Contributors: Miriam Kosmale, Kari Luojus, Mikko Moisander, Matias Takala, Pinja Venäläinen (Finnish Meteorological Institute)
Issued by: Finnish Meteorological Institute / Miriam Kosmale, Kari Luojus
Date:
Ref: C3S2_313d_ENVEO.WP3-DDP-SWE-v1.0_202506_SWE_PQAR
Official reference number service contract: 2024/C3S2_313d_ENVEO/SC1
History of modifications
List of datasets covered by this document
Acronyms
General definitions
Accuracy: is defined as the “closeness of the agreement between a measured quantity value and a true quantity value of the measurand”. The concept “measurement accuracy” is not a quantity and is not given a numerical quantity value. (GCOS-200, p293)
Brightness Temperature is the measurand of “passive“ microwave remote sensing system (radiometers). Brightness temperature (in degree Kelvin) is a function of kinetic temperature and emissivity. Depending on the depth of a snowback the brightness temperature observed from the satellite and channel varies. The emissivity of a snowpack originates from its snow surface, characteristics of its deeper snow layers and from the ground beneath. Attenuation in emissivity and hence brightness temperature depending on the underlying snow depth is lower at lower frequencies. Means of a brightness temperature gradient between two channels allow for estimating snow characteristics.
Backscatter is the measurand of “active” microwave remote sensing systems (radar). As the energy pulses emitted by the radar hit the surface, a scattering effect occurs and part of the energy is reflected back.
Bias: “Bias is defined as an estimate of the systematic measurement error” (GCOS-200, p293)
Correlation: In statistics the correlation coefficient measures the linear correlation between two sets of data. It can range with values from -1 to 1 that indicates how strongly two sets of ranks are correlated.
Error: “The term error refers to the deviation of a single measurement (estimate) from the true value of the quantity being measured (estimated), which is always unknown” (Gruber et al., 2020)
Freshet: Freshet is used to describe snowmelt an annual high water event on rivers resulting from snow and river ice melting
GCOS Breakthrough (B): An Essential Climate Variable (ECV) requirement level set by Global Climate Observing System (GCOS) which "[...], if achieved, would result in a significant improvement for the targeted application. The breakthrough value may also indicate the level at which specified uses within climate monitoring become possible. It may be appropriate to have different breakthrough values for different uses." (GCOS-244)
GCOS Goal (G): An Essential Climate Variable (ECV) requirement level set by Global Climate Observing System (GCOS) which "[...], if achieved, would be an ideal requirement above which further improvements are not necessary." (GCOS-244)
GCOS Threshold (T): An Essential Climate Variable (ECV) requirement level set by Global Climate Observing System (GCOS) which "[...], if achieved, would be the minimum requirement to be met to ensure that data are useful." (GCOS-244)
Radiometer: Spaceborne radiometers are satellite-carried sensors that measure energy in the microwave domain emitted by the Earth. The amount of radiation emitted by an object in the microwave domain (~1-20 GHz). The observed quantity is called “brightness temperature” and depends on kinetic temperature of an object and its emissivity. Due to the high emissivity of water compared to dry matter, radiometer measurements of Earth’s surface contain information in the water content in the observed area.
RIHMI-WDC: RIHMI is a Federal State Budget Institution operating under operating under Federal Service on Hydrometeorology and Environmental Monitoring of Russian Federation (ROSHYDROMET). RIHMI is hosting World Data Centers in Meteorology and in Oceanography.
RMSE: Root Mean Square Error is the quadratic mean of the differences between observed values and predicted ones. It is a measure of accuracy of a model or predicted value.
Satellite processing level 0: Raw measured signal at satellite: Reconstructed, unprocessed instrument and payload data at full resolution, with any and all communications artifacts (e.g., synchronization frames, communications headers, duplicate data) removed.
Satellite processing level 1: Level 1A (L1A) data are reconstructed, unprocessed instrument data at full resolution, time-referenced, and annotated with ancillary information. L1C data are L1B data that include new variables to describe the measured spectra.
Satellite processing level 2: Derived geophysical variables at the same resolution and location as L1 source data. For SWE products derived from PMR satellite observations the projection is the Equal-Area Scalable Earth (EASE) Grid, which is a system of projections that is used by NASA for distribution of some of the polar orbiting satellite sensors. (ref to EPSG definition: https://epsg.io/6933 and https://epsg.io/6933 and https://nsidc.org/data/user-resources/help-center/guide-ease-grids). EASE grids utilise an equal-area projection which minimises the amount of distortion over the poles, using the Northern and Southern Hemisphere projections, and on other key areas of the globe, using the temperate and global projections.
