Mapping Roughness Length in Urban and Suburban Areas using ERS Interferometric SAR Data
 

A contribution to subproject SATURN

Gerhard Smiatek and Wu Xiaoling

Institute for Atmospherical Environmental Research Kreuzeckbahnstr. 19, 82467 Garmisch-Partenkirchen


 


  1. Summary
For an area near Bonn, Germany first tests with synthetic ERS interferometry data have been performed. They show a high correlation between the standard deviation of the relative phase and the roughness for all major land use categories with exception of waters.

2. Aim of the Research

Transport velocity and deposition rates of atmospheric pollutants, such as SOx, NHx, NOy, vary with the structure of the surface. This structure, called roughness length, is required by environmental pollutant transport and deposition models. Until now land use data from optical remote sensing systems were used to map roughness length. The disadvantage of this method is the uncertainty in the image classification, and the high cost of data processing. The data source for roughness length can be the interferometry with Synthetic Aperture Radar (INSAR). The principal aim of the research is the development of a mapping procedure for roughness length from synthetic ERS interferometry data.
 
 

3. Activities during the year

During the year two tests with synthetic interferometry data were performed using ERS1 and ERS2 image pairs covering an area near Bonn, Germany. The results have been compared with the roughness length data derived from land use data.

4. Principal results

The Synthetic Aperture Radar (SAR) maps the three-dimensional earth surface onto the two-dimensional image according to their slant range to the SAR. Each cell of that image is represented by a pixel. Its brightness is proportional to the power of the SAR's echo. The INterferometry with Synthetic Aperture Radar (INSAR) provides information about the third dimension. The height of each resolution cell is calculated by signal correlation from data of two SAR images representing the same area. These images are acquired from