【作者】 盛传贞；

【导师】 甘卫军；

【作者基本信息】 中国地震局地质研究所 ， 固体地球物理学， 2013， 博士

【摘要】 自上世纪90年代初以来，GPS空间大地测量学的迅速发展和全面应用，为全球范围各种规模尺度的地壳运动和构造形变观测提供了革命性的手段，使高精度、大范围、全天候、低成本的大地测量变成了现实。为及时把握GPS空间对地观测技术为防震减灾应用带来的机遇，我国先后于1997－2000年和2007－2012年实施了国家重大科学工程“中国地壳运动观测网络”和“中国大陆构造环境监测网络”，在中国大陆及周边建立了由260个连续GPS观测站和2056个非连续GPS观测站构成的高精度、高密度观测网络，为精细而定量地研究中国大陆不同构造区域的现今地壳运动方式和构造形变演化态势提供了至关重要的基础平台。近年来，随着GPS星座系统、观测设备、处理软件和全球服务的进一步完善和发展，在正常观测环境和规范测量方式下，基于约24小时的连续静态观测和标准化的GPS数据处理策略，已经实现以“数毫米”量级的精度（水平向优于3mm、垂直向优于6mm）获得站点在全球参考框架下的单日平均坐标。因此，通过对“中国地壳运动观测网络”等工程的各GPS站点进行长年累月的连续观测或间歇性的非连续观测，可获得其高精度坐标位置随时间变化的序列过程。据此，我们不仅能够估计各站点的平均运动速度，而且能够解析其运动变化过程，并直接获取到一系列量值较为显着的非线性变化现象，如同震位移、震后驰豫形变、活动断裂的慢地震滑移等等。但是，当我们关注和地震危险性相关的构造运动或地壳形变时，GPS站点坐标变化时间序列中所包含的一些系统性误差，尤其是一些非构造形变的干扰将不容忽视，因为许多非构造干扰（如时变的大气潮汐、区域海洋潮汐、陆地水负荷等等）所产生的地壳形变或站点位移，往往与构造形变处于同一量级。因此，如何从GPS的坐标变化时间序列中有效地分离和剔除各种非构造形变的影响，使其更好地服务于地震危险性监测与分析，一直是GPS大地测量领域兼具重要科学意义和迫切实际需求的研究课题。本论文以非构造负荷形变对中国大陆GPS坐标时间序列的影响为主题，以中国大陆及周边高精度连续GPS观测资料为基础，对不同区域时变大气潮汐、区域海洋潮汐和陆地水负荷所引起的非构造形变特征开展定量研究，并将改正模型应用于中国大陆连续GPS观测站和非连续GPS观测站的坐标变化时间序列中，获得了更加接近真实构造形变的结果，使其更加有效地反映构造形变。具体的研究和探讨主要有以下几个方面：（I）归纳总结国内外高精度GPS数据处理优化方法和先进策略，采用美国JPL的GPS前处理软件GIPSY和后处理平差软件QOCA，对中国大陆各区域典型连续GPS观测站数据进行了严密处理，获得了高精度坐标变化时间序列，为进一步分离和研究不同区域非构造形变特征提供了基础数据GPS坐标变化时间序列中既包含着一系列随机误差、模型偏差、参数误差，也包含地表负荷（如大气潮汐、区域海洋潮汐及陆地水负荷）引起的非构造形变信息。为了从高精度GPS时间序列中有效获取中国大陆不同区域的非构造负荷形变，我们选取了具有区域代表性的连续GPS观测站，并在严密的数据处理中采用了GIPSY软件的精密单点定位（PPP）策略和基于固定点法则的整网模糊度解算方法（Ambizap），同时，选用了国际上最新的先验对流层延迟模型ECMWF（European Centre for Medium-RangeWeather Forecasts）和对流层投影函数VMF1(Vienna Mapping Function1)、并纳入高阶电离层影响改正，对GPS数据处理的关键模型进行了优化。最后，通过联合平差与地壳形变分析软件QOCA，扣除了GPS时间序列中的构造形变（构造运动速率、地震影响等）及站点异常影响，突出了中国大陆各区域非构造形变。（II）分析研究了中国大陆及周边大气潮汐负荷的非构造形变影响,并针对目前国际上通用大气压数据中包含有许多虚假“潮汐分量”信号的实际情况,设计编制了一套20阶的Butterworth低通滤波器,有效地剔除了这种虚假“潮汐分量”的影响,并较好地应用于中国大陆各区域的GPS时间序列中大气受到日月引力潮作用和局部热力作用产生大气潮汐和质量的重新分布，由此可引起地壳的非构造负荷形变。基于GPS的高精度观测，可探测和分辨大气负荷形变的存在。大气负荷包含潮汐负荷和非潮汐负荷，总体的大气负荷形变在地球表面的最大量值可达厘米，主要出现在高纬度区域。我们的研究表明，对于中国大陆区域，大气潮汐现象所引起的垂向形变通常为亚毫米级，而北向和东向分量仅为垂向分量的1/10左右；其中，半日潮汐(S1)负荷形变振幅具有明显的纬度相关性，即同一纬度负荷形变的振幅基本一致，但相位随经度变化而不同。大气非潮汐负荷形变的计算主要利用实测大气压力资料，并通过格林函数积分获得。然而，考虑到目前国际上的大气压变化数据主要来源于ECMWF和NCEP等气象中心提供的重分析产品，由于模型缺陷、数据噪声和采样率等引起的混频效应，使这些大气压数据中包含有许多虚假的“潮汐分量”信号。理论计算表明，这种虚假的“潮汐分量”会对GPS时间序列产生垂向0.5-1mm、水平向0.1-0.2mm的偏差，使GPS时间序列产生周期为5-6天、半年和整年的多种波动，在GPS时间序列中不容忽视。为此，本文设计了一套阻带频率为1周/每天、阻带衰减为35dB的20阶的Butterworth低通滤波器，有效地剔除了这种虚假“潮汐分量”的影响，并获取了更加可靠的大气非潮汐分量。（III）通过对比多种全球海洋潮汐模型在中国海域的差异性，提出采用高分辨率区域海洋潮汐模型进行潮汐负荷形变的改正，可有效避免GPS时间序列中长周期的非构造形变虚假信号。并通过对目前卫星测高资料的分析，说明当下海洋非潮汐改正的实用效果尚不可靠，需进一步提高海洋非潮汐模型精度中国大陆东濒太平洋、南临印度洋，海岸线蜿蜒悠长，海水的潮汐变化和非潮汐变化均会引起沿海及内陆一定范围的非构造形变。由于几种全球海洋潮汐模型在中国海域的分辨率不同，我们发现纳入高分辨率的区域海洋潮汐模型，可对沿海GPS站垂向M2潮汐波产生平均量值达1.1mm的改善，最大的改正值可达5mm；若不考虑高分辨率的区域海洋潮汐模型，则会在GPS时间序列中产生“虚假”的长周期信号；而对于海洋非潮汐负荷形变，我们基于目前高精度卫星测高资料对沿海区域GPS观测站时间序列进行海洋非潮汐负荷形变改正后，仅少数站点的时间序列有所改善，究其原因在于目前的卫星测高资料仍不能完全满足高精度海洋非潮汐负荷形变计算和改正的要求。考虑到海洋非潮汐形变的最大振幅小于1mm，因此，我们建议在GPS数据处理和时间序列分析中，可暂不考虑海洋非潮汐负荷形变的影响。（IV）结合GRACE时变重力场和全球陆地水资料NCEP的各自优势，尝试研究了一种数据同化方法，获得了中国大陆兼顾时－空分辨率的陆地水负荷时变资料，并据此计算了不同区域的陆地水迁徙负荷形变。在此基础上，分析评估了该方法对GPS时间序列的改正的效果陆地水负荷所产生的地壳形变，在所有非构造形变中量值最为显着。中国大陆不同区域的陆地水负荷形变差异，可达厘米量级。GRACE时变重力场和全球陆地水资料（NCEP）为目前计算陆地水负荷形变最重要的基础资料，两者均能反映陆地水的短期变化。其中，GRACE时变重力场不仅反映土壤湿度影响，而且还反映地表水和地下水影响，因而具有较高的精度，但存在时间分辨率较低的问题。全球陆地水资料（NCEP）提供0-2m深度范围的土壤湿度和积雪负荷，具有较高的时间分辨率，但不包含地表和地下水资料。针对两种资料各自的长短优劣，我们尝试研究了一种数据同化方法：以GRACE资料为主，以NCEP作为陆地水短时变化的有效补充，使陆地水负荷资料兼具GRACE在长时间尺度上的高精度和NCEP在短时间尺度上高分辨。基于同化结果，我们就中国大陆不同区域进行了验证分析，并利用小波分析方法讨论了陆地水迁徙负荷形变与GPS时间序列在低频分量的相关性及改善程度。