE) in temporal testing improved from 0.92 to 0.97 with 30-day-lagged SWE integration, and root-mean-square error (RMSE) and the difference between estimated and observed peak SWE values dmax were reduced by 41% and 57%, respectively. DI effectively mitigated accumulated model and forcing errors that would otherwise be persistent. Moreover, by applying DI to different observations (30-day-lagged, 7-day-lagged), we revealed the spatial distribution of errors with different persistent lengths. For example, integrating 30-day-lagged SWE was ineffective for ephemeral snow sites in the southwestern United States, but significantly reduced monthly-scale biases for regions with stable seasonal snowpack such as high-elevation sites in California. These biases are likely attributable to large interannual variability in snowfall or site-specific snow redistribution patterns that can accumulate to impactful levels over time for nonephemeral sites. These results set up benchmark levels and provide guidance for future model improvement strategies.

Measuring global snow water equivalent from passive microwave remote sensing: opportunities and challenges

Measuring global snow water equivalent from passive microwave remote sensing: opportunities and challenges

Thermal equilibrium of the tractor and vehicle engines’ cooling systems in agriculture technological processes

The article shows how the heat introduced into the engine is consumed into the coolant. Factors influencing the temperature regime of the tractor and vehicle cooling systems during operation. Necessary dependencies for constructing the heat balance of the cooling system of an automobile and autotractor engine. The use of heat balance makes it possible to determine a criterion for assessing the efficiency of the engine cooling system. Experimental analysis of the thermal balance of the cooling system is based on original equations characterizing the heat transfer of the engine into the coolant, water equivalents of air and water flows through the radiator, and can be used as the basis for a calculation method for determining the characteristics of a cooling system with various radiators.

Persistent net release of carbon dioxide and methane from an Alaskan lowland boreal peatland complex.

Permafrost degradation in peatlands is altering vegetation and soil properties and impacting net carbon storage. We studied four adjacent sites in Alaska with varied permafrost regimes, including a black spruce forest on a peat plateau with permafrost, two collapse scar bogs of different ages formed following thermokarst, and a rich fen without permafrost. Measurements included year-round eddy covariance estimates of net carbon dioxide (CO2 ), mid-April to October methane (CH4 ) emissions, and environmental variables. From 2011 to 2022, annual rainfall was above the historical average, snow water equivalent increased, and snow-season duration shortened due to later snow return. Seasonally thawed active layer depths also increased. During this period, all ecosystems acted as slight annual sources of CO2 (13-59 g C m-2 year-1 ) and stronger sources of CH4 (11-14 g CH4 m-2 from ~April to October). The interannual variability of net ecosystem exchange was high, approximately ±100 g C m-2 year-1 , or twice what has been previously reported across other boreal sites. Net CO2 release was positively related to increased summer rainfall and winter snow water equivalent and later snow return. Controls over CH4 emissions were related to increased soil moisture and inundation status. The dominant emitter of carbon was the rich fen, which, in addition to being a source of CO2 , was also the largest CH4 emitter. These results suggest that the future carbon-source strength of boreal lowlands in Interior Alaska may be determined by the area occupied by minerotrophic fens, which are expected to become more abundant as permafrost thaw increases hydrologic connectivity. Since our measurements occur within close proximity of each other (≤1 km2 ), this study also has implications for the spatial scale and data used in benchmarking carbon cycle models and emphasizes the necessity of long-term measurements to identify carbon cycle process changes in a warming climate.

RESULTS OF MODELLING OF SNOW WATER EQUIVALENT WITH USING MODSNOW MODEL IN ESIL RIVER MANAGEMENT BASIN

The article presents the results of modeling of snow water equivalent using the V03 module of the MODSNOW model, based on the empirical modeling method, for the territory of the Esil water management basin for 1980…2021 periods. The multi-year period was divided into two equal parts: 1980...1999 for model calibration, 2000...2021 for model validation. For this purpose, 20 meteorological stations located in the section of the water management basin, which have continuous input daily data of average daily air temperature, precipitation, snow cover height and ten-day data of snow water equivalent, were selected. The results of model calibration and validation for reproducibility were evaluated by Nash-Sutcliffe, RSR, and PBIAS criteria. Good reproducibility of modeled snow water equivalent at the meteorological stations Blagoveshenka, Balkashino, Sergeevka, Stepnogorsk, Yavlenka, Bulayevo was revealed; for these stations the results corresponded to the «good» reproducibility rating. It is concluded that the model for these stations can be used for short, medium- and long-term forecasting of snow water equivalent.

