青藏高原东北缘断层活动变形的模拟研究

分析和讨论了位错理论模型的发展历史和应用领域，采用走滑和倾滑位错理论模型对青藏高原东北缘的多组断层活动产生的三维形变场进行了数值模拟研究，并将数值模拟结果与该地区1999～2001年间的GPS观测结果进行了对比分析。得出了甘青块体顺时针转动等一些较重要的结论和认识。     

Virtual Field Geologic Trip System

Virtual Field Geologic Trip System (VFGTS) constructed by the technique of visualization can efficiently present geologic field information and widely used in the field of geologic education. This paper introduces the developing thinking of VFGTS and discusses the main implement processes. Building VFGTS mainly includes systemically gathering of field geological data, the building of virtual geological world, and displaying of virtual geologic world and human-computer interaction.     

基于密度的线数据分组算法研究

目前,地理空间数据面临着由于数据量膨胀和计算量高速增长而引起算法效率低的问题,采用"分而治之"的数据分组策略提高运算效率已成为研究的热点。面向分布不均匀的线数据,本文提出了基于密度的线数据分组算法(简称LGAD)。首先,算法通过查找高密度区提取样本线段,保证了分组算法的起点落到高密区;其次,考虑线空间拓扑关系的复杂性,引用水平、垂直和夹角距离度量线段间距离,创建样本线段与其他线段的距离矩阵;最后,以距离矩阵和最优选择方法实现数据负载均衡分组。实验结果显示,对数据分组和分组后数据进行线段聚类的2个过程中,该算法体现了较好的时间优势,与串行计算相比,在分组数为2-12 时,平均比率达4.3,提高了应用的响应速度,具有较好的实际意义。     

空间同位模式支持下城市服务业关联发现及特征分析

空间同位模式分析是数据挖掘中一种常见的方法,可有效挖掘城市设施在空间位置上的关联特征,进而发现城市设施的分布规律.本文基于POI数据同位模式挖掘用来获取城市服务业空间关联结构:首先,通过邻近实例获取、同位候选模式存储与筛选,得到城市服务业二阶同位模式;然后,据此构造产业空间关联图,得到产业间的关联结构;最后,分别构造了...     

顾及地理-语义动态的城市热点预测框架

城市热点时空预测是城市管理和智慧城市建设的一项长期而富有挑战性的任务。准确地进行城市热点时空预测可以提高城市规划、调度和安全保障能力并降低资源消耗。现有的区域级深度时空预测方法主要利用基于地理网格的图像、给定的网络结构或额外的数据来获取时空动态。通过从原始数据中挖掘出潜在的自语义信息,并将其与基于地理空间的网格图像融合,也可以提高时空预测的性能,基于此,本文提出了一种新的深度学习方法地理语义集成神经网络(GSEN),将地理预测神经网络和语义预测神经网络相叠加。GSEN模型综合了预测递归神经网络(PredRNN)、图卷积预测递归神经网络(GCPredRNN)和集成层的结构,从不同的角度捕捉时空动态。并且该模型还可以与现实世界中一些潜在的高层动态进行关联,而不需要任何额外的数据。最终在3个不同领域的实际数据集上对本文提出的模型进行了评估,均取得了很好的预测效果,实验结果表明GSEN模型在不同城市热点时空预测任务中的推广性和有效性,利用该模型可以更好地进行城市热点时空预测,解决一系列如犯罪、火灾、网约车预订等等现代城市发展中亟需解决的相关问题。     

基于DEM小流域复杂网络的黄土高原地貌自动识别研究

地貌识别,对于人类建设,地质构造研究,环境治理等相关领域都有着重要意义。传统的基于像素单元或面向对象的地貌识别方法存在局限性。由于流域小单元具有表面形态的完整性,在地貌演化中具有明确的地理意义,基于流域小单元的地貌识别成为了该领域的一个新热点。然而,基于传统地形因子的地貌识别方法使用的因子往往较为单一或者在地学描述上存在重复性,目前尚无针对流域小单元进行空间结构描述和拓扑关系特征量化的地貌识别研究。基于此,本文基于DEM进行水文分析并通过坡谱方法解决了小流域稳定面积难以确定的问题,在黄土高原样区提取了181个稳定小流域。根据复杂网络理论和地貌学原理提出了流域加权复杂网络的概念和相应的8个定量指标用于流域空间结构的模拟和量化描述。最后采用了基于决策树的XGBoost机器学习算法进行地貌识别,实验对于黄土高原主要地貌类型的识别显现出较好的效果,Kappa系数为86.00%,总体精度达到了88.33%。对于地貌形态特征明显的地貌,复杂网络方法其顾及空间结构和拓扑特征的特性导致了其较高的识别性能,精度和召回率都在90%～100%之间。通过与前人的研究进行对比,其识别结果亦呈现出较高的精度,这些...     

