Overview
This document provides an overview of the snow depth gridded dataset created by the Canadian Forest Service, Natural Resources Canada. The dataset provides gridded snow depth from 1955 to 2017 at the daily, pentad, monthly and 30-year average time scales over Canada. It was generated using ANUSPLIN with a 60 arc-second (approximately 2 km) and with a 300 arc-second (approximately 10 km) Digital Elevation Model (DEM; Lawrence et al., 2008) and station observations from Environment and Climate Change Canada (ECCC) and the National Oceanic and Atmospheric Administration (NOAA).
Provider's contact information
This spatial dataset was developed by researchers from the Integrated Ecology and Economics Division at Canadian Forest Service at Natural Resources Canada (Dr. Dan McKenney, Dr. Heather MacDonald, John Pedlar, Kevin Lawrence, Pia Papadopol, Kaitlin de Boer from the CFS and Dr. Michael Hutchinson from Fenner School of Environment and Society, Australian National University). If you have questions about this dataset, please contact Dr. Dan McKenney (dan.mckenney@canada.ca)
Licensing
Freely available. Contact Dr. Dan McKenney, Canadian Forest Service, Natural Resources Canada, dan.mckenney@canada.ca
The following data citations should be used if you use these data in publications:
MacDonald, H., and D. McKenney, 2021: Canadian Snow Depth Spatial Models, 1950-2017. OSF. July 7. doi:10.17605/OSF.IO/UZAC9.
MacDonald, H., and D. McKenney, 2021: Comparison of Snow Depth Spatial Models Generated for Year 2000 Using Global Historical Climate Network Daily (GHCN-D) and Environment and Climate Change Canada (ECCC) in Situ Station Records.” OSF. July 6. osf.io/abhng. DOI 10.17605/OSF.IO/ABHNG
Variable name and units:
Daily - Snow depth in centimetres (cm);
Pentad – Average 5-day snow depth in cm (cm);
Monthly – Maximum snow depth in centimetres (cm);
30-year Average – monthly average snow depth over 30 years in centimetres (cm)
Spatial coverage and resolution:
Two products are available over Canada:
1) at 60 arc-second (approximately 2 km),
2) at 300 arc second spatial resolution (approximately 10 km)
Temporal coverage and resolution:
Daily – 1955-2017
Pentad – 1955-2017
Monthly – 1955-2017
30-year Average – 1961/1990, 1971/2000
The data will continue to be updated regularly.
Information about observations (number, homogeneity)
The spatial gridded data of 30-year averages were constructed using 1224 stations for the 1961-1990 period and 1723 stations for the 1971-2000 period. For monthly, pentad and daily data, the number of ECCC observations ranged from 220 in 1955 to more than 2000 observations in the 1990s, declining to fewer than 2000 stations post-1996 (Figure 1). The data are not homogenized for time series analysis and are not evenly distributed in space (many stations are located in southern Canada and fewer in northern Canada). Brown et al. (2021) provide an analysis of in situ data over Canada. In-house analyses at CFS indicated that there are differences between the spatial gridded models developed in ANUSPLIN using just ECCC stations versus those using just GHCD stations (the surface means for the data developed using GHCD data are higher than those created using ECCC data). Based on Brown et al. (2021) analysis of ECCC in-situ measurements and the in-house comparison, it was decided to construct the present dataset using ECCC stations for 1955-2002 period and NOAA’s GHCD station observations for the 2000-2017 period.
Figure 1. Number of station observations with snow depth measurements from ECCC and GHCN.
Methodology
Snow depth station observations were used to create thin-plate spline models in ANUSPLIN (Hutchinson & Xu, 2013). Snow depth observations from ECCC were available from 1955-2002 period. Those were completed with snow depth observations from the National Oceanic and Atmospheric Administration (NOAA)’s GHCD[1] from 2000-2017 period. Station records were not joined, and instead each record was linked to the exact observation location (see description in MacDonald et al., 2021). Details on methodology are presented in the T1 technical documentation (provided by CFS/NRcan) at the end of this document.
Information about the technical and scientific quality
The data were quality controlled and use up-to-date spatial modelling methods as detailed in MacDonald et al., 2021. See also Brown et al. (2021) for a description of the application and assessment of snow depth estimates generated from these models.
