Abstract
During the last decades, timeseries interferometric synthetic aperture radar (InSAR) has emerged as a powerful technique to measure various surface deformation phenomena of the earth. Early generations of timeseries InSAR methodologies, i.e. Persistent Scatterer Interferometry (PSI), focused on point targets, which are mainly manmade features with a high density in urban areas and associated infrastructure. Later, methodologies were introduced aiming to extract information from other targets known as distributed scatterers (DS), which are associated with ground resolution cells occurring mainly in rural areas. Unfortunately, the underlying properties and assumptions behind various DSphase estimation methodologies are sometimes subjective and incomparable, which hampers the objective application of the different methods. Moreover, for some terrain types, such as agricultural terrain or pastures, the feasibility of DSmethodologies is not straightforward.
In view of these challenges, the two main objectives of this study are (i) to formulate and implement the estimation methodology of DSpixels in a standard geodetic framework and to compare it with other existing methods, and (ii) to assess the feasibility of exploiting distributed scatterers for deformation monitoring over agricultural and pasture areas.
We review stateoftheart timeseries InSAR methodologies with special attention to
processing aspects related to distributed scatterers. From an estimation theory perspective, the key processing step to extract information from DSpixels is the equivalent singlemaster (ESM) phase estimation. To situate this estimation in a geodetic framework, a mathematical model is proposed in the form of a GaussMarkov model. To evaluate the stochastic part of the model, a numerical MonteCarlo methodology as well as an analytical approach are introduced. Regarding the functional part, the ESMphase estimation is formulated in the form of a hybrid linear system of observationequations with both realvalue and integer unknowns. The solution of the proposed model is given by the integer leastsquares (ILS) estimator. The properties of such an estimator for ESMphase estimation are described and demonstrated using synthetic and real datasets. Furthermore, to provide a theoretical comparison between the proposed ILS estimator and other existing ESMphase estimators, a unified mathematical model in the form of a system of observation equations is proposed. Evaluating all the existing DSmethods shows that, although they all provide specific solutions, their fundamental difference is in how they assign weights to the interferometric observations.
The feasibility of exploiting PS, DS, and their combination over agricultural and rural
landscapes is assessed via a case study on a subsidence area near city of Veendam,
the Netherlands, based on the coherence behavior of different types of land use. It is
shown that, under the condition of using the entire timeseries, agricultural and pasture areas show only limited improvement in point density compared to the results of PSonly processing. This is due to the seasonal behavior of the temporal coherence, which causes an almost complete drop in coherence during summer periods, mainly as a result of tillage, crop growth and harvesting.
To model this periodicity, a new analytical model is introduced. In this model, the hypothetical movements of elementary scatterers within DS resolution cells are modeled as a stochastic process with nonstationary but periodic increments. The parameters of this model are estimated for pasture areas, and are subsequently used to assess the feasibility of exploiting DSpixels in agricultural areas by different satellite missions. The results confirm that, assuming a threeyear stack of data, the information content in DSpixels from current Cband and Xband missions is not enough for the successful utilization of their entire timeseries. However by using intermittent series, e.g., by processing individual coherent periods, the results indicate that DSpixels can be exploited: based on the proposed decorrelation model, the short repeat times of Sentinel1 (6 or 12 days) results in a sufficient number of coherent interferograms over each winter period, enabling DS exploitation even over agricultural and pasture areas.
In view of these challenges, the two main objectives of this study are (i) to formulate and implement the estimation methodology of DSpixels in a standard geodetic framework and to compare it with other existing methods, and (ii) to assess the feasibility of exploiting distributed scatterers for deformation monitoring over agricultural and pasture areas.
We review stateoftheart timeseries InSAR methodologies with special attention to
processing aspects related to distributed scatterers. From an estimation theory perspective, the key processing step to extract information from DSpixels is the equivalent singlemaster (ESM) phase estimation. To situate this estimation in a geodetic framework, a mathematical model is proposed in the form of a GaussMarkov model. To evaluate the stochastic part of the model, a numerical MonteCarlo methodology as well as an analytical approach are introduced. Regarding the functional part, the ESMphase estimation is formulated in the form of a hybrid linear system of observationequations with both realvalue and integer unknowns. The solution of the proposed model is given by the integer leastsquares (ILS) estimator. The properties of such an estimator for ESMphase estimation are described and demonstrated using synthetic and real datasets. Furthermore, to provide a theoretical comparison between the proposed ILS estimator and other existing ESMphase estimators, a unified mathematical model in the form of a system of observation equations is proposed. Evaluating all the existing DSmethods shows that, although they all provide specific solutions, their fundamental difference is in how they assign weights to the interferometric observations.
The feasibility of exploiting PS, DS, and their combination over agricultural and rural
landscapes is assessed via a case study on a subsidence area near city of Veendam,
the Netherlands, based on the coherence behavior of different types of land use. It is
shown that, under the condition of using the entire timeseries, agricultural and pasture areas show only limited improvement in point density compared to the results of PSonly processing. This is due to the seasonal behavior of the temporal coherence, which causes an almost complete drop in coherence during summer periods, mainly as a result of tillage, crop growth and harvesting.
To model this periodicity, a new analytical model is introduced. In this model, the hypothetical movements of elementary scatterers within DS resolution cells are modeled as a stochastic process with nonstationary but periodic increments. The parameters of this model are estimated for pasture areas, and are subsequently used to assess the feasibility of exploiting DSpixels in agricultural areas by different satellite missions. The results confirm that, assuming a threeyear stack of data, the information content in DSpixels from current Cband and Xband missions is not enough for the successful utilization of their entire timeseries. However by using intermittent series, e.g., by processing individual coherent periods, the results indicate that DSpixels can be exploited: based on the proposed decorrelation model, the short repeat times of Sentinel1 (6 or 12 days) results in a sufficient number of coherent interferograms over each winter period, enabling DS exploitation even over agricultural and pasture areas.
Original language  English 

Qualification  Doctor of Philosophy 
Awarding Institution 

Supervisors/Advisors 

Award date  31 May 2017 
Print ISBNs  9789461868060 
DOIs  
Publication status  Published  2017 
Keywords
 satellite radar interferometry
 InSAR
 surface deformation
 geodetic estimation
 distributed scatterers