葵花-8卫星AOD资料在CMA-MESO/CUACE CW 3DVar同化系统中的个例应用研究

CMA-MESO/CUACE CW化学天气耦合模式是自主研发的大气化学耦合模式,目前CMA-MESO/CUACE CW3DVar同化系统实现了地面气溶胶观测可吸入颗粒物PM_(2.5)和PM10的同化,为增强耦合同化系统非常规观测的同化能力,文中在CMA-MESO大气化学天气耦合三维变分同化框架基础上,利用查表法获得气溶胶消光系数,然后建立气溶胶光学厚度(AerosolOpticalDepth,AOD)和气溶胶组分之间关系的观测算子、切线性观测算子和伴随观测算子,实现AOD观测资料的同化应用。针对2016年12月18-20日华北、黄淮地区一次污染天气过程进行同化预报试验,试验结果表明同化葵花-8卫星(Hi‐mawari-8)气溶胶光学厚度观测后,PM_(2.5)分析的重污染区范围有所扩大,山西东南部分析与实况分布更为接近,但是山东大部地区PM_(2.5)分析偏强,与观测相比PM_(2.5)质量浓度存在高估。同时同化Himawari-8 AOD观测和地面气溶胶站点观测的PM_(2.5)分析最优,分析与观测距平相关系数最高,平均偏差、均方根误差及标准差最小。重污染区的PM_(2.5)预报检验结果表明,同化Himawari-8AOD观测对大于350µg·m^(-3)量级PM_(2.5)预报正贡献可以持续到48 h,但整体来说,同时同化Himawari-8AOD观测和地面气溶胶站点观测对各个量级的PM_(2.5)质量浓度预报质量最优。

基于Depth-wise卷积和视觉Transformer的图像分类模型

图像分类作为一种常见的视觉识别任务,有着广阔的应用场景。在处理图像分类问题时,传统的方法通常使用卷积神经网络,然而,卷积网络的感受野有限,难以建模图像的全局关系表示,导致分类精度低,难以处理复杂多样的图像数据。为了对全局关系进行建模,一些研究者将Transformer应用于图像分类任务,但为了满足Transformer的序列化和并行化要求,需要将图像分割成大小相等、互不重叠的图像块,破坏了相邻图像数据块之间的局部信息。此外,由于Transformer具有较少的先验知识,模型往往需要在大规模数据集上进行预训练,因此计算复杂度较高。为了同时建模图像相邻块之间的局部信息并充分利用图像的全局信息,提出了一种基于Depth-wise卷积的视觉Transformer(Efficient Pyramid Vision Transformer,EPVT)模型。EPVT模型可以实现以较低的计算成本提取相邻图像块之间的局部和全局信息。EPVT模型主要包含3个关键组件:局部感知模块(Local Perceptron Module,LPM)、空间信息融合模块(Spatial Information Fusion,SIF)和“+卷积前馈神经网络(Convolution Feed-forward Network,CFFN)。LPM模块用于捕获图像的局部相关性;SIF模块用于融合相邻图像块之间的局部信息,并利用不同图像块之间的远距离依赖关系,提升模型的特征表达能力,使模型学习到输出特征在不同维度下的语义信息;CFFN模块用于编码位置信息和重塑张量。在图像分类数据集ImageNet-1K上,所提模型优于现有的同等规模的视觉Transformer分类模型,取得了82.6%的分类准确度,证明了该模型在大规模数据集上具有竞争力。

AOD炉钼精矿直接合金化钼收得率影响因素分析

以钢液中C、Si、Mn、Fe等元素还原MoO_(3)的冶金热力学为基础,在AOD炉中进行钼精矿直接合金化试验,并分析加入时机、加入方式、熔炼温度、钢液成分对AOD炉钼的收得率影响。结果表明:采用加氧化钼矿球后其上覆盖碱性炉渣的加料方式钼元素收得率最高,可达到98.33%;在加入量相同的情况下,C质量分数为0.8%时加入钼精矿,钼元素的收得率较高;[C]质量分数为4.00%的纯铁液中,1600℃时[C]还原氧化钼的速率明显高于1500℃,纯铁液中不同质量分数的[Si]还原氧化钼的速率均高于碳元素。