Satellite processing level 3: Variables mapped on uniform space-time grid scales; for C3S SWE is a regular 0.1 degree latitude-longitude grid and daily temporal resolution.
SMMR: Scanning Multichannel Microwave Radiometer sensor onboard Nimbus-7 satellites in space between 1979-1987. The passive microwave radiometer data is used to derive Snow Water Equivalent.
SSM/I: Special Sensor Microwave/Imager is a seven-channel, four-frequency, linearly polarized passive microwave radiometer system. It is flown on board the United States Air Force Defense Meteorological Satellite Program running from 1987-2008. The passive microwave radiometer data is used to derive Snow Water Equivalent.
SSMIS: Special Sensor Microwave Imager / Sounder is a 24-channel, 21-frequency, linearly polarized passive microwave radiometer system. The instrument is flown on board the United States Air Force Defense Meteorological Satellite Program F-16, F-17, F-18 and F-19 satellites, which were launched in October 2003, November 2006, October 2009, and April 2014, respectively. It is the successor to the Special Sensor Microwave/Imager (SSM/I). The passive microwave radiometer data is used to derive Snow Water Equivalent.
Snow Water Equivalent retrieval algorithm: The here described snow water equivalent (SWE) retrieval algorithm combines information from satellite-based microwave radiometer and optical spectrometer observations with ground-based weather station snow depth (SD) measurements and produces daily hemispherical scale SWE estimates.
Snow Water Equivalent (SWE): SWE represents the resulting water column of a snowpack melt in place. SWE is the product of depth and density of a snowpack. => Snow water equivalent (SWE) = snow depth (SD) * snow density (ρ)
Snowpack: An accumulation of precipitated snow at the surface that compresses with time and melts seasonally. Snowpacks are an important water resource for the surrounding area.
Stability may be thought of as the extent to which the uncertainty of measurement remains constant with time. [...] “stability” refer to the maximum acceptable change in systematic error, usually per decade. (GCOS-200, p293)
Uncertainty: Uncertainty (of measurement) non-negative parameter, associated with the result of a measurement that characterizes the dispersion of the values that could reasonably be attributed to the measurand. (GCOS-200). Satellite retrievals of SWE usually contain considerable systematic errors which, especially for model calibration and refinement, provide better insight when estimated separate from random errors. Therefore, we use the term bias to refer to systematic errors only and the term uncertainty to refer to random errors only, specifically to their standard deviation (or variance).
Executive summary
This report is providing a quality assessment on Snow Water Equivalent (SWE) data records within C3S. The Copernicus Climate Change Service (C3S) on cryosphere offers Climate Data Record (CDR) and consecutively updated Intermediate Climate Data Record (ICDR) products on Snow Water Equivalent (SWE).
The SWE estimates, derived from passive microwave satellite data and ground-based snow depth observations via Bayesian assimilation, are provided on a 0.1° grid with mountainous areas masked. More details about the product is given in the Snow Water Equivalent (SWE) version 1.0: Product User Guide and Specification (PUGS). The product is based on passive microwave radiometer measurements from the Scanning Multichannel Microwave Radiometer sensor (SMMR) aboard NIMBUS-7, and Special Sensor Microwave/Imager (SSM/I) and Special Sensor Microwave Imager / Sounder (SSMIS) sensors aboard DMSP 5D F-series satellites. A detailed description on the algorithm and satellite input datasets is given in the Snow Water Equivalent (SWE) version 1.0: Algorithm Theoretical Basis Document (ATBD).
The validation is conducted on the Snow Water Equivalent (SWE) Climate Data Record (CDR v1.0), covering 1979–2023 over the Northern Hemisphere at daily temporal resolution. Validation is based on in-situ snow course measurements from Russia, Finland, Canada, and the USA. Key validation statistics are unbiased RMSE, bias, and correlation. For SWE ≤150 mm, typical values are: uRMSE ~36 mm, bias ~–1.5 mm, and correlation ~0.65. Validation statistics are presented for the Northern Hemisphere, Eurasia and North America. Additionally, annual and monthly validation statistics are shown in the following sections.
Section 1 gives a short overview of the validation methodology, followed by the validation results presented in section 2. In section 3 and 4 a short discussion is given on climate change assessment and compliance with user requirements defined by GCOS, respectively.
