Simulating snow maps for Norway: Description and statistical evaluation of the seNorge snow model

73Citations
Citations of this article
62Readers
Mendeley users who have this article in their library.

Abstract

Daily maps of snow conditions have been produced in Norway with the seNorge snow model since 2004. The seNorge snow model operates with 1 × 1 km resolution, uses gridded observations of daily temperature and precipitation as its input forcing, and simulates, among others, snow water equivalent (SWE), snow depth (SD), and the snow bulk density (ρ). In this paper the set of equations contained in the seNorge model code is described and a thorough spatiotemporal statistical evaluation of the model performance from 1957-2011 is made using the two major sets of extensive in situ snow measurements that exist for Norway. The evaluation results show that the seNorge model generally overestimates both SWE and ρ, and that the overestimation of SWE increases with elevation throughout the snow season. However, the R2-values for model fit are 0.60 for (log-transformed) SWE and 0.45 for ρ, indicating that after removal of the detected systematic model biases (e.g. by recalibrating the model or expressing snow conditions in relative units) the model performs rather well. The seNorge model provides a relatively simple, not very data-demanding, yet nonetheless process-based method to construct snow maps of high spatiotemporal resolution. It is an especially well suited alternative for operational snow mapping in regions with rugged topography and large spatiotemporal variability in snow conditions, as is the case in the mountainous Norway. © Author(s) 2012. CC Attribution 3.0 License.

Figures

  • Table 1. The 15 seNorge snow model parameters with their default values and units. The values have been set on the basis of literature, expert judgement as well as model evaluation against observations (see Engeset et al., 2004b).
  • Table 2. Features of the two snow data sets used in the seNorge model evaluation (data since 1 September 1957 is considered). SD, SWE and ρ denote snow depth, snow water equivalent and bulk density, respectively, and “m a.s.l.” elevation (meters above sea level).
  • Fig. 1. Spatiotemporal distribution of the snow measurements in the (a) met.no-data (392 547 samples in total, varying from 10 to 1170 per station) and (b) HPC-data (32 256 samples in total, varying from 1 to 239 per station) from 1957–2011. The colours in the map denote the number of observations per station, as indicated in the legend. The contours in (b) show the number of samples per 10× 10 km grid cells (dashed contours denote 15 samples, solid contours 50, 100 and 150 samples), and the city of Trondheim marks the division line between southern and northern Norway.
  • Fig. 2. (a) Seasonal changes in the number of observations, in their median and 25/75 % percentile elevation, as well as in the observed median snow depth (SD), snow water equivalent (SWE) and bulk snow density (ρ) in the met.no- nd HPC-data from 1957-2011. The dashed line denotes the “low SD” subset of the met.no-data where the simulated and/or observe snow depth (SD) is < 10 cm. (b) Seasonal changes in the median and 5, 25, 75 and 95 % percentiles of the distribution of the difference (1SD) and log10-transformed ratio (1SD ∗) between the simulated and observed SD from 1957–2011 in the met.no-data. (c) Same as (b), but now for the HPC-data, for log10-transformed ratio (1SWE∗, 1SD∗) between the simulated and observed SWE and SD, as well as for the difference (1ρ) between the simulated and observed ρ. Only values based on at least 250 measurements (50 in HPC-data) per date are shown. Note that some of the 5 % percentile values in (b) are zero, and thus cannot be properly plotted on log-scale. The total number of observations in the figure is (b) ∼ 392 000 and (c) ∼ 32 000.
  • Fig. 3. Spatial distribution (with elevation and station position) of the station-wise mean log10-transformed ratio (1SD ∗) between simulated and observed snow depth at two selected dates at met.no-stations from 1957–2011. The green, red and blue circles indicate stations with at least 15 years of observations, where the station-wise median1SD∗ is not (green; “good match”), or is detected to be statistically significantly smaller (red; “underestimation”) or larger (blue; “overestimation”) than the threshold window of −23 to +30 % (corresponding to a factor larger than 1.3 deviation from a “perfect match”). A two-sided p-value < 0.05 is applied as he significance level. In the upper panel, the solid and dashed lines denote the median and the 10/90 % percentiles of the station-wise means in each 200-m elevation bin, respectively. The total number of stations is 539 and 252 on 30 March and 29 April, respectively.
  • Fig. 4. Caption on next page.
  • Fig. 4. Median (solid lines) and 10/90 % percentiles (dashed lines) of the difference (grey dots) between the seNorge model-simulated and observed (a) snow water equivalent (SWE), (b) snow depth (SD), and (c) bulk snow density (ρ) at four different dates in snow seasons 1957–2011 in the HPC-data. Statistics are calculated in 200 m elevation-bins (with at least 50 observations; centers of the bins are denoted by black “+” markers). The panels (a) and (b) show the relative log-transformed difference (1SD∗, 1SWE∗) and (c) the absolute difference (1ρ). The coloured circles in (a) and (b) denote the st ions with “good match” (green), “overestimation” (blue) and “underestimation” (red), as explained in Fig. 3.

References Powered by Scopus

A manifesto for the equifinality thesis

1831Citations
N/AReaders
Get full text

Development and test of the distributed HBV-96 hydrological model

871Citations
N/AReaders
Get full text

Estimating snow water equivalent using snow depth data and climate classes

360Citations
N/AReaders
Get full text

Cited by Powered by Scopus

Changing Arctic snow cover: A review of recent developments and assessment of future needs for observations, modelling, and impacts

174Citations
N/AReaders
Get full text

Evidence for changes in the magnitude and frequency of observed rainfall vs. snowmelt driven floods in Norway

126Citations
N/AReaders
Get full text

Transient thermal modeling of permafrost conditions in Southern Norway

101Citations
N/AReaders
Get full text

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Cite

CITATION STYLE

APA

Saloranta, T. M. (2012). Simulating snow maps for Norway: Description and statistical evaluation of the seNorge snow model. Cryosphere, 6(6), 1323–1337. https://doi.org/10.5194/tc-6-1323-2012

Readers over time

‘12‘13‘14‘15‘16‘17‘18‘19‘20‘21‘22‘23‘24‘25036912

Readers' Seniority

Tooltip

PhD / Post grad / Masters / Doc 28

64%

Researcher 12

27%

Professor / Associate Prof. 4

9%

Readers' Discipline

Tooltip

Earth and Planetary Sciences 27

59%

Engineering 9

20%

Environmental Science 9

20%

Computer Science 1

2%

Save time finding and organizing research with Mendeley

Sign up for free
0