Statistical Distribution Characteristics Research Articles

Quantized Graph Neural Networks for Image Classification

Researchers have resorted to model quantization to compress and accelerate graph neural networks (GNNs). Nevertheless, several challenges remain: (1) quantization functions overlook outliers in the distribution, leading to increased quantization errors; (2) the reliance on full-precision teacher models results in higher computational and memory overhead. To address these issues, this study introduces a novel framework called quantized graph neural networks for image classification (QGNN-IC), which incorporates a novel quantization function, Pauta quantization (PQ), and two innovative self-distillation methods, attention quantization distillation (AQD) and stochastic quantization distillation (SQD). Specifically, PQ utilizes the statistical characteristics of distribution to effectively eliminate outliers, thereby promoting fine-grained quantization and reducing quantization errors. AQD enhances the semantic information extraction capability by learning from beneficial channels via attention. SQD enhances the quantization robustness through stochastic quantization. AQD and SQD significantly improve the performance of the quantized model with minimal overhead. Extensive experiments show that QGNN-IC not only surpasses existing state-of-the-art quantization methods but also demonstrates robust generalizability.

Construction of college music information teaching mode under the background of Internet

Abstract The traditional music teaching method in the informationization era has been difficult to adapt to the needs of modern teaching and must be reformed in the direction of informationization. In this paper, based on the closure, inflection point and outer enclosing box features of the stroke line element, the recognition of handwritten notes is carried out from the three categories of straight line segments, folded line segments and quadratic curves. Meanwhile, for the binarized music score image, the multi-directional LBP features for spectral line detection are improved, and the computation method of multi-scale spectral line detection LBP features is established. The Manhattan distance is used to evaluate and select the features, which are inputted into XGBoost for classification and recognition training based on the statistical distribution characteristics of the features. Note recognition and spectral line recognition are applied to college music teaching, and the effectiveness of teaching is explored. In the rhythm-recognition path, the recognition teaching based on multi-scale and multi-directional LBP features led to an increase in students’ mastery of the musical score by 2.8 and in the phrasing and segmentation path by 3.5. Informational teaching led to a deepening of students’ mastery of the notes and musical scores.

Characteristics of hemoglobin distributions in preschool children and non-pregnant women of reproductive age and their implications for establishing quality control criteria for hemoglobin data in field surveys: evidence from 483 surveys conducted in refugee settings worldwide

BackgroundCurrently, there is a lack of clear guidance on hemoglobin (Hb) data quality parameters and plausible flagging ranges for population-representative surveys. There is a need to determine which properties of Hb data indicate lower data quality and increased measurement error and which represent intrinsic statistical properties of Hb distributions rather than quality problems.MethodsWe explored statistical characteristics of Hb distributions and plausible exclusion ranges in population-representative surveys of non-pregnant women of reproductive age (WRA) (15–49 years, n = 401 surveys) and children (6–59 months, n = 461 surveys) conducted in refugee settings by the United Nations High Commissioner for Refugees (UNHCR). Hb distribution characteristics [standard deviation (SD), skewness and kurtosis] were compared to those from Demographic and Health Surveys (DHS).ResultsOverall, 0.08% of child and 0.14% of WRA Hb values were outside of the previously proposed 4.0–18.0 g/dL plausible range. Surveys conducted in Uganda tended to have unusually high SD compared with surveys from other settings, possibly an indication of problematic measurement quality. We therefore used summary results on SD, skewness and kurtosis excluding surveys from Uganda when comparing with DHS results or proposing plausible ranges. Both WRA and child Hb distributions tended to be left-skewed and had excess positive kurtosis. Mean survey-level SD was greater, mean skewness more negative, and mean kurtosis more positive in WRA surveys compared to child surveys. All these findings were broadly similar to those from DHS surveys. Mean SD in DHS surveys was higher than that in our data for both children (1.48 vs. 1.34) and WRA (1.58 vs. 1.43).ConclusionsWe observed several statistical characteristics of Hb distributions that may not necessarily be indicative of data quality problems and bear strong similarities with the characteristics found in DHS surveys. Hb distributions tended to be negatively skewed and positively kurtotic, and SD in many surveys exceeded 1.5 (previously proposed upper plausible range). Based on our empirical evidence, surveys with skewness above + 0.2 and kurtosis below -0.5 or Hb SD outside the range of 1.1–1.55 g/dL for children (6–59 mo) or 1.1–1.65 g/dL for non-pregnant WRA (15–49 y) may require further quality investigation.

