Characteristics Of Snowpack In Major Forest Types And Remote Sensing Estimation Of Northern Daxing’anling Mountains

Relations between the forest and water cycle was a hot topic of current research. Snow was solid water, hence snow was an important part of the forest water cycle. Meanwhile, the effect of snow cover on the climate, the natural environment and human activities could not be ignored due to its ecological significance. Daxing’anling Mountains was located in high latitudes, which was one of the major state-owned forest region. It was the natural barrier of Northeast China and North China Plain, and has great significance to maintain the ecological balance in Northeast Asia.This study was supported by the MOHE Forest Ecosystem Research Station. The study area was in Mohe forests of northern Daxing’anling. In order to study the impact of three major forest types on snowfall and snowpack, and get a deeper understanding of the relationship between forests and snow, and further explore the forest snow hydrology process mechanism, we observed the snow hydro logical variables of snowfall amount, snow physical characteristics, and snow surface evaporation in the Daxing’anling Mountains in 2014 winter. To provide a scientific basis of water regulation mechanism of forest vegetation, Larix gmelinii forest, Pinus sylvestris var. mongolica forest and Betula platyphylla and Populus davidiana forest were studied in northern Daxing’anling Mountains.Simultaneously, in order to explore the method of inversion of snow water equivalent(SWE) using optical remote sensing, and estimate the region spatial distribution of SWE in northern Daxing’anling Mountains, and implement of monitoring snow hydrological processes within the forest area, a systematic inversion model of SWE was established by using the method of traditional multiple linear regression, partial least squares regression(PLS) and BP neural network. Comparing the results of three models, the optimal model was selected to inverse and estimate the snow hydrological variables, we selected to. The results indicated that:1) Interception of Forest snowfall was mainly affected by canopy density and composition of the forest. The snow retention of different forest types in the same grade were ranked as follows:Pinus sylvestris var. mongolica forest> Larix gmelinii forest> Betula platyphylla and Populus davidiana forest. With increasing in the levels of snowfall, snow forests interception showed a gradual decreasing trend. Pinus sylvestris var. mongolica forest had the maximum amount of snowfall interception and rate with the values of 11.24 mm and 22.54%, respectively, which were almost 2 times of Larix gmelinii forest, and 5 times of Betula platyphylla and Populus davidiana forest.2) Snow had regulatory effects on snow hydro logical processes. The presence of forest was beneficial to reserve ground water resource. Due to the evaporation, the loss of snow in open field was 22.49 mm throughout the observation period, which was 45%of the amount of snow. It is much higher than that in the forest.3) There were significant differences among the snow depth features of different forest types. The snow depth of evergreen forest (Pinus sylvestris var. mongolica forest) was less than deciduous orests, of which Larix gmelinii forest had the maximum snow depth of 27.92 cm, Pinus sylvestris var. mongolica forest had the minimum snow depth of 23.56 cm.4) Forest types, snowfall, the external environment and other factors directly affect the snow density variations within forest. The snow density variations of Larix gmelinii forest and Betula platyphylla and Populus davidiana forest have the similar trends, but the snow density of Pinus sylvestris var. mongolica forest have the smaller changes.5) During the entire period of observation, the SWE in different forest types all showed significant differences (P< 0.05). Forest snow evaporation loss was a factor resulting in such difference, and the difference of intraforest snowfall input is the dominant factor. The SWE in Pinus sylvestris var. mongolica forest was the minimum with an amount of 26.49 mm, while the Betula platyphylla and Populus davidiana forests had the maximum SWE of 39.18 mm.6) Comparing the results of different inversion models, PLS model was superior to traditional multiple linear regression model. According to the results of regression level, model stability and prediction accuracy, BP neural network model was better than the linear model. Its RMSE, rRMSE, average fitting accuracy, average prediction accuracy and highest prediction accuracy was up to 2.83 mm,0.23,79.9%,79.3%, and 81.9%, respectively.7) The SWE spatial distribution of study area was inversed by using the optimal model. The results indicated that the average ground SWE was 11.8 mm and the spatial distribution of SWE was related to topography and land use.