Search

Aiming at the prediction of passenger traffic volume with small samples, multi-dimension and nonlinearity, v-support vector regression (v-SVR) is introduced to forecast passenger traffic volume. To seek the optimal forecast accuracy and generalization performance of v-SVR, chaos adaptive genetic algorithm (CAGA) is used to optimize the parameter, which is based on chaos mapping and adaptive mechanism. Then, a new passenger traffic volume forecasting model of v-SVR named by CAGA-v--SVR is proposed. The model is applied to forecasting passenger traffic volume with data of 1978-2008. Compared with RBF neural network model, GA-SVR model and GA-v-SVR model, it is concluded that CAGA-v-SVR prediction model has higher prediction precision, and can effectively predict passenger traffic volume with less than 2. 3% of mean absolute relative error. [ABSTRACT FROM AUTHOR]

MERSI is the keystone payload of FengYun-3 and there have been two sensors operating on-orbit since 2008. The on-orbit response changes obviously at reflective solar bands (RSBs) and must be effectively monitored and corrected. However MERSI can not realize the RSBs onboard absolute radiometric calibration. This paper presents a new vicarious calibration (VC) method for RSBs based on in-situ BRDF model, and vector radiometric transfer model 6SV with gaseous absorption correction using MOTRAN. The results of synchronous VC experiments in 4 years show that the calibration uncertainties are within 5% except for band at the center of water vapor absorption, and 3% for most bands. Aqua MODIS was taken as the radiometric reference to evaluate the accuracy of this VC method. By comparison of the simulated radiation at top of atmosphere (TOA) with MODIS measurement, it was revealed that the average relative differences are within 3% for window bands with wavelengths less than 1 microm, and 5% for bands with wavelengths larger than 1 microm (except for band 7 at 2.1 microm). Besides, the synchronous nadir observation cross analysis shows the excellent agreement between re-calibrated MERSI TOA apparent reflectance and MODIS measurements. Based on the multi-year site calibration results, it was found that the calibration coefficients could be fitted with two-order polynomials, thus the daily calibration updates could be realized and the response variation between two calibration experiments could be corrected timely; there are large response changes at bands with wavelengths less than 0.6 microm, the degradation rate of the first year at band 8 (0.41 microm) is about 14%; the on-orbit response degradation is maximum at the beginning, the degradation rates slow down after one year in operation, and after two years the responses even increase at some band with wavelengths larger than 0.6 microm.