Limitations and strengths for application in North Canada
These spatial models are based on in-situ stations. As noted by Brown et al. (2021), the station network exhibits “greater reliance on automated sensors over northern regions of Canada (e.g., over 80% of the Canadian surface SD-observing network north of 55°N was equipped with sonic sensors).” With respect to data quality of models based on in-situ observing stations, Brown et al. (2021) conclude that “the spatial and temporal coverage is insufficient to provide a complete picture of snow cover trends across Canada, but it does provide information for the most populated regions and at the community level across northern Canada.”
References to documents describing the methodology or/and the dataset
Brown, R. D., C. Smith, C. Derksen, and L. Mudryk, 2021: Canadian in situ snow cover trends for 1955-2017 including an assessment of the impact of automation. Atmosphere-Ocean, 59(2), 77-92. https://doi.org/10.1080/07055900.2021.1911781
Lawrence, K. M., M. F. Hutchinson, and D. W. McKenney, 2008: Multi-scale digital elevation models for Canada. Natural Resources Canada, Great Lakes Forestry Centre Frontline Tech. Note 109, 4 pp., https://d1ied5g1xfgpx8.cloudfront.net/pdfs/31499.pdf.
Hutchinson, M. F., and T. Xu, 2013: ANUSPLIN version 4.4 user guide. Australian National University, Fenner School of Environment and Society Doc., 55 pp., https://fennerschool.anu.edu.au/files/anusplin44.pdf.
MacDonald, H., D.W. McKenney, P. Papadopol, K. Lawrence, J. Pedlar, and M. F. Hutchinson, 2020: North American historical monthly spatial climate dataset, 1901–2016. Scientific Data, 7(1), 411-411. https://doi.org/10.1038/s41597-020-00737-2
*MacDonald, H., D. W. McKenney, X. L. Wang, J. Pedlar, P. Papadopol, K. Lawrence, Y. Feng, and M. F. Hutchinson, 2021: Spatial models of adjusted precipitation for Canada at varying time scales. Journal of Applied Meteorology and Climatology, 60(3), 291-304. https://doi.org/10.1175/JAMC-D-20-0041.1
*Describes spatial models at the pentad time scale for adjusted precipitation, ANUSPLIN-AdjPdly
Technical documentation:
See T1 below detailing comparison of ECCC and GHCD spatial models for 2000.
Link to download the data and format of data:
All data is available in asci format at Canadian Forest Service, NRCan (contact Dr. Dan McKenney, dan.mckenney@canada.ca if you are interested in the asci datasets)
Monthly dataset with 300 arc-second spatial resolution in netCDF format is available at Canadian Center for Climate Services, ECCC (contact the Climate Services Support Desk, https://climate-change.canada.ca/support-desk/Inquiry, if you are interested in this netCDF subset of data)
Publications including dataset evaluation or comparison with other data in northern Canada
This dataset is a new product and was not used yet in scientific publications over northern Canada.
T1. Technical documentation Comparison of Snow Depth Spatial Datasets developed from station observations from Environment and Climate Change Canada (ECCC) and from Global Historical Climate Data (GHCD) for the year 2000
We extended a series of snow depth spatial models developed using station observation data from 1955 to 2002 from Environment and Climate Change Canada using Global Historical Climate Data (GHCD) from 2000 to 2017. This technical documentation is comparing ANUSPLIN-generated statistics for the year 2000 that were obtained for spatial gridded models of snow depth developed using (1) ECCC and (2) GHCD stations. The analysis includes:
- Root GCV (RTGCV), or the square root of the generalized cross validation (GCV). The GCV is calculated for each value of the smoothing parameter ρ by implicitly removing each data point and calculating the residual from the omitted data point of a surface fitted to all other data points (Wahba 1990).
- Ratio of signal to the number of knots (NK) (the S:NK ratio). The signal is a measure of the complexity of the fitted surface that ranges between a small positive integer and the number of stations used to create the model (Wahba 1990). S:NK ratios greater than 0.8 or less than 0.2 represent problematic spatial surfaces (i.e. >0.8 or <0.2; Hutchinson & Xu, 2013).
Number of Stations in the Comparison
The ECCC dataset comprised 1984 observing stations overall compared to 1963 Canadian stations in the GHCD (see tables 2 and 3).