基于AOD-Net的雾天高速公路能见度动态检测方法

单纯以图像帧差特征为主的公路能见度动态检测方法,缺乏足够多样且具有代表性的真实数据集,使得能见度的智能检测精度下降,为此,利用生成对抗网络的博弈迭代计算优势,设计一种基于AOD-Net(An All-in-One Network)的雾天高速公路能见度动态检测。首先,采用生成对抗网络中的带雾图像生成算法,主要是判断生成的带雾图像的真实程度,旨在使判别器准确地区分真实图像和生成的图像。然后,捕获不同尺度下的特征,从而更准确地估计雾霾参数,利用AOD-Net完成雾霾动态识别与能见度检测。最后,构建团雾分级预警模型,以实现团雾智能预警。通过对比实验证明,所提检测方法可以实现对雾天高速公路能见度动态高精度检测,检测结果与实际能见度偏差不超过5m,具备较高的应用价值。

基于改进AOD-Net的图像去雾算法

为了更好解决图像去雾后颜色失真、去雾不彻底和耗时等问题,提出了一种基于改进AOD-Net的图像去雾算法。首先,在原有的卷积模块中引入残差连接,并保留了第二个特征融合层第一层的特征信息,以增强网络的特征提取能力。其次,在第三个特征融合层后引入注意力模块,强化雾图中的关键特征信息,抑制无关背景干扰。最后,采用新的复合损失函数进行训练。实验结果表明,改进算法在公共数据集上的峰值信噪比提高了3.8 dB,结构相似性达到了93.6%。相较于其他去雾算法,该算法在去雾精度和处理效率方面均表现出色。

基于改进FeatDepth的足球运动场景无监督单目图像深度预测

为了在降低成本的同时提高图像深度信息预测的精确度,并将深度估计应用于足球运动场景,提出一种基于改进FeatDepth的足球运动场景无监督单目图像深度预测方法。首先,对原FeatDepth引入注意力机制,使模型更加关注有效的特征信息;其次,将FeatDepth中的PoseNet网络和DepthNet网络分别嵌入GAM全局注意力机制模块,为网络添加额外的上下文信息,在基本不增加计算成本的情况下提升FeatDepth模型深度预测性能;再次,为在低纹理区域和细节上获得更好的深度预测效果,由单视图重构损失与交叉视图重构损失组合而成最终的损失函数。选取KITTI数据集中Person场景较多的部分进行数据集制作并进行仿真实验,结果表明,改进后的FeatDepth模型不仅在精确度上有所提升,且在低纹理区域及细节处拥有更好的深度预测效果。最后,对比模型在足球场景下的推理效果后得出,改进后的模型在低纹理区域(足球、球门等)及细节处(肢体等)有更好的深度预测效果,实现了将基于无监督的单目深度估计模型应用于足球运动场景的目的。

Spatio-Temporal Variations in the Associations between Hourly PM_2.5and Aerosol Optical Depth (AOD) from MODIS Sensors on Terra and Aqua

Recent studies have explored the relationship between aerosol optical depth (AOD) measurements by satellite sensors and concentrations of particulate matter with aerodynamic diameters less than 2.5 μm (PM2.5). However, relatively little is known about spatial and temporal patterns in this relationship across the contiguous United States. In this study, we investigated the relationship between US Environmental Protection Agency estimates of PM2.5 concentrations and Moderate Resolution Imaging Spectroradiometer (MODIS) AOD measurements provided by two NASA satellites (Terra and Aqua) across the contiguous United States during 2005. We found that the combined use of both satellite sensors provided more AOD coverage than the use of either satellite sensor alone, that the correlation between AOD measurements and PM2.5 concentrations varied substantially by geographic location, and that this correlation was stronger in the summer and fall than that in the winter and spring.

基于Stacking模型融合的AOD终点碳温预测

为提高AOD不锈钢冶炼终点碳温预测的准确性和可靠性,提出一种基于多个机器学习算法(RF、XGBoost、AdaBoost、KNN、SVR和岭回归)嵌入的Stacking模型融合的AOD终点碳温预测方法。基于理论基础和相关性分析确定了模型的输入特征变量,利用箱线图法对历史数据进行预处理,结合5折交叉验证和贝叶斯优化算法确定了模型的最优参数,通过对上述6种机器学习算法的逐层筛选,构建了RF+XGBoost+KNN—RF二层Stacking多模型融合的终点碳含量预测模型、RF+AdaBoost+KNN—RF+XGBoost+KNN—XGBoost三层Stacking多模型融合的终点温度预测模型。预测结果表明,RF+XGBoost+KNN—RF二层Stacking模型在终点碳质量分数误差为±0.01%的命中率为87.86%,RF+AdaBoost+KNN—RF+XGBoost+KNN—XGBoost三层Stacking模型在终点温度误差为±15℃的命中率为94.22%,相较于单一的机器学习模型,预测精度显著提高。

