基于K2结构学习算法的多光谱影像贝叶斯网络分类器

用基于启发式搜索的结构学习算法，学习得到多光谱影像的贝叶斯网络结构，分析了TM的波段（特征）间条件独立性假设的合理性，给出了贝叶斯推理中后验概率计算的公式，并通过和最大似然法对比试验分析了简单贝叶斯网络应用于多光谱影像分类的优势。     

基于力学模式的大地测量反演理论及应用

本文的研究工作主要集中在大地测量反演的模型建立、求解方法及应用的相关问题的探讨.首先,在综合分析大地测量反演理论与应用现状的基础上,明确了大地测量反演的定义并对反演方法作了分类,提出了深化大地测量反演研究的若干观点. 根据有代表性的准则函数推演出大地测量反演参数可辨识条件,给出其适用范围;概括典型的大地测量力学模型,讨论了其反演参数可辨识性问题.尝试性地提出大地测量反演的模型辨识这样的一个新课题,并初步探讨了大地测量反演模型辨识的定义、原则和方法,基本描绘了现代大地测量反演研究的系统化、模型化和智能化的架构. 探讨了适合于大地测量反演区域变形的位移数据处理方案:①基于相似变换的"特定基准"确定;②基于一种线性模型的"异常位移"定位;③局部构造有限元的处理.方案体现了一种用"拟准检定法"进行"异常位移"定位的新思想.研究了正法反演结果的几种定性检验方案和统计假设检验定量检验方法,引入一种基于概率距离的定量评价新方法. 比较了约束和非约束的各种局部最优化方法(如单纯形法、模式搜索法、鲍威尔法、变量轮换法、混合罚函数法、复合形法等)和全局搜索算法(如蒙特卡罗、遗传算法),通过反演试验,探讨了不同最优化算法在大地测量反演中的有效性.探索了遗传算法在大地测量反演应用研究中的参数设置和算法的实施步骤,反演实例得到了较好的结果. 初步研究了利用包括大地测量位移在内的2种以上的观测物理量进行联合反演的模型和实施步骤,特别提出在联合反演中引入相对权比,并将它作为待反演参数,同其他反演参数一同反演求解的理论和方法.模拟算例结果显示出这种方法具有优越性和合理性. 在阐述了中国大陆及邻域板块边界构造作用力和构造应力场反演研究现状的基础上,本文利用自己开发编制的大地测量反演程序和中国地壳观测网络GPS观测资料和主应力方向数据进行了中国大陆及邻域边界力位移反演和川滇地区边界力的联合反演研究.反演结果表明:①形成现今中国大陆内部位移场和构造应力场的主要力源仍是来自印度板块和欧亚板块的陆-陆碰撞;②太平洋板块和菲律宾海板块作用力虽然比印度板块作用力小,但仍是影响中国大陆东部的重要因素;③日本海弧后盆地现在继续着引张活动,这可能与洋壳后撤所形成的对大陆的拖曳有关;④来自北部的力承载着中国大陆北向运动作用,表现出相当大的阻力;⑤嫩江断裂和西拉木伦河断裂等似有活动迹象,活动方式不能确定;⑥南北构造带在现今构造运动中起着明显的削减作用.     

结合非线性贝叶斯反演算法的油田储层参数反演

基于地球物理反演算法，利用地表沉降信息快速获取油田储层状态信息，对于油田储层安全的实时监控具有重要意义。目前国内外对油田储层参数反演的研究相对较少，因此本文以典型的因石油开采引起地表沉降的辽河盘锦地区作为研究区域，以InSAR沉降监测结果作为观测数据，以Okada模型作为反演模型，首次将非线性贝叶斯反演算法引入到油田储层参数反演中。在试验过程中，发现油田地下存在不止一个油层，首次将双Okada模型引入到油田储层参数反演中。试验表明：①非线性贝叶斯反演算法不仅能够获取模型参数的最优值，同时可以对反演结果的不确定性做出解释。②与单源Okada模型相比，双Okada模型更符合该地区油田储层参数变化与地表沉降之间的映射关系，基于双Okada模型反演得到的油田储层参数更可靠。     

