Radiation Harmonization Of Multi-source Optical Satellite Data And Application Of Crop Phenology Monitoring

Crop phenology is a key indicator for crop field management,crop yield estimation,and regional food security monitoring because it indicates the regularity and complex changes in crop growth.Crop phenology,on the other hand,is characterized by a high rate of change because it is influenced by a variety of factors such as land management and climate change.Cultivated land landscapes are also characterized by high spatial variability,making coarse spatial resolution data ineffective for monitoring crop growth status.High spatial resolution data contains a wealth of spatial detail,which is critical for crop phenology mapping at high spatial resolution.However,because of sensor limitations and cloud influence,it is often difficult for a single high-resolution satellite to acquire useful data during the crucial time of crop growth due to the long revisit period.The use of multi-source high spatial resolution data to reconstruct the phenological mapping of high time and high spatial resolution remote sensing data has become a hotspot for study.However,due to differences in radiometric calibration processes and methods among sensors,as well as differences in spectral response function and the illumination-viewing geometries among different sensors,vegetation index estimation and phenological information extraction would be unpredictable,affecting crop phenological monitoring performance.At the moment,the radiometric calibration and normalization methods for various sensor data are mostly being designed for sensors with a narrow field of view,a small view zenith angle,and relatively close spectral band settings,as well as for the harmonization of a wide field of view and broadband sensors.The scope of the research is limited.This paper develops a series of radiometric normalization methods for multi-source remote sensing data,as well as various spectral settings and observations,in order to minimize the variation between different sensor data induced by radiation calibration,spectral environment,and solar incidence-satellite observation angle during imaging.To minimize the disparity in the observed effects of different sensors,the reflectance under the conditions is transferred to the same spectral channel of the regular sensor and the reflectance under the observation geometric conditions.The spatiotemporal data fusion model is used to produce a high-precision and high-temporal-resolution data set,and winter wheat is developed.Corn and rice phenological testing and mapping analysis The below are the thesis’ s key research conclusions:(1)The radiometric cross-calibration method proposed in this paper can better fulfill the high-accuracy near-real-time radiometric calibration of multispectral satellite sensors with wide FOV and high spatial resolution.This method does not require the observation of pseudoinvariant calibration sites(PICS),nor does it rely on light matching or simultaneous nadir observations(SNOs).It only needs the atmospheric and ground conditions,including aerosol optical depth(AOD)and surface bi-directional reflectance distribution function(BRDF),and other auxiliary data.The radiometric cross-calibration method can be applied to the other optical satellite sensors when the reference sensor is available.(2)The application results of the piecewise linear fitting method proposed in this paper show that this method can better improve the consistency of the reflectance data under different illumination-viewing geometry and eliminate the influence of the surface BRDF effect.The relationship of two directional reflectances under different illumination-viewing geometries is nonlinear,and the huge the angle difference,the more obvious the nonlinear distribution.Due to the non-Lambertian characteristics of the surface,the reflectance variations of the same ground type under different illumination-viewing geometries often show nonlinear characteristics.Therefore,the relationship between two directional reflectances under different illumination-viewing geometries in a specific scenario can not be established by a single linear model.This paper realized the harmonization between the GF-1 data which has a wide FOV and the nadir view satellite through the piecewise linear fitting method.(3)The phenological monitoring findings of common crops show that the data normalization approach introduced in this paper helps to increase the accuracy of data with different spatial resolutions and can,to some extent,eradicate the inconsistency of vegetation index data under different observation angles.Simultaneously,multi-source remote sensing data can be analyzed using the radiation normalization introduced in this paper to increase the accuracy of the spatiotemporal data fusion model’s performance results.Finally,high spatial resolution phenology products have richer spatial information than coarse spatial resolution phenology products,and can effectively minimize product noise while still avoiding the ambiguity of phenological mapping induced by the mixed pixel problem.