基于Prophet-LSTM+P-Bshade融合模型的中国陆地MAIAC AOD补值研究

因受云层等气象条件干扰和卫星传感器性能影响,卫星获取的气溶胶光学厚度(aerosol optical depth,AOD)存在数据缺失的情况,为了提高AOD数据的覆盖度,利用AOD的时空相关性,提出Prophet-LSTM+P-Bshade融合模型,对2016年3月—2021年2月我国1 km×1 km分辨率的多角度大气校正算法(multiangle implementation of atmospheric correct, MAIAC) AOD数据进行时空补值,并与地面监测站点AERONET AOD进行对比验证。结果表明:提出的融合模型补值成果精度明显优于其他对比模型,验证精度指标R、MAE和RMSE分别为0.905、0.237和0.375。2016年3月—2021年2月中国陆地AOD值分布时间上季度性变化较明显,整体呈现冬季>春季>夏季>秋季的季节分布特征,空间上总体呈现东高西低和塔里木盆地局部较高的特点。

AOD脱铁提镍工艺用镁碳砖的侵蚀机制

为了探究AOD(氩氧精炼)脱铁提镍工艺用镁碳砖的损毁原因,从AOD脱铁提镍炉下线炉衬渣线镁碳砖取样,分析了AOD脱铁提镍工艺用后渣线镁碳砖的化学和物相组成及显微结构,并采用FactSage热力学软件模拟了脱铁渣对镁碳砖的侵蚀,并对比其与镁钙砖炉衬对钢水纯净度的影响。结果表明:在冶炼机械应力和耐火材料毛细管力的双重作用下,高FeO含量脱铁渣进入方镁石相界面;由于FeO和MgO可以无限固溶,相界面的黏结相和部分方镁石相会溶解进脱铁渣中,导致耐火材料中方镁石骨料剥落,且高温液相黏度降低,耐火材料缓冲应力能力下降。在机械应力的作用下,脱铁渣对耐火材料剧烈扰动,形成机械剥落,造成用后耐火材料工作面较为平缓且光滑,分形维数在1.1~1.2。高温液相进入脱铁渣过程中,使得镁碳砖中石墨相在强氧化环境下得以大量保存,同时降低了耐火材料对镍铁水的影响;与使用镁钙砖相比,镁碳砖炉衬寿命大幅提升,镍铁水纯净度方面达到了镁钙砖相同的效果。

降低太钢AOD炉和钢包用耐火材料单耗的控制措施

介绍了在耐火材料区域整体承包模式下,钢铁企业关注的指标不仅仅只是耐火材料吨钢成本,更要重点关注耐火材料吨钢消耗指标,降低耐火材料吨钢消耗,既能提高炉包寿命,降低耐火材料吨钢成本,又能保证耐火材料产品不污染钢水,提高钢水纯净度,实现耐火材料企业和钢铁企业双赢。以太钢不锈钢冶炼耐火材料消耗占比较大工序AOD炉和钢包为例,对不锈钢冶炼用AOD炉和钢包冶炼环境进行了分析,对降低耐火材料单耗的方法进行了介绍。通过采用优质耐火材料原料改进材质、加强砌筑管理、优化生产节奏、采用在线热态喷补技术、提高透气砖寿命等,使不锈钢冶炼用耐火材料单耗和成本最低。

Measurement of Metallogenic Depth and Deep Second Enrichment Belt Forecasting in Jiaodong Gold Deposit

The depth is important for ore finding in Jiaodong gold deposit. However, many geologists are still discussing how to confirm the depth for the tectonic and metallogenesis formation. The authors of this paper propose a new method-the correction of metallogenic depth via its structure to calculate the depth. This method, based on the crust rock in a solid stress state, emphasizes the elastic pattern rather than the static fluid pattern. In addition, this method is more appropriate to the actual situation in the crust than the method of weight/special weight. The authors of this paper illustrating, with the Jiaodong gold deposit as an example, the metallogenic depth correction via structure conclude that the depth of the most deposits, lower than 4-6 km, is often 2.5 km. Therefore, the authors suggest that there exists a second enrichment belt and that ore resources are more potential at the belt of Jiaodong area. These results have been demonstrated by years of exploration.

Edge enhanced depth perception with binocular meta-lens

The increasing popularity of the metaverse has led to a growing interest and market size in spatial computing from both academia and industry.Developing portable and accurate imaging and depth sensing systems is crucial for advancing next-generation virtual reality devices.This work demonstrates an intelligent,lightweight,and compact edge-enhanced depth perception system that utilizes a binocular meta-lens for spatial computing.The miniaturized system comprises a binocular meta-lens,a 532 nm filter,and a CMOS sensor.For disparity computation,we propose a stereo-matching neural network with a novel H-Module.The H-Module incorporates an attention mechanism into the Siamese network.The symmetric architecture,with cross-pixel interaction and cross-view interaction,enables a more comprehensive analysis of contextual information in stereo images.Based on spatial intensity discontinuity,the edge enhancement eliminates illposed regions in the image where ambiguous depth predictions may occur due to a lack of texture.With the assistance of deep learning,our edge-enhanced system provides prompt responses in less than 0.15 seconds.This edge-enhanced depth perception meta-lens imaging system will significantly contribute to accurate 3D scene modeling,machine vision,autonomous driving,and robotics development.

基于MAIAC AOD时空补值数据的PM2.5浓度估算研究

气溶胶光学厚度被广泛应用于PM2.5浓度估算中,受极端气候影响以及卫星传感器影响,AOD数据存在大量缺失,本文提出Prophet-LSTM + P-Bshade时空补值模型对MAIAC AOD数据进行补值并使用Catbooost模型结合AOD数据以及ERA5气象数据对中国2020年陆地区域的PM2.5浓度进行估算。结果表明:① Prophet-LSTM + P-Bshade时空补值模型精度明显优于传统补值方法,R、MASE和MAE分别为0.891、0.275和0.183。② Catboost模型在PM2.5浓度估算中比常用的其他机器学习等模型显示更高的估算精度,R、MASE和MAE分别为0.93、15.89 μg∙m−3和10.54 μg∙m−3。③ 中国陆地区域2020年的PM2.5浓度在季节尺度分布上明显,整体呈现冬季 > 春季 > 秋季 > 夏季的季节分布特点。在空间分布上,PM2.5浓度整体呈现东部地区较高,塔里木盆地区域局部较高的特点。

Applications of Satellite Data for Aerosol Optical Depth (AOD) Retrievals and Validation with AERONET Data

An understanding of the amount and type of aerosols present in the atmosphere is required for the atmospheric correction of satellite imagery. A sensitivity analysis of the atmospheric inputs to the MOD09 software has shown that uncertainty in the estimation of Aerosol Optical Depth (AOD) has the greatest impact on the accuracy of atmospheric correction of MODIS data. The MOD09 software used for the generation of surface reflectance products estimates the AOD of the atmosphere at the time of image acquisition. AOD measurements retrieved from MODIS were compared with near-simultaneous AErosol RObotic NETwork (AERONET) data over three sites in Australia, using time-series of MODIS surface reflectance products. The results of the study provide an important independent validation of ACRES MODIS Surface Reflectance Products. This procedure may be applied to long time series MODIS data for estimating the accuracy of MOD09 retrieved AOD.

Responses of the Southern Ocean mixed layer depth to the eastern and central Pacific El Niño events during austral winter

Based on the Ocean Reanalysis System version 5(ORAS5)and the fifth-generation reanalysis datasets derived from European Centre for Medium-Range Weather Forecasts(ERA5),we investigate the different impacts of the central Pacific(CP)El Niño and the eastern Pacific(EP)El Niño on the Southern Ocean(SO)mixed layer depth(MLD)during austral winter.The MLD response to the EP El Niño shows a dipole pattern in the South Pacific,namely the MLD dipole,which is the leading El Niño-induced MLD variability in the SO.The tropical Pacific warm sea surface temperature anomaly(SSTA)signal associated with the EP El Niño excites a Rossby wave train propagating southeastward and then enhances the Amundsen Sea low(ASL).This results in an anomalous cyclone over the Amundsen Sea.As a result,the anomalous southerly wind to the west of this anomalous cyclone advects colder and drier air into the southeast of New Zealand,leading to surface cooling through less total surface heat flux,especially surface sensible heat(SH)flux and latent heat(LH)flux,and thus contributing to the mix layer(ML)deepening.The east of the anomalous cyclone brings warmer and wetter air to the southwest of Chile,but the total heat flux anomaly shows no significant change.The warm air promotes the sea ice melting and maintains fresh water,which strengthens stratification.This results in a shallower MLD.During the CP El Niño,the response of MLD shows a separate negative MLD anomaly center in the central South Pacific.The Rossby wave train triggered by the warm SSTA in the central Pacific Ocean spreads to the Amundsen Sea,which weakens the ASL.Therefore,the anomalous anticyclone dominates the Amundsen Sea.Consequently,the anomalous northerly wind to the west of anomalous anticyclone advects warmer and wetter air into the central and southern Pacific,causing surface warming through increased SH,LH,and longwave radiation flux,and thus contributing to the ML shoaling.However,to the east of the anomalous anticyclone,there is no statistically significant impact on the MLD.

改进AOD-Net的道路交通图像去雾算法

针对现有图像去雾算法在处理道路交通图像时无法同时兼顾去雾效果和实时性的问题,本文以快速一体化网络去雾算法(AOD-Net)为基础进行改进.首先,在AOD-Net中添加SE通道注意力,以自适应的方式分配通道权重,关注重要特征;其次,引入金字塔池化模块,扩大网络的感受野,并融合不同尺度特征,更好地捕捉图像信息;最后,使用复合损失函数同时关注图像像素信息和结构纹理信息.实验结果表明,改进后的AOD-Net算法对道路交通图像去雾后的峰值信噪比提升了2.52 dB,结构相似度达到了91.2%,算法复杂度和去雾耗时略微增加,但仍满足实时要求.

Emplacement Depths and Exhumation of Mesozoic Granitoids Responding to Lithospheric Destruction in the Jiaodong Peninsula, North China Craton

1 Introduction The North China Craton(NCC)has experienced lithospheric destruction in Mesozoic accompanied with crustal exhumation.Fission track or(U-Th)/He dating of zircon and apatite for the Mesozoic granitoids in the

Three-dimensional crystal defect imaging by STEM depth sectioning

One of the major innovations awaiting in electron microscopy is full three-dimensional imaging at atomic resolution.Despite the success of aberration correction to deep sub-angstrom lateral resolution,spatial resolution in depth is still far from atomic resolution.In scanning transmission electron microscopy(STEM),this poor depth resolution is due to the limitation of the illumination angle.To overcome this physical limitation,it is essential to implement a next-generation aberration corrector in STEM that can significantly improve the depth resolution.This review discusses the capability of depth sectioning for three-dimensional imaging combined with large-angle illumination STEM.Furthermore,the statistical analysis approach remarkably improves the depth resolution,making it possible to achieve three-dimensional atomic resolution imaging at oxide surfaces.We will also discuss the future prospects of three-dimensional imaging at atomic resolution by STEM depth sectioning.

Investigation of Groundwater Quality with Borehole Depth in the Basin Granitoids of the Ashanti Region of Ghana

The dependence of groundwater quality on borehole depth is usually debatable in groundwater studies, especially in complex geological formations where aquifer characteristics vary spatially with depth. This study therefore seeks to investigate the relationship between borehole depth and groundwater quality across the granitoid aquifers within the Birimian Supergroup in the Ashanti Region. Physicochemical analysis records of groundwater quality data were collected from 23 boreholes of public and private institutions in the Ashanti Region of Ghana, and the parametric values of iron, fluoride, total hardness, pH, nitrate, and nitrite were used to study the groundwater quality-depth relationship. The results showed that the depth-to-groundwater quality indicated a marginal increase in water quality in the range of 30 to 50 m, which is mathematically represented by the low-value correlation coefficient (r2 = 0.026). A relatively significant increase occurs in the depth range of 50 to 80 m, which is given by a correlation coefficient of r2 = 0.298. The mean percent parameter compatibility was 74%, 82%, 89%, and 97% at 50, 60, 70, and 80 m depths, respectively. The variations in groundwater quality per depth ratio ranged from 1.48, 1.37, 1.27, and 1.21 for 50, 60, 70, and 80 m depth, respectively. The recommended minimum borehole depth for excellent groundwater quality is suggested with a compatibility per meter depth ratio of 1.37. This results in a range between 50 and 70 m as the most desirable drilling depth for excellent groundwater quality within the granitoids of the Birimian Supergroup of the Ashanti Region in Ghana.