海表净辐射是海洋表面接收到的净短波和净长波辐射的代数和，它描述了海洋表面辐射能量收支平衡。海表净辐射是海气相互作用的关键参量，对地球系统的能量平衡有重要影响，并且是全球气候模式和海洋模型的重要输入参数。但净辐射或辐射四分量并非海表常规观测参量，而现有可用的海表辐射产品普遍存在产品精度差异大、空间分辨率较粗及持续时间有限等问题，并且尚无针对海表净辐射估算的经验模型，因此发展具有较好的普适性且高精度的海表净辐射估算模型十分有必要。本研究基于全球 66 个海表系泊浮标站的辐射和气象观测数据，充分探讨了海表常规观测下行短波辐射和下行长波辐射与净辐射的关联规律，在此基础上发展了基于日照时长的海表天均净辐射直接估算经验模型（Case1和 Case2）。其中， Case1 模型适用于日照时长大于 9.6 小时情况，此时海表净辐射主要由短波辐射决定，而 Case2 模型适用于日照时长小于或等于 9.6 小时，此时海表净辐射则主要由长波下行辐射决定决定。站点数据直接验证结果显示，本论文发展的海表净辐射经验估算模型精度令人满意，其总体验证精度 R2为 0.972， RMSE 为 9.768 Wm-2， Bias 为 -0.092Wm-2，且 Case1 模型和 Case2 模型各自的 R2 分别为 0.972/0.979， RMSE 分别为 9.805/2.824Wm-2， Bias 分别为 -0.095/0.346 Wm-2。总体而言，受样本数量不足和大气状况差异等影响，新发展的模型在沿海海域和高纬度海域表现不佳，且模型的表现主要受海表下行短波和长波辐射影响。

与此同时，鉴于目前海表长波下行辐射算法及产品精度普遍不佳，本研究基于观测数据和再分析数据，尝试发展了可用于瞬时及天均海表长波下行辐射估算的经验模型，并与现有的 8 种海表长波下行辐射算法进行了比较。站点实测的直接验证及比较结果显示，本论文新发展的海表长波下行辐射估算模型在晴空或全天空情况下估算的天均及瞬时海表长波下行辐射的表现均优于其他模型，其在全天空情况下天均及瞬时模型的验证精度分别为： R2  = 0.892/0.828， RMSE= 9.738/12.684 Wm-2， Bias= 0.054/0.018 Wm-2，已满足气候模型模拟预测需求。在此基础上，本论文生产了 2000-2018 年海表天均长波下行辐射数据集，与 CERES 比较的结果表明该数据集精度更高，且提供了更加丰富的空间细节信息。两种模型的发展说明了：（1）影响海表净辐射的主要辐射分量由白天时长决定；（2）海表大气 有云天空最常见，因此海表长波下行辐射估算需要充分考虑云及其相关参量的影响，云液态/冰水含量及晴朗指数可有效地表达云对海表长波下行辐射的贡献。总而言之，本论文成功发展了海表净辐射及长波下行辐射经验估算模型，两种模型的表现均令人满意，因此新模型在实际应用及产品生产等方面具有很好的应用潜力。

Net radiation (Rn) at ocean surface is the algebraic sum of the net shortwave and net longwave radiation received on the ocean surface, which characterizes the radiative balance at ocean surface.Ocean surface Rn is one of the essential parameters in the air-sea interaction, and it has the significant influence on the energy balance of the earth system, hence, it is an important input todrive the global climate models and ocean models. However, Rn or the four radiative components are not routinely measured at ocean surface, and the available ocean surface radiation products generally have the drawbacks in accuracy, coarse spatial resolution, and limited duration and so on. Moreover, there is no model for ocean surface Rn estimation by taking it as an integrate variable.Therefore, it is necessary to develop a robustness model for estimating the ocean surface Rn accurately. Based on the collected radiation and meteorological measurements from 66 globaldistributed moored buoys, this dissertation successfully developed the daytime duration based empirical models for ocean surface Rn estimation after fully exploring the correlations between theradiative components and ocean surface Rn . The empirical models include two conditional models named as Case1 and Case2. Specifically, Case1 model is determined by the downward shortwave radiation when the daytime duration was larger than 9.6 hours, while Case2 model taking thedownward longwave radiation as the major input when the daytime duration was less than 9.6 hours. The validation results based on buoy measurements showed that the overall accuracy of thenew models were satisfactory with an R2 of 0.972, RMSE of 9.768 Wm-2, and Bias of -0.092 Wm-2, and the accuracies for Case1 and Case2 model were both very well with the R2 values of 0.972and 0.979, RMSEs of 9.805 and 2.824 Wm-2, and Biases of -0.095 and 0.346 Wm-2, respectively.
However, the new developed model performed relatively worse in coastal and high-latitude seas than in the open seas due to the insufficient samples and their differences in the atmospheric conditions, besides, the performances of the models were mainly determined by the ocean surface downward shortwave and longwave radiation. At the same time, this dissertation also attempted to develop an empirical model for estimating the ocean surface instantaneous and daily downward longwave radiation based on the moored measurements and model reanalysis data. After validating against the measurements and comparing with eight algorithms from previous studies, the results demonstrated that the new developed ocean surface downward longwave radiation estimation model performed superior to all other eight models at both daily and instantaneous scales under all-sky condition with the R2 values of 0.892 and 0.828, RMSEs of 9.738 and 12.684 Wm-2, and Biases of 0.054 and 0.018 Wm-2, respectively. Afterwards, a dataset of the global ocean surface daily downward longwave radiation from 2000 to 2018 has been generated with the new model, which provides more spatial details and better quality comparing other available products. During the modeling, several points could be drawn:

(1) The ocean surface net radiation is determined bythe downward shortwave or longwave radiation according to the daytime duration.

(2) The cloudy sky is very usual over ocean surface, the three parameters (total column cloud liquid water, totalcolumn cloud ice water, and clearness index) used in this study could express the contribution ofthe cloud to the downward longwave radiation very well. In summary, this dissertation has been successfully developed the empirical models for ocean surface net radiation and downward longwave radiation estimation, and the performances of the new models were satisfactory, so the new models have a good potential in scientific studies and practical applications.