地壳形变数据反演

本论文旨在研究利用地壳形变数据来研究地壳形变和地震的物理机制。受到板块构造运动的驱使,全球地壳形变最大和地震发生最频繁的区域位于板块的边界处。参照地震孕育的过程,地壳形变可以划分为4个阶段：震间形变、震前形变、同震形变和震后形变。本论文主要研究震间和同震形变,提出了一种刚体运动加断层位错的板块边界断层运动模型。该模型假定板块间的相对运动长期（数千年到数百万年）不变,其大小可以通过数年的VLBI测量和全球板块运动模型来确定。板块间的相对运动在其边界的断层面上受到阻碍,从而引起周围地壳的形变,产生应变和应力累积。当累积的应力超过断层岩石的强度时,就会产生破裂和滑动,即地震。传统的边角测量方法和现代的空间大地测量方法,如GPS、VLBI等,都可以用来监测地壳形变。利用观测到的形变数据可以反演地震时地下断层面上的滑动分布,以便进一步研究地震的物理机制;也可以反演震间形变所对应的断层带上的应变和应力累积速率（通过反演负位错模型参数求得）,从而给出活动断层中长期地震预报。所涉及的形变数据反演一般都是非线性反演问题,其解不惟一。为此论文详细讨论了顾及模型参数先验信息的贝叶斯反演方法,针对大地测量数据反演提出了两种构造先验信息矩阵的方法,经验统计法和物理约束法,较好地克服了反演问题的不惟一性。运用上述刚体运动加位错模型和贝叶     

用大地测量资料反演青藏高原构造应力场的初步尝试

本文讨论了大地测量反演构造应力场的理论与方法。结合青藏高原地质、地球物理资料、利用限单元法，初步建立了青藏高原构造力场三维弹性数值分析模型。用地表水准和ＧＰＳ资料提供的位移值作为地表边界约束条件，反演计算了青藏高原变场和应力场，数值结果表明，青藏高原构造应力场以南北向挤压为主，东西向拉张为辅，欧亚板块与印度板块相撞作用仍控制着高原现今构造应力场。     

贝叶斯网络支持的地表参数混合反演模式研究

基于贝叶斯网络理论，建立用于植被地表参数估计的混合反演模式，结合遥感物理模型实现了冬小麦叶片叶绿素含量（Cab）和冠层叶面积指数（LAI）的反演。用模型模拟数据以及2001年顺义遥感实验数据验证结果表明，LAI和Cab均有较好的反演精度。针对含噪声模拟数据反演结果中约有10％的噪声数据反演失败的情况，用不确定知识的处理方法有效地降低了失败点的比例。混合反演模式本质上是一个融合先验知识与观测数据的知识推理方案，本文实现了对反演过程中参数后验概率更新算法并引入热力学中的信息熵概念实现了参数后验信息动态定量计算，同时简单探讨了现阶段定量评价遥感反演过程中信息流控制存在的难点问题。     

基于改进人工蜂群算法的概率积分法参数反演

针对传统算法在反演概率积分法参数时易发散且难以获得全局最优解的问题,提出利用自适应人工蜂群算法反演概率积分法参数。根据该算法在求解过程中收敛速度快,获得全局最优解的特点,将参数反演问题转化为组合优化问题,建立了自适应人工蜂群算法的概率积分法预计参数反演流程,并将计算结果与实际值进行对比分析。通过理论分析与实验证明,自适应人工蜂群算法反演概率积分法参数精度高,较最小二乘法和模矢法拟合效果好,可应用于矿山开采沉陷预计。     

整周模糊度搜索方法的效率比较和分析

GPS整周模糊度的快速、正确求解是利用GPS载波相位进行高精度定位中最为关键的问题，目前应用最广泛的模糊度解算方法是搜索法。本文比较了不同搜索法的搜索效率，并对搜索的算法和过程进行内在原因的分析。     

遗传算法及其在GOMS模型反演中的应用效果分析

几何光学交互遮蔽模型（GOMS）是一种重要的遥感前向模型,它较好解释了“热点”现象,具有较强的前向模拟能力。但由于其固有的非线性性,给反演带来困难。本文尝试采用近年来兴起的并行随机全局寻优算法－遗传算法对GOMS进行反演,并针对传统遗传算法的不足进行了改进。在使用相同先验知识的条件下,将该算法与目前最有效的约束非线性最优化确定性搜索算法－逐步二次规划法对GOMS模型的反演效果进行了比较,结果表明,逐步二次规划法搜索效率较高,但结果受初值的影响很大,初值选择不当,易收敛于局部最优解,而遗传算法具有全局最优的收敛效果,但局部搜索效率较差。在某些对精度要求不高,而对搜索效率要求较高的场合,可以采用遗传算法与确定性搜索算法相结合的混合遗传算法,以提高算法的搜索效率,获得较为满意的效果。     