城市用地功能精细化识别方法：时序动态图嵌入深度学习模型

多源大数据融合背景下的城市功能区识别是复杂非线性系统的模式识别问题,如何有效地从大规模的轨迹数据中提取出多粒度连续性时变和多尺度空间相互作用的信息是进行城市区域功能识别的关键。本研究设计实现了一种基于时序动态图嵌入的深度学习模型,在融合滴滴出行及兴趣点数据（Point of Interest, POI）基础上,提取城市区域存在的时间和空间上的隐式特征,结合聚类分析实现城市用地功能的语义识别。结果表明,成都市中心的用地功能趋向复合多样化的发展,且用地属性随时间发生作用范围和用地类型的变化,呈现出功能随着城市群体活动而变化的时空规律。与相关文献的对比实验表明,本文提出方法在更细粒度的时间段下进行功能区识别,得到的同一类功能区域内集聚度更高,能够更好的捕获复合型区域在不同时间模式下呈现出的用地功能变化。本研究为城市用地功能识别研究提供了新的技术方法,为城市规划研究人员全面理解城区结构属性提供了有效手段,对推动城市空间得到更合理高效的利用具有一定的价值。     

A two-zone model for the broad iron line emission in MCG–6-30-15

We reanalyse the ASCA and BeppoSAX data of MCG–6-30-15, using a double-zone model for the iron line profile. In this model, the X-ray source is located around ≈10 Schwarzschild radii and the regions interior and exterior to the X-ray source produce the line emission. We find that this model fits the data with a similar reduced χ ² to the standard single-zone model. Thus we show that the presence of a broad iron line feature does not necessarily require that the X-ray source be located close to the last stable orbit or in the disc rotation axis.
Within the framework of this model, the best-fitting inclination angle of the source     for the intermediate-intensity ASCA data set is compatible with that determined by earlier modelling of optical lines. The observed variability of the line profile with intensity can be explained as variations of the X-ray source size. That several active galactic nuclei with broad lines have the peak centroid near 6.4 keV can be explained under certain conditions.
We also show that the simultaneous broad-band observations of this source by BeppoSAX rule out the Comptonization model which was an alternative to the standard inner-disc one. We thereby strengthen the case that line broadening occurs as a result of the strong gravitational influence of a black hole.     

The role of non-thermal electrons in the hydrogen and calcium lines of stellar flares

There is observational evidence showing that stellar and solar flares occur with a similar circumstance, although the former are usually much more energetic. It is expected that the bombardment by high-energy electrons is one of the chief heating processes of the flaring atmosphere. In this paper we study how a precipitating electron beam can influence the line profiles of Ly α , H α , Ca  ii K and λ 8542. We use a model atmosphere of a dMe star and make non-LTE computations taking into account the non-thermal collisional rates owing to the electron beam. The results show that the four lines can be enhanced to different extents. The relative enhancement increases with increasing formation height of the lines. Varying the energy flux of the electron beam has different effects on the four lines. The wings of Ly α and H α become increasingly broad with the beam flux; change of the Ca  ii K and λ 8542 lines, however, is most significant in the line centre. Varying the electron energy (i.e. the low-energy cut-off for a power-law beam) has a great influence on the Ly α line, but little on the H α and Ca  ii lines. An electron beam of higher energy precipitates deeper, thus producing less enhancement of the Ly α line. The Ly α /H α flux ratio is thus sensitive to the electron energy.     

Further constraints on variation of the fine-structure constant from alkali-doublet QSO absorption lines

Comparison of quasar (QSO) absorption-line spectra with laboratory spectra provides a precise probe for variability of the fine-structure constant, α , over cosmological time-scales. We constrain variation in α in 21 Keck/HIRES Si  iv absorption systems using the alkali-doublet (AD) method in which changes in α are related to changes in the doublet spacing. The precision obtained with the AD method has been increased by a factor of 3:     . We also analyse potential systematic errors in this result. Finally, we compare the AD method with the many-multiplet method, which has achieved an order of magnitude greater precision, and we discuss the future of the AD method.