基于SBDART和BP的夜间雾遥感检测和能见度反演

为了挖掘卫星影像中的雾信息为雾预测预报服务,利用双红外亮温差值作为雾与地物识别分类标志,建立基于SBDART辐射传输模型和BP神经网络的夜间雾遥感检测和能见度反演模型。对2007年11月24日我国华北地区的一次陆地辐射雾MODIS卫星数据进行雾检测,同时反演雾区能见度。根据陕西省气象局提供的地面气象观测数据对模型雾检测结果和能见度反演结果进行验证,该次实验夜间雾检测的准确率为79.2%,地面观测能见度和反演能见度一元线性回归分析方程斜率为1.006,相关系数为0.8498。实验结果表明,模型具有较高的雾识别率和雾能见度反演结果,可为夜间雾识别和生消发展规律探讨提供一定的帮助。

Nighttime radiation fog detection and visibility retrieval based on SBDART and BP neural networks

Fog is a disastrous weather phenomenon, which seriously influences people's daily life and economic activities.A nighttime fog detection and visibility retrieval model was established based on SBDART(santa barbara DISORT atmospheric radiative transfer)and BP(back propagation) Neural Networks using brightness temperature information and difference information between MIR and TIR. The model was tested in the North China Plain on November 24, 2007, with 1 km resolution MODIS/ TERRA images, and the fog detection accuracies and visibility retrieval result were evaluated using the observation results from Shanxi Meteorological Bureau. Results show that the accuracy of fog detection is 79. 2%, the slope is 1. 006 and the correlation coefficient is 0. 8498 of unary linear regression between the visibility observation values and the inversion visibility values. The model can provide strong support for the fog detection and visibility retrieval, which can contribute to further understanding of the development and extinction discipline of nighttime radiation fog.