Statistical characteristics of raindrop size distribution during rainy seasons in complicated mountain terrain

Abstract. In order to improve our understanding of the characteristics of raindrop size distribution (DSD) over complex mountainous terrain, the differences in DSD over the southern slopes, northern slopes, and interior of the Qilian Mountains were analyzed using 6 months of observations. For all rainfall events, the number concentrations of small and large raindrops in the interior and on the southern slopes were greater than on the northern slopes, but midsize raindrops were less. The DSD spectrum of the interior was more variable and differed significantly from that of the northern slopes. The differences in the normalized intercept parameters of the DSD for stratiform and convective rainfall were 8.3 % and 10.4 %, respectively, and those of the mass-weighted mean diameters were 10.0 % and 23.4 %, respectively, while the standard deviations of DSD parameters at interior sites were larger. The differences in the coefficient and exponent of the Z–R relationship were 2.5 % and 10.7 %, respectively, with an increasing value of the coefficient from the southern to the northern slopes for stratiform rainfall but the opposite for convective rainfall. In addition, the DSD characteristics and Z–R relationships were more similar at the ipsilateral sites and had smaller differences between the southern slopes and interior of the mountains.

Raindrop Size Distribution Characteristics of Heavy Precipitation Events Based on a PWS100 Disdrometer in the Alpine Mountains, Eastern Tianshan, China

As a key component of the hydrological cycle, knowledge and comprehension of precipitation formation and evolution are of leading significance. This study investigates the statistical characteristics of raindrop size distribution for heavy precipitation events with observations collected by a Present Weather Sensor (PWS100) disdrometer located in the alpine area of eastern Tianshan, China. The characteristics are quantified based on heavy rain, heavy snow, and hail precipitation events classified using the rainfall intensity and the precipitation-related weather codes (US National Weather Service). On average, the heavy precipitation events in the headwaters of the Urumqi River are dominated by medium-sized (2–4 mm) raindrops. As well, we investigate mass-weighted mean diameter–normalized intercept parameter scatterplots, which demonstrate that the heavy precipitation events in alpine regions of the Tianshan Mountains can be identified as maritime-like clusters. The concentration of raindrops in heavy precipitation is the highest overall, while the concentration of raindrops in heavy snow is the lowest when the diameter is lower than 1.3 mm. The power–law relationships of radar reflectivity (Z) and rain rate (R) [Z = ARb] for the heavy rain, heavy snow, and hail precipitation events are also calculated. The Z–R relationship of heavy rain and heavy snow in this work has a lower coefficient value of A (10 and 228.7, respectively) and a higher index value of b (2.6 and 2.1, respectively), and the hail events are the opposite (A = 551.5, b = 1.3), compared to the empirical relation (Z = 300R1.4). Furthermore, the possible thermodynamics and general atmospheric circulation that cause the distinctions in the raindrop size distribution characteristics between alpine areas and other parts of the Tianshan Mountains are also debated in this work. The headwaters of the Urumqi River in alpine areas have relatively colder and wetter surroundings in the near-surface layer than the foothills of the Tianshan Mountains during the precipitation process. Meanwhile, a lower temperature, a higher relative humidity, a more efficient collision coalescence mechanism, and glacier local microclimate effects (temperature jump, inverse glacier temperature, glacier wind) at the headwaters of the Urumqi River during the precipitation process are probably partly responsible for more medium- and large-size drops in the mountains.

Semi-supervised fuzzy clustering algorithm based on prior membership degree matrix with expert preference

Existing pre-processing methods for the prior membership degree matrix suffer from the following issues: (1) The labeling constraints for prior membership degree matrix have an effect on the expert’s judgment on the prior membership degree, which easily causes the distortion problem of the prior membership degree labeling information; (2) There exists the problem of inconsistency between the filling information and the labeling information in the prior membership degree matrix to be filled in the missing values with zeros. To address these problems, we propose an unconstrained labeling idea for the prior membership degree matrix and the corresponding pre-processing method for the missing values by introducing the statistical characteristics of extreme value distribution and simultaneously apply it to the semi-supervised fuzzy clustering algorithm. More specifically, we focus on learning an expert preference value from the prior membership degree matrix and filling in the missing values with the expert preference value. Thus, we propose an unconstrained pre-processing method for the prior membership degree matrix by filling in missing values with an expert preference to keep the filling information consistent with the labeling information in the prior membership degree matrix as much as possible. In addition, we design a semi-supervised fuzzy clustering algorithm based on an unconstrained prior membership degree matrix with expert preference (SFCM-EP) by introducing the K-L divergence to improve the applicability, utility and running performance of semi-supervised fuzzy clustering algorithm. Our experimental results on the simulation dataset and the UCI datasets show the feasibility and effectiveness of the proposed pre-processing method of the prior membership degree matrix with encouraging results.