遗传算法及其在GOMS模型反演中的应用效果分析

几何光学交互遮蔽模型（GOMS）是一种重要的遥感前向模型,它较好解释了“热点”现象,具有较强的前向模拟能力。但由于其固有的非线性性,给反演带来困难。本文尝试采用近年来兴起的并行随机全局寻优算法－遗传算法对GOMS进行反演,并针对传统遗传算法的不足进行了改进。在使用相同先验知识的条件下,将该算法与目前最有效的约束非线性最优化确定性搜索算法－逐步二次规划法对GOMS模型的反演效果进行了比较,结果表明,逐步二次规划法搜索效率较高,但结果受初值的影响很大,初值选择不当,易收敛于局部最优解,而遗传算法具有全局最优的收敛效果,但局部搜索效率较差。在某些对精度要求不高,而对搜索效率要求较高的场合,可以采用遗传算法与确定性搜索算法相结合的混合遗传算法,以提高算法的搜索效率,获得较为满意的效果。     

红河断裂带闭锁程度和滑动亏损分布特征研究

利用1999~2013年青藏高原东南缘的GPS速度场观测数据,采用DEFNODE程序反演红河断裂带走滑速率、三维闭锁程度和滑动亏损分布。反演结果表明:红河断裂带北、中、南段右旋走滑速率分别约为(5.9±1.2)mm/a,(4.8±0.6)mm/a和(4.3±0.4)mm/a,在地表以下6km的闭锁程度分别为0.43,0.22和0.25,其滑动亏损速率分别为3.3mm/a,2.6mm/a和2.3mm/a,红河断裂带闭锁程度和应变积累程度都比较低,与近年来红河断裂带整体活动微弱的现状相吻合。     

陕西地区垂直形变分析

利用中国陕西地区三期国家一等水准网的平差成果、两期二等水准网成果、两期陕西省高程控制网成果、两期全球导航卫星系统（global navigation satellite system, GNSS）大地控制网成果、陆态网络等工程连续运行基准站及流动站成果、地理国情监测成果,分析陕西地区最新垂直运动趋势。结果表明:GNSS垂直运动与水准垂直运动趋势整体一致;20世纪70年代至2018年,关中平原垂直运动呈下降趋势,关中平原周边山区及高原垂直运动呈上升趋势;陕北地区地壳垂直运动总体上呈现上升趋势,平均上升速度约为4 mm/a;关中平原总体呈下降趋势,下降最大速度为6 mm/a;陕南地区也呈上升趋势,平均上升速度约为3 mm/a;陕北、西安局部区域的不均匀地面沉降灾害严重。     

Contemporary deformation in the Kashmir–Himachal,Garhwal and Kumaon Himalaya: significant insights from 1995–2008 GPS time series

