We present performance of the physical snowpack model SMAP, which was applied to Japan Meteorological Agency (JMA)ʼs AMeDAS (Automated Meteorological Data Acquisition System) sites in Niigata, Japan during the 2015-2016 winter to simulate temporal evolution of snow depth. Input precipitation, air temperature, and wind speed data for the SMAP model were obtained from in-situ AMeDAS data, where precipitation were corrected considering catch efficiency of each rain gauge. Before performing this correction, it is necessary to discriminate measured precipitation into snow and rain. In the present study, air temperature was employed as a criterion for the discrimination (Tdisc), and three input precipitation data were created by modulating Tdisc within the realistic range: 0, 0.5, and 1℃.Other necessary data to drive the SMAP model (downward shortwave and longwave radiant fluxes, cloud fraction, relative humidity, and air pressure) were prepared by running the JMA Nonhydrostatic Mesoscale Model. The best performance of the SMAP model in terms of snow depth at Nagaoka was obtained when Tdisc was set to be 0.5℃; however, the best scores were obtained when Tdisc was 0℃ at most of the sites. We also found that the SMAP model simulations in Niigata during the period were very sensitive to the choice of Tdisc: average differences in simulated snow depths for the cases Tdisc=0 and 1℃ reached 0.41m at Koide. On the other hand, the difference was at most 0.10m at Sekiyama, suggesting that sensitivities to Tdisc of model simulations are different from place to place. By investigating the winter precipitation curve at each site, it was clarified that the above-mentioned sensitivities were controlled by the winter total precipitation amount, as well as the representative air temperature at which precipitation frequently occurred during the winter period.
View full abstract