Soil freezing depth forecast using simple regression model

Features of the dynamics of the seasonally frozen layer and climatic characteristics of the cold period of the year were studied using Bakchar weather station observation data for 1963 - 2011. The analysis of monthly data has shown that significant trends in air temperature found for October (0.6 °C/10 yr), February (0.93 °C/10 yr), March (0.7 °C/10 yr) and May (0.55 °C/10 yr). Statistically significant trends in monthly precipitation amounts were not found. Snow depth increases from February to April and in October and November. Discovered tendencies in winter temperature and snow cover lead to changes in soil temperature and the depth of soil freezing. Tendency to decrease the depth of freezing was obtained for February, March and April. To identify factors affecting the depth of soil freezing the correlation analysis was performed. Significant relationships exist between the freezing depth and air temperature in November (correlation coefficient R = -0.55) and March (R = -0.38), and between the freezing and snow depth in December (R = -0.30), January (R = -0.39), February (R = -0.53) and March (R = -0.54). Development of the seasonal frozen layer in different months is influenced by different factors, we proposed a multiple regression model for the prediction of average thickness of the frozen layer. In general, the model includes the average monthly air temperature and snow depth of the current and previous month as well as the freezing depth of the previous month. The model parameters were determined using software package Statistica 8.0 using monthly depths of freezing, air temperature and snow depth for 1963-2011. Set of independent variables was determined for each month individually. For example, the depth of freezing in November depends on the air temperature and snow cover in November. Regression models for January - April have high coefficients of determination (R = 0.79 ^ 0.95). The proposed method for predicting soil freezing depth using regression relations based on monthly air temperature and snow depth can be used to determine the average values of freezing depth for long-term intervals.

Keywords

Authors

References