基于近红外体散射数据的气溶胶尺度谱正则化反演方法

基于0.86 um 波段气溶胶的体散射数据,提出了一种新的尺度谱反演方法。根据大气气溶胶尺度谱特征,将尺度谱函数n(r)分解为趋势变化函数H(r)和细节变化函数(r),并构造了一组新的基函数对(r) 进行参数化逼近,然后严格照Mie 散射理论,采用Tikhonov 正则化对尺度谱函数进行了反演。采用城市型、乡村型和海洋型气溶胶的尺度谱实测数据进行反演仿真,结果表明,在粒径0.2~10 um 区间、噪声不大于50%的条件下,实际与反演的尺度谱曲线相关系数高于0.98,表现出良好的抗噪声能力;针对小尺寸段(r0.2 um)反演结果的不稳定性,提出了小尺寸段的荣格分布修正法与细模态参数补偿法,模拟结果表明,两种方法对尺度谱修正效果较理想,在0.1~10 um 区间,实际与反演尺度谱曲线相关系数大于0.97。与基于遗传算法的尺度谱反演方法相比,该方法效率高,耗时短,且对尺度谱函数细节变化特征反演较好。

Regularized inversion method for retrieving aerosol size distribution based on volume scattering function data at near-infrared waveband

Based on volume scattering data at 0.86 um, a new method for retrieving aerosol size distribution functions was put forward. According to the characteristics of aerosol size distribution, size distribution function n (r) was broken into two part, whole-trend function H (r) and detail-describing function (r), and a new series of basis functions were advanced and employed to approach (r). To overcome ill-posed nature in retrieval process, Tikhonov regularization method was combined with Mie scattering theory to strengthen the capabilities to void the influence of measurement noise and errors caused by numerical integration. Retrieval simulations are performed with size distribution data measured by Anhui Institute of Optics and Fine Mechanics, which represents three different kinds of aerosol, urban, rural and oceanic. Results show that, when radius of particles is larger than 0.2 um, the curve of retrieved size distribution function nearly coincides with that of actual size distribution, and their correlation coefficient is larger than 0.98 on condition that measurement noise isn't larger than 50%. Good robustness is also exhibited in the model put forward here; however when radius is smaller than 0.15 um, there is some deviation when measurement noise is large. To overcome this problem, Junge-correction method and accumulation-mode-compensation method were put forward, corrected results of retrieval size distributions show great consistency with actual size distribution. Compared with retrieval method based on intelligent algorithm, method here is more effective and less time-consuming, and has advantage in retrieving the detail characteristics of aerosol size distribution.