SWE modelling: comparison between different approaches applied to Germany

The characterization of the snow cover by snow water equivalent (SWE) is fundamental in several environmental applications, e.g., monitoring mountain water resources or defining structural design standards. However, SWE observations are usually rare compared to other snow measurements as snow depth (HS). Therefore, model-based methods have been proposed in past studies for estimating SWE, in particular for short timescales (e.g., daily). In this study, we compare two different approaches for SWE-data modelling. The first approach, based on empirical regression models (ERMs), provides the regional parametrization of the bulk snow density, which can be used to estimate SWE values from HS. In particular, we investigate the performances of four different schemes based on previously developed ERMs of bulk snow density depending on HS, date, elevation, and location. Secondly, we apply the semi-empirical multi-layer &#916;snow model, which estimates SWE solely based on snow depth observations. The open source &#916;snow model has been recently used for deriving a snow load map for Austria, resulting in an improved Austrian standard. A large dataset of HS and SWE observations collected by the National Weather Service in Germany (DWD) is used for calibrating and validating the models. This dataset consists of daily HS and three-times-a-week SWE observations from in total ~1000 stations operated by DWD over the period from 1950 to 2020. A leave-one-out cross validation is applied to evaluate the performance of the different model approaches. It is based on 185 time series of HS and SWE observations that are representative of the diversity of the regional snow climatology of Germany. Cross validation reveals for all ERMs: 90% of the modelled SWE time series have a root mean square error (RMSE) and a bias lower than 45 kg/m&#178; and 2 kg/m&#178;, respectively. The &#916;snow model shows the best performance with 90% of the modelled SWE time series having an RMSE lower than 30 kg/m&#178; and bias similar to the ERMs. This comparative study provides new insights on the reliability of model-based methods for estimating SWE values. The results show that the &#916;snow model and, to a lower degree, the developed ERMs can provide satisfactory performances even on short timescales. This suggest that these models can be used as reliable alternative to more complex thermodynamic snow models, even more if long-term meteorological observations aside HS are scarce.

Download Full-text

Assimilation of snow cover and snow depth into a snow model to estimate snow water equivalent and snowmelt runoff in a Himalayan catchment

The Cryosphere ◽

10.5194/tc-11-1647-2017 ◽

2017 ◽

Vol 11 (4) ◽

pp. 1647-1664 ◽

Cited By ~ 29

Author(s):

Emmy E. Stigter ◽

Niko Wanders ◽

Tuomo M. Saloranta ◽

Joseph M. Shea ◽

Marc F. P. Bierkens ◽

...

Keyword(s):

Kalman Filter ◽

Data Assimilation ◽

Snow Cover ◽

Climate Sensitivity ◽

Snow Depth ◽

Snow Water Equivalent ◽

Snowmelt Runoff ◽

Snow Water ◽

Snow Model

Abstract. Snow is an important component of water storage in the Himalayas. Previous snowmelt studies in the Himalayas have predominantly relied on remotely sensed snow cover. However, snow cover data provide no direct information on the actual amount of water stored in a snowpack, i.e., the snow water equivalent (SWE). Therefore, in this study remotely sensed snow cover was combined with in situ observations and a modified version of the seNorge snow model to estimate (climate sensitivity of) SWE and snowmelt runoff in the Langtang catchment in Nepal. Snow cover data from Landsat 8 and the MOD10A2 snow cover product were validated with in situ snow cover observations provided by surface temperature and snow depth measurements resulting in classification accuracies of 85.7 and 83.1 % respectively. Optimal model parameter values were obtained through data assimilation of MOD10A2 snow maps and snow depth measurements using an ensemble Kalman filter (EnKF). Independent validations of simulated snow depth and snow cover with observations show improvement after data assimilation compared to simulations without data assimilation. The approach of modeling snow depth in a Kalman filter framework allows for data-constrained estimation of snow depth rather than snow cover alone, and this has great potential for future studies in complex terrain, especially in the Himalayas. Climate sensitivity tests with the optimized snow model revealed that snowmelt runoff increases in winter and the early melt season (December to May) and decreases during the late melt season (June to September) as a result of the earlier onset of snowmelt due to increasing temperature. At high elevation a decrease in SWE due to higher air temperature is (partly) compensated by an increase in precipitation, which emphasizes the need for accurate predictions on the changes in the spatial distribution of precipitation along with changes in temperature.

