The Impact Of Snowpack On Soil Hydrothermal In Typical Cropland Regions Of Northeast China

Northeast China is the most potential food production area in China.The high organic matter content of the soil,fertile soil and the wide and concentrated distribution of cultivated land have made Northeast China the largest commercial grain base in China.As one of the three major seasonal snow areas in China,Northeast China is covered with snow for a long time and has a wide distribution of permafrost.Snowpack is essential for the maintenance and development of regional climate by regulating the regional energy balance and water cycle.The presence of snowpack directly affects soil temperature and the timing of soil freezing,thus affecting the metabolic activity of microorganisms in early spring.Moreover,as one of the important sources of freshwater resources in the Northeast China,snowmelt directly affects spring moisture conditions,and in turn has an impact on spring cultivation and crop growth.Northeast China is a climate change sensitive region,and climate warming will change the spatial distribution of snowpack and affect hydrothermal soil conditions,which has a significant impact on terrestrial ecosystems and regional agricultural and forestry production.Therefore,accurate monitoring of changes in the spatial pattern of hydrothermal conditions of cropland areas in Northeast China with climate change background and understanding the impact of snowpack on soil hydrothermal conditions are essential to ensure the safety of food production,optimize the agricultural planting structure,develop sustainable agriculture and maintain the balance of agroecosystems.To understand the changes of near-surface soil thermal state under the impact of snowpack in the cropland,this study used the Snow Thermal Model and introduced the concept of spatial and temporal velocity,combined with the regional reanalysis data and meteorological station observation data to analyze the spatial and temporal changes of near-surface soil temperature and the spatial pattern changes in the cropland of Northeast China,and explored the response relationship between snowpack and near-surface soil temperature.In order to understand the changes of near-surface soil moisture under the influence of snowpack,this study selected typical farmland areas in the Northeast China and build a spring soil moisture prediction model based on the currently used machine learning algorithms,combining climate,vegetation,topography,soil and other factors,to investigate the effect of winter snowfall on spring soil moisture in cropland.The main work and conclusions are as follows.（1）Snow thermal model（SNTHERM）was used to explore the effect of snowpack on snow-soil interface temperatures.In this paper,we localize the plasticity index in the SNTHERM model based on the soil physicochemical properties in Northeast China,and set the initial snow layer number,snow temperature,liquid water content and snow grain size in the SNTHERM model based on the snow characteristics in Northeast China,and initial the SNTHERM model multilayer soil temperature and soil moisture using soil temperature and moisture data from the GLDAS dataset,and three-hour steps of air temperature,relative humidity,wind speed,incident solar radiation,reflection solar radiation,precipitation data from the China Meteorological Forcing Dataset were used as meteorological driving data to obtain Daily snow-covered soil temperatures（i.e.,snow-soil interface temperatures）in the cropland of Northeast China.This study solved the problem of extending the SNTHERM model from point scale to surface scale.Validation of the simulation results using observations of snow–soil interface temperatures from 36 meteorological stations with the agricultural field as the subsurface from 2005-2018.The results show that the SNTHERM model can better simulate changes in snow-soil interface temperature in the cropland of Northeast China（BIAS=0.43°C,RMSE=3.48°C）.We analyzed the daily snow-soil interface temperature and air temperature datasets from 1979-2018 in the cropland of Northeast China,and the results showed that the snow-soil interface temperature in 99%of the regions and the snow insulation capacity（the difference between snow-soil interface temperature and air temperature）in 98%of the regions showed an increasing trend,with more significant increasing trends in the cropland of Sanjiang Plain and Songnen Plain,and slower increasing trends in the cropland of the Liaohe River Plain.Using the air temperature and snow depth data from 1979-2018,the Pearson correlation coefficients of snow-soil interface temperature with air temperature,snowpack days,maximum snow depth,and average snow depth were calculated per pixel,and the results showed that both snowpack and air temperature had a certain degree of influence on the change of snow-soil interface temperature in the cropland of Northeast China.However,average snow depth and maximum snow depth had the most significant effect on snow-soil interface temperature,and the correlation coefficients were mainly distributed in the range of 0.3 to 0.8.The correlation with air temperature was the weakest,and the correlation coefficients were mainly distributed between-0.3 and 0.3.