利用遥感数据估算四川省PM_2.5的4种模型对比

Comparative Analysis of Four Models for Estimating PM_2.5 in Sichuan Province Using Remote Sensing Data

摘要: 基于2015年四川省PM_2.5地面监测数据、MODIS 3 km气溶胶光学厚度(AOD)数据和地面气象站点数据，采用简单线性回归模型、多元线性回归模型、BP神经网络模型、线性混合模型预测近地面PM_2.5浓度，并利用地面监测站点数据评估模型的拟合结果。同时利用GIS分析技术，得到四川省2015年空间连续的PM_2.5年均、季均浓度分布。结果表明：(1) 利用线性混合模型反演的PM_2.5浓度精度最高、效果最好，其可以解释四川省PM_2.5浓度75.77%的变异。(2) 线性混合模型预测的PM_2.5浓度与地面实测PM_2.5浓度在时空变化趋势上基本一致，即东高西低，其中成都平原经济区、川南经济区的PM_2.5浓度最大，其次为川东北经济区，最低的为攀西经济区和川西北经济区。PM_2.5浓度大小关系为：冬季 > 春季 > 秋季 > 夏季。

Abstract: Based on the PM_2.5 ground monitoring data of Sichuan Province, MODIS 3 km aerosol optical depth (AOD) data and surface meteorological station data in 2015, the near surface PM_2.5 concentration was predicted by simple linear regression model, multiple linear regression model, BP neural network model and linear mixed model, and the fitting results of the model were evaluated by using the ground monitoring station data. At the same time, using GIS analysis technology, the spatial continuous annual and seasonal average concentration distribution of PM_2.5 in Sichuan Province in 2015 was obtained. The results show that: (1) PM_2.5 concentration retrieved by linear mixed model has the highest accuracy and the best effect, which can explain 75.77% variation of PM_2.5 concentration in Sichuan Province. (2) The PM_2.5 concentration predicted by the linear mixed model is basically consistent with the measured PM_2.5 concentration on the ground, which is higher in the East and lower in the West. The PM_2.5 concentration in Chengdu Plain Economic Zone and southern Sichuan Economic Zone is the largest, followed by Northeast Sichuan Economic Zone, and the lowest is Panxi Economic Zone and Northwest Sichuan Economic Zone. The relationship of PM_2.5 concentration was as follows: winter > spring > autumn > summer.