Download Full-text

Improved Northern Hemisphere Snow Water Equivalent product from passive microwave remote sensing and in situ data

10.5194/egusphere-egu21-1325 ◽

2021 ◽

Author(s):

Colleen Mortimer ◽

Lawrence Mudryk ◽

Chris Derksen ◽

Kari Luojus ◽

Pinja Venalainen ◽

...

Keyword(s):

Spatial Resolution ◽

Snow Depth ◽

Snow Water Equivalent ◽

Microwave Remote Sensing ◽

Passive Microwave ◽

Snow Density ◽

Snow Season ◽

Snow Water ◽

Snow Mass

The European Space Agency Snow CCI+ project provides global homogenized long time series of daily snow extent and snow water equivalent (SWE). The Snow CCI SWE product is built on the Finish Meteorological Institute's GlobSnow algorithm, which combines passive microwave data with in situ snow depth information to estimate SWE. The CCI SWE product improves upon previous versions of GlobSnow through targeted changes to the spatial resolution, ancillary data, and snow density parameterization.Previous GlobSnow SWE products used a constant snow density of 0.24 kg m-3 to convert snow depth to SWE. The CCI SWE product applies spatially and temporally varying density fields, derived by krigging in situ snow density information from historical snow transects to correct biases in estimated SWE. Grid spacing was improved from 25 km to 12.5 km by applying an enhanced spatial resolution microwave brightness temperature dataset. We assess step-wise how each of these targeted changes acts to improve or worsen the product by evaluating with snow transect measurements and comparing hemispheric snow mass and trend differences.Together, when compared to GlobSnow v3, these changes improved RMSE by ~5 cm and correlation by ~0.1 against a suite of snow transect measurements from Canada, Finland, and Russia. Although the hemispheric snow mass anomalies of CCI SWE and GlobSnow v3 are similar, there are sizeable differences in the climatological SWE, most notably a one month delay in the timing of peak SWE and lower SWE during the accumulation season. These shifts were expected because the variable snow density is lower than the former fixed value of 0.24 kg m-3 early in the snow season, but then increases over the course of the snow season. We also examine intermediate products to determine the relative improvements attributable solely to the increased spatial resolution versus changes due to the snow density parameterizations. Such systematic evaluations are critical to directing future product development.

Download Full-text

Statistical Stability and Spatial Instability in Mapping Forest Tree Species by Comparing 9 Years of Satellite Image Time Series

Remote Sensing ◽

10.3390/rs11212512 ◽

2019 ◽

Vol 11 (21) ◽

pp. 2512 ◽

Cited By ~ 7

Author(s):

Nicolas Karasiak ◽

Jean-François Dejoux ◽

Mathieu Fauvel ◽

Jérôme Willm ◽

Claude Monteil ◽

...

Keyword(s):