Spatial Estimation of Snow Water Equivalent for Glaciers and Seasonal Snow in Iceland Using Remote Sensing Snow Cover and Albedo

Efficient water resource management in glacier- and snow-dominated basins requires accurate estimates of the snow water equivalent (SWE) in late winter and spring and melt onset timing and intensity. To understand the high spatio-temporal variability of snow and glacier ablation, a spatially distributed energy balance model combining satellite-based retrievals of albedo and snow cover was applied. Incoming short-wave energy, contributing to daily estimates of melt energy, was constrained by remotely sensed surface albedo for snow-covered surfaces. Fractional snow cover was used for non-glaciated areas, as it provides estimates of snow cover for each pixel to better constrain snow melt. Thus, available daily estimates of melt energy in a given area were the product of the possible melt energy and the fractional snow cover of the area or pixel for non-glaciated areas. This provided daily estimates of melt water to determine seasonal snow and glacier ablation in Iceland for the period 2000–2019. Observations from snow pits on land and glacier summer mass balance were used for evaluation, and observations from land and glacier-based automatic weather stations were used to evaluate model inputs for the energy balance model. The results show that the interannual SWE variability was generally high both for seasonal snow and glaciers. For seasonal snow, the largest SWE (>1000 mm) was found in mountainous and alpine areas close to the coast, notably in the East- and Westfjords, Tröllaskaga, and in the vicinity of glacier margins. Lower SWE values were observed in the central highlands, flatter inland areas, and at lower elevations. For glaciers, more SWE (glacier ablation) was associated with lower glacier elevations while less melt was observed at higher elevations. For the impurity-rich bare-ice areas that are exposed annually, observed SWE was more than 3000 mm.

How to handle glacier area change in geodetic mass balance

Abstract Innovations in geodesy enable widespread analysis of glacier surface elevation change and geodetic mass balance. However, coincident glacier area data are less widely available, causing inconsistent handling of glacier area change. Here we quantify the bias introduced into meters water equivalent (m w.e.) specific geodetic mass balance results when using a fixed, maximum glacier area, and illustrate the bias for five North American glaciers. Sites span latitudes from the northern U.S. Rocky Mountains (48°N) to the Central Alaska Range (63°N) between 1948 and 2021. Results show that fixed (maximum) area treatment subdues the m w.e. mass change signal, underestimating mass balance by up to 19% in our test cases. This bias scales with relative glacier area change and the mass balance magnitude. Thus, the bias for specific geodetic mass balances will be most pronounced across rapidly deglaciating regions. Our analysis underscores the need for temporally resolved glacier area in geodetic mass balance studies.

Six Consecutive Seasons of High‐Resolution Mountain Snow Depth Maps From Satellite Stereo Imagery

AbstractFine‐scale seasonal snow depth observations can improve estimates of snow water equivalent at critical times of year. Airborne lidar is the current gold standard for snow depth measurement, but it involves high costs and relatively limited coverage. Using very‐high‐resolution satellite stereo images from WorldView‐2, WorldView‐3, and Pléiades‐HR 1A/1B, we produced a six‐year time series (2017–2022) of spatially continuous digital elevation models for an 874 km2 study area over Grand Mesa, Colorado. We generated high‐resolution stereo snow depth maps that capture intra‐ and interannual variability and span multiple anomalous years (58%–158% of median peak SNOTEL snow depth). Comparisons with near‐contemporaneous airborne lidar snow depth measurements showed good agreement, with median offset of −0.13 m, precision of 0.19 m and accuracy of 0.31 m. Our results suggest that satellite stereo can provide snow depth observations with the spatiotemporal coverage needed to improve operational forecast models and inform adaptive management strategies.