1. Product validation methodology
1.1. Validated products
The C3S SWE dataset is a long-term, daily gridded record of SWE across the Northern Hemisphere, spanning from 1979 to 2023. Extensions to the dataset will be released quarterly to complement the timeseries. The SWE estimates are derived by combining passive microwave radiometer data from SMMR, SSM/I, and SSMIS sensors with ground-based snow depth observations using a Bayesian data assimilation approach. The dataset is provided on a 0.1° × 0.1° grid and includes daily and monthly products. Areas with complex terrain, land ice, water bodies, and coastal Greenland are masked due to known retrieval limitations.
More information on the theoretical and algorithmic basis of the SWE algorithm is presented in the Algorithm theoretical baseline document (ATBD) document, with further details on the product provided in the Snow Water Equivalent (SWE) version 1.0: Product User Guide and Specification (PUGS).
1.2. Ground reference snow course data
Snow courses, also called snow transects, involve manual gravimetric snow measurements taken at multiple points along a predefined path. Individual measurements are averaged to derive a single SWE value for the transect on a given date. The ground reference dataset includes snow course data from Russia, Finland, Canada, and the United States (Figure 1).
Snow course measurements are typically conducted several times per snow season along the same transects. However, transect length, number of sampling points, and aggregation methods vary across the different monitoring programs. Snow transect observations are independent of the snow depth measurements assimilated into the SWE product, but they are not entirely independent of the density measurements used to derive the dynamic density fields. Reported measurement uncertainty for various snow samplers ranges from approximately 3% to 13% (Dixon and Boon, 2012; López-Moreno, 2020).
In Finland, the snow course network is maintained by the Finnish Environment Institute (SYKE), and it includes around 200 spatially distributed transects, each ranging from 2 to 4 km in length. Observations are typically taken once per month (mid-month) between November and April. The dataset spans 1979–2023.
Across the Russia more than 500 snow courses are monitored, with course lengths varying from 1–2 km in open areas to approximately 500 m in forested regions. The data is available via the RIHMI-WDC data centre operating under Federal Service on Hydrometeorology and Environmental Monitoring of Russia. Measurement frequency is every 5 to 10 days in open areas and once per month in forests starting in January (Bulygina et al., 2011). This dataset covers 1979–2024.
For Canada, data are sourced from CanSWE v6 (Vionnet et al., 2021). The network is concentrated in southern population centres, with limited coverage in the tundra and northern boreal regions. Snow course protocols and measurement frequencies vary by agency and location, but typically include two observations per month (early/mid-month and end of month) during the snow season. The Canadian dataset spans 1979–2024.
In the United States, snow course data are provided by the Natural Resources Conservation Service (NRCS) of the U.S. Department of Agriculture. Measurements are generally collected once per month, beginning in January, with the dataset covering the period 1979–2021.
Figure 1: Snow course locations.
1.3. Validation methodology
The reference data is spatially and temporally matched with the modelled SWE grid cells. The mean of all reference observations falling within a SWE product grid cell will be compared to the corresponding co-located SWE estimate. Validation metrics will include the unbiased root-mean-squared error (RMSE), bias, and correlation coefficient. SWE values of zero will be excluded from these computations.
RMSE (Eq. 1), bias (Eq. 2), and correlation (Eq. 3) are defined as:
Eq. 1
Eq. 2
Eq. 3
Eq. 4
where xi and x̄ represent, respectively, the retrieved and mean of the retrieved SWE values and yi and ȳ represent the same for the reference data. The definition of the unbiased RMSE (uRMSE) is given in Eq. 4. As correlation the Pearson correlation is used in this method.
All available matched pairs will be used to calculate the validation statistics, providing a comprehensive assessment of retrieval accuracy. The results will be summarised as follows:
- Spatial Scope: Separate evaluations for the Northern Hemisphere, North America, and Eurasia.
- Temporal Scope:
- Full period: To assess overall product accuracy.
- Monthly: To evaluate changes in accuracy throughout the snow season (temporal stability).
- Annually: To assess inter-annual variability in accuracy.
Monthly and annual statistics will be restricted to the December–May period. For evaluations that require assessment of temporal stability, a temporally consistent subset of in situ observations will be employed. This means only locations that are present every year are included in the evaluation.
