| The DOAS technology proposed in the 1970s has become a research hotpot in the field of atmospheric optical measurement after years of development.By building multi-platform air pollution three-dimensional monitoring systems including groundbased,airborne,and space-borne,the spatial and temporal distribution of air pollutants can be effectively observed,and the evolution process of air pollution formation and transmission can be analyzed.This development provides a favorable means for China’s atmospheric research and pollution control.Imaging differential optical absorption spectroscopy is a combination of imaging spectroscopy and differential optical absorption spectroscopy(DOAS).Imaging DOAS instruments can collect hyperspectral data cubes with one spectral dimension and two spatial dimensions.After DOAS analysis of the hyperspectral data,the two-dimensional trace gas distribution can be resolved.The ground-based imaging DOAS instruments realize whisk-broom imaging through the horizontal rotation of the installation platform,and the airborne or space-borne imaging DOAS instruments realize push-broom imaging through the movement of the mounting platform.The core of imaging DOAS spectral inversion is the characteristic absorptions of trace gases based on the Lambert-Beer law.However,the inversion results are also affected by preprocessing steps such as spectral calibration and trace gas absorption cross-section convolution.In addition,similar to other imaging spectroscopic monitors,imaging DOAS instruments are prone to stripe noise,producing corresponding pseudo-structures and affecting subsequent information extraction and data analysis.Based on airborne and ground-based imaging DOAS instruments,this paper studies the spectral calibration algorithm using solar spectrum,the fast absorption cross-section convolution algorithm and the weighted variational destripe algorithm from the aspects of speed and accuracy.Accurate spectral calibration is the prerequisite of the accurate spectral inversion.Ground-based or airborne imaging DOAS instruments usually do not have standard light sources for wavelength calibration during measurement.Meanwhile,the standard light sources have limited characteristic peaks in the wavelength range of the instruments.Considering this situation,an effective spectral calibration algorithm is proposed,using high resolution solar spectrum based on control points and spline curve interpolation.This algorithm uses cubic spline curves to characterize parameters such as surface reflectance,iteratively compares the difference between the measured spectrum and the convoluted high-resolution solar spectrum through the least squares algorithm,to obtain the wavelength shift of the measured spectrum and the full width at half maximum of the instrument slit function.Compared with polynomial fitting,splines are more suitable for parameterization of complex wavelength characteristic variables.Moreover,without dividing into sub-windows,the information of the entire spectrum can be used for fitting,which is more suitable for broad band spectral calibration.For imaging DOAS data processing,it is necessary to produce trace gas absorption cross-sections and Ring effect spectra by convolution with the instrument slit function,at each field of view in the spatial dimension.The time required to make multiple convoluted cross-sections is lengthy when the instrument is constantly undergoing spectral drift or working for long periods of time.A simplified convolution algorithm of gas absorption cross-section and Ring effect spectrum is proposed.This algorithm directly constructs the slit functions at all wavelength positions of the instrument and performs convolutions with the high-resolution input spectrum,which can reduce two re-sampling operations and greatly reduce the number of convolution operations.Meanwhile the ring effect spectrum calculation requires two rounds of convolution operations,so using this algorithm can save more computing time.In practical applications,the amount of computation can be greatly reduced with little loss of precision.After testing,the algorithm can at least reduce the computing time of trace gas cross-sections and Ring effect spectra to 1/10 and 1/20,and the relative deviation is generally less than 2×10-3.Stripe noise is a data degradation problem of remote sensing with obvious spatial distribution characteristics.At present,algorithms based on homogeneous reference areas are commonly used in de-striping processing for space-borne imaging DOAS sensors.However,it may not be applicable to ground-based imaging DOAS instruments.Therefore,a variational algorithm based on the inherent characteristics of the stripe noise is developed.The algorithm first obtains the weight matrix representing the abnormal area through block adaptive threshold segmentation,then utilizes the unidirectionality and sparsity of the stripe noise to establish an anisotropic variational model,and solve the model iteratively using an alternating direction method of multiplier in the end.To test the performance of this de-striping algorithm,simulated and real data experiments were performed.Corresponding results prove that this algorithm can effectively remove stripe noise without over-smoothing.The test results of the airborne and spaceborne imaging DOAS data prove that this algorithm is also suitable for airborne and spaceborne platforms.Field experiments were carried out with self-developed airborne and ground-based imaging DOAS instruments,respectively.The measured data in these experiments verified the applicability and reliability of above algorithms.The experiment results also showed that the self-developed imaging DOAS instruments have sufficiently high spectral and spatial resolution to capture different pollution sources and monitor pollution transmission.It has been proved that the imaging DOAS instruments of multiplatform are effective tools for atmospheric research and pollution prevention and control in China. |
PDF Full Text Request