Scale-by-scale analysis of probability distributions for global MODIS-AQUA cloud properties: how the large scale signature of turbulence may impact statistical analyses of clouds
Scale-by-scale analysis of probability distributions for global MODIS-AQUA cloud properties: how the large scale signature of turbulence may impact statistical analyses of clouds
复制标题
DOI：
10.5194/acp-11-2893-2011
复制
发表时间：
2010-09
影响因子：
6.3
通讯作者：
M. Juárez;A. Davis;E. Fetzer
中科院分区：
文献类型：
--
作者：
M. Juárez;A. Davis;E. Fetzer
抽象的。平均值，标准偏差，均匀性参数的基础上，它们的比例模型中使用的，和概率分布函数（PDF）的云属性从中分辨率红外光谱仪（MODIS）的全球估计的平均尺度从5到500公里的函数。云的分数，液滴有效半径和液态水的路径-所有的云气候不确定性量化和减少努力的问题。确认全球平均值和标准差随规模变化。对于所考虑的尺度范围，全球平均值的变化范围仅为云分数的3%，液态水路径的7%，云粒子有效半径的0.2%。这些规模依赖性导致其全球预算的不确定性。尺度依赖的标准偏差和广义平坦度相比，湍流系统的预测。分析表达式被确定为最适合每个观察到的PDF。虽然每个变量的最佳分析PDF拟合不同，但当平均值通过每个平均域内的标准差归一化时，所有PDF都可以通过对数正态PDF很好地描述。重要的是，对数正态分布在所有尺度上都比高斯分布更适合观察结果。这表明一种可能的方法，这些变量在所有尺度的子网格和统一的随机建模。结果还强调，需要建立一个足够的空间分辨率的双流辐射研究云气候相互作用。
Abstract. Means, standard deviations, homogeneity parameters used in models based on their ratio, and the probability distribution functions (PDFs) of cloud properties from the MODerate resolution Infrared Spectrometer (MODIS) are estimated globally as function of averaging scale varying from 5 to 500 km. The properties – cloud fraction, droplet effective radius, and liquid water path – all matter for cloud-climate uncertainty quantification and reduction efforts. Global means and standard deviations are confirmed to change with scale. For the range of scales considered, global means vary only within 3% for cloud fraction, 7% for liquid water path, and 0.2% for cloud particle effective radius. These scale dependences contribute to the uncertainties in their global budgets. Scale dependence for standard deviations and generalized flatness are compared to predictions for turbulent systems. Analytical expressions are identified that fit best to each observed PDF. While the best analytical PDF fit to each variable differs, all PDFs are well described by log-normal PDFs when the mean is normalized by the standard deviation inside each averaging domain. Importantly, log-normal distributions yield significantly better fits to the observations than gaussians at all scales. This suggests a possible approach for both sub-grid and unified stochastic modeling of these variables at all scales. The results also highlight the need to establish an adequate spatial resolution for two-stream radiative studies of cloud-climate interactions.