结果表明，当陆地水迁徙负荷形变量大于GPS观测噪声时，改正效果明显。而对于信噪比较小的区域，改正效果较差甚至负面。（V）以连续GPS观测资料和GRACE同化资料为基础，采用支持向量回归方法(SVR)联合反演区域非构造负荷形变模型，并针对中国大陆地表水变化显着性差异较大的两个典型区域，验证分析了该方法对连续GPS和流动GPS观测站时间序列的改善效果高精度GPS可直接探测到地表的非构造负荷形变，但往往包含有站点局域干扰和环境模型误差等方面的影响；而基于地球物理模型亦可获得各种地表负荷所引起的“理论形变”，但地球物理模型参数的精度缺陷和输入资料的时空分辨率可能导致理论计算结果的偏差。为此，我们尝试以连续GPS观测资料作为基本观测量，以GRACE同化资料为约束，采用SRV (Support Vector Regressing)方法联合反演区域非构造负荷形变。在滇西地区和陇中黄土高原区域实际应用和效果验证表明：联合反演方法兼顾了GPS和GRACE资料的特点，并纳入了物理意义明确的模型约束，可避免对连续GPS的单一依赖，且联合反演结果能明显改善非连续GPS观测的时间序列。更多还原

【Abstract】 Since the early1990s the space geodetic technology GPS has been developed rapidlyand widely applied to various fields. As a revolutionary tool for detecting crustalmovement and tectonic deformation on various scales all over the world, the GPS is veryconvenient for surveying with high accuracy, large scope, all-weather and low cost. Totimely grasp the opportunity of application of GPS to monitoring the earth for earthquakeprevention and disaster reduction, the national major scientific project “ObservationalNetwork for Crustal Movement of China” and “Monitoring Network for ContinentalTectonic Environment of China” have been implemented in our country from1997to2000and2007to2012, respectively. The observation network, which is of high-accuracy andhigh-density, consisting of260continuous GPS stations and2056non-continuous GPSstations built around China mainland, has become an important infrastructure platform fordetecting the present crustal movement with high accuracy and analyzing the evolutionprocess of tectonic deformation quantitatively.In recent years, with further optimization and development of the GPS constellationsystem, observation devices, data processing software and IGS (International GNSSService), given the data with continuous24hours under well observation circumstance andstandard measurement mode, it is possible to achieve daily coordinates of the station inglobal reference framework with an accuracy of several millimeters by using the routineGPS data processing（the accuracy can be better than3mm and6mm in horizontal andvertical direction respectively）. By using years of continuous observation data ornon-continuous observation data of different GPS stations from CMONOC and otherprojects, one can get the GPS time series with high-accuracy. Based on these time series,one can not only estimate the velocity of each station, but can also analyze obviousnon-linear variations, such as co-seismic displacement, post-earthquake relaxationdeformation, and slow slip earthquakes of active faults. However, when the study focuseson the tectonic movement or crustal deformation related with the possibility of occurrenceof earthquakes, some systematic errors in GPS site coordinate time series cannot beignored, especially some non-tectonic deformation, because crustal deformation or sitedisplacement caused by non-tectonic load, such as time-varying atmospheric tide, regionalocean tide, and land water migration, always has the same magnitude as tectonicdeformation. Therefore, it is an important issue how to effectively separate and eliminatevarious non-tectonic deformation from GPS time series, so as to better serve detections andanalysis of earthquake risks. In this thesis, the effect of non-tectonic load deformation on GPS time series aroundChina is discussed firstly. Then the characteristics