Water Defense System Monitoring using SAR Interferometry Ramon F. Hanssen, Freek J. van Leijen Delft Institute of Earth Observation and Space Systems Delft University of Technology Kluyverweg 1, 2629HS, Delft, The Netherlands r.f.hanssen@tudelft.nl, f.j.vanleijen@tudelft.nl Abstract Monitoring the safety of water defense systems is crucial for life in the Netherlands. Conventional monitoring of structures such as dams and dikes is often limited to visual inspection, with additional in situ measurements if deemed necessary. Here we suggest that advanced satellite radar technology can be used to obtain weekly updates on dike stability for a significant part of all dikes in the Netherlands. This may have a significant impact on safety assessment and hazard mitigation in the Netherlands. 1 Introduction The majority of the Dutch population is living on land reclaimed from the sea, below the high water levels of the sea, large rivers and lakes. Seventy percent of the gross national product is earned in these vulnerable areas[3]. Therefore, the safety of the water defense sys- tems (WDS) is of paramount importance to sustain Dutch society. Failure can have catastrophic humanitarian and socio-economic consequences. The primary water defense systems form a protection against flooding from the sea, the main rivers, and the large lakes, for which failure would have dramatic con- sequences In autumn 2006, the inspection authority in the Netherlands concluded that 24% of these primary water defense systems does not satisfy the legally adopted stan- dards, and that for another 33% the status of the WDS is not known [6]. Monitoring the status of WDS is particularly difficult, partly because of the their large extent: the Netherlands has 17000 km, of which 4300 km are primary, of WDS. The inspection methods rely largely on expert observers, who perform yearly manual (visual) inspections, a method that has been unchanged since the centuries [4]. Conse- quently, such observations are infrequent, subjective and qualitative. Moreover, even expert observers cannot see the minute changes in the dike volume that may eventu- ally lead to failure, making their observations not precise enough. Apart from evident system failure modes such as overtop- ping during extremely high water events, structural fail- ure is of great concern. Failure of earthworks can be due to many different causes such as sliding slopes, loss of bearing capacity, hydraulic loading, or structural weaken- ing due to draining [5]. Some of these events will come without any precursory structural change. Other failure modes will be preceded by slow and minute structural or geometric changes, which can be potentially measured as displacements. It is for the latter situation that satellite In- SAR based methods have enormous potential, due to their frequent revisits, wide areal coverage, and high precision displacement monitoring. 2 Processing approach A wide class of interferometric SAR processing method- ologies can be characterized as time series SAR interfer- ometry, using many or all of the available radar acqui- sitions [2]. Perhaps the most effective subclass of these methods is referred to as persistent scatterer interferom- etry (PSI), due to its ability to work with single pixels or scatterers as a function of time [1]. PSI methods at- tempt to solve two problems simultaneously. First, they need to identify coherent scatterers, whose phase history is dominated by the geometry between satellite and scatterer, rather than physical changes within the scatterers resolu- tion cell. Second, for scatterers deemed coherent, various parameters need to be reliably estimated, such as their ge- ometric height, their displacement behavior in time, atmo- spheric delay factors, and integer phase ambiguities. The main problem in PSI is that identification and esti- mation usually need to be performed in concert, as it is not known beforehand which of the millions of observa- tions will behave coherently. Inevitably, this will result in errors. We distinguish type-I errors—coherent scatterers which are not identified as being coherent—and type-II er- rors, which are incoherent scatterers which are erroneously not rejected (false detections). In most PSI approaches, such errors are practically unavoidable, due to the wide spatial extent, the huge number of observations, and the impossibility to check every possible pair (arc) of points due to numerical constraints. Therefore, type-I errors will lead to undetected points.