Wave heights from sea surface and bottom measurements: Variations with different water depths

Wave instruments equipped with pressure sensors have long been recognized as an effective means of obtaining statistical characteristics of wave distributions in complex sea conditions. However, the commonly used theory based on pressure response functions may not accurately capture surface height signals due to the nonlinearity of wave propagation at different water depths. To address this issue, field measurements were conducted at five stations with water depths ranging from 2.4 m to 54.8 m. The validity threshold of the linear wave theory was investigated using surface buoy and bottom pressure measurements. Statistical analyses showed that the classical linear theory performed well in shallow water, but its credibility decreased with increasing depth. In addition, the measured data showed agreement with the Weibull distribution for wave pressure and the Rayleigh distribution for surface wave height. For applications in ocean and coastal engineering, we propose two empirical formulas for estimating wave heights based on seabed or sea surface measurements at intermediate water depths. These formulas take into account the ratio of wave height to water depth, which reflects the nonlinear effects in shallow water, thereby improving the accuracy of surface wave height estimation (with an accuracy of ±10%) compared to the traditional linear theory method.

Modeling of System Availability and Bayesian Analysis of Bivariate Distribution

To meet the desired standard, it is important to monitor and analyze different engineering processes to obtain the desired output. The bivariate distributions have received a significant amount of attention in recent years due to their ability to describe randomness of natural as well as artificial mechanisms. In this article, a bivariate model is constructed by compounding two independent asymmetric univariate distributions and by using the nesting approach to study the effect of each component on reliability for better understanding. Furthermore, the Bayes analysis of system availability is studied by considering prior parametric variations in the failure time and repair time distributions. Basic statistical characteristics of marginal distribution like mean median and quantile function are discussed. We used inverse Gamma prior to study its frequentist properties by conducting a Monte Carlo Markov Chain (MCMC) sampling scheme.

Neural computing for grey Richards differential equation to forecast traffic parameters with various time granularity

The existing traffic parameter prediction methods generally adopt a single prediction model, but the fusion of different theories and methods can complement each other and improve the prediction performance of the model. Starting from the statistical distribution characteristics of traffic flow, this work introduces the Richards equation to conduct grey modeling, which is used to simulate the development trend of traffic parameters; and then fuses the abilities of error feedback adjustment and complex nonlinear fitting of neural network to estimate the parameters of the grey model and forecast the volatility of traffic flow respectively. At the same time, a dynamic prediction framework of real-time data update is built for the new model, and finally establish the dynamic grey Richards neural network model (DGR-NN).Apply the new model to different traffic parameters (Speed; Volume; Jam mileage) and data resolutions (5 min; 15 min; 1 h), the modeling effect of DGR-NN is significantly improved compared to the grey Richards model (GRM), and the training and testing errors of the model in the three forecast scenarios are reduced to varying degrees, where the testing MAPE, RMSE and STD are reduced by 1.80% ∼ 10.87%, 2.55% ∼ 7.26% and 8.08–25.56% respectively. Furthermore, the results of the new model were verified and analyzed with the other five comparison models, among which the boxplot of APE shows that the error distribution of DGR-NN prediction data is concentrated and the value level is relatively low. It can be seen that DGR-NN can accurately and stably forecast traffic parameters of different time granularities.

