Research On The Cloud Base Height Over Eastern China

As an important feedback factor of the climate system,clouds play a critical role in adjusting the global radiation budget and hydrological cycle.The cloud base height(CBH),which is an important parameter of the cloud vertical profile,largely determines the energy exchanges between the clouds and surface.Accordingly,determining the CBHs is extremely critical for weather forecasting and ensuring flight safety.In this study,based on Lidar,balloon sounding and aircraft soundings,we investigated the features of the CBH and determined the thresholds of the environmental relative humidity(RH)corresponding to the observed CBHs over Eastern China.Furthermore,cloud parameters provided by multi-source satellite such as FY-4A,Himawari-8 and ground observations of the MICAPS system were selected to construct the Deep Neural Network(DNN)model under two input layer schemes,which realized the simulation of the CBH.The main conclusions obtained are as follows:(1)Features of the CBH over Eastern China based on multi-source ground-based observationsAmong the three ground-based observations,CBHs derived from Lidar and aircraft are highly correlated and passed the mutual verification,which can be collectively regarded as the accurate value of the CBH.And results detected by the pilot balloon are significantly higher than that from aircraft.In general,most of the CBHs were below 2000 m and the mean value were commonly higher in cold months,while lower in warm months.From the perspective of diurnal variation,CBH was lower in the daytime than the nocturnal ones.The lowest value appeared at 10 am,and the highest value appeared at 21 pm,and the significant diurnal variation of the CBH is about 1400 m.In addition,the diurnal variation of the CBH and the cloud thickness(CT)is also related to the spatial distribution of observation sites.Single-layer clouds above the observational sites appear most frequently,and the frequency of multi-layer clouds decreases with the increase of cloud layer.When there is a cloudy sky,the total frequency of various cloud formations is highest in winter and lowest in summer.In terms of time changes,the frequency of cloud layers is significantly lower in the daytime than the nocturnal ones.(2)Determination and analysis of the RH threshold of the CBH over Eastern ChinaThe cloud base pressure(CBP)from CERES product were consistently smaller than those calculated from the Lidar/aircraft measurements for most samples,indicating that CERES has a certain degree of overestimation of the cloud base information.This study determined the critical altitude caused by the difference between the two CBPs: for clouds higher than 1100 m,two data showed great agreement;and for clouds lower than 1100 m,CERES overestimates the cloud base locations to a certain extent.Combined with the ERA-Interim RH profile,the RH threshold at the cloud base was determined,and it was found that the threshold changed significantly with the CBH.When the CBH was below 1000 m,it corresponded to a stable RH threshold of approximately 80%.With the increase of CBH,RH threshold begins to decrease.As the CBH increased from 2000 m,the RH threshold began to decrease to smaller than 60%.In terms of seasonal variation,the RH threshold for determining the CBH was the highest and lowest in the summer and winter,respectively.(3)Simulation of the CBH based on DNN modelPrincipal component analysis(PCA)achieves the successful dimension reduction of the original data under two input schemes,and obtains the corresponding eigenvalues and variance contribution results.The cumulative contribution rate of the variance of the first 4 principal components in Input-P1 is 78%,and result of the first3 principal components in Input-P2 can reach 97%.It is found that the simulation results under two input schemes are in good agreement with the CBH from ground-based observations,and the result of Input-P2 with FY-4A cloud parameters and MICAPS observations is better.Furthermore,DNN model has a better simulation effect for high clouds,while the results for low and medium clouds are still unsatisfactory.Comparing the mean absolute error(MAE)between the simulation results of two input schemes and the observed values,it is found that the MAEs under two schemes are roughly distributed normally,and the distribution of Input-P2 is more concentrated.Finally,samples of the MAE in positive is significantly higher than the negative interval,indicating that the CBH predicted by the model is generally higher.