基于随机森林的湖北雪密度预测模型及其在雪压分析中的应用

雪密度、雪压等积雪参数资料的缺乏是南方地区雪灾精细化防御研究的难点之一，通过历史地面积雪气象观测资料来反演测站及周边的雪密度，是对现有积雪监测资料的有益补充。本文利用湖北省76站的逐日气象观测资料，分析并选取了积雪期的积雪日数、积雪深度、气温、日照等8个影响雪密度的自变量因子，构建了雪密度的随机森林回归(RF)模型，并通过RF模型反演数据，分析了湖北省雪密度和雪压分布情况。结果表明：①雪密度RF模型预测的均方根误差为0.04 g/cm3左右，可以用于湖北省雪密度资料反演。②湖北省平均雪密度在0.14～0.20 g/cm3之间，从中部以0.17 g/cm3为〖JP2〗界分为东西两个区，东部区雪密度较大。③湖北省近60年来最大雪压值在1.3～6.7 g/cm2之间，不同重现期最大雪压分布存在鄂西北和鄂东两个高值区，且鄂东区的中北部基本雪压值更大。〖JP〗

Snow Density Prediction Model in Hubei Based on Random Forest and Its Application in Snow Pressure Analysis

The lack of snow parameters such as snow density and snow pressure is one of the difficulties in the study of snow disaster prevention in the Southern China. It is a helpful supplement to the existing snow monitoring data to invert the snow density of the station and its surroundings through historical ground snow meteorological observation data. Based on the daily meteorological observation data of 76 stations in Hubei Province, this paper analyzes and selects eight independent variable factors affecting snow density, such as the number of snow days, snow depth, air temperature and sunshine, and constructs a Random Forest (RF) regression model of snow density. Through the inversion data of the RF model, the distributions of snow density and snow pressure in Hubei Province are analyzed. The results show that: (1) The root mean square error predicted by the snow density RF model is about 0.04 g/cm³, which can be used for the inversion of snow density data in Hubei Province. (2) The average snow density in Hubei Province is between 0.14 and 0.20 g/cm³, and is divided into east and west regions based on the value of 0.17 g/cm³, with the larger snow density in the eastern region. (3) The maximum snow pressure in Hubei Province in recent 60 years is between 1.3 and 6.7 g/cm². The distributions of maximum snow pressure in different return periods have two high-value areas in northwest and east of Hubei Province, and the basic snow pressure value in the north central part of the east of Hubei Province is greater.