Time Series ◽

Tree Species ◽

Cross Validation ◽

Satellite Image ◽

Support Vector ◽

Validation Data ◽

Spatial Instability ◽

Classification Errors ◽

Leave One Out ◽

The Impact

Mapping forest composition using multiseasonal optical time series remains a challenge. Highly contrasted results are reported from one study to another suggesting that drivers of classification errors are still under-explored. We evaluated the performances of single-year Formosat-2 time series to discriminate tree species in temperate forests in France and investigated how predictions vary statistically and spatially across multiple years. Our objective was to better estimate the impact of spatial autocorrelation in the validation data on measurement accuracy and to understand which drivers in the time series are responsible for classification errors. The experiments were based on 10 Formosat-2 image time series irregularly acquired during the seasonal vegetation cycle from 2006 to 2014. Due to lot of clouds in the year 2006, an alternative 2006 time series using only cloud-free images has been added. Thirteen tree species were classified in each single-year dataset based on the Support Vector Machine (SVM) algorithm. The performances were assessed using a spatial leave-one-out cross validation (SLOO-CV) strategy, thereby guaranteeing full independence of the validation samples, and compared with standard non-spatial leave-one-out cross-validation (LOO-CV). The results show relatively close statistical performances from one year to the next despite the differences between the annual time series. Good agreements between years were observed in monospecific tree plantations of broadleaf species versus high disparity in other forests composed of different species. A strong positive bias in the accuracy assessment (up to 0.4 of Overall Accuracy (OA)) was also found when spatial dependence in the validation data was not removed. Using the SLOO-CV approach, the average OA values per year ranged from 0.48 for 2006 to 0.60 for 2013, which satisfactorily represents the spatial instability of species prediction between years.

Download Full-text

Spatiotemporal Characteristics of Snowpack Density in the Mountainous Regions of the Western United States

Journal of Hydrometeorology ◽

10.1175/2008jhm981.1 ◽

2008 ◽

Vol 9 (6) ◽

pp. 1416-1426 ◽

Cited By ~ 45

Author(s):

Naoki Mizukami ◽

Sanja Perica

Keyword(s):

United States ◽

Snow Depth ◽

Snow Water Equivalent ◽

Winter Season ◽

Western United States ◽

Densification Rate ◽

Snow Density ◽

Mountainous Regions ◽

Spatiotemporal Characteristics ◽

Snow Water

Abstract Snow density is calculated as a ratio of snow water equivalent to snow depth. Until the late 1990s, there were no continuous simultaneous measurements of snow water equivalent and snow depth covering large areas. Because of that, spatiotemporal characteristics of snowpack density could not be well described. Since then, the Natural Resources Conservation Service (NRCS) has been collecting both types of data daily throughout the winter season at snowpack telemetry (SNOTEL) sites located in the mountainous areas of the western United States. This new dataset provided an opportunity to examine the spatiotemporal characteristics of snowpack density. The analysis of approximately seven years of data showed that at a given location and throughout the winter season, year-to-year snowpack density changes are significantly smaller than corresponding snow depth and snow water equivalent changes. As a result, reliable climatological estimates of snow density could be obtained from relatively short records. Snow density magnitudes and densification rates (i.e., rates at which snow densities change in time) were found to be location dependent. During early and midwinter, the densification rate is correlated with density. Starting in early or mid-March, however, snowpack density increases by approximately 2.0 kg m−3 day−1 regardless of location. Cluster analysis was used to obtain qualitative information on spatial patterns of snowpack density and densification rates. Four clusters were identified, each with a distinct density magnitude and densification rate. The most significant physiographic factor that discriminates between clusters was proximity to a large water body. Within individual mountain ranges, snowpack density characteristics were primarily dependent on elevation.

Download Full-text

Investigating the dynamics of bulk snow density in dry and wet conditions using a one-dimensional model

The Cryosphere ◽

10.5194/tc-7-433-2013 ◽

2013 ◽

Vol 7 (2) ◽

pp. 433-444 ◽

Cited By ~ 22

Author(s):

C. De Michele ◽

F. Avanzi ◽

A. Ghezzi ◽

C. Jommi

Keyword(s):

Bulk Density ◽

Liquid Water ◽

Snow Depth ◽

Snow Water Equivalent ◽

Temporal Dynamics ◽

Dimensional Model ◽

Snow Density ◽

One Dimensional ◽

Hourly Data ◽

Dry And Wet Conditions

Abstract. The snowpack is a complicated multiphase mixture with mechanical, hydraulic, and thermal properties highly variable during the year in response to climatic forcings. Bulk density is a macroscopic property of the snowpack used, together with snow depth, to quantify the water stored. In seasonal snowpacks, the bulk density is characterized by a strongly non-linear behaviour due to the occurrence of both dry and wet conditions. In the literature, bulk snow density estimates are obtained principally with multiple regressions, and snowpack models have put the attention principally on the snow depth and snow water equivalent. Here a one-dimensional model for the temporal dynamics of the snowpack, with particular attention to the bulk snow density, has been proposed, accounting for both dry and wet conditions. The model represents the snowpack as a two-constituent mixture: a dry part including ice structure, and air; and a wet part constituted by liquid water. It describes the dynamics of three variables: the depth and density of the dry part and the depth of liquid water. The model has been calibrated and validated against hourly data registered at three SNOTEL stations, western US, with mean values of the Nash–Sutcliffe coefficient ≈0.73–0.97 in the validation period.