Spatiotemporal snow water storage uncertainty in the midlatitude American Cordillera

Abstract. This work quantifies the uncertainty of accumulation-season peak snow water storage in the portions of the midlatitude American Cordillera where snow is a dominant driver of hydrology. This is accomplished through intercomparison of commonly used global and regional products over the Western United States (WUS) and Andes domains, which have similar hydrometeorology but are disparate with respect to the amount of available in situ information. The recently developed WUS Snow Reanalysis (WUS-SR) and Andes Snow Reanalysis (Andes-SR) datasets, which have been extensively verified against in situ measurements, are used as baseline reference datasets in the intercomparison. Relative to WUS-SR climatological peak snow water equivalent (SWE) storage (269 km3), high- and moderate-resolution products (i.e., those with resolutions less than ∼10 km) are in much better agreement (284±14 km3; overestimated by 6 %) compared to low-resolution products (127±54 km3; underestimated by 53 %). In comparison to the Andes-SR peak snow storage (29 km3), all other products show large uncertainty and bias (19±16 km3; underestimated by 34 %). Examination of spatial patterns related to orographic effects showed that only the high- to moderate-resolution Snow Data Assimilation System (SNODAS) and University of Arizona (UA) products show comparable estimates of windward–leeward SWE patterns over a subdomain (Sierra Nevada) of the WUS. Coarser products distribute too much snow on the leeward side in both the Sierra Nevada and Andes, missing orographic and rain shadow patterns that have important hydrological implications. The uncertainty of peak seasonal snow storage is primarily explained by precipitation uncertainty in both the WUS (R2=0.55) and Andes (R2=0.84). Despite using similar forcing inputs, snow storage diverges significantly within the ECMWF Reanalysis v5 (ERA5) (i.e., ERA5 vs. ERA5-Land) products and the Global Land Data Assimilation System (GLDAS) (modeled with Noah, Variable Infiltration Capacity (VIC), and Catchment model) products due to resolution-induced elevation differences and/or differing model process representation related to rain–snow partitioning and accumulation-season snowmelt generation. The availability and use of in situ precipitation and snow measurements (i.e., in WUS) in some products adds value by reducing snow storage uncertainty; however, where such data are limited, i.e., in the Andes, significant biases and uncertainty exist.

Runoff response to the uncertainty from key water-budget variables in a seasonally snow-covered mountain basin

Study regionThe American River Basin, a seasonally snow-covered mountain basin in the northern Sierra Nevada of California, USA. Study focusPrecipitation phase (rain, snow, and mixed rain-snow) partitioning in snowy basins is important for runoff forecasting. This study focuses on how predicted runoff response is affected by uncertainty in rain-snow-transition elevation and other water-budget variables. Using a distributed hydrologic model, we examined predicted runoff differences across 33 extreme-precipitation events and dry seasons during 1982–2020, through a set of modeling experiments with different sources of transition elevation and values of water-budget variables. New hydrological insights for the regionDifferent transition elevations can result in noticeable differences in predictions of peak snow water equivalent (SWE) and dry-season runoff. Transition elevation modulates runoff response in a nonlinear way, with runoff uncertainty increasing during larger precipitation events. In terms of the relative impacts of water-budget variables on runoff response, precipitation has the greatest impact, followed by antecedent soil moisture, transition elevation, and antecedent SWE. A 2 °C warming would push the rain-snow-transition elevation ∼300-m higher, increasing the number and severity of extreme events. Compared to the historical period, the same transition-elevation uncertainty in climate warming conditions nearly triples its impacts on runoff uncertainty, posing additional challenges to flood control. This study provides implications for addressing uncertainty in flood-control and water-supply decisions confronted with hydrologic extremes and climate warming.

Evaluation of snow cover properties in ERA5 and ERA5-Land with several satellite-based datasets in the Northern Hemisphere in spring 1982–2018