of deformation induced by time-varyingatmospheric tide, regional ocean tide, and land water load are analyzed quantitativelybased on high-accuracy continuous GPS observation data. At last correction tonon-tectonic load deformation from GPS continuous and non-continuous time series ofChina is made by using different geophysical models. The specific contents are concernedwith following several aspects:I. By deducing and summarizing the optimized method and advanced strategy inGPS data processing, the GPS data processing software GIPSY and post-processingbalancing software QOCA from JPL are adopted to process continuous GPS data ofdifferent area around China in order to acquired the GPS time series withhigh-accuracy, which provides a foundation for further separation and analysis ofcharacteristics of non-tectonic deformation in different areas.Not only the biases from random errors, model biases and parameter errors，but alsonon-tectonic deformation caused by time-varying surface loads, such as atmospheric tide,regional ocean tide, land water load, can affect GPS time series. To effectively getnon-tectonic deformation of different areas around China from high-accuracy GPS timeseries, this work selected some typical regional continuous GPS stations, and used precisepoint positioning (PPP) of GPSY software and Ambiguity resolution for the whole network(Ambizap) based on principal of fixed point in GPS data processing. Meanwhile, in orderto optimize the key model of GPS data processing, the international advanced prioritroposphere delay model ECMWF （European Centre for Medium-Range WeatherForecasts）and troposphere projection function VMF1(Vienna Mapping Function1) areemployed, and the higher-order ionosphere model is adopted to correct the ionosphereeffect in GPS data processing. Ultimately, tectonic deformation (like the rate of tectonicmovement, seismic influence and etc) and site abnormal effect of each site are eliminatedby using QOCA (Quasi-Observation Combination Analysis). Thus, non-tectonicdeformation of each area around China is highlighted.II. Effects of atmosphere tide load on non-tectonic deformation around Chinahave been analyzed and studied. Considering partial “tide” atmosphere component inatmosphere pressure data, which are commonly used all over the world, a low passfilter with20orders is designed to eliminate this effect and well adopted in GPS timeseries around China.Ocean water redistribution, which is originated from sun/lunar gravitation tides andlocal thermodynamic processes, will lead to non-tectonic load deformation. Based on high-accuracy GPS time series, one can observe and detect the atmosphere loaddeformation. The atmospheric load contains tidal load and non-tidal load. The maximumvalues of atmospheric load deformation on the earth can reach centimeters, which mainlyappears in high-latitude areas.The study shows that the vertical deformation induced by atmosphere tide is ofsubmilimeter level around China, while the north and east component are smaller,equivalent to10percent of vertical component. Among them, amplitude of loaddeformation induced by semidiurnal tidal (S1) is related with latitude, i.e., amplitudes ofload deformations of same latitude are identical, but the phases are different with longitude.This work mainly integrates atmosphere pressure data with