Effects of vegetation distribution along river transects on the morphology of a gravel bed braided river

The interaction between vegetation, sediment, and water flow creates various fluvial landscapes. Hydrological conditions and flood disturbances, as well as the habitat preference of vegetation, regulate its spatial distribution. To describe the spatial distribution of vegetation cover, here, we focus on vegetation distributions along river transverse transects that define vertical and horizontal distances from water areas during low flow periods. As one of the most dynamic river types, braided rivers can be significantly influenced by vegetation encroachment. However, the effects of vegetation distributions along river transects on braided river morphology remain unknown. To study the potential influence of vegetation distribution along river transects, a depth-averaged, hydro-morphodynamic model was employed. Using the model, we investigated a medium-sized, braided river with a gravel bed affected by riparian vegetation. The following scenarios of vegetation transect distributions were examined: (1) vegetation established near or covering the low water channel, and (2) vegetation established on bar tops and kept at a distance from the low water channel. The model successfully reproduced a reduction in the braiding index for a vegetated braided river. Depending on the transect distribution scenarios employed, significantly different effects for river morphology were obtained. For example, compared to vegetation on bar tops, vegetation located near the low water channel played a more critical role for changing river morphology, redirecting water flow, and changing the statistical characteristics of the riverbed elevation distribution. Vegetation near the low water channel not only concentrated water flow to low water channels but also redirected flow to the high elevation area by reducing low water channel flow capacity. The revealed effects of the vegetation transect distribution on river morphology development helped to determine effective management protocols for reducing the negative impact of vegetation encroachment.

Influence of Support Parameter and Excavation Methods on Statistical Distribution Characteristics of Surrounding Rock Pressure in Shallow-Buried Metro Tunnel

For large-spanshallow-buried tunnels in composite strata of urban metro, the study on the distribution characteristics of surrounding rock pressure is a necessary prerequisite for the reliability analysis of tunnel structure. In this paper, through numerical simulation, the random joint file generated by MATLAB is combined with the discrete element software to establish the corresponding random network model. According to the network model calculation, the influence of surrounding rock pressure on the support structure system is discussed, and the reasonable support parameter types are given. Comparing the whole section excavation with the subsectional excavation, the subsectional excavation can effectively control the internal stress self-adjustment of the surrounding rock and effectively restrain the plastic zone. It provides some theoretical guidance and reference for tunnel reliability design.

A Hierarchical Met-Ocean Data Selection Model for Fast O&M Simulation in Offshore Renewable Energy Systems

In this research, a hierarchical met-ocean data selection model is proposed to reduce the computational cost in stochastic simulation of operation and maintenance (O&M) and enable rapid evaluation of offshore renewable energy systems. The proposed model identifies the most representative data for each calendar month from the long-term historical met-ocean data in two steps, namely the preselection and the refined selection. The preselection incorporates three distinct metrics to evaluate the characteristics of statistical distributions, including the Jensen–Shannon divergence, the encapsulation of extreme met-ocean conditions, as well as the overall vessel accessibility. For the refined selection, a component of temporal synchrony is devised to emulate dynamic changes of met-ocean conditions. As such, a met-ocean reference year comprising twelve representative historical months is subsequently produced and deployed as the input for O&M stochastic simulation. While this research focuses on the development of a generalised methodology for selecting representative met-ocean data, the proposed statistical method is validated empirically using a case study inspired by real-life floating offshore wind installations in Scotland, e.g., Hywind and Kincardine projects. According to the O&M simulation results with five capacity scenarios, the proposed data selection model reduces the computational cost by up to 97.65% while emulating the original results with minor deviations, i.e., within ±5%. The simulation speed is therefore 43 times quicker. Overall, the proposed met-ocean data selection model attains an excellent trade off between computational efficiency and accuracy in O&M stochastic simulation.

Deep Learning-Based Channel Modeling for Free Space Optical Communications

Instead of conventional knowledge-driven modeling, this study introduced data-driven channel modeling methods based on deep learning (DL) in a free-space optical (FSO) communication system to achieve low complexity, less modeling difficulty, high fidelity, and differentiability. The FSO channel was modeled via three DL algorithms: generative adversarial network (GAN), bidirectional long short-term memory (BiLSTM), and Bayesian neural network (BNN). Consequently, their performance was comprehensively compared in terms of optical amplitude waveforms, Kullback-Leibler (KL) divergence, 95% confidence interval, data distribution, and fitted probability density functions. The results indicated that the GAN-based channel exhibited optimal performance, implying its suitability for learning the stochastic distribution of the FSO channel. In addition, the performance of GAN-based channels was examined under different conditions: turbulence intensities (weak, moderate, and strong), transmission distances (3, 4, and 5 km), launch powers (0, 10, and 20 dBm), modulation formats (on-off keying, pulse position modulation, and pulse amplitude modulation 4), and the adaptability to channel variation. The test results were satisfactory and confirmed the ability of the GAN to learn the approximate distribution function of the FSO channel. Thus, the primary aim was to prove that for FSO channels with statistical distribution characteristics, a data-driven DL model can not only serve as an effective surrogate model for specific scenarios but also provide a potential tool for future modeling methods.