Abstract. Seasonal snow cover of the Northern Hemisphere (NH) greatly influences surface energy balance; hydrological cycle; and many human activities, such as tourism and agriculture. Monitoring snow cover at a continental scale is only possible from satellites or using reanalysis data. This study aims to analyze the time series of snow water equivalent (SWE), snow cover extent (SCE), and surface albedo in spring in ERA5 and ERA5-Land reanalysis data and to compare the time series with several satellite-based datasets. As reference data for the SWE intercomparison, we use bias-corrected SnowCCI v1 data for non-mountainous regions and the mean of Brown, MERRA-2, and Crocus v7 datasets for the mountainous regions. For surface albedo, we use the black-sky albedo datasets CLARA-A2 SAL, based on AVHRR data, and MCD43D51, based on MODIS data. Additionally, we use Rutgers and JAXA JASMES SCE products. Our study covers land areas north of 40∘ N and the period between 1982 and 2018 (spring season from March to May). The analysis shows that both ERA5 and ERA5-Land overestimate total NH SWE by 150 % to 200 % compared to the SWE reference data. ERA5-Land shows larger overestimation, which is mostly due to very high SWE values over mountainous regions. The analysis revealed a discontinuity in ERA5 around the year 2004 since adding the Interactive Multisensor Snow and Ice Mapping System (IMS) from the year 2004 onwards considerably improves SWE estimates but makes the trends less reliable. The negative NH SWE trends in ERA5 range from −249 to −236 Gt per decade in spring, which is 2 to 3 times larger than the trends detected by the other datasets (ranging from −124 to −77 Gt per decade). SCE is accurately described in ERA5-Land, whereas ERA5 shows notably larger SCE than the satellite-based datasets. Albedo estimates are more consistent between the datasets, with a slight overestimation in ERA5 and ERA5-Land. The negative trends in SCE and albedo are strongest in May, when the albedo trend varies from −0.011 to −0.006 per decade depending on the dataset. The negative SCE trend detected by ERA5 in May (-1.22×106 km2 per decade) is about twice as large as the trends detected by all other datasets (ranging from −0.66 to -0.50×106 km2 per decade). The analysis also shows that there is a large spatial variability in the trends, which is consistent with other studies.

Combined GNSS reflectometry–refractometry for automated and continuous in situ surface mass balance estimation on an Antarctic ice shelf

Abstract. Reliable in situ surface mass balance (SMB) estimates in polar regions are scarce due to limited spatial and temporal data availability. This study aims at deriving automated and continuous specific SMB time series for fast-moving parts of ice sheets and shelves (flow velocity > 10 m a−1) by developing a combined global navigation satellite system (GNSS) reflectometry and refractometry (GNSS-RR) method. In situ snow density, snow water equivalent (SWE), and snow deposition or erosion are estimated simultaneously as an average over an area of several square meters and independently on weather conditions. The combined GNSS-RR method is validated and investigated regarding its applicability to a moving, high-latitude ice shelf. A combined GNSS-RR system was therefore installed in November 2021 on the Ekström ice shelf (flow velocity ≈ 150 m a−1) in Dronning Maud Land, Antarctica. The reflected and refracted GNSS observations from the site are post-processed to obtain snow accumulation (deposition and erosion), SWE, and snow density estimates with a 15 min temporal resolution. The results of the first 16 months of data show a high level of agreement with manual and automated reference observations from the same site. Snow accumulation, SWE, and density are derived with uncertainties of around 9 cm, 40 kg m−2 a−1, and 72 kg m−3, respectively. This pilot study forms the basis for extending observational networks with GNSS-RR capabilities, particularly in polar regions. Regional climate models, local snow modeling, and extensive remote sensing data products will profit from calibration and validation based on such in situ time series, especially if many such sensors will be deployed over larger regional scales.

Assimilation of airborne gamma observations provides utility for snow estimation in forested environments

Abstract. An airborne gamma-ray remote-sensing technique provides a strong potential to estimate a 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 (LSM) to improve SWE estimates in forested areas in the northeastern US. Here, we demonstrate that the airborne gamma SWE observations add value to the SWE estimates from the Noah LSM with multiple parameterization options (Noah-MP) via assimilation despite the limited number of measurements. Improvements are witnessed during the snow accumulation period, while reduced skills are seen during the snowmelt 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 at reducing the SWE estimation errors for areas with higher topographic heterogeneity. The gamma SWE data assimilation (DA) also shows a potential to extend 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 distance 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.

Assessing robustness in global hydrological predictions by comparing modelling and Earth observations

ABSTRACTHydrological modelling to support hypotheses on Earth system boundaries or the accelerating water crisis is nowadays done at the global scale, with difficulties associated to model uncertainties. Here we bring a robustness analysis of internal model variables as an additional tool for model evaluation using data from six Earth observation products and the global catchment model World-Wide HYPE in a comparative study. The assessment shows that: (i) variables have high agreement in mid-latitude temperate regions; (ii) the variables with higher agreement, and associated with good model performance in streamflow, were actual evapotranspiration, fractional snow cover and snow water equivalent; and (iii) changes in total water storage showed very poor agreement, probably due to an insufficient number of aquifers in the model set-up. We propose this procedure as a standard complementary method in global hydrological modelling, highlighting the importance of justifying models before using them for scenario analysis or water accounting.

