Ice clouds play a critical role in the balance of the earth–atmosphere radiation system, but there are some limitations in the existing remote sensing methods for ice clouds. Terahertz wave is expected to be the best waveband for retrieving ice clouds, with terahertz wavelengths in the order of the size of typical ice cloud particles. An inversion method for the remote sensing of ice clouds at terahertz wavelengths based on genetic algorithm is proposed in this paper. First, suitable channel sets in the terahertz band, which are mainly a combination of absorption lines and window regions, are determined. Then, to improve the efficiency of the generation of the retrieval database, based on the brightness temperature simulated by the atmospheric radiative transfer simulator (ARTS) for different cloud parameters, a fast forward operator is constructed using three-dimensional interpolation to simulate the brightness temperature difference between clear sky and a cloudy scene. Finally, an inversion model to retrieve the ice cloud base height, the effective particle diameter and the ice water path is established based on the genetic algorithm, and an analysis of the inversion errors is performed. The results show that the forward operator, constructed by the nearest interpolation, can accurately calculate the brightness temperature difference at a high speed. The proposed inversion method at terahertz wavelengths based on the genetic algorithm can achieve the expected scientific requirement. The absolute error of the cloud height is around 0.2 km, and the absolute error of the low ice water path (below 20 g/m²) is small, while the relative error of the high ice water path is generally maintained at around 10%, and the absolute error of the effective particle diameter is mostly around 4 μm.

Currently, terahertz remote sensing technology is one of the best ways to detect the microphysical properties of ice clouds. Influenced by the representativeness of the ice crystal scattering (ICS) model, the existing terahertz ice cloud remote sensing inversion algorithms still have significant uncertainties. In this study, based on the Voronoi ICS model, we developed a terahertz remote sensing inversion algorithm of the ice water path (IWP) and median mass diameter (D_me) of ice clouds. This study utilized the single-scattering properties (extinction efficiency, single-scattering albedo, and asymmetry factor) of the Voronoi, sphere, and hexagonal column ICS models in the terahertz region. Combined with 14 408 groups of particle size distributions obtained from aircraft-based measurements, we developed the Voronoi, sphere, and column ICS schemes based on the Voronoi, sphere, and column ICS models. The three schemes were applied to the radiative transfer model to carry out the sensitivity analysis of the top-of-cloud (TOC) terahertz brightness temperature differences between cloudy and clear skies (BTDs) on the IWP and D_me. The sensitivity results showed that the TOC BTDs between 640 and 874 GHz are functions of the IWP, and the TOC BTDs of 380, 640, and 874 GHz are functions of the D_me. The Voronoi ICS scheme possesses stronger sensitivity to the D_me than the sphere and column ICS schemes. Based on the sensitivity results, we built a multi-channel look-up table for BTDs. The IWP and D_me were searched from the look-up table using an optimal estimation algorithm. We used 2000 BTD test data randomly generated by the RSTAR model to assess the algorithm's accuracy. Test results showed that the correlation coefficients of the retrieved IWP and D_me reached 0.99 and 0.98, respectively. As an application, we used the inversion algorithm to retrieve the ice cloud IWP and D_me based on the Compact Scanning Submillimeter-wave Imaging Radiometer (CoSSIR) airborne terahertz radiation measurements. Validation against the retrievals of the Bayesian algorithm reveals that the Voronoi ICS model performs better than the sphere and hexagonal column ICS models, with enhancement of the mean absolute errors of 5.0 % and 12.8 % for IWP and D_me, respectively. In summary, the results of this study confirmed the practicality and effectiveness of the Voronoi ICS model in the terahertz remote sensing inversion of ice cloud microphysical properties.

A retrieval methodology based on the Bayesian neural network (BNN) is presented that inverts the ice water path (IWP), mean mass-weighted diameter (D _me ), and cloud height of ice clouds from sub-millimeter radiometer observations. The training dataset was created using collecting cloud profiles from the DARDAR (raDAR/liDAR) database and running simulations by the Atmospheric Radiative Transfer Simulator (ARTS) model. Since the effective radius (r _e ) is the size descriptor of ice particles in the DARDAR database, a look-up table of ice water content (IWC), D _me , and r _e was constructed to convert r _e profiles into D _me profiles. In addition, random noises corresponding to the measurement uncertainties of the Compact Scanning Submillimeter-wave Imaging Radiometer (CoSSIR) during the TC4 experiment were added to the simulated brightness temperatures before training the BNN. The proposed retrieval method was first applied to the simulated testing database, and then to the observations of CoSSIR. Moreover, the retrieved IWP and D _me were compared to the retrievals of the Bayesian Monte Carlo Integration (BMCI) method. The retrieved cloud height was assessed by cloud height extracted from the reflectivity data of the Cloud Radar System (CPS) flow on the same aircraft with CoSSIR. The comparison showed that the correlation coefficients of the retrieved IWP and D _me from the two methods are above 0.8, and the retrieved cloud height also showed good agreement with that extracted from the CPS.

