Paris       Barcelona       Las Vegas     Munich       Delft       MERIS instrument

The objective of the Delft corner reflector experiment was to simulate a set of stable scattererers whose phase history can be validated by additional measurements.

Comparison of the results for the ERS2 and the tandem Envisat acquisitions should give a better insight in the comparability of the two sensors. For this reason accurate monitoring of the corner reflector height is necessary, with a temporal frequency at least identical to the radar acquisition frequency. A period of over one year should be analyzed to gather enough statistics to draw conclusions on the quality of the phase time histories. As an additional objective, phase differences with adjacent ‘natural’ stable scattererers (buildings, infrastructure) should be computed to address the problem of the selection of potential stable scattererers with only a limited amount of Envisat data available.

................................

The Delft university campus area is chosen because of logistic reasons: the reflectors need to be leveled every 35 days, they need to be checked before every satellite acquisition, and they should be relatively secured from interference or tampering by people. Figure 20 shows the location of the reflectors.



Figure 20: Corner reflector #3 (28-Jan-2004)



Figure 20b: Location of the corner reflectors

................................

In order to obtain an estimate of the corner reflectors peak phase at a sub-pixel level the following procedure is followed. First, the area of interest is "cropped" from the interferogram, an area of approximately 256x256 complex pixels. Second, the crop is harmonically interpolated by the factor of 16. Third, an algorithm for the automatic phase extraction of the reflector, based on the Canny edge detection algorithm (Canny, 1986), is applied to the interpolated crop. Finally, the corner reflector phase is calculated using complex harmonic interpolation.

Since March 2003, 12 leveling campaigns of the five corner reflectors have been performed. Their heights have been determined relative to a well-founded benchmark located in a concrete highway bridge. The leveling network has been set up introducing redundant measurements, which makes it possible to detect outliers and give a quality description of the estimated heights.

One of the alternatives to evaluate the interferometric observations is to look at the time evolution of a single reflector, relative to another. This approach is comparable with the evolution of stable scattererers as a function of time. The following figures show these analyses. The four plots show the ‘deformation’ history of reflector 1, 3, 4, and 5, relative to reflector 2. The bold solid black line connects the leveling values, the dashed red line the ERS2 values, and the dash-dot blue line the Envisat values. The time span covered is more than one year.



Figure 21: Time series of reflector 1 relative to 2. With a-priori standard deviations of 7 mm for both ERS2 and Envisat, both sensors ‘catch’ the seasonal amplitude of 20 mm reasonably well.



Figure 22: Time series of reflector 3 relative to 2. For this double difference, the Envisat observations are much closer to the leveling results compared to the ERS2 ones. This is reflected in the a priori standard deviations of 3 and 7 mm, respectively. The seasonal signal obtained from the ERS2 results seems to be overestimated by about 50%.



Figure 23: Time series of reflector 4 relative to 2. It is interesting that the main anomaly, at +70 days is apparent both in the ERS2 and in the Envisat series. Note that both radar values are actually close together, due to the ambiguity of 30 mm. One possibility for such a systematic difference with the leveling might be a thermal effect of the reflector. The slave date is 19 November 2003, which was not an extremely warm or cold day. Nevertheless, a seasonal signal is visible in both time series.



Figure 24: Time series of reflector 5 relative to 2. Both ERS2 and Envisat show a very nice match with the deformation as observed by leveling, although there seems to be a systematic effect here. The seasonal trend is captured very well though.