| Sea surface salinity (SSS) data has significant meaning in predicting global climate change, global water cycle pattern and so on. Satellite remote sensing is the most effective method to obtain large scale and long term sea surface salinity data. With the successful launch of two salinity remote sensing satellite-SMOS and Aquarius, and acquiring of global sea surface salinity data, many remote sensing experts and marine scientists have opened a new chapter in the domain of sea surface salinity data processing and oceanography application. In this paper, the influence of auxiliary data errors on the salinity retrieval errors are studied, and related theory, retrieval algorithm and application of salinity remote sensing in two complex sea conditions of high wind speed and raining are investigated.First, it is found that ancillary sea surface temperature (SST) errors, radiometer measurement errors and wind speed errors have less impact on salinity retrieval, when SST and SSS are high, and the influence raises rapidly in the case of low SST and low SSS. At the same time, the discrepancy between the Monte-Carlo method estimated error and the sensitivity method estimated error is large in the case of low SST and low SSS. A more precise empirical model of brightness temperature (TB) for Aquarius satellite is used in computing the influence of wind speed error on salinity retrieval error. The results reveal that the salinity error generated by wind speed error of 2 m/s is small in the case of 3-7 m/s wind speed, and is large in the case of low and high wind speed; meanwhile the error is larger in H-polarization than V-polarization. A matchup dataset consisting of Aquarius data, HYCOM model data and Argo data is built. When we analysis the accuracy of satellite salinity data, it is found that the root mean square error (RMSE) between Aquarius and the two other dataset is about 0.9 psu in general. The bias of satellite salinity remote sensing is reduced with increasing temperature, while the RMSE is reduced from 1.5 psu to 0.8 psu as the temperature increases. The bias of satellite salinity remote sensing will increase with increasing wind speed, which may be larger than 0.5 psu when wind speed is higher than 20m/s. The RMSE is no larger than 0.8 psu in the case of 3-10m/s wind speed; however, it will be larger than 1.5 psu in the case of high wind speed. The bias and RMSE of satellite salinity remote sensing are small in low latitude (<30°) area, however, it is large in high latitude area where the SST is low and the wind speed is high. Besides, there is larger error in area near land due to radio frequency interference (RFI).The salinity remote sensing in a complex sea surface condition of high wind speed is studied. First, the sea surface emissivity increments obtained by using the Small Slope Approximation (SSA) and by the Aquarius satellite are compared. It shows that although a variety of ocean wave spectrum models were used, there is still a large discrepancy between simulation and measured data in the case of middle and high wind speed, which may be caused by the foam generated by breaking waves. The influence of foam can be separated by using foam emissivity and foam coverage. With considering the vertical structure inside the foam and foam thickness distribution, the foam emissivity is simulated by a physical model. An empirical foam emissivity model is also used for comparison. The foam coverage model and foam emissivity model are added into the SSA model to simulate sea surface emissivity increment. It is found that, the RMSE caused by foam coverage under the two polarizations and three incidence angles of Aquarius can reach a relatively small value when the empirical foam emissivity model or physical model with Vaf (void fraction in foam-air interface) of 75% is used. An improved sea surface TB model is obtained by combining the SSA model, the modified foam coverage model and the foam emissivity model together. The bias of sea surface TB simulated by the improved model is relatively small when compared with Aquarius measurement, and the RMSE can be less than 0.5 K.Finally, in this paper, three kinds of influence of rainfall on salinity remote sensing are analyzed separately. First of all, the liquid water in the atmosphere has very small effects on the radiation received by radiometer when there is rain, which may cause a TB increment of less than 0.5K in 25mm/hr rain rate, while the error may be less than 0.02K, which can be neglected. Secondly, the fresh water of rain can change the sea surface salinity, the bias of salinity between satellite measurement and HYCOM model may reach 2psu in 25mm/hr rain rate. The spatial distribution of rain rate is well corresponding to salinity error; for example, in the area of ITCZ with larger rain rate, large salinity bias and RMSE may arise. For single rainfall process, the salinity bias is well correlated with rain rate; rainfall has large effect on H-polarization TB than V-polarization TB. The increment of wind speed may defect the influence of rain. Finally, without considering the effect of rainfall on salinity change, the rain wave spectrum is added to the rough sea surface TB model to simulate the rain effect on sea surface roughness. Some coefficients in the rain wave spectrum are modified by fitting them with measured data. By comparing the new model output and Aquarius data, it is found that the rain effect is large in the case of low wind speed and low rain rate, and is small in the case of high wind speed. The increase rate of emissivity with rain rate is large in the case of low rain rate and is small in the case of high rain rate. The salinity bias is reduced when we use the retrieval model with considering the rain wave spectrum, the RMSE is also reduced. |
PDF Full Text Request