Green Function to computenon-tectonic atmosphere load deformation. Nevertheless, current international atmospherepressure data are mainly from re-analysis products provided by different weather centers(ECMWF and NCEP). Due to complicated signals related with model defect, noise andsampling rate, these atmosphere pressure data contain partial "tide" atmosphere in it.Theoretical calculation shows that this kind of partial "tide" atmosphere will lead to biasesin GPS time series with the magnitude of0.5-1mm and0.1-0.2mm in vertical andhorizontal direction, respectively, thus generating spurious signal in GPS time series withperiod of5to6days and seminal and annual, which can not be ignored in the GPS dataprocessing. To this end, a low pass filter Butterworth with the order of20is designed;which has a cutoff frequency of1week/day and a stop-band attenuation of35dB. Byadopting this filter, the partial "tide" atmosphere can be effectively eliminated, thusatmosphere non-tide component can be acquired with good reliability.III. By comparing the differences of several global ocean tide models aroundChina’s seas, a regional ocean tide model with high spatial resolution is adopted tocorrect tide load deformation, which is an effective way to eliminate spuriousnon-tectonic deformation signal of long-period in GPS time series. Meanwhile, byanalyzing the current satellite altimetry data, this work concludes that non-tide oceanloading deformation is not reliable enough now and accuracy of the non-tide oceanmodel needs to be enhanced further.With a long coastline, China faces the Pacific Ocean in the east and Indian Ocean inthe south. Both the tidal and non-tidal variations of oceans will lead to non-tectonicdeformation around China’s seas, especially in coastal areas. Because several global oceantide models have different spatial resolutions in Chinese seas, it is found that a regionalocean tide model with high spatial resolution is incorporated, the coastal GPS verticaldeformation induced by M2tide wave will be improved by a magnitude of1.1mm on average, and the maximal correction value can reach5mm. If this regional ocean tidemodel with high spatial resolution is not taken into consideration, then spuriouslong-period signal will be generated in GPS time series. And as for ocean non-tidal loaddeformation, only few costal GPS time series can be improved if ocean non-tidal loaddeformation from GPS time series is corrected based on currently high-accuracy satellitealtimetry. It is the reason that currently satellite altimetry data is still not accurate enough,thus elimination of ocean non-tidal load deformation in GPS time series is still on the way.In view of maximum amplitude of ocean non-tidal deformation is lower than1mm; thisthesis suggests ignoring ocean non-tidal load deformation currently when performing GPSdata processing and time series analysis.IV. Combining respective advantages of the GRACE time-varying gravity fieldand global land water data NCEP, this work attempts to design a kind of dataassimilation method and get the land water data around China. This method has theadvantage in both time and space resolution. Based on the above, the land watermigration load deformation of different regions is computed. Then, analysis andevaluation of this method in GPS time series are performed.The magnitude