2. Validation of Snow Water Equivalent product
2.1. Full period validation
Using in-situ snow course data (Figure 1) and the validation protocol explained above, the CDR v1.0 SWE dataset has a bias of -1.48 mm, an unbiased RMSE of 35.8 mm, and a correlation of 0.65 for SWE values ≤150 mm (Table 1). Performance declines with higher SWE thresholds (up to 500 mm). This behaviour is expected as the signal starts to saturate when SWE values are larger than 150 mm. Because of known limitations with high snow depths especially occurring in alpine terrain, a complex-terrain mask is applied. Additionally the native satellite pixel size of approximate 25km is not able to resolve such small-scale features in elevation change. A high-resolution digital elevation model excludes therefore mountain pixels for the retrieval. Retrieval performs better in Eurasia, where snow packs are typically more moderate than in North America. Additionally, the North American reference dataset contains measurements from high mountain plateaus that are not masked out from the SWE retrieval but have very high SWE values. Very small amounts of snow (SWE <50 mm) cannot be reliable retrieved because the brightness temperature difference between the two frequencies, falls within 2 K detection precision of the radiometer instruments used. Even relatively small amount of liquid water will contaminate the signal making retrieval of SWE difficult.
Table 1: Validation statistics for 1980-2022 for Northern Hemisphere, Eurasia and North America. Statistics are calculated separately for SWE < 150 mm and SWE < 500 mm.
| uRMSE (mm) | Bias (mm) | Correlation | ||
|---|---|---|---|---|
| Northern Hemisphere | <150 mm | 35.8 | -1.48 | 0.65 |
| <500 mm | 47.5 | -9.4 | 0.68 | |
| Eurasia | <150 mm | 34.1 | -0.89 | 0.67 |
| <500 mm | 44.2 | -7.5 | 0.70 | |
| North America | <150 mm | 41.0 | -3.5 | 0.53 |
| <500 mm | 56.8 | -15.7 | 0.62 |
Figures 2 and 3 illustrate the validation results for v1.0 of the product. Similarly to Table 1, the performance of the product is better in Eurasia. Performance starts to decline for SWE > 200 mm.
Figure 2: Scatter plots of the estimated versus snow course measured SWE for the Northern Hemisphere, Eurasia, and North America for 1979-2023.
Figure 3: Accuracy plot of the v1.0 SWE product for the Northern Hemisphere, Eurasia, and North America.
2.2. Monthly validation
Figure 4 displays validation statistics for each month from December to May. Statistics are calculated separately for SWE < 150 mm and SWE < 500 mm. uRMSE increases as the snow season progresses and snowpacks get deeper, and there is more liquid water present. Similarly, bias becomes largely negative during the late winter (April and May). Correlation is relatively stable during mid-winter and slightly smaller during early and late winter.
Figure 4: Monthly validation statistics for the Northern Hemisphere 1980-2022 based on the full reference dataset (Figure 1). Left shows statistics for SWE<150mm and right shows statsitical values for SWE < 500mm.
2.3. Annual validation
Figure 5 shows validation statistics for each year between 1989 and 2022. Statistics are shown separately for SWE < 150 mm and SWE < 500 mm. These statistics have been calculated using a temporally consistent set of validation data, meaning only reference snow courses producing data every year are included in the validation. Long-term trends show decreasing uRMSE and increasing correlation. The CDR v1.0 utilises passive microwave data from three different sensors (details on the input data in the Snow Water Equivalent (SWE) version 1.0: Algorithm Theoretical Basis Document (ATBD)), which explains some of the year-to-year variability.
Figure 5: Annual validation statistics for the Northern Hemisphere 1980-2022 based on a temporally consistent reference dataset. Left shows statistics for SWE<150mm and right shows statistical values for SWE < 500mm.
2.4. Summary
The C3S Snow Water Equivalent satellite-based product has been validated against independent snow transect data from Canada, Russia and Finland, to give a good overview on the quality of this long term time series for SWE on the Northern Hemisphere.
The presented validation results show good correlation with respect to in-situ observations during the peak winter seasons, which for the Northern hemisphere are typically between January and March (figure 4). Moreover the lowest bias and RMSE values are reported between December and March. With this the SWE product demonstrates a sufficient ability to observe snow conditions during the winter season.
Validation results decline for deep snow conditions (SWE values above 300 mm). This behaviour is related to saturation of the observed signal from the satellite sensor. For these high SWE conditions the retrieval underestimates the SWE and they have to be assessed with care. Mountainous areas are flagged out from the final product due to retrieval challenges in complex terrain and thick snow packs in those areas. Very wet snow conditions and snow conditions with low snow depths need to be treated carefully as well. Such conditions typically occur during the spring snow melt season and early winter season with low snow depths. Regardless of these known limitations, overall the SWE product performs quite well with respect to in-situ measurements, as shown in this report. The retrieval algorithm performs best for SWE values between 50 to 300 mm and offers a reliable product on Snow Water Equivalent for Northern Hemispheric winters.