Robust Human Upper-Limbs Trajectory Prediction Based on Gaussian Mixture Prediction

Accurate prediction of human motion trajectory can improve the security of human-robot cooperation. Due to the unstructured nature of collaborative workspace and the uncertainty of sensor sensing data, the trajectory prediction accuracy of traditional prediction algorithms is low, and the uncertainty is difficult to estimate. Aiming at the complex characteristics of human upper limb movement patterns, this paper proposes a robust upper limb end trajectory prediction algorithm. The robust Gaussian mixture model was used to model the trajectory of human upper limb movement, and the statistical values of the future trajectory were obtained by combining Gaussian mixture regression. The advantage of this algorithm is that the prediction result is not only the predicted value of the position, but also the probability distribution of all possible future motion trajectories of the upper limb. The position prediction information in a specific motion mode can be obtained by using probability and statistical distribution characteristics. The algorithm is tested on both public and private datasets. Experimental results show that this method can predict human trajectories well.

Statistical Characteristics of Raindrop Size Distribution in the South China Monsoon Region (Guangdong Province)

Statistical Characteristics of Raindrop Size Distribution in the South China Monsoon Region (Guangdong Province)

Характеристика розподілу хімічних елементів у вертикальному розрізі торфів за допомогою рентген-флуоресцентного аналізу (родовище Гончари, Львівська область)

This article discusses the features of peat analysis using X-ray fluorescence (XRF) analysis in order to study its qualitative and quantitative elemental composition, including heavy metals. The distribution of chemical elements is an indicator of various processes in geochemical and biological systems, by using of which it is possible to reproduce the conditions of accumulation of mineral deposits. This analysis is an important component of a comprehensive study of peat formation features, the environmental friendliness of peat extraction, and also for determining the suitability of peat for industrial use. We analyzed the content of chemical elements in peat samples taken at different depths using a portable X-ray fluorescence spectrometer. The article considers the main characteristics of the spectrum of individual elements, depending on the atomic number. In order to establish the general regularity of the distribution of 20 chemical elements in peat samples, we performed a mathematical and statistical analysis of the obtained data: calculation of the main statistical characteristics of chemical elements distribution (average, minimum and maximum values, median, variance, coefficient of variation, etc.), calculation of correlation matrices, selection of typomorphic geochemical associations of chemical elements using cluster and factor analyses. We singled out two types of factors that are decisive and influence the accumulation of chemical elements in the investigated peat: “organogenic” and “natural” (lithological), which are decisive, and a secondary factor –anthropogenic. We compared the obtained results with the average values obtained from the results of spectral semi-quantitative analysis of peat ash samples taken at depths of 0.1–7 m in the same region. We evaluated the possibility and efficiency of using a portable X-ray fluorescence spectrometer for the analysis of the macro- and microelement composition of peats with different ash content. It has been established that portable X-ray fluorescence analysis is a powerful tool for fast and high-quality elemental analysis of peat, and the range of its application depends on specific research goals and tasks.

Topic Controlled Steganography via Graph-to-Text Generation

Generation-based linguistic steganography is a popular research area of information hiding. The text generative steganographic method based on conditional probability coding is the direction that researchers have recently paid attention to. However, in the course of our experiment, we found that the secret information hiding in the text tends to destroy the statistical distribution characteristics of the original text, which indicates that this method has the problem of the obvious reduction of text quality when the embedding rate increases, and that the topic of generated texts is uncontrollable, so there is still room for improvement in concealment. In this paper, we propose a topic-controlled steganography method which is guided by graph-to-text generation. The proposed model can automatically generate steganographic texts carrying secret messages from knowledge graphs, and the topic of the generated texts is controllable. We also provide a graph path coding method with corresponding detailed algorithms for graph-to-text generation. Different from traditional linguistic steganography methods, we encode the secret information during graph path coding rather than using conditional probability. We test our method in different aspects and compare it with other text generative steganographic methods. The experimental results show that the model proposed in this paper can effectively improve the quality of the generated text and significantly improve the concealment of steganographic text.

