Variability of snow depth at the plot scale: implications for mean depth estimation and sampling strategies
- 1Instituto Pirenaico de Ecología, CSIC, Campus de Aula Dei, P.O. Box 202 Zaragoza, 50080, Spain
- 2Watershed Science Program, Warner College of Natural Resources, Colorado State University, Fort Collins, Colorado 80523-1472, USA
- 3Estación Experimental de Aula Dei, CSIC, Campus de Aula Dei, Avda Montañana 1005, Zaragoza 50.016, Spain
- 4Hydrology and Erosion Group, Institute of Environmental Assessment and Water Research (IDǼA-CSIC), Solé i Sabarís, s/n. 08028-Barcelona, Spain
Abstract. Snow depth variability over small distances can affect the representativeness of depth samples taken at the local scale, which are often used to assess the spatial distribution of snow at regional and basin scales. To assess spatial variability at the plot scale, intensive snow depth sampling was conducted during January and April 2009 in 15 plots in the Rio Ésera Valley, central Spanish Pyrenees Mountains. Each plot (10 × 10 m; 100 m2) was subdivided into a grid of 1 m2 squares; sampling at the corners of each square yielded a set of 121 data points that provided an accurate measure of snow depth in the plot (considered as ground truth). The spatial variability of snow depth was then assessed using sampling locations randomly selected within each plot. The plots were highly variable, with coefficients of variation up to 0.25. This indicates that to improve the representativeness of snow depth sampling in a given plot the snow depth measurements should be increased in number and averaged when spatial heterogeneity is substantial.
Snow depth distributions were simulated at the same plot scale under varying levels of standard deviation and spatial autocorrelation, to enable the effect of each factor on snowpack representativeness to be established. The results showed that the snow depth estimation error increased markedly as the standard deviation increased. The results indicated that in general at least five snow depth measurements should be taken in each plot to ensure that the estimation error is <10 %; this applied even under highly heterogeneous conditions. In terms of the spatial configuration of the measurements, the sampling strategy did not impact on the snow depth estimate under lack of spatial autocorrelation. However, with a high spatial autocorrelation a smaller error was obtained when the distance between measurements was greater.