（2）The concept of velocity was introduced to analyze the speed and direction of change in the spatial distribution pattern of near-surface soil freezing and thawing state in the cropland of Northeast China from 1979 to 2020.This study shows that in the past41 years,the spatial distribution patterns of onset date of soil freeze,frozen days and number of soil freeze/thaw cycles in spring in Northeast China are shifting northward（>60%）,among which t the spatial distribution patterns of onset date of soil freeze,frozen days and number of soil freeze/thaw cycles in spring in cropland are changing faster than other land use types,with speeds of 1.63 km yr^-1,2.84 km yr^-1 and 1.55 km yr^-1,respectively.The spatial pattern of offset date of soil freeze changed at a speed of1.55 km yr^-1,which is slightly slower than other land types,especially forested regions.The spatial patterns of near-surface soil freezing and thawing states in the cropland of the Sanjiang Plain,Songnen Plain and Liaohe Plain were calculated on a regional scale,and the results showed that the spatial patterns of near-surface soil freezing and thawing states in the Sanjiang Plain,Songnen Plain and Liaohe River Plain would be completely changed within only about 100 years.As the largest agricultural distribution area in Northeast China,the spatial pattern of near-surface soil freezing and thawing states in the Sanjiang Plain has the most rapid change speed,in which the spatial pattern of onset date of soil freeze,frozen days and number of soil freeze/thaw cycles in spring all change faster than 4 km yr^-1.The spatial pattern of near-surface soil freezing and thawing states in the Sanjiang Plain and Liaohe River Plain changes at a lower speed than in Songnen Plain,mainly staying around 2 km yr^-1.However,both smaller farmland areas relative to the Songnen Plain also caused a complete change in the spatial pattern of near-surface soil freezing and thawing states in the short term.We analyzed the consistency of air temperature,snowpack days,maximum snow depth,and average snow depth with spatial pattern changes in the near-surface soil freezing and thawing states.We also assessed the effects of multiple environmental factors on the spatial pattern changes of near-surface soil freezing and thawing states in the cropland area using ANOVA in a general linear model.The results showed that air temperature and snowpack days were the major factors influencing the speed of change in the near-surface soil freezing and thawing states in the cropland.The most important factor influencing the velocity change of offset date of soil freeze is the snowpack days,while air temperature is the most important factor influencing the velocity change of frozen days,onset date of soil freeze and number of soil freeze/thaw cycles in spring.（3）Winter climate conditions have a great effect on spring soil moisture in cold regions.In this study,we selected the western of Jilin Province as a typical cropland,and used the 0–10 cm soil moisture data from 81 agro-meteorological stations and winter climate data（e.g.,air temperature,snowfall）,soil properties,vegetation and topography in April from 2017 to 2019 based on three currently popular machine learning algorithms（Multiple Stepwise Regression,Random Forest and Support Vector Machine）to predict the spatial pattern of average soil moisture for the period of April1–15 and April 16–30 with 1 km resolution in the dry cropland.The results showed that the model constructed based on the Random Forest algorithm had the best prediction accuracy(April 1–15:r=0.74,RMSE=0.050 m³ m^-3;April 16–30:r=0.73,RMSE=0.051m³ m^-3).Comparing the predicted soil moisture data from western of Jilin Province with the soil moisture data from the currently widely used reanalysis dataset ERA5＿Land and the remote sensing dataset SMAP,the spring soil moisture prediction model constructed in this study not only achieves the soil moisture spatial prediction one month in advance based on environmental factors,but also has higher inversion accuracy and provides finer spatial resolution（1 km）.In addition,the importance ranking of environmental factors calculated based on the Random Forest algorithm,combined with correlation analysis,showed that the most important factors affecting soil moisture in April were climatic factors,mainly snowfall and snow depth,and soil texture also significantly affected soil moisture.Based on the soil moisture spatial prediction distribution map of the dry cropland region in west of Jilin Province,the agricultural drought grade prediction was achieved using soil relative humidity in combination with the spatial distribution map of percentage content of sand and clay in west of Jilin Province.Compared with the currently available drought monitoring results,the drought grade prediction map of dry cropland areas in the west of Jilin Province obtained in this study significantly improved the spatial resolution,avoided classifying some cropland as cropland,and obtained more accurate drought prediction results.The drought prediction results indicate that Baicheng and Songyuan were the hardest hit areas in terms of drought occurrence,the main reason being that the area is one of the most saline areas in China with low percentage clay and low precipitation and poor soil water holding capacity,which makes it more important to develop irrigation plans based on drought grades and provides important information for the agricultural management to prepare for spring irrigation.