We present new insights on the time-averaged surface velocities, convergence and extension rates along arc-normal transects in Kumaon, Garhwal and Kashmir–Himachal regions in the Indian Himalaya from 13 years of high-precision Global Positioning System (GPS) time series (1995–2008) derived from GPS data at 14 GPS permanent and 42 campaign stations between $29.5{-}35^{\circ }\hbox {N}$ and $76{-}81^{\circ }\hbox {E}$ . The GPS surface horizontal velocities vary significantly from the Higher to Lesser Himalaya and are of the order of 30 to 48 mm/year NE in ITRF 2005 reference frame, and 17 to 2 mm/year SW in an India fixed reference frame indicating that this region is accommodating less than 2 cm/year of the India–Eurasia plate motion ( ${\sim }4~\hbox {cm/year}$ ). The total arc-normal shortening varies between ${\sim }10{-}14~\hbox {mm/year}$ along the different transects of the northwest Himalayan wedge, between the Indo-Tsangpo suture to the north and the Indo-Gangetic foreland to the south indicating high strain accumulation in the Himalayan wedge. This convergence is being accommodated differentially along the arc-normal transects; ${\sim } 5{-}10~\hbox {mm/year}$ in Lesser Himalaya and 3–4 mm/year in Higher Himalaya south of South Tibetan Detachment. Most of the convergence in the Lesser Himalaya of Garhwal and Kumaon is being accommodated just south of the Main Central Thrust fault trace, indicating high strain accumulation in this region which is also consistent with the high seismic activity in this region. In addition, for the first time an arc-normal extension of ${\sim }6~\hbox {mm/year}$ has also been observed in the Tethyan Himalaya of Kumaon. Inverse modeling of GPS-derived surface deformation rates in Garhwal and Kumaon Himalaya using a single dislocation indicate that the Main Himalayan Thrust is locked from the surface to a depth of ${\sim }15{-}20~\hbox {km}$ over a width of 110 km with associated slip rate of ${\sim }16{-}18~\hbox {mm/year}$ . These results indicate that the arc-normal rates in the Northwest Himalaya have a complex deformation pattern involving both convergence and extension, and rigorous seismo-tectonic models in the Himalaya are necessary to account for this pattern. In addition, the results also gave an estimate of co-seismic and post-seismic motion associated with the 1999 Chamoli earthquake, which is modeled to derive the slip and geometry of the rupture plane.     

时变重力测量确定青藏高原地壳隆升与增厚速率

联合绝对重力和重力反演与气候实验卫星（gravity recovery and climate experiment,GRACE）重力多年观测数据,获得了青藏高原多个基准站区域的地壳垂直形变速率。研究结果表明,绝对重力呈明显的负变化,绝对重力和卫星重力的时变系统差也呈较一致的负值,鼎新（DXIN）、德令哈（DLHA）、西宁（XNIN）、拉萨（LHAS）和仲巴（XZZB）5个基准站的区域地壳垂直形变呈明显的隆升状态,即拉萨块体、祁连块体和阿拉善块体处于地壳隆升状态,隆升速率分别约为2.01±0.15 mm/a、1.88±0.19mm/a、1.91±0.10 mm/a。在印度板块和欧亚板块的双向挤压下,青藏高原的地壳在不断的隆升与增厚,平均隆升速率约为1.94±0.17 mm/a,平均增厚速率约为2.35±3.30 mm/a。     

尼泊尔Mw 7.9级地震同震垂直位移与断层运动模型

2015年4月25日尼泊尔地区发生了Mw 7.9级地震,发震断层位于印度板块与欧亚板块碰撞边界带,此次地震是一次典型的板块逆冲型事件。利用中国境内加密的GPS同震观测资料,融合ALOS-2卫星L波段的InSAR（interferometric synthetic aperture radar）同震形变数据,基于最小二乘方法获得了此次地震的同震垂直位移场。同震垂直位移结果表明,此次地震造成尼泊尔加德满都地区抬升约0.95 m,珠穆朗玛峰地区受地震的影响有所下降,其主峰的沉降量为2~3 cm,中国境内的希夏邦马主峰沉降约为20 cm。地区利用改进的二维弹性半空间位错模型反演了发震断层运动参数,本文模型显示此次地震的断层面破裂宽度约为60 km,平均滑动量达到4 m,相当于Mw 7.89级。     

Determination of Postglacial Land Uplift in Fennoscandia from Leveling, Tide-gauges and Continuous GPS Stations using Least Squares Collocation

Precise leveling, tide-gauge recordings and time series from continuous GPS stations are all sources of information about the ongoing postglacial land uplift in Fennoscandia. This article describes how to gather these three sources of data together in a least-squares collocation adjustment in order to calculate uplift rates, as well as new heights for the benchmarks in the leveling network. The estimated reliability of the resulting uplift model is in general better than 0.4 mm/year, decreasing in areas of fewer data.A smoother and extended version of the land uplift model described here is used in the adjustment of the new height system in Sweden (RH 2000) under the name RH 2000 LU. Later on, the name was changed to NKG2005LU. The Finnish Geodetic Institute (FGI) has decided to use the same model when they calculate their new height system, N2000.     

