Representativeness of local snow data for large scale hydrologic investigations Daqing Yang 1 and Ming-Ko Woo 2 * 1 Water and Environmental Research Center, University of Alaska, Fairbanks, AK 99775, USA 2 School of Geography and Geology, McMaster University, Hamilton, Ontario, Canada L8S 4K1 Abstract: Arctic snow cover usually attains maximum values at the end of winter and such information is important for hydrological investigations because most ¯oods are associated with spring snowmelt. Snow data from weather stations or collected at some local sites are often extrapolated to large areas, but without verifying that the upscaling procedureyields correct results. This study compares maximum snow cover data gathered over two large target areas (170 to 300 km 2 ) with weather station snow course measurements to determine the representativeness of local-scale data for areas typically occupied by large grid cells of macro-hydrological models. The ®eld snow survey results con®rmed the controlling role of terrain on snow distribution in the High Arctic. The variability of areal mean snow water equivalence for a grid cell (with dimensions of 1  1 km 2 to 13  13 km 2 ) increases with terrain complexity but decreases with grid size. Although point data do not represent the snow cover over an area, an attempt was made to upscale the weather station data to the target areas using an index method. Test results show that this index approach works well in the areawith a shallow snow cover, but the error increases for an area with relatively deep snow. More eort is needed to re®ne this method, perhaps in conjunction with remote sensing, so that point data can be upscaled to yield snow information suitable for large-scale hydrological models or land surface schemes. Copyright # 1999 John Wiley & Sons, Ltd. KEY WORDS Arctic; extrapolation error; snow distribution; snow survey; snow index method; upscaling INTRODUCTION For the Arctic, the end-of-winter snow measurements are particularly important because the precipitation stored over the long winter months is often released rapidly by spring melt to generate the peak ¯ows of the season. For regional hydrological models and land surface schemes that model snowmelt of large areas, snow information on a large scale is needed (Kirnbauer et al., 1994). Information collected by weather stations and conventional snow surveys are point data incompatible with the scale of these models. Site-scale measurements must be extrapolated over an area (Davis et al., 1995) and several factors should be con- sidered: (1) complexity of the terrain in the study area, (2) snow cover variability between years, and (3) representativeness of the microscale snow data for the meso-scale domain. There are several options in obtaining snow information for the initialization of distributed models for snowmelt computations: (1) apply weather station snow data directly to the target area without reference to the spatial variability of snow distribution, an approach that is sometimes practised by lumped models; (2) conduct an extensive ®eld survey to map the snow water equivalent, traditionally for drainage basins, CCC 0885±6087/99/121977±12$17 . 50 Received 30 April 1998 Copyright # 1999 John Wiley & Sons, Ltd. Revised 2 October 1998 Accepted 18 March 1999 HYDROLOGICAL PROCESSES Hydrol. Process. 13, 1977±1988 (1999) *Correspondence to: Dr Ming-Ko Woo, School of Geography and Geology, McMaster University, Hamilton, Ontario, Canada L8S 4K1. Contract grant sponsor: Atmospheric Environment Service (AES) of Environment Canada.