Download Full-text

What drives basin scale spatial variability of snow water equivalent during two extreme years?

The Cryosphere Discussions ◽

10.5194/tcd-7-2943-2013 ◽

2013 ◽

Vol 7 (3) ◽

pp. 2943-2977

Author(s):

G. A. Sexstone ◽

S. R. Fassnacht

Keyword(s):

Linear Regression ◽

Spatial Variability ◽

Multiple Linear Regression ◽

Snow Depth ◽

Snow Water Equivalent ◽

Density Model ◽

Snow Density ◽

Depth Measurement ◽

Basin Scale ◽

Snow Water

Abstract. This study uses a combination of field measurements and Natural Resource Conservation Service (NRCS) operational snow data to understand the drivers of snow water equivalent (SWE) spatial variability at the basin scale. Historic snow course snowpack density observations were analyzed within a multiple linear regression snow density model to estimate SWE directly from snow depth measurements. Snow surveys were completed on or about 1 April 2011 and 2012 and combined with NRCS operational measurements to investigate the spatial variability of SWE. Bivariate relations and multiple linear regression models were developed to understand the relation of SWE with terrain and canopy variables (derived using a geographic information system (GIS)). Calculation of SWE directly from snow depth measurement using the snow density model has strong statistical performance and model validation suggests the model is transferable to independent data within the bounds of the original dataset. This pathway of estimating SWE directly from snow depth measurement is useful when evaluating snowpack properties at the basin scale, where many time consuming measurements of SWE are often not feasible. During both water year (WY) 2011 and 2012, elevation and location (UTM Easting and UTM Northing) were the most important model variables, suggesting that orographic precipitation and storm track patterns are likely consistent drivers of basin scale SWE variability. Terrain characteristics, such as slope, aspect, and curvature, were also shown to be important variables, but to a lesser extent at the scale of interest.

Download Full-text

The role of liquid water percolation representation to estimate snow water equivalent in a Mediterranean mountain region (Mount Lebanon)

10.5194/hess-2019-514 ◽

2019 ◽

Author(s):

Abbas Fayad ◽

Simon Gascoin

Keyword(s):

Snow Cover ◽

Liquid Water ◽

Snow Depth ◽

Snow Water Equivalent ◽

Mountain Regions ◽

Snow Cover Area ◽

Water Percolation ◽

Snow Water ◽

Snow Model ◽

Snow Mass

Abstract. In many Mediterranean mountain regions, the seasonal snowpack is an essential yet poorly known water resource. Here, we examine, for the first time, the spatial distribution and evolution of the snow water equivalent (SWE) during three snow seasons (2013–2016) in the coastal mountains of Lebanon. We run SnowModel (Liston and Elder, 2006a), a spatially-distributed, process-based snow model, at 100 m resolution forced by new automatic weather station (AWS) data in three snow-dominated basins of Mount Lebanon. We evaluate a recent upgrade of the liquid water percolation scheme in SnowModel, which was introduced to improve the simulation of the snow water equivalent (SWE) and runoff in warm maritime regions. The model is evaluated against continuous snow depth and snow albedo observations at the AWS, manual SWE measurements, and MODIS snow cover area between 1200 m and 3000 m a.s.l.. The results show that the new percolation scheme yields better performance especially in terms of SWE but also in snow depth and snow cover area. Over the simulation period between 2013 and 2016, the maximum snow mass was reached between December and March. Peak mean SWE (above 1200 m a.s.l.) changed significantly from year to year in the three study catchments with values ranging between 73 mm and 286 mm we (RMSE between 160 and 260 mm w.e.). We suggest that the major sources of uncertainty in simulating the SWE, in this warm Mediterranean climate, can be attributed to forcing error but also to our limited understanding of the separation between rain and snow at lower-elevations, the transient snow melt events during the accumulation season, and the high-variability of snow depth patterns at the sub-pixel scale due to the wind-driven blown-snow redistribution into karstic features and sinkholes. Yet, the use of a process-based snow model with minimal requirements for parameter estimation provides a basis to simulate snow mass SWE in non-monitored catchments and characterize the contribution of snowmelt to the karstic groundwater recharge in Lebanon. While this research focused on three basins in the Mount Lebanon, it serves as a case study to highlight the importance of wet snow processes to estimate SWE in Mediterranean mountain regions.

