جستجوی مقالات مرتبط با کلیدواژه « GPS Time Series » در نشریات گروه « عمران »

Due to the wide application of GPS time series such as the study of tectonic movements, land crust change and earthquake dynamics, etc., it is important to provide a method to increase the speed in the analysis of variance components. Researchers have shown that nearly 90% of GPS time series have a combination of white+flickr noise and a lower percentage have a combination of white+random-walk noise. Researchers have used a trigonometric relationship inside the maximum likelihood estimation (MLE) method, to reduce the dimensions; instead of calculating 3 unknowns (white, flicker and random-walk noise components) they calculated 2 unknowns, a phase such as phi introducing relationship between white noise and colored noise and a phase such as theta introducing relationship between two colored noise (flicker and random-walk noises). On the other hand, researchers have been able to estimate white noise by empirical mode decomposition (EMD). In this research first we try to estimate the white noise using EMD method and then with the mentioned trigonometric relation, we can estimate the flicker or random-walk noise. It is expected that if this project is successful, the noise components will be estimated immediately. Finally, the results of the proposed method are compared to the results of least square variance estimation (LS-VCE) method. In another new method, since white noise can be extracted and separated by EMD, first white noise is extracted and then GPS time series components including linear trend, periodic movements with annual and semi-annual frequencies are extracted by least squares. The type of colored noise (flicker or random noise) can be determined by the Hurst parameter and assuming that the residual is flicker or random-walk noise, the statistical information of colored noise can be estimated (standard deviation method). The methods were first tested on simulated series and after its success, real GPS time series were used for verification. In this study, extracting white noise by EMD has been fundamentally reviewed; including how to decompose and how to detect white noise through intrinsic mode functions (IMFs) by Hurst parameter. Among Hurst parameter estimation methods, 12 methods were evaluated and boxed-periodogram method had better results. The results of the new methods are very efficient for simulated and GPS time series with white+flicker noise, but challenging for series with white+random-walk noise.

More than two decades, continuous observations of permanent GPS stations used extensively for behavior of geodynamic phenomena have found such as deformation of the earth's crust, moving tectonic plates and fault. Daily position of permanent GPS stations are considered independent of each other in statistic. In other hand, some errors such as the satellite orbit modeling errors, determining parameters of rotation's earth, parameters of atmospheric modeling and etc; cause color noise or the correlation between daily positions of stations. Another important systematic error in GPS time series is offset. Offset by factors such as earthquakes, replace the GPS antenna, human and environmental errors are generated. Offset in GPS time series's functional model causes the bias estimation of unknown parameters Therefor, we require exact statistical model and function model of GPS time series to exact estimation of the speed parameter. For this purpose, multivariate noise analysis on 38 permanent GPS stations of Iran, were carried out by the period of 7 years (2006–2012), in this study. These stations are scattered throughout the country. In this analysis, statistical model of data presented of by incorporating white noise, flicker noise and random Walk noise and noise components estimation has taken place by method of " least squares variance component estimation ". Also the effect of offset in the data were examined on estimation of noise and station's speed parameters by method of "least squares". The temporal correlation with multivariate noise analysis was performed for both single-station and multi- station. The results showed that noise amounts in coordinate series distributed due to processing all stations with each other in multi-station mode. In multivariate analysis single-station, Random Walk noise amounts after offset removal get zero in some stations, but in multi- station mode any stations were not zero for Random Walk noise. So single- station mode is more realistic than multi- station mode. Noises estimation from data with offset and compare them with results from data without offset shows offset effect on the amounts of noise, especially Random Walk noise that the greater part of it consist of offset. In addition to consideration of offset effect on noise, its effect on the speed parameters were also assessed. This assessment shows changes of length and direction of speed vector, after removing offset that examine necessity of offset's studying at data . So having the correct functional model and statistical model and without offset in the exact determination of parameters such as speed is essential. The studying of spatial correlation showed that there are significant correlation for components location North - North, East - East and height - height after removing offset, but correlation results are not significant between the various components of the coordinate, North - East, North - height and East - height. After removing offset of data, difference of speed vector from its average value, was calculated. It's found, speed of stations in tectonic boundaries (Saudi Arabia in the south and Eurasia tectonic plates in the north of Iran) more than center of Iran.

GPS time series consists of a linear trend, harmonic signals, probable offsets and also noise which is described as a stochastic part. Because of various applications of GPS time series such as plate tectonics, crustal deformation and earthquake dynamics studies, these time series should be modeled with high accuracy. For this purpose, systematic effects in functional model should be determined with high accuracy. In this paper the effect of earthquakes is also considered in the functional model in addition to mentioned behaviors. Because earthquakes cause crustal deformations, their effects can be observed in the shape of offsets (as coseismic effects) and (or) rate changes (as coseismic or postseismic effects) in the time series. Neglecting these effects lead to biased estimation of noise amplitudes. To discover the effect of earthquakes, a manual solution is used for each station. Effects are detected graphically by comparison of behavior of time series and epoch of occurred earthquakes in the region. The earthquakes which considering their effects, lead to the best fitting of functional model to time series, are selected as effective ones. Because the Alborz range is the most seismically active region in the Northern Iran, 25 permanent GPS stations with the time span between 2005 and 2013 in this area are selected for this study. Analysis of time series indicates similar behavior of time series with the same offset times and common earthquake effects for most stations (also for those which are located in far distances from epicenters). This result means that systematic effects may propagate from one station to the others during the processing and the network adjustment. Furthermore, noise analysis of time series using least squares (co)variance components estimation method, shows that neglecting seismic effects can result in the presence of random walk noise in 88%,12% and 60% of north, east and up components, respectively. However, considering the seismic effects causes positive estimation of variances of random walk noise in 12%,12% and 36% of north, east and up components, respectively. Finally, due to similar behavior of time series, a reprocessing of them could be suggested.