Principles of Modern Radar - Volume 2 1891121537

378 CHAPTER 8 Interferometric SAR and Coherent Exploitationas 50 m to 1100 m. In this case, γ spatial is as low as 0.757. Additional analysis and experimentaldecorrelation data are available in [54].Another benefit of a large critical baseline is that it permits a larger actual baselinewithout suffering significant baseline decorrelation. As was seen in Section 8.3.5, a largebaseline improves the height measurement sensitivity.8.6.1.3 Temporal DecorrelationTemporal decorrelation occurs when there are physical changes in the scattering surfacebetween the two SAR data collections used to form the pair of images. While the timelapse between SAR images is a nonissue in one-pass operation, it can be significant inrepeat-pass operation. Pairs of orbital tracks that provide a suitable baseline over a desiredportion of the earth’s surface can occur days or weeks apart, giving ample time for thesurface to change due to weather, man-made disturbances such as construction or vehiclemovement, natural disturbances such as subsidence, glacial movement, or earthquakes,and similar effects. Decorrelation time scales as observed from spaceborne systems aretypically on the order of several days [16].Bamler and Hartl [36] give a model for γ temporal and report that water surfaces decorrelatein milliseconds so that γ temporal ≈ 0 over water in any repeat-pass InSAR system,while forests tend to have γ temporal in the vicinity of 0.2 at C band due to leaf and branchmovement between passes. While temporal decorrelation can be used to positive effect forterrain motion mapping and change detection (see Section 8.8) as well as classification ofscatterer types, it is an undesirable effect for elevation profile mapping.One technique for combating temporal decorrelation is reliance on persistent scatterers,which are scatterers in a scene that are exceptionally stable over very long periods oftime. Persistent scatterers are typically found in urban areas and rocky terrain. Once identified,InSAR processing can be applied to develop reliable elevation estimates at thosepoints or to perform terrain motion mapping (see Section 8.8.1). Elevation accuracies ofones of meters using images spanning many years have been reported [55].Another approach avoids temporal decorrelation altogether by using two collectionradars flying in close formation, effectively creating the equivalent of one-pass operation.The obvious cost of this approach is the need for two radar satellites. Figure 8-24 illustratesone configuration, the TanDEM-X mission based on the TerraSAR-X radar system [56].This system uses a pair of orbits that move in a helical relationship to one another, resultingin a InSAR baseline that varies in length and orientation (value of B and β) at differentpoints in the orbit, as shown in Figure 8-24b. Adjustment of the orbit spacing using onboardpropulsion units allows fine tuning of the baseline length over a range of a few hundredmeters to several kilometers. A dual-frequency GPS is used for fine orbit determination andPRF synchronization, and an X band intersatellite data link provides phase synchronizationof the two transmitters.8.6.1.4 Atmospheric and Weather EffectsAtmospheric conditions can affect estimation of the IPD in several ways. Changes inwater vapor content and distribution in the troposphere can lead to variations in signalpropagation delays through the atmosphere and thus in radar phase measurements. Sincerepeat-pass systems may combine measurements taken days apart or longer, the delaysmay change between acquisitions and the resulting delay differences will introduce errorsinto the IPD [57]. As discussed in Section 8.4.1, this effect can produce significant DEM