Estimating snow microphysical properties using collocated multisensor observations
Estimating snow microphysical properties using collocated multisensor observations
复制标题
DOI：
10.1002/2013jd021303
复制
发表时间：
2014-07
期刊：
影响因子：
--
通讯作者：
N. Wood;T. L’Ecuyer;A. Heymsfield;G. Stephens;D. Hudak;P. Rodriguez
中科院分区：
文献类型：
--
作者：
N. Wood;T. L’Ecuyer;A. Heymsfield;G. Stephens;D. Hudak;P. Rodriguez
使用贝叶斯最优估计反演方法检查地面原位和遥感观测约束干雪微物理特性的能力。描述质量和水平投影面积随颗粒尺寸和与颗粒形状相关的参数的变化的幂函数从近瑞利雷达反射率、颗粒尺寸分布、降雪率和尺寸分辨的颗粒下落速度中检索。算法性能的背景下，加拿大CloudSat CALIPSO验证项目期间部署的仪器进行了探讨，但该算法是适应其他类似的传感器组合。观测和正演模型的不确定性的关键估计的开发和使用量化的方法，从实际观测的雪事件的合成情况下开发的性能。除了说明的技术，结果表明，这种组合的传感器提供了有用的约束的质量参数和面积幂函数的系数，但只有弱约束的面积幂函数的指数和形状参数。信息内容度量表明，约两个独立的量测量的一套意见，该方法是能够解决约八个不同的实现的状态向量包含的质量和面积的幂函数参数。关于观测和正演模型不确定性的替代假设表明，粒子下落速度的改进建模可能有助于大幅提高该方法的性能。
The ability of ground‐based in situ and remote sensing observations to constrain microphysical properties for dry snow is examined using a Bayesian optimal estimation retrieval method. Power functions describing the variation of mass and horizontally projected area with particle size and a parameter related to particle shape are retrieved from near‐Rayleigh radar reflectivity, particle size distribution, snowfall rate, and size‐resolved particle fall speeds. Algorithm performance is explored in the context of instruments deployed during the Canadian CloudSat CALIPSO Validation Project, but the algorithm is adaptable to other similar combinations of sensors. Critical estimates of observational and forward model uncertainties are developed and used to quantify the performance of the method using synthetic cases developed from actual observations of snow events. In addition to illustrating the technique, the results demonstrate that this combination of sensors provides useful constraints on the mass parameters and on the coefficient of the area power function but only weakly constrains the exponent of the area power function and the shape parameter. Information content metrics show that about two independent quantities are measured by the suite of observations and that the method is able to resolve about eight distinct realizations of the state vector containing the mass and area power function parameters. Alternate assumptions about observational and forward model uncertainties reveal that improved modeling of particle fall speeds could contribute substantial improvements to the performance of the method.