| Seasonal snow is one of the most active components of cryosphere.It has high albedo and low conductivity,which not only change local hydrologic cycle,but also affect energy budget.We focus on seasonal snow in Northern Hemisphere?NH?,and investigate the variation and distribution of snow from 1980 to 2014 by analyzing MERRA2 reanalysis data and CMIP6 historical simulation data.We also compare the future snow process under different emission scenarios in CMIP6 future predicted data.Thus,we combine the land cover type data and NDVI data from MODIS with snow data to investigate the interaction between snow and vegetation in middle and high latitudes of NH.And the simulated snow process in CMIP6 is also evaluated.Finally,we analyze the differences between the simulation result and reanalysis data,and make corrections to the simulation result.The main conclusions are as follows:?1?In the period of 1980-2014,the seasonal snow in reanalysis data mainly distributes in middle and high latitudes of NH.Both snow cover and snow depth show a decrease trend with the rate of-2.85×10-2%/a and-2.91×10-4 m/a,respectively.Meanwhile,the variations of snow cover and snow depth have significant seasonal characteristics.Snow cover shows the most significant declines in spring and autumn,with the rate of-3.10×10-2%/a and-4.48×10-2%/a,respectively.Snow depth rapidly decreases in spring and winter,with the rate of-4.08×10-4 m/a and-3.48×10-4 m/a,respectively.The areas with the most significant decline in snow cover and snow depth are found in the Rocky Mountains,Europe and the Tibetan Plateau.However,the trends of snow cover and snow depth in historical simulation are higher than those in reanalysis data.This is because that the model does not reproduce the increase trend in middle latitudes.Under different emission scenarios,both snow cover and snow depth will continue to reduce in the future.Under the SSP1-2.6,snow cover and snow depth slightly decrease from 2015 to 2040,and have change little after 2040.Under the SSP5-8.5,snow cover and snow depth obviously decrease,and are only half in 2100.?2?Snow cover and snow depth show different characteristics under different vegetation types during snow accumulation stage and melt stage.In the middle and high latitudes of the NH,snow cover and snow depth first begin to accumulate and reach maximum in open shrubland,followed by evergreen needleleaf forest,and mixed forest.As a result,the curves of relationship between snow cover and snow depth have different patterns.With the climate changes,the relationship between snow cover and snow depth also changes,which affects vegetation by changing the growth season and life cycle.From 2001 to 2012,evergreen needleleaf forest and mixed forest in the middle and high latitudes of NH continue to expand northward,while the open shrubland shrink overall,which leads to a corresponding increase in NDVI.However,the CMIP6 models do not reproduce the change in the snow process,especially in snow accumulation and melt stage,which is significantly different from the reanalysis data.In different emission scenario,the snow cover and snow depth will continue to decrease,and the snow process will also change in the future.?3?The CMIP6 historical simulation shows different average and trend,comparing with MERRA2 reanalysis data.The areas with the most obvious difference in snow cover and snow depth are the Rocky Mountains and the Tibet Plateau,where the largest differences in annual average are more than 20%and 0.1 m,respectively.And the seasons with the most obvious differences in average and trend are spring and winter.We use the annual average to establish a linear regression,and then correct the model simulation,which lead to the decreasing difference of annual average between model simulation and reanalysis data.However,the seasonal average and trend are still significant.After correcting each grid by using seasonal data,the difference of seasonal average is further reduced,but the difference of trend is not much change.Therefore,the inter-decadal component and long-time trend in model are replaced by the component in reanalysis data by using EEMD.As a result,the differences of average and trend significantly decrease in most regions. |
PDF Full Text Request