Ground-based All-sky Imaging System, Cloud And Aerosol Parameters Inversion And Its Application

Cloud plays an important role in the meteorological research. The parameterization of cloud in climate model is still a difficult problem needed to be resolved. Satellite and ground-based observation of cloud play significant role in the study of global and regional cloud characteristics as well as relations of clouds and aerosols. The automatic all-sky imager observation system developed and operated successfully provided us the high space-time resolution images in visible band. In view of the latent superiority of all-sky imager system, it is needed to develop a cloud recognition algorithm for the all-sky images in order to fully and entirely get the quantified cloud /aerosol information. The main goal of this paper is to establish a cloud recognition algorithm as well as an inversion algorithm for aerosol optical thickness on the basis of study of all-sky image data and simulations of radiative transfer model. The work provides a new method and analysis tool for ground-based macroscopic cloud research and the aerosol observation. The main findings of paper are as follows:1. Calibration experiment of the digital camera is performed. The camera dark current characteristic, the uniform characteristic, the linearity characteristic and the fish-eye lens cosine-characteristic are separately examined and analyzed. And the geometric calibration as well as the establishment of inversion algorithm from the gray value to relative radiance also is made. Analysis results show that the dark current characteristic made a less impact on the all-sky images. There exits some fluctuations between the gray values under the same exposure conditions. Except the blue-channel characteristic influenced by the source light in experiment, the standard deviation of uniform characteristics of the red and green channel is below 3.0. An inversion algorithm from gray to relative radiance is developed through controlling camera shutter speed, aperture as well as source light intensity. Using the all-sky image data and the same time radiance observed by CIMEL instrument, we made the contrast analysis and examined this algorithm. Results show that the ratio of relative radiance from images to absolute radiance (CIMEL observed) almost keeps stable. Fish-eye lens cosine characteristic calibration is also made and the fish-eye lens cosine attenuation coefficients of the three channels. 2. Numerical simulations and analysis of cloud detection algorithm are made using LIBRADTRAN radiative transfer model. In the work, distribution of sky radiance and blue/red radiatvie ratio is analyzed with different optical thickness cloud and free-cloud conditions. Further, impacts of different solar zenith angle and aerosol optical depth on the threshold-method of cloud detection are also analyzed. Results show that the solar elevation angle as well as the aerosol optical thickness affect in a big way regarding the cloud detection research work. The threshold-method of cloud detection has a high precision under low aerosol optical thickness. Aerosol easily causes to mistake cloud for sky (or sky for cloud) under high optical thickness aerosol conditions as well as those sky field near horizontal. Consequently, the sole threshold-method carries on cloud recognition have certain limitation. The enhancement of precision to cloud recognition algorithm also needs to make with the aid of other methods.3. Distribution of radiance and blue/red radiatvie ratio was analyzed through numerical simulations using a 3D radiative transfer model-SHDOM under different cloudy situations. Especially, the different radiative character between the cloudy day and the free-cloud day were analyzed in detail. One of the goals is to provide the judgment experience for the cloud determination algorithm, on the other hand, is to understand the feasibility that the results gotten from the 1D-radiative transfer model can be used in the cloudy situations. Distribution of radiative characters in cloudy day is connected with the references of AOD, cloud optical depth, cloud cover and sun angle. Compared with the radiance in free-cloud sky, the radiance changes bigger with the increase of the AOD and cloud cover under cloudy situation, and it will change greatly when the sun is obscured by the cloud. As the impacts made by the cloud to the"cloud-free"fields are concerned, the variation near the cloud is bigger than those far away from the cloud; and the radiance of those fields facing the sun will increase while it will decrease when back to the cloud. In most conditions, the variation is between±5%. Under the same condition, the variance of radiative ratio is less than the radiance itself. Except those points near the cloud, variation of other points is also between±5%.4. On the basis of numerical simulation results, we establish a cloud-detection algorithm with the FFT algorithm, symmetrical methods and threshold method. This algorithm has a more powerful capability to the cloud and non-cloud recognition and can make an accurate discrimination to most of cloud element. The cloud recognition precision is high under the high-visibility atmosphere condition. The capability of determination of the haze layer or the fog is enhanced. But this algorithm is needed to be improved somewhat for certain cirrus and the sky near the horizon in heavy dust conditions.5. The relations between aerosol optical depth (AOD) and blue/red radiative ratio was analyzed by numerical model simulations. We find the radiative ratio value indicates sensitivity to the aerosol optical depth under different aerosol types and there exits an exponential relations between them. Then, we set up an inversion formula for aerosol optical depth based on the results of model simulation. Using the formula, we compare the AOD inversed from all-sky images to the values obtained from the CIMEL instrument. Results show that there exits differences between the inversion value and the AOD provided by the CIMEL. Though, this result indicates that it is possible and feasible to obtain the AOD from the radiative ratio of the all-sky image.