Download Full-text

Climate sensitivity of snow water equivalent and snowmelt runoff in a Himalayan catchment

10.5194/tc-2016-216 ◽

2016 ◽

Cited By ~ 1

Author(s):

Emmy E. Stigter ◽

Niko Wanders ◽

Tuomo M. Saloranta ◽

Joseph M. Shea ◽

Marc F.P. Bierkens ◽

...

Keyword(s):

Kalman Filter ◽

Snow Cover ◽

Climate Sensitivity ◽

Snow Depth ◽

Snow Water Equivalent ◽

Snowmelt Runoff ◽

Snow Water ◽

Snow Model ◽

Increasing Temperature

Abstract. Snow is an important component of water storage in the Himalayas. Previous snowmelt studies in the Himalayas have predominantly relied on remotely sensed snow cover. However this provides no information on the actual amount of water stored in a snowpack i.e. the snow water equivalent (SWE). Therefore, in this study remotely sensed snow cover was combined with in situ meteorological observations and a modified version of the seNorge snow model to estimate climate sensitivity of SWE and snowmelt runoff in the Langtang catchment in Nepal. Landsat 8 and MOD10A2 snow cover maps were validated with in situ snow cover observations provided by surface temperature and snow depth measurements resulting in classification accuracies of 85.7 % and 83.1 % respectively. Optimal model parameter values were obtained through data assimilation of MOD10A2 snow maps and snow depth measurements using an Ensemble Kalman filter. The approach of modelling snow depth in a Kalman filter framework allows for data-constrained estimation of SWE rather than snow cover alone and this has great potential for future studies in the Himalayas. Climate sensitivity tests with the optimized snow model show a strong decrease in SWE in the valley with increasing temperature. However, at high elevation a decrease in SWE is (partly) compensated by an increase in precipitation, which emphasizes the need for accurate predictions on the changes in the spatial distribution of precipitation along with changes in temperature. Finally the climate sensitivity study revealed that snowmelt runoff increases in winter and early melt season (December to May) and decreases during the late melt season (June to September) as a result of the earlier onset of snowmelt due to increasing temperature.

Download Full-text

Ensemble-based assimilation of fractional snow covered area satellite retrievals to estimate snow distribution at a high Arctic site

10.5194/tc-2017-109 ◽

2017 ◽

Author(s):

Kristoffer Aalstad ◽

Sebastian Westermann ◽

Thomas Vikhamar Schuler ◽

Julia Boike ◽

Laurent Bertino

Keyword(s):