Cryosphere changes and their impacts on regional water resources in the Chinese Altai Mountains from 2000 to 2021

The cryosphere has an important impact on regional water resources and ecosystems in the Chinese Altai Mountains and its piedmont zone. Using the latest remote-sensing datasets of cryosphere changes and combining with in-situ observation data from glacier monitoring stations and snow cover surveys, the main cryosphere elements including glaciers, snow cover, and permafrost are investigated with emphasis on their changes since 2000 and the current situation. Their water resource effects are also discussed. The results indicate that although the glaciers in the region have experienced continuous and intensive melting, mass loss has slowed because both glacier area shrinkage and thickness reduction were larger during 2000–2010 than during 2010–2021. Snow-cover water equivalent (w.e.) has increased due to obvious increases in snow depth, although snow-cover area has decreased slightly. Permafrost has been degrading. Overall, cryosphere contributions to the regional water resource are approximately 40.9% since 2000, among which snow-cover melting is the largest, contributing 37.1% to water resources in the Irtysh River Basin and significantly more in the mountainous sub-basins with increased snowfall. Glacier melting contributes 2.9%∼3.4%, lower than earlier estimations of 3.4%∼3.6% for the late 20th century. Permafrost thaw caused by active layer thickening contributes approximately 0.59%. Meteorological data shows a warming and wetting trend, but summer temperature has a much lower increase rate and a slowing increase trend after 2013. Moreover, snowfall frequency has increased. In the future, glacier water resource contribution will continue to decrease, but the water resource effects of snow-cover melting and permafrost degradation would increase.

SNOOPI: Demonstrating Earth remote sensing using P-band signals of opportunity (SoOp) on a CubeSat

SigNals Of Opportunity: P-band Investigation (SNOOPI) will be the first in-space demonstration of P-band (240–380 MHz) signals of opportunity (SoOp) remote sensing and validation of a prototype P-band SoOp instrument. This technique has the potential to enable remote sensing of root-zone soil moisture (RZSM) and snow water equivalent (SWE). SNOOPI technology validation goals will be met by targeting observations within 9 km of SMAP calibration/validation sites in the continental United States. A second priority is collection of continuous phase data over snow-covered regions. These goals are evaluated under constraints of a limited data budget and mission lifetime, with a launch expected in early 2024. SNOOPI will capture the autocorrelation of a signal formed by the interference between the direct and reflected signals. The complex surface reflection coefficient (phase and magnitude) can be extracted from this observation. Definition of critical instrument and mission requirements for SNOOPI are described. Analyses and simulations used to justify these requirements are presented, including orbit coverage, uncertainty in the open-loop delay-Doppler prediction, observations error, and mission planning operations.

Performance of precipitation phase partitioning methods and their impact on snowpack evolution in a humid continental climate

AbstractAccurate estimations of the precipitation phase at the surface are critical for hydrological and snowpack modelling in cold regions. Precipitation phase partitioning methods (PPMs) vary in their ability to estimate the precipitation phase at around 0°C and can significantly impact simulations of snowpack accumulation and melt. The goal of this study is to evaluate PPMs of varying complexity using high‐quality observations of precipitation phase and to assess the impact on snowpack simulations. We used meteorological data collected in Edmundston, New Brunswick, Canada, during the 2021 Saint John River Experiment on Cold Season Storms (SAJESS). These data were combined with manual observations of snow depth. Five PPMs commonly used in hydrological models were tested against observations from a laser‐optical disdrometer and a Micro Rain Radar. Most PPMs produced similar accuracy in estimating only rainfall and snowfall. Mixed precipitation was the most difficult phase to predict. The multi‐physics model Crocus was then used to simulate snowpack evolution and to diagnose model sensitivity to snowpack accumulation processes (PPM, snowfall density, and snowpack compaction). Sixteen snowpack accumulation periods, including nine warm accumulation events (average temperatures above −2°C) were observed during the study period. When considering all accumulation events, simulated changes in snow water equivalent (SWE) were more sensitive to the type of PPM used, whereas simulated changes in snow depth were more sensitive to uncertainties in snowfall density. Choice of PPM was the main source of model sensitivity for changes in SWE and snow depth when only considering warm events. Overall, this study highlights the impact of precipitation phase estimations on snowpack accumulation at the surface during near‐0°C conditions.

