%0 Conference Proceedings
%4 sid.inpe.br/mtc-m19/2010/09.09.19.26
%2 sid.inpe.br/mtc-m19/2010/09.09.19.26.54
%T Cloud Classification in Satellite Imagery applied to Assessment of Sunshine Duration and Global Solar Radiation
%D 2010
%A Ceballos, Juan Carlos,
%A Rodrigues, Marcos Lima,
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@electronicmailaddress juanc_ceballos@yahoo.com.br
%@electronicmailaddress marcos.rodrigues@cptec.inpe.br
%B The Meeting of the Americas.
%C Foz do Iguaçu, BR
%8 08-12 aug 2010
%S Abstracts
%K Cloud/radiation interaction, radiation, transmission and scattering, radiative processes, remote sensing.
%X Geostationary satellites perform high-frequency observation of VIS reflectance. The so-called cloud index CI is estimated as CI= (L - Lmin)/(Lmax - Lmin) = (R - Rmin)/(Rmax - Rmin). L is radiance and R is reflectance (assuming isotropic outgoing radiance); min and max refer to extreme values expected in VIS channel. The index is an estimate of effective cloud cover; Rmin is easily assessed by cloud masking, but definition of Rmax is not so clearly stated in literature. Value CI < 1 is meaningful only in Cu-covered scenes (CI=1 in other cloud cover situations), so identification of cloud-type is a previous step for CI assessment. For cloud classification four variables (available in GOES and Meteosat) were used: VIS R, brightness temperature Tb in 10.7µm window, and textures XV and XT defined as R and Tb standard deviation in a 3x3-pixel set. Dynamical clustering by minimal Euclidean distance defined a set of N=32 reference centroids; identification of cloud-type versus centroid was made by visual nephanalysis. This set provides the reference for image classification (see operational results at http: //satelite.cptec.inpe.br). Distribution function F(R) of cumulative frequencies of reflectance for different cloud-types shows that Rmax= 0.465 may be adopted as transition value between Cu- and St-type pixels. The same result is found for different regions in South America, suggesting an easy way of assessing Cloud Index (being CI=1 for Rmax>0.465). Values (1-CI) are taken as clear-sky time fraction during interval between two images. Daily integration provides fairly good results for sunshine, shown by comparison with heliograph data. For solar radiation assessment, CPTEC model GL (http: //satelite.cptec.inpe.br/radiacao/) splits solar spectrum in a UV+VIS band (only stratospheric ozone depletion is considered and radiative transfer within troposphere is conservative) and a solar IR band (where transmittance by cloud is negligible and direct beam is affected mainly by H2O and CO2 depletion). Global radiation (GL) in UV+VIS band is only a matter of energy balance, while GL in solar IR is mainly related to direct beam weighted with clear-sky fraction (1-CI). Recent comparisons of monthly mean of GL model x Brazilian network (about 400 stations) were made in Feb. 2010. GOES 12 imagery calibration induces mean deviation of only +10 W.m-2. Conclusion: cloud cover estimated through Cloud Index CI provides clear and valuable information for sunshine duration and solar energy assessments.
%@language en