一种基于TMI观测结果的海表温度反演算法

An algorithm for sea surface temperature retrieval based on TMI measurements

中文关键词:  海表温度，反演，微波成像仪，半经验统计算法

英文关键词:SST, Retrieval, TMI, Semistatistical algorithm

中文摘要:

      基于星载微波仪器观测结果反演海表温度，能很好地克服云对反演结果的干扰，实现对海表温度全天候的监测。文中利用热带测雨卫星所搭载的微波成像仪的观测结果，建立了一种新的适用于非降水条件下的海表温度反演算法。作为一种半经验统计算法，它以辐射传输方程为基础，通过理论模拟计算，建立海表温度与微波成像仪多通道亮温之间的关系，较好地反演海表温度。该算法最大的特点是选择了合适的5个微波成像仪通道，并通过这5个通道亮温的对数线性组合方式提取海表温度信息，从而有效地避开了其他环境参数对反演结果的影响。海表温度的反演结果与地基浮标观测结果的比较表明，二者间的均值相差0.116 K、均方根误差为0.665 K。在不同的风速、风向及天空状态（有无云）情况下，二者的相关系数均在0.95以上，均值差异小于0.2 K，均方根误差在0.65 K左右。在全球尺度上海表温度的反演结果与现有海表温度产品的比较显示，二者的差异一般不超过1 K，且差异不随其他环境参数发生明显变化；与多年月平均海表温度产品对比研究结果表明，本算法反演获得的海表温度在全球大部分地区（除高风速高水汽区外）与其他海表温度资料的差异在1 K范围以内。上述结果表明，该反演算法不仅适用于实时反演，亦能用于气候尺度研究。

英文摘要:

      Based on the observations of the microwave instruments on satellites, sea surface temperature (SST) can be accurately obtained regardless of clouds, which is helpful to globally monitor changes in SSTs under any conditions. In this study, a new, as well as simple and accurate, algorithm is proposed for retrieving SSTs in the absence of rain by using the Tropical Rainfall Measuring Mission Microwave Imager (TMI) measurements. Applying a log linear relationship between the brightness temperatures and the main environmental parameters and based on the data from the selected five channels, this algorithm, as a semistatistical technique, was actualized through the radiative transfer model simulations. The unique advantage of this algorithm is that the retrieved SST is able to refrain from the influence by the other environmental parameters, due to the appropriate 5 channel log linear combination. In order to verify the accuracy of the current algorithm, some validations are made in this study. First, the results of case analysis show that there are no unreasonable SST distributions, no matter whether clouds are present or not. On the other hand, the retrievals are also compared against the in situ buoy observations, which showed a good agreement with a bias of 0.116 K and a root mean square error (RMSE) of about 0.665 K. For the different wind speeds, wind directions and or sky conditions, there are high correlation coefficients exceeding 0.95 between these two with, the low bias less than 0.2 K and RMSE of about 0.65 K. Additionally, comparison is also made with the SST retrieved from the Remote Sensing Systems (RSS) products, which showed that there are more than half grids whose differences lower than 1 K, and the variation of the bias between them is not dramatic with changes in other environmental parameters. It is worth noticing that the departure can hold 0.5 K with increasing cloud liquid water path, which implies the influence of clouds can be well eliminated by the current algorithm. However, when high wind speeds the departure is extremely large because of the limitation of the microwave theory on high wind speeds which can produce SST uncertainties not only in the current algorithm, but also in the RSS products. Finally, the similarity of SST global distributions based on this algorithm to those from the other SST data (e.g. the OISST) is shown with the difference less than 1 K in most of the globe, except for few regions having high wind speed related to the error in the current algorithm, or high water vapor content associated with the error in the OISST products. Generally, the above results suggest that this SST algorithm can be well applied to the real time retrievals, as well as the climatology investigations.