Grading Evaluation and Ranking of CO2 Sequestration Capacity in Place(CSCIP) in China's Major Oil Basins: Theoretical,Effective, Practical and CCUS-EOR

Because it is necessary to focus on differences in regional oil reservoirs and determine the priority of the CCUSEOR(Carbon capture, utilization, and storage-enhanced oil recovery) deployment under China’s net-zero CO₂ emission target, systematic and regional evaluations of CO₂ sequestration capacity in major oil basins are needed considering the geofluid properties―carbon sequestration capacity in place(CSCIP)―where the ’in place’ indicates actual geological formation cond...     

卷积神经网络和视觉注意力语义分割模型在高分辨率遥感影像分类中的性能分析

随着深度学习语义分割的快速发展,基于计算机视觉语义分割模型的高分辨率遥感影像分类方法也大量涌现。为系统定量地研究经典的和先进的视觉语义分割模型在遥感影像分类中的性能,在总结深度学习语义分割进展的基础上,选择9种基于卷积神经网络(CNN)和视觉注意力的语义分割算法,对米级和厘米级2个尺度的遥感数据集进行分析研究。在模型构建上基于计算机视觉通用的语义分割框架,训练时采用红绿蓝3波段遥感图像并基于ImageNet预训练权重进行迁移学习训练。研究结果表明：通用的语义分割模型通过常规训练设置进行训练能取得较好的遥感影像分类效果,部分地物的交并比(IoU)可以达到90%以上;基于视觉注意力的遥感影像分类模型的精度普遍高于基于CNN的模型,且MaskFormer能更有效地提取离散的地物信息;不同类别的精度最高值并不全在总体最优模型中,部分会存在于次优模型中;类似的地物在更高分辨率遥感数据集中可以获得更高的精度。     

基于SWAN雷达拼图产品在暴雨过程中的对流云降水识别及效果检验

本文基于SWAN雷达拼图产品,选取了组合反射率因子、组合反射率因子水平梯度、回波顶高及垂直累积液态水含量作为识别参数,采用模糊逻辑法对暴雨过程中的对流云与层状云降水进行了分类试验,对发生在重庆的12次区域性暴雨天气过程分类结果进行了验证。并以ADTD地闪资料作为对流云降水的实况观测数据,分别采用了四种不同半径的空间匹配与四种不同时间匹配方式对识别出的对流云降水产品进行了定量检验。检验结果显示:随着空间匹配半径的增大,正确率明显提高,而6 min地闪相对于6 min拼图产品提前6、3、0 min及滞后3 min四种时间匹配方式,其正确率变化很小。对于12次暴雨过程的总体评分较高,检验方法具有清楚的物理意义,在不同的时空匹配方式下的评分结果符合实际情况,同时说明对流云与层状云降水分类效果较好,也是对对流云降水识别定量检验的一次探索。     

基于胶囊网络的跨领域情感分类方法

情感分析主要基于文本数据研究人们对于商品、服务、事件等对象的情感、意见或者态度.标记数据稀缺是情感分析领域面临的巨大挑战.在有监督的情感分类任务中,标记数据稀少会导致分类器的效果下降.跨领域的方法能够在一定程度上帮助解决该问题,但领域间往往存在差异.因此在利用领域适应方法进行情感分类时,分类器对目标领域的效果会变差.本文提出利用少量的目标领域标记信息来提高目标领域适应效果的思想.特别地,本文提出了一个基于胶囊网络的跨领域情感分类模型,在此基础框架上,设计了额外的胶囊网络层辅助目标领域的适应.在真实数据集上的实验结果表明,本文提出的模型效果优于以往的研究方法.     

2017年上海臭氧污染气象条件分析及臭氧污染天气分型研究

根据2016-2017年上海空气污染物和相关气象要素数据,分析上海2017年O₃变化特征以及气象影响因素,并对造成O₃污染的天气系统进行主观分型。结果表明：2017年上海市空气质量优良天数为275 d,占全年天数的75.3%,O₃-8h污染天数为55 d。上海市O₃-8h年均浓度115 μg·m^-3,同比增幅超过10%,为2013年以来的最高值,主要体现为夏半年O₃浓度的上升。与2016年相比,2017年夏半年西太平洋副热带高压偏西偏强,上海地区风速风向、温度、水汽、光照、辐射条件均有利于O₃浓度上升。造成2017年高浓度O₃污染主要有4种天气类型：副热带高压控制型（SH）、地面高压型（G）、均压场型（J）和低压型（D）。其中副热带高压控制型是典型的O₃-8h污染天气型,占总污染日数的29.1%,且污染程度较重;低压型出现次数较少;地面高压型的臭氧平均污染程度最弱。     

