Retrieval Of Cirrus Clouds Properties Based On Active And Passive Remote Sensing

Cirrus clouds are an important component of the Earth’s atmosphere,and their micro-physical and optical characteristics have significant impacts on global climate change,space communication,and airplane flight safety.In this paper,data from A-Train satellite constellation,including CPR,CALIOP,MODIS,and ECMWF are employed to perform a detailed analysis on the distribution and parameters retrieval of cirrus clouds over South China Sea.The contents of this thesis are as follows:1.ECMWF reanalysis data assimilate a large number of ground observation data and satellite remote sensing data,and is widely used for its high spatial and temporal resolution and long time span.However,due to the sparsity of thin cirrus clouds in the upper atmosphere,many observations missed many thin cirrus clouds.The spaceborne dual-band lidar CALIOP has quite high detection accuracy for thin cirrus cirrus,enabling us to comprehend the distribution and properties of thin cirrus clouds.Therefore,the high cloud product from ECMWF is compared with CALIOP observations over the South China Sea from December 2006 to February 2019.The annual average of cirrus cloud amount is analyzed and compared between CALIOP and ECMWF day and night,and the monthly comparison is also performed.The results suggest that the distributions of cirrus clouds of CALIOP and ECMWF are in fairly good agreement over the study area both day and night.The correlation coefficient between the two products is 0.79 during daytime,0.75 nighttime.50.8%of high cloud amount from ECMWF is 8.8%lower than 59.6%from CALIOP during daytime.A larger difference of 16.2%occurred at night with 53.9%from ECMWF,and 70.1%from CALIOP.The possible reason for the difference is that ECMWF missed the detection of sub-visual cirrus clouds(SVC,optical depth<0.03).Therefore,after removing SVC,the results show that the difference between the two data decreased from 8.8%to 2.3%in the daytime and from 16.2%to 5.2%at night,and the correlation coefficient also increased to 0.82 both day and night.2.There is a significant error in estimating the amount of cirrus clouds in many observations due to the invisibility of thin cirrus clouds.This paper uses CALIOP data to help understand the distribution and trend of subvisual cirrus clouds over the South China Sea.The results show that the occurrence of cirrus clouds is relatively high in lower latitude regions,reaching up to 16%during the day and 24%at night.As latitude increases,the occurrence of cirrus clouds decreases,possibly due to a decrease of strong convection weather events in the Intertropical Convergence Zone(ITCZ).The monthly variation of cirrus cloud amount shows an M-shaped pattern,with the highest occurrence in May,reaching 17%during the day and 27%at night.The annual change shows a slight downward trend,reaching its maximum in 2007 with 15.2%during the day and 25.1%at night.The frequency of cirrus clouds at night(21.2%)is higher than the day(13.4%).In addition,the mean cloud top height of SVC is 15.77 km,the average base height of SVC is 14.96 km,and the average thickness is 0.81 km in the daytime.At night,the average cloud top height is 15.88 km,the base height is 14.8 km,and the average thickness is 1.08 km.The average thickness at night is greater than the day.The cloud top height of cirrus clouds is very close to the average height of the tropopause(16.93 km).3.Detection and retrieval of various properties of single-layer transparent cirrus clouds over the South China Sea were carried out by combining the advantages of active and passive remote sensing sensors.Transparent cirrus clouds are defined as those clouds with cloud top pressure less than 440hPa and optical depth less than 3.Based on the matched data set,this thesis first proposes a classification neural network for identifying Transparent cirrus clouds,and then uses two regression neural networks to retrieve Transparent cirrus clouds optical depth and top height.The results show that the accuracy of the classification neural network can reach 84%,the detection rate is 79%,the false alarm rate is 9.8%,and the AUC is 0.92,indicating that the classification neural network works well.The probability of detection(POD)rapidly increases with the increase of the optical depth of transparent cirrus clouds.When the optical depth is less than 0.03,POD is only 36%,as SVC are difficult to distinguish from clear sky.When the optical depth is in the range of 0.03-0.1,POD rapidly increases to 73%.When the optical depth is greater than 0.4,POD reaches over 95%,and when the optical depth is greater than 1,POD is 100%,indicating that the neural network can effectively detect the transparent cirrus clouds in this area.However,it has poor performance in detecting SVC.4.In this paper,two regression neural networks are employed to retrieve the cloud top height and optical depth of the transparent cirrus clouds detected by the classification neural network.The results show good consistency between the retrieved values and the true values.The correlation coefficient of the cloud top height reaches 0.87 with a mean absolute error of 0.61 km,while the correlation coefficient of the optical depth is 0.79 with a mean absolute error of 0.2.Moreover,this study collects many MODIS clear-sky pixels over the South China Sea,and put these pixels into the classification neural network.The outputs are colocated compared with CALIOP observations,and the results show that they are in a good agreement,proving that MODIS indeed missed many thin cirrus clouds.Then,regression neural networks are used to retrieve the optical depth and top height of these thin cirrus clouds undetected by MODIS.Some of the retrieval results are compared with CALIOP data.The results suggest that the retrieval results are reliable in a certain extent.5.Based on the neural network algorithm,the retrieval of the optical depth of cirrus clouds over South China Sea is carried out using the matched dataset of the joint product 2C-ICE of CloudSat and CALIPSO and MODIS radiation product MYD02.Due to the better agreement of the retrieval results with MODIS official product in daytime,the optical depth of cirrus clouds in the nighttime was studied using MODIS nighttime radiation data.The results show that there is a good linear relationship between the retrieved values and the true values,with a correlation coefficient R of 0.81,a root mean square error RMSE of 13.9,a mean absolute error MAE of 4.8.6.It is found that there are significant differences of the retrieval accuracy in different optical depth ranges.For the cirrus clouds with optical depth less than 30,the correlation coefficient R between the retrieved values and the true values reaches 0.88,MAE is only 1.2,and RMSE is 2.6;while for the part with optical depth more than 30,R is 0.59,MAE is 29,and RMSE is 37.The validation analysis shows that the consistency between the retrieval results and the true values is very good and the reliability is high,which means that the retrieval results can basically capture the distribution characteristics of the optical depth of cirrus clouds over South China Sea.