Assessment of a multidimensional satellite rainfall error model for ensemble generation of satellite rainfall data
Assessment of a multidimensional satellite rainfall error model for ensemble generation of satellite rainfall data
复制标题
DOI：
10.1109/lgrs.2006.873686
复制
发表时间：
2006-07-01
影响因子：
4.8
通讯作者：
Anagnostou, Eimmanouil N.
中科院分区：
文献类型：
--
作者：
Hossain, Faisal;Anagnostou, Eimmanouil N.
这封信提出了对以下科学问题的初步见解：“通过多维卫星降雨误差模型整体生成卫星降雨数据的真实性如何？”作者首先评估了 NASA 戈达德太空飞行中心开发的两种卫星降雨算法的尺度相关多维误差结构，即：1) 称为 3B41RT 产品的红外 (IR) 估计，以及 2) 称为 3B42RT 产品的组合无源微波 (PMW) 和 IR 估计。来自美国南部平原的地面雷达（WSR-88D）降雨场被用作参考。接下来，通过反转多维卫星降雨误差模型（SREM2D，由 Hossain 和 Anagnostou 开发）产生的参考雨场和损坏雨场的定义，作者推导了 WSR-88D 降雨场相对于卫星降雨估计算法的逆多维误差结构。然后将 SREM2D 应用于实际卫星降雨数据以及相关的逆误差参数，以生成参考 WSR-88D 降雨场最可能实现的集合。然后将模拟的系综与从更简单的（二维）逆误差建模方法得出的系综进行比较。据观察，SREM2D 降雨集合的精度高于 3B41RT 产品的更简单的误差建模方案。 3B42RT 产品没有观察到明显的改进，这是由于 3B42RT 数据统计的异构性，而逆 SREM2D 方法中没有考虑到这一点。总体结论是，SREM2D 等多维误差建模方法有可能生成真实的卫星降雨场集合，这应被视为对更广泛使用的更简单误差建模方案的改进。因此，多维误差模型与顺序误差修正方案的结合使用可能会改善基于卫星降雨的全球水和能源循环可预测性的诊断。
This letter presents preliminary insights from the pursuit of the following scientific query: "How realistic is ensemble generation of satellite rainfall data by a multidimensional satellite rainfall error model?" The authors first evaluated the scale-dependent multidimensional error structure for two satellite rainfall algorithms developed at the NASA Goddard Space Flight Center, namely: 1) the infrared (IR) estimates known as the 3B41RT product and 2) the combined passive microwave (PMW) and IR estimates known as the 3B42RT product. Ground radar (WSR-88D) rainfall fields from the Southern Plains of the U.S. were used as reference. Next, by reversing the definition of reference and corrupted rain fields produced by a multidimensional satellite rainfall error model (SREM2D, developed by Hossain and Anagnostou), the authors derived the inverse multidimensional error structure of WSR-88D rainfall fields with respect to the satellite rainfall estimation algorithms. SREM2D was then applied on actual satellite rainfall data with the pertinent inverse error parameters to generate an ensemble of most likely realizations of the reference WSR-88D rainfall fields. The simulated ensemble was then compared with that derived from a simpler (bidimensional) inverse error modeling approach. The accuracy of the SREM2D rainfall ensemble was observed to be higher than the, simpler error-modeling scheme for the 3B41RT product. No tangible improvement was observed for the 3B42RT product, which is attributed to the heterogeneous nature of 3B42RT data statistics that was not accounted for in the inverse SREM2D approach. The overall conclusion is that a multidimensional error modeling approach such as SREM2D has the potential to generate realistic ensembles of satellite rainfall fields, which should be considered as an improvement over the more widely used simpler error-modeling scheme. A combined use of the multidimensional error model with a sequential error correction scheme could therefore potentially improve the diagnosis of satellite rainfall-based predictability of the global water and energy cycle.