Time Series ◽

Data Assimilation ◽

Coefficient Of Variation ◽

Snow Water Equivalent ◽

Mean Squared Error ◽

High Arctic ◽

Sensing Applications ◽

Snow Covered Area ◽

Covered Area ◽

Snow Model

Abstract. Snow, with high albedo, low thermal conductivity and large water holding capacity strongly modulates the surface energy and water balance, thus making it a critical factor in high-latitude and mountain environments. At the same time, already at medium spatial resolutions of 1 km, estimating the average and subgrid variability of the snow water equivalent (SWE) is challenging in remote sensing applications. In this study, we demonstrate an ensemble-based data assimilation scheme to estimate peak SWE distributions at such scales from a simple snow model driven by downscaled reanalysis data. The basic idea is to relate the timing of the snow cover depletion (that is accessible from satellite products) to pre-melt SWE, while at the same time obtaining the subgrid scale distribution. Subgrid SWE is assumed to be lognormally distributed, which can be translated to a modeled time series of fractional snow covered area (fSCA) by means of the snow model. Assimilation of satellite-derived fSCA hence facilitates the constrained estimation of the average SWE and coefficient of variation, while taking into account uncertainties in both the model and assimilated data sets. Our method makes use of the ensemble-smoother with multiple data assimilation (ES-MDA) combined with analytical Gaussian anamorphosis to assimilate time series of MODIS and Sentinel-2 fSCA retrievals. The scheme is applied to high-Arctic sites near Ny Ålesund (79° N, Svalbard, Norway) where in-situ observations of fSCA and SWE distributions are available. The method is able to successfully recover accurate estimates of peak subgrid SWE distributions on most of the occasions considered. Through the ES-MDA assimilation, the root mean squared error (RMSE) for the fSCA, peak mean SWE and subgrid coefficient of variation is improved by around 75 %, 60 % and 20 % respectively when compared to the prior, yielding RMSEs of 0.01, 0.09 m water equivalent (w.e.) and 0.13 respectively. By comparing the performance of the ES-MDA to that of other ensemble-based batch smoother schemes, it was found that the ES-MDA either outperforms or at least nearly matches the performance of the other schemes with regards to various evaluation metrics. Given the modularity of the method, it could prove valuable for a range of satellite-era hydrometeorological reanalyses.

Download Full-text

Evaluating methods for reconstructing large gaps in historic snow depth time series

10.5194/gi-2021-16 ◽

2021 ◽

Author(s):

Johannes Aschauer ◽

Christoph Marty

Keyword(s):

Time Series ◽

Missing Data ◽

Snow Depth ◽

Coefficient Of Determination ◽

Gap Filling ◽

Suitable Alternative ◽

Temperature Index ◽

Distance Weighted ◽

Snow Model ◽

Evaluating Methods

Abstract. Historic measurements are often temporally incomplete and may contain longer periods of missing data whereas climatological analyses require continuous measurement records. This is also valid for historic manual snow depth (HS) measurement time series, where even whole winters can be missing in a station record and suitable methods have to be found to reconstruct the missing data. Daily in-situ HS data from 126 nivo-meteorological stations in Switzerland in an altitudinal range of 230 to 2536 m above sea level is used to compare six different methods for reconstructing long gaps in manual HS time series by performing a "leave-one-winter-out" cross-validation in 21 winters at 33 evaluation stations. Synthetic gaps of one winter length are filled with bias corrected data from the best correlated neighboring station (BSC), inverse distance weighted (IDW) spatial interpolation, a weighted normal ratio (WNR) method, Elastic Net (ENET) regression, Random Forest (RF) regression and a temperature index snow model (SM). Methods that use neighboring station data are tested in two station networks with different density. The ENET, RF, SM and WNR methods are able to reconstruct missing data with a coefficient of determination (r2) above 0.8 regardless of the two station networks used. Median RMSE in the filled winters is below 5 cm for all methods. The two annual climate indicators, average snow depth in a winter (HSavg) and maximum snow depth in a winter (HSmax), can be well reproduced by ENET, RF, SM and WNR with r2 above 0.85 in both station networks. For the inter-station approaches, scores for the number of snow days with HS ≥ 1 cm (dHS1) are clearly weaker and except for BCS positively biased with RMSE of 18–33 days. SM reveals the best performance with r2 of 0.93 and RMSE of 15 days for dHS1. Snow depth seems to be a relatively good-natured parameter when it comes to gap filling of HS data with neighboring stations in a climatological use case. However, when station networks get sparse and if the focus is set on dHS1, temperature index snow models can serve as a suitable alternative to classic inter-station gap filling approaches.

Download Full-text