With a correlation around 0.68 on average throughout the full climate data record period (figure 5) this C3S product provides stable observations on SWE from the existing passive microwave satellite sensors. As the method utilises passive microwave radiometer data for retrieval of the snow, it is completely unaffected by clouds or polar night conditions, which is an additional advantage for the spatial coverage and applicability of this product.
This long-term data record on Snow Water Equivalent by C3S can help facilitate new application developments related to winter snow conditions as well as support further climate research.
More details on potential applications and usage of such long-term SWE observations, including strengths and limitations of the product, are also summarised in the Snow Water Equivalent (SWE) version 1.0: Product User Guide and Specification (PUGS).
3. Climate Change Assessment
The here presented C3S SWE product is available as long-term climate data record based on satellite observations going back to 1979. Having more then 40 years of continuous satellite observations on snow parameter like SWE allows for relevant climate studies. Figure 6 shows the total snow mass for Northern Hemisphere 1980-2022 derived from C3S SWE v1.0 product. The maximum snow mass concentration occurs in late winter, typically between February and April for the Northern hemisphere. It shows on average decreasing snow mass concentrations for the latest years compared to observations derived from earlier years before 1999.
Similar observations were also summarised in Nature by Pulliainen et al. 2020. Continuous and long-term data records, such as those provided within C3S on Snow Water Equivalent, support these and other research activities for climate assessment.
Figure 6: Total snow mass for Northern Hemisphere 1980-2022 derived from C3S SWE v1.0 product
4. Compliance with user requirements concerning data quality
Table 2 assembles the C3S ECV Snow Water Equivalent product target requirements adopted from the Global Climate Observing System (GCOS)-245 requirements and shows to what extent these requirements are currently met by the products. As one can see, the CDR product currently provided by the system are compliant in most cases with the GCOS target requirements.
Table 2: Compliance with GCOS-245 Requirements, 2022 edition - updated in 2025 for the ECV product "Snow-Water Equivalent".
Requirement | GCOS-245 Requirement | Reported value | ||
G | B | T | ||
Horizontal Resolution (km) | 0.5 | 5 | 25 | Target: 0.1° (ca 10 km) |
Vertical Resolution | - | - | - | N/A |
Temporal Resolution | 3h | Daily | Monthly | Breakthrough: daily |
Required Measurement Uncertainty (mm) | 10 or 10%, whichever is greater | 20 or 20%, whichever is greater | 40 or 40%, whichever is greater | Target: 95 mm (< 40 %) |
Stability (m/decade) | 0.01 | 0.05 | 0.1 | These values still lack justification in the scientific literature. |
References
Bulygina, O. N., Groisman, P. Ya., Razuvaev, V. N., and Korshunova, N., N. (2011). Changes in snow cover characteristics over Northern Eurasia since 1966, Environmental Research Letters, 6(4): 045204
Dixon D. and Boon, S. (2012). Comparison of the SnowHydro snow sampler with existing snow tube designs, Hydrol. Process., 26, 2555–2562, https://doi.org/10.1002/hyp.9317
GCOS-200, The Global Observing System for Climate, Implementation Needs, https://library.wmo.int/idurl/4/55469
GCOS-244, Climate Observing System (GCOS), ECVs on snow, https://gcos.wmo.int/site/global-climate-observing-system-gcos/essential-climate-variables/snow
GCOS-245, The 2022 GCOS ECVs Requirements, 2022 edition - Updated in 2025
López Moreno, J.I., Leppänen, L., Luks, B., Holko, L., Picard, G., Sanmiguel Vallelado, A., Alonso González, E., Finger, D.C., Arslan, A.N., Gillemot, K., Sensoy, A., Sorman, A., ErtaÅ, M.C., Fassnacht, S.R., Fierz, C., and Marty, C. (2020). Intercomparison of measurements of bulk snow density and water equivalent of snow cover with snow core samplers: Instrumental bias and variability induced by observers, Hydrol. Process., 34, 3120–3133, https://doi.org/10.1002/hyp.13785
Pulliainen, J., Luojus, K., Derksen, C. et al. (2020). Patterns and trends of Northern Hemisphere snow mass from 1980 to 2018, Nature 581, https://doi.org/10.1038/s41586-020-2258-0
Vionnet, Vincent, Mortimer, Colleen, Brady, Mike, Arnal, Louise, & Brown, Ross. (2021). Canadian historical Snow Water Equivalent dataset (CanSWE, 1928-2020) (Version v1), Zenodo, http://doi.org/10.5281/zenodo.4734372