Results of the Comparison
The Root GCV (spatially averaged standard error, see Hutchinson & Xu, 2003) is somewhat lower for the spatial models using ECCC compared to GHCD, but comparable overall. The surface mean tends to be higher for spatial models built from GHCD data compared to those from ECCC data (Figure 2).
The S:NK ratio was problematic for ECCC snow depth surfaces for April, May and September 2000. For the GHCD, the April 2000 surface was problematic (red text).
Conclusion
There are differences between the spatial models developed in ANUSPLIN using ECCC versus GHCD data. Specifically, the surface means for the spatial models developed using GHCD data had higher surface means compared to those created using ECCC data. The spatial models presented here have not been homogenized for time series analysis. However, Brown et al. (2021) provide analysis of in situ snow cover trends over Canada. Future efforts will include accessing GHCD snow depth data for the entire time period.
Table 1: Statistics for Spatial Model (year=2000) using GHCD snow depth
| Month | Number of Points | Error* | Signal* | Ratio* | Rt GCV (cm) | Surface Mean (cm) |
|---|---|---|---|---|---|---|
| 1 | 1084 | 661.8 | 422.2 | 64.36 | 11.73 | 25.55 |
| 2 | 1147 | 696.2 | 450.8 | 65.33 | 11.47 | 17.28 |
| 3 | 1229 | 688.7 | 540.3 | 75.57 | 5.51 | 8.79 |
| * 4 | 1457 | 748.7 | 708.3 | 83.53 | 1.45 | 2.70 |
| 5 | 1529 | 908 | 621 | 68.62 | 0.24 | 0.35 |
| 9 | 1590 | 950.6 | 639.4 | 68.83 | 0.03 | .09 |
| 10 | 1408 | 833.6 | 574.4 | 71.44 | 0.44 | 0.77 |
| 11 | 1137 | 700.1 | 436.9 | 65.8 | 3.67 | 6.66 |
| 12 | 1076 | 722.1 | 353.9 | 55.38 | 10.48 | 25.88 |
Table 2: Statistics for Spatial Model (year=2000) using ECCC snow depth
| Mth | Number of Points | Error* | Signal* | Ratio* | Rt GCV (cm) | Surface Mean (cm) |
|---|---|---|---|---|---|---|
| 1 | 1402 | 885.5 | 516.5 | 62.53 | 11.35 | 22.76 |
| 2 | 1406 | 902.3 | 503.7 | 61.05 | 10.91 | 15.84 |
| 3 | 1435 | 822.2 | 612.8 | 73.21 | 6.11 | 8.62 |
| * 4 | 1530 | 708.2 | 821.8 | 92.03 | 1.28 | 3.32 |
| * 5 | 1577 | 654.3 | 922.7 | 98.9 | 0.10 | 0.63 |
| * 9 | 1634 | 677.8 | 956.2 | 99.92 | 0.002 | 0.12 |
| 10 | 1527 | 953 | 574 | 65.08 | 0.70 | 0.98 |
| 11 | 1440 | 1010.1 | 429.9 | 51.86 | 4.63 | 6.50 |
| 12 | 1382 | 980.5 | 401.5 | 49.63 | 11.37 | 24.00 |
- Signal to the number of knots (S:NK) ratio. . S:NK Ratios >80 or <20 represent problematic surfaces (Hutchinson & Xu, 2013).
Comparison of Snow Depth Spatial Model Surface Means by Data Source
References
Brown, R. D., C. Smith, C. Derksen, and L. Mudryk, 2021: Canadian in situ snow cover trends for 1955-2017 including an assessment of the impact of automation. Atmosphere-Ocean, 59(2), 77-92. https://doi.org/10.1080/07055900.2021.1911781
Hutchinson, M. F., and T. Xu, 2013: ANUSPLIN version 4.4 user guide. Australian National University, Fenner School of Environment and Society Doc., 55 pp., https://fennerschool.anu.edu.au/files/anusplin44.pdf.
Wahba, G., 1990: Spline Models for Observational Data. CBMS-NSF Regional Conference Series in Applied Mathematics, Vol. 59, SIAM, 161 pp., https://doi.org/10.1137/1.9781611970128.