Abstract
Abstract. The traditional statistical methods and radiation transfer theory methods for cloud detecting have a high adaptability just only in those areas with a uniform surface coverage and noncomplex terrain. Therefore, adapted to large spatial and temporal scales, in this work a cloud detection method is developed, seeking the main influencing factors of the change of Brightness Temperature(BT) of clear sky and their relationships, researching the change regularity and normal fluctuation range of BT on the basis of function fitting, setting the cloud detecting dynamic threshold depending on the cloud spectral characteristics, and making accuracy assessment in order to ensure higher adaptability and accuracy of this cloud detecting method. In this paper, a dynamic threshold algorithm is presented for cloud detection using daytime imagery from the VISSR sensor on board FY-2C/D/E, which is the first generation geostationary satellite. And the land surface/brightness temperature influence functions are analysis and established, including latitude, longitude, altitude, time, land cover. The theoretical temperature value of clear sky can be calculated through these influence functions. Then, the dynamic threshold cloud detection model is proposed based on the high temporal resolution of VISSR data. Meanwhile, the land surface emissivity is considered as the main factor to the change range of brightness temperature which determines the dynamic threshold for cloud detection. Finally, the dynamic threshold cloud detecting model is evaluated using FY-2C/D/E VISSR data covering China, and the Kappa of dynamic method is maximum, equalling 0.6195, which is much higher than the indexes for the reflectivity and BT fixed methods, equalling 0.4511 and 0.403, respectively. Consequently, the dynamic threshold cloud detecting method provides an important improvement because the spatial, temporal and geographic characteristics were considered into the model.
Highlights
1.1 BackgroundClouds present a variety of shapes and sizes, covering, at any time, more than 50% of the Earth surface (Saunders, et al, 1988)
Cloud detection techniques from remote sensing imagery can be roughly classified in three main categories (Goodman, et al, 1988): threshold methods, statistical approximations and those techniques based in radiation transfer computations
Remote sensing image cloud detection is usually conducted by setting the corresponding thresholds based on the two cloud characteristics of high reflectivity and low brightness temperature, but these two factors change with different environment under different time and space conditions
Summary
Clouds present a variety of shapes and sizes, covering, at any time, more than 50% of the Earth surface (Saunders, et al, 1988). Cloud detection techniques from remote sensing imagery can be roughly classified in three main categories (Goodman, et al, 1988): threshold methods, statistical approximations and those techniques based in radiation transfer computations. The first type of method is based on the adequate selection of thresholds in the different spectral bands to distinguish cloudy pixels from clear ones These thresholds can be applied to a combination of spectral bands or to new variables obtained from them, such as some measurements related to the space consistency and phase correlation. Typical examples of these techniques include the International Satellite Cloud. Based on the highly temporal resolution of the geostationary orbit meteorological satellite, the brightness temperature or temperature time series imageries can be utilized in the cloud detection of nominal imageries, and in identifying clouds that are developing rapidly or located at the boundaries of the moving clouds (Yang, et al, 2008)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