Ocean loading effects on the prediction of Antarctic glacial isostatic uplift and gravity rates

The effect of regional ocean loading on predicted rates of crustal uplift and gravitational change due to glacial isostatic adjustment (GIA) is determined for Antarctica. The effect is found to be significant for the ICE-3G and ICE-5G loading histories (up to ?8 mm/year and ?3 mm/year change in uplift rate and ?3 cm/year and ?1 cm/year equivalent water height change (EWHC) of surface mass, respectively). The effect is smaller (+1 mm/year; +0.25 cm/year) for the IJ05 loading history. The impact of ocean loading on the rate of change of the long-wavelength zonal harmonics of the Earth’s gravitational field is also significantly smaller for IJ05 than ICE-3G. A simple analytical formula is derived that is accurate to about 3% in a root-mean-square sense that relates predicted or observed gravitational change at the surface of the Earth (r = a) to the EWHC. A fundamental difference in the definition of the load histories accounts for the differing sensitivities to ocean loading. IJ05 defines its surface load history relative to the present-day surface load, rather than specifying an absolute loading history, and thus implicitly approximates the temporal and spatial mass exchange between grounded ice and open ocean. In contrast, ICE-3G and ICE-5G specify an absolute load history and explicit regional ocean loading substantially perturbs predicted GIA rates. Conclusions of previous studies that used IJ05 predictions without ocean loading are relatively robust.     

Assessment of Land Surface Subsidence Due to Underground Metal Mining Using Integrated Spaceborne Repeat-Pass Differential Interferometric Synthetic Aperture Radar (DInSAR) Technique and Ground Based Observations

Differential interferometric synthetic aperture radar (DInSAR) is a novel remote sensing technique to measure earth surface deformation. It is capable of obtaining dense information related to the deformation of a large area efficiently, economically and effectively. Therefore, DInSAR is a promising technology for monitoring the earth surface deformation related to some natural hazardous events, such as earthquake, volcano eruption, land subsidence, landslide. In present study, Conventional DInSAR technique have been applied to a mineral rich zone, coming under the Khetri copper belt, a part of Northern Aravali range of hillocks in India, predominant with mining activities since late 1960’s to address the possibility of deformation phenomena due to hard rock underground metal mining. Four interferometric SAR data sets of Radarsat-2 was used for the study area to address the subsidence/uplift phenomena. Further, results obtained from conventional DInSAR technique using Radarsat-2 data sets compared with results obtained from ground based observation technique for its validity. In both the techniques, deformation results obtained in terms of average subsidence rate in mm (quarterly basis) of points under study within mining zone of Mine-A has well agreed to each other. Further, it has been observed that average subsidence rate in mm (quarterly basis) obtained from space based observation and ground based observation are 5.6 and 6.67, respectively over the points under study in mining zone of Mine-A.     

Estimation and validation of snow surface temperature using modis data for snow-avalanche studies in NW-Himalaya

Snow avalanche studies require different snow-meteorological parameters for avalanche forecasting. Snow surface temperature is one of the major parameters, which is responsible for the evolution of snow pack characteristics. In the present paper, the snow surface temperature was estimated using TERRA satellite based — Moderate resolution imaging spectroradiometer (MODIS) sensor for NW-Himalayas. Ground data observed by automatic weather stations (AWS) was used to calibrate the brightness temperature obtained by MODIS thermal bands data into the actual snow surface temperature data through regression analysis. A split window technique has been implemented for the estimation of snow surface temperature. The multi-date satellite derived snow surface temperature was validated with ground data of winter 2004–05 and 2005–06 collected at various observation stations located in different ranges of NW-Himalaya. Good correlations were observed for Upper Himalaya (0.98, 0.98), Middle Himalaya (0.92, 0.96) and Lower Himalaya (0.88, 0.82) for 2004–05 and 2005–06 winter respectively. Further, estimated snow surface temperature was also verified with snow-cover information collected by manned observatories and area delineated by thematic maps of snow surface temperature was validated with the different snow climatic zones of NW-Himalaya.     

小江断裂带地壳运动特征及地震危险性研究

利用1999~2013年间川滇地区的GPS速度场观测数据和块体模型研究小江断裂带地壳运动特征及地震危险性。结果表明:小江断裂带闭锁程度较高,应变量积累较大,在地表以下6km的闭锁程度约为0.94,滑动亏损速率约为(9.2±0.4)mm/a,整条断裂带左旋走滑速率约为(10.8~12.1)mm/a。结合小江断裂带上的历史地震数据估算,小江断裂带未来地震最大震级为7.3级,与近年来断裂带活动较为活跃相吻合。