of crustal deformation induced by land water is very considerableamong all values of non-tectonic deformation. The magnitude of deformation differenceinduced by land water load can reach centimeters around China. The GRACE time-varyinggravity field and global land water model （NCEP）are currently most important data forstudying land water load deformation. Both of them can reflect short-term variation of landwater. Among them, the GRACE time-varying gravity field contains not only snow/soilhumidity but also surface water and ground water variation. Thus it is of higher accuracy.But GRACE has disadvantage of low time resolution. The global land water model（NCEP）can provide soil humidity and snow load in a depth range from0to2meter andis of higher time resolution, but not includes surface water and ground water data.According to the advantage and disadvantage of these two kinds of data, this work tries todesign a data assimilation method by taking NCEP as effective complementation forGRACE in short term, in this way the land water load data with advantages ofhigh-accuracy of GRACE on a long-time scale and high resolution of NCEP on short timescale can be jointly employed. Based on the assimilation result, this thesis analyzes thecharacteristics of land water deformation over China. And by adopting the wave analysismethod, it discusses the correlation between land water load deformation and GPS timeseries, and then analyzes the improvement of GPS time series corrected by the simulationresult in low frequency component. The results show that when the land water load deformation is larger than GPS noise, the correction coefficient is larger. But for the regionwith lower SNR (signal noise ratio), the correction coefficient is smaller, even negative.V. Based on continuous GPS observation data and GRACE assimilation data,this work adopts the Support Vector Regression (SVR) to jointly invert the regionalnon-tectonic load deformation model, and this method is validated by analyzing theimprovement of continuous GPS and non-continuous GPS time series correctednon-tectonic load deformation based on the assimilation results in two typical regionswith larger significant differences of surface water variations over China.High-accuracy GPS can directly detect the surface non-tectonic load deformation, butit usually also contains deformation from local effects and environment model errors andother factors. Based on the geophysical model，"theoretical deformation" induced byvarious kinds of surface loads can also be acquired, but the poor accuracy of thegeophysical model and time-space resolution of input data may lead to deformation biases.To solve this problem, this work tries to take the continuous GPS observation data as basicobservations and GRACE assimilation data as a prior constraint, and adopts SVR (SupportVector Regressing) to jointly invert the regional non-tectonic load deformation model. Theapplication and validation in western Yunnan and the Longzhong plateau area areperformed. The result shows that the joint inversion method can utilize advantages of bothGPS and GRACE data, with a prior constraint from the known physical model, which doesnot solely originate from continuous GPS. The joint inversion results can obviouslyimprove the time series of continuous and non-continuous GPS stations.更多还原