Modeling the total outflow of reservoirs using Wavelet-developed approaches: a case study of the Mahabad Dam reservoir, Iran

Abstract Lack of water reserves in artificial reservoirs poses serious challenges in meeting various human requirements, especially during periods of water scarcity. In the current research, the Total Outflow (TO) of the Mahabad Dam reservoir has been estimated under six scenarios including the Monthly Cumulative Rainfall (MCR), Snow Water Equivalent (SWE), Stream Flow (SF), Mean Temperature (T), Pan Evaporation (Ep), Sediment Flushing Gate Outlet (SFGO), Penstock Outflow (PO), Evaporation Losses (EL), Cumulative Non-Scheduled Discharge (CNSD), Live Storage Volume (LSV), Water Surface Area (WSA), Monthly Water Level (MWL), Total Allocated Water (TAW), and Generated Power (GP) variables for the 2001–2021 period. Estimation of TO is accomplished via individual and wavelet-developed (W-developed) data-mining approaches, including Artificial Neural Networks (ANNs), wavelet-ANNs (WANNs), adaptive neuro-fuzzy inference system (ANFIS), wavelet-ANFIS (WANFIS), Gene Expression Programming (GEP), and wavelet-GEP (WGEP). The obtained values of RMSE for WGEP1–WGEP6 models account for 5.917, 2.319, 4.289, 8.329, 10.713, and 9.789 million cubic meters (MCM), respectively, based on the following scenarios: reservoir inlet elements, reservoir outlet elements, consumption, storage characteristic, climate, and energy. This research revealed that combining the wavelet theory (WT) with individual models can be a powerful method to improve the modeling performance in the TO estimation.

Improving prediction of mountain snowfall in the southwestern United States using machine learning methods

AbstractSnowfall forecasting has historically been an area of difficulty for operational meteorologists, particularly in regions of complex terrain, such as the western United States. Attempts at improving forecasts have been made, but skill is still poor, with snowfall routinely overpredicted. A major reason for this overprediction has been the failure to accurately predict snow–liquid ratios (SLR) ahead of major events. This research proposes, develops, and tests multiple machine learning methods for dynamic SLR prediction for the Sky Islands of southeast Arizona by objectively comparing a multiple linear regression (MLR) against several more complex and flexible machine learning methods. Input parameters for each method were chosen based on variables found by previous studies to have a regression‐based relationship with SLR, with a focus on the lower mid‐levels of the troposphere. These parameters were also used to construct the MLR model, and its performance was compared objectively with the machine learning methods. When tested on historical events, a very high percentage of the network‐predicted SLR values fall within the margin of error of observed SLRs, which were calculated using gridded snow depth and snow water equivalent (SWE) data from the University of Arizona daily 4‐km SWE, SD, and SCE dataset (UASnow). A support vector machine (SVM), a k‐nearest neighbor (KNN) algorithm, and a random forest also showed high accuracies when tested on the dataset, and each showed a significant gain in skill compared with the MLR model, with skill being evaluated by multiple metrics.

Morphological indexes to describe snow-cover patterns in a high-alpine area

Abstract The spatiotemporal distribution of snow affects hydrological and climatological processes at different scales. Accordingly, quantifying geometric features of snow-cover patterns is important, providing a valuable complement for snow water equivalent (SWE) modelling. This study on satellite-based morphological analysis originally uses two types of geometric indexes: (1) MN, Minkowski numbers (area (MN1), perimeter (MN2), Euler number (MN3)), and (2) CL, average chord length, to describe the morphology of Sentinel-2-derived snow-covered areas (SCAs), within the high-alpine site Zugspitze for a 5 year period. Results indicate that they capture the seasonal variability of snow-cover patterns, particularly during accumulation and ablation. Being to some degree independent from each other, MN2, MN3 and CL provide additional information upon shape, connectivity and length scale of snow cover, compared to most used indexes (e.g. fractional SCA). Correlation values up to +0.7 for MN2, +0.58 for MN3 and +0.46 for CL were observed with selected topographic characteristics, suggesting a close connection between geometric features of snow cover and ground features. Comparing in situ SWE measurements with MN and CL shows a correlation between −0.5 and +0.5. These indexes can hence be applied in combination with in situ data and/or modelling approaches to improve spatially distributed SWE in high-alpine catchments.