Correction of multi-scale sunglint reflections from the water surface in airborne high-spatial resolution optical images.

Airborne optical images (AOI) are often with complex sunglint reflections, which brings a certain influence to watercolor retrieval. This includes the sunglint reflection with water surface statistical distribution characteristics caused by imaging viewing angles differences, with high spatial resolution surface discrete characteristics sharing similar viewing angles, and the surface Fresnel reflection sunglint differences caused by the skylight difference during the flight of unmanned aerial vehicles. Aiming at the multiscale optical characteristics of sunglint reflection in high spatial resolution AOI, based on multi-path optical radiation transmission, the sunglint reflection interference from three different imaging processes is clarified. We developed a correction method to eliminate these different sunglint reflections on water surfaces and improve the reflectivity accuracy. The comparison with the in situ measured remote sensing reflectance of water indicated that the root mean square error (RMSE) was reduced from 0.0009 sr-1 to 0.0004 sr-1, and the mean relative error (MRE) decreased from 21.8% to 15.7%. This method has also been applied to correct the Airborne Visible Infrared Imaging Spectrometer (AVIRIS) images, showing good applicability. The method is fast, effective, and without auxiliary parameters, which provides a correction reference for different surface sunglint corrections of various AOI.

Assessing Air Pollution and Determining the Composition of Airborne Dust in Urbanized Areas: Granulometric Characteristics

The article focuses on the problem of atmospheric pollution in urban areas caused by particulate matter, which is a problem for most industrial cities. The work aims to assess air pollution by particulate matter PM10 and PM2.5; determine the composition of airborne dust in a zone free of any industrial activity; obtain statistical characteristics of dust distribution in the atmospheric air of urbanized areas according to granulometric characteristics. The issues of PM10 and PM2.5 presence in the atmospheric air of urban and industrial centers in major European cities were analyzed. The degree of air pollution by PM10 and PM2.5 in the autumn-winter season 2020–2021 in Chelyabinsk city was assessed by gravimetry, considering the air quality standards valid in Russia and the European Union. It was found that the average concentration of particulate matter in the air of the urbanized areas of Chelyabinsk in the six months of the autumn-winter period 2020–2021 was 83 µg/m3 and exceeded the MPC (maximum permissible concentration) 2.1 times. The average value of the concentration of fine dust fractions for six months of the study period was 67 µg/m3, exceeding the MPC by 2.7 times. The obtained results indicate that there is a big problem of pollution, therefore it is important to apply the necessary actions to restore the state of air quality standards, which do not cause health risks to the population exposed to ultrafine particulate matter.

Statistical Characteristics of Warm Season Raindrop Size Distribution in the Beibu Gulf, South China

Raindrop size distribution (DSD) can be used to improve the accuracy of radar quantitative precipitation estimation (QPE) and further understand the microphysical process of precipitation; however, its spatio-temporal characteristics vary with different climates, rain types, and geographical locations. Due to the lack of observations, the DSD characteristics in the Beibu Gulf, especially at the rainfall center of Guangxi in South China, is poorly understood. In this paper, these regional DSD characteristics were analyzed during the warm season with an upgraded version of the OTT Particle Size Velocity (Parsivel) (OTT2) disdrometer. The DSD datasets from June to October 2020 and March to May 2021 were grouped into convective and stratiform precipitation by rain rate (R). The rainfall parameters were calculated from DSDs to further understand the rain characteristics. The results showed that: (1) the regional DSDs feature the lowest concentration of largest-sized drops when compared with the statistical results for other areas such as Zhuhai in South China, Nanjing in East China, Hubei province in Central China and Beijing in North China; (2) the raindrop spectra have an excellent fit with the three-parameter gamma distribution, particularly in regard to the medium-size raindrops; (3) the μ–Λ relation is closer to the coastal regions than the inland area of South China; (4) the localized Z−R relations differ greatly for convective rainfall (Z = 202.542 R1.553) and stratiform rainfall (Z = 328.793 R1.363). This study is the first study on DSDs in the Beibu Gulf region. The above findings will provide a better understanding of the microphysical nature of surface precipitation for different rain types along the Beibu Gulf in southern China, which may improve precipitation retrievals from remote sensing observations.