Infrared hyperspectral remote sensing has been widely used in the field of meteorology. Many scientists have carried out research on inversion methods of meteorological elements such as thermodynamic profile, boundary layer height, cloud base height, etc. In this study, a method based on machine learning for cloud detection using ground-based infrared hyperspectral radiation data is proposed. The features of outliers, the cloudy and cloud-free data of Atmospheric Emitted Radiance Interferometer (AERI) radiation are extracted. The “reference values” of cloudy and cloud-free are determined based on the observation data of Vaisala CL31 ceilometer within the time range of 8 min before the corresponding time of AERI. A support vector machine (SVM) algorithm is used for training. The dataset comes from the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site and North Slope Alaska (NSA) site from 2015 to 2017, and the ARM West Antarctic Radiation Experiment (AWARE) site in 2016 is also analyzed. The instruments used in this paper include AERI, ceilometer, etc. The experimental results reveal that the agreement of cloud detection results between the proposed algorithm and ceilometer is about 93% at each site. However, for high clouds or optically thin clouds, the agreement will decrease.

The invariant imbedding (IIM) T-matrix method has shown great potential in light scattering field. However, the T-matrix need to be calculated through the matrix recurrence formula derived from the Helmholtz equation, thus its computational efficiency is much lower than Extended Boundary Condition Method (EBCM). To alleviate this problem, the Dimension-Variable Invariant Imbedding (DVIIM) T-matrix method is presented in this paper. Compared with the traditional IIM T-matrix model, the dimensions of the T-matrix and relevant matrices are gradually increasing as the iteration performed step by step, thus the unnecessary operations of large matrices can be avoided in early iterations. To optimally determine the dimension of these matrices in each iterative calculation, the spheroid-equivalent scheme (SES) is also proposed. The effectiveness of the DVIIM T-matrix method is validated from the modeling accuracy and calculation efficiency. The simulation results show that compared with traditional T-matrix method, its modeling efficiency can be improved notably, especially for the particles with large size and aspect ratio, where for the spheroid with a aspect ratio of 0.5, the computational time is cut down by 25%. Though the dimension of the T matrix is cut down in the early iterations, the computational precision of DVIIM T-matrix model is not decreased notably, and a good agreement is achieved between the calculation results of DVIIM T-matrix method, IIM T-matrix method and other well-validated models (like EBCM and DDACSAT), where the relative errors of the integral scattering parameters (e.g., extinction, absorption, scattering cross sections) are generally less than 1%.

This article presents a new method for retrieving the Ice Water Path (IWP), the median volume equivalent sphere diameter (D_me) of thin ice clouds (IWP < 100 g/m², D_me < 80 μm) in the Terahertz band. The upwelling brightness temperature depressions caused by the ice clouds at 325.15, 448.0, 664.0 and 874.0 GHz channels are simulated by the Atmospheric Radiative Transfer Simulator (ARTS). The simulated forward radiative transfer models are taken as historical data for the M5 model tree algorithm to construct a set of piecewise functions which represent the relation of simulated brightness temperature depressions and IWP. The inversion results are optimized by an empirical relation of the IWP and the D_me for thin ice clouds which is summarized by previous studies. We inverse IWP and D_me with the simulated brightness temperature and analyze the inversion performance of selected channels. The 874.4 ± 6.0 GHz channel provides the most accurate results, because of the strong brightness temperature response to the change of IWP in the forward radiative transfer model. In order to improve the thin ice clouds IWP and D_me retrieval accuracy at the middle-high frequency channels in Terahertz band, a dual-channel inversion method was proposed that combines the 448.0± 3.0 GHz and 664.0 ± 4.2 GHz channel. The error analysis shows that the results of the 874.4 ± 6.0 GHz channel and the dual-channel inversion are reliable, and the IWP inversion results meet the error requirement range proposed by previous studies.