6 CONTENTS
Chapter 1IntroductionSyn<strong>the</strong>tic aperture radar (SAR) <strong>in</strong>terferometry (<strong>InSAR</strong>) processes a pair <strong>of</strong> satellite SAR images. Theresult may be a digital elevation model (DEM) or a map <strong>of</strong> Earth-crust deformations <strong>in</strong> <strong>the</strong> processedarea.The north-Bohemian coal bas<strong>in</strong> is a largely unstable area. In addition to many huge open m<strong>in</strong>es, itconta<strong>in</strong>s also deep m<strong>in</strong>es, and some <strong>of</strong> <strong>the</strong>m are very old and abandoned and may possess a potentialdanger for <strong>the</strong> people liv<strong>in</strong>g <strong>in</strong> <strong>the</strong> area.However, as a result <strong>of</strong> <strong>the</strong> previous ESA project, it is not reliable to estimate a deformation (or even itsvelocity) from one pair <strong>of</strong> SAR images, especially if <strong>the</strong> time between <strong>the</strong>ir acquisitions is ra<strong>the</strong>r short.On <strong>the</strong> o<strong>the</strong>r hand, if <strong>the</strong> time is long, <strong>the</strong> <strong>in</strong>terferogram created from <strong>the</strong> images gets decorrelated dueto many effects (most <strong>of</strong>ten <strong>the</strong> vegetation change) and <strong>the</strong> <strong>in</strong>formation gets lost.The purpose <strong>of</strong> <strong>the</strong> ESA project nr. 3423 is to estimate <strong>the</strong> velocity <strong>of</strong> <strong>the</strong> subsidences <strong>in</strong> <strong>the</strong> area <strong>of</strong>north Bohemia us<strong>in</strong>g a stack <strong>of</strong> SAR images from different seasons with<strong>in</strong> <strong>the</strong> 1996-2004 period.Due to <strong>the</strong> fact that <strong>the</strong> process<strong>in</strong>g is time and memory requir<strong>in</strong>g, we do not process <strong>the</strong> whole scene,only specific areas (usually towns and cities). These areas were selected as coherent <strong>in</strong> an <strong>in</strong>terferogram<strong>of</strong> about 13 months long temporal basel<strong>in</strong>e. In some o<strong>the</strong>r <strong>in</strong>terferograms, <strong>the</strong>se areas are not coherent– <strong>the</strong>se <strong>in</strong>terferograms were excluded from <strong>the</strong> postprocess<strong>in</strong>g. In addition, <strong>the</strong> urban areas are <strong>the</strong> mostimportant to be monitored for subsidences.The stack method <strong>in</strong>volves process<strong>in</strong>g <strong>of</strong> several <strong>in</strong>terferograms with a common master, with respect towhich <strong>the</strong> deformations are related. In order to achieve a larger number <strong>of</strong> <strong>in</strong>terferograms, some <strong>of</strong> <strong>the</strong>slave scenes are resampled <strong>in</strong> order to correspond exactly (i.e. with subpixel accuracy) to <strong>the</strong> master.Then, any <strong>of</strong> <strong>the</strong>se resampled scenes can be used as a master for creat<strong>in</strong>g o<strong>the</strong>r <strong>in</strong>terferograms, although<strong>the</strong> basel<strong>in</strong>e parameters etc. do not change by resampl<strong>in</strong>g.The larger number <strong>of</strong> <strong>in</strong>terferograms is not only usable <strong>in</strong> <strong>the</strong> case when some <strong>in</strong>terferograms must beexcluded due to a bad coherence (<strong>the</strong> bad coherence may be also crop-depend<strong>in</strong>g, i.e. different for differentcrops <strong>of</strong> <strong>the</strong> scene), but also for reduc<strong>in</strong>g errorneous <strong>in</strong>fluences <strong>in</strong> <strong>the</strong> deformation adjustment and forambiguity resolution, which is an important part <strong>of</strong> <strong>the</strong> postprocess<strong>in</strong>g.In <strong>the</strong> <strong>InSAR</strong> method, all measurements are relative. In deformation mapp<strong>in</strong>g, a reference po<strong>in</strong>t (or area)must be said to be stable. The deformations can only be assessed with regard to a reference po<strong>in</strong>t. Wedo not know <strong>the</strong> areas <strong>in</strong> detail, and <strong>the</strong>refore are not able to determ<strong>in</strong>e <strong>the</strong> stable po<strong>in</strong>t; on <strong>the</strong> o<strong>the</strong>rhand, it may be decorrelated <strong>in</strong> some <strong>of</strong> <strong>the</strong> <strong>in</strong>terferograms. That is why we select <strong>the</strong> stable po<strong>in</strong>t as<strong>the</strong> most coherent po<strong>in</strong>t <strong>in</strong> all <strong>in</strong>terferograms. In this case, if an unstable po<strong>in</strong>t is selected, it looks stablebut <strong>the</strong> surround<strong>in</strong>g, which really is stable, looks unstable.However, for assess<strong>in</strong>g <strong>the</strong> quality and reliability <strong>of</strong> <strong>the</strong> results, <strong>the</strong> knowledge <strong>of</strong> <strong>the</strong> area and a stablepo<strong>in</strong>t is neccessary. A priori, we consider all po<strong>in</strong>ts to be stable, and only those po<strong>in</strong>ts, where <strong>the</strong>deformation standard deviation is much larger than <strong>the</strong> deformation itself, we may consider unstable. Infuture, we plan to perform geocod<strong>in</strong>g <strong>of</strong> <strong>the</strong> <strong>in</strong>terferograms and contact a person who knows <strong>the</strong> area <strong>in</strong>detail.7