甘肃舟曲县牙豁口滑坡发育特征与成因分析

2019年7月16日,位于舟曲—化马深大活动断裂带的舟曲县牙豁口发生滑坡,滑坡长1 920 m,宽50~190 m,体积212×10⁴ m³,为典型的推移式大型长条状断层破碎带堆积层滑坡。滑坡自上而下沿原老滑道分级分块逐步发育形成,滑动历时近50天,最大滑距500 m,滑坡破坏X414县道、养鸡场等,并使岷江呈半堵塞状态,影响汛期行洪安全。通过现场调查、监测、无人机测量、岩土试验和收集历史滑坡资料,对牙豁口滑坡形成的地质环境、滑坡发育历史、滑坡分级分块特征、滑动过程和成因进行了系统研究。牙豁口滑坡分为上、下相对独立的两级滑坡。上级滑坡又可分为南支滑块和北支滑块两部分,在老滑坡的基础上长期孕育发展,上级滑坡北支滑块在2013年前就已活动,2019年7月16日上级滑坡南支滑块开始剧烈滑动,并推挤北支滑块前部一同复活滑动。上级滑坡约10×10⁴ m³滑体下滑加载于下级滑坡,下级滑坡于7月24日由上向下逐步扩展变形和滑动,下级滑坡又可分为上、中、下三段滑体,滑坡前缘于8月10日达到岷江边,使岷江呈半堵状态。牙豁口滑坡总体具有孕育历史悠久、滑动历时长、滑动速度较低、分级分块差异性滑动的特点。长期活动的深大断裂带及凹槽状地形、软弱易滑的灰黑色风化破碎炭质板岩、地下水和大气降水的强烈作用、两侧崩滑体的堆积加载是滑坡发生的主要原因。     

基于赫尔默特方差分量估计的水准网平差方法研究

复杂地区水准网平差时传统定权方法存在不合理性。根据高差观测误差与地形起伏的关系,建立高差分类新标准,推导出各类高差观测值权的计算公式,并引入赫尔默特方差估计法来合理匹配各类高差观测值的权比关系,最后通过算例验证本文方法的优越性。     

条带消除对于高光谱图像分类效果的影响——以CHRIS/PROBA图像为例

以CHRIS/PROBA高光谱图像数据为例,使用非监督分类的ISODATA和监督分类的最大似然法、支持向量机等3种经典的图像分类算法,对消条带前后的图像分别开展分类实验,并对分类结果做了分析。实验结果表明,消条带处理可以较好地改善高光谱图像分类结果的目视效果,能够消除不同类别斑块边缘因条带而产生的"毛刺"现象,这对地物斑块的形状及几何分布敏感的研究(如景观生态学)至关重要;但消条带处理对于提高分类精度的效果并不显著,精度提高最大值不到2%。     

Waveform-based point cloud classification in land-cover identification

Full-waveform topographic LiDAR data provide more detailed information about objects along the path of a laser pulse than discrete-return (echo) topographic LiDAR data. Full-waveform topographic LiDAR data consist of a succession of cross-section profiles of landscapes and each waveform can be decomposed into a sum of echoes. The echo number reveals critical information in classifying land cover types. Most land covers contain one echo, whereas topographic LiDAR data in trees and roof edges contained multi-echo waveform features. To identify land-cover types, waveform-based classifier was integrated single-echo and multi-echo classifiers for point cloud classification.The experimental area was the Namasha district of Southern Taiwan, and the land-cover objects were categorized as roads, trees (canopy), grass (grass and crop), bare (bare ground), and buildings (buildings and roof edges). Waveform features were analyzed with respect to the single- and multi-echo laser-path samples, and the critical waveform features were selected according to the Bhattacharyya distance. Next, waveform-based classifiers were performed using support vector machine (SVM) with the local, spatial features of waveform topographic LiDAR information, and optical image information. Results showed that by using fused waveform and optical information, the waveform-based classifiers achieved the highest overall accuracy in identifying land-cover point clouds among the models, especially when compared to an echo-based classifier.     

Operationalizing measurement of forest degradation: Identification and quantification of charcoal production in tropical dry forests using very high resolution satellite imagery

Quantification of forest degradation in monitoring and reporting as well as in historic baselines is among the most challenging tasks in national REDD+ strategies. However, a recently introduced option is to base monitoring systems on subnational conditions such as prevalent degradation activities. In Tanzania, charcoal production is considered a major cause of forest degradation, but is challenging to quantify due to sub-canopy biomass loss, remote production sites and illegal trade. We studied two charcoal production sites in dry Miombo woodland representing open woodland conditions near human settlements and remote forest with nearly closed canopies. Supervised classification and adaptive thresholding were applied on a pansharpened QuickBird (QB) image to detect kiln burn marks (KBMs). Supervised classification showed reasonable detection accuracy in the remote forest site only, while adaptive thresholding was found acceptable at both locations. We used supervised classification and manual digitizing for KBM delineation and found acceptable delineation accuracy at both sites with RMSEs of 25–32% compared to ground measurements. Regression of charcoal production on KBM area delineated from QB resulted in R²s of 0.86–0.88 with cross-validation RMSE ranging from 2.22 to 2.29 Mg charcoal per kiln. This study demonstrates, how locally calibrated remote sensing techniques may be used to identify and delineate charcoal production sites for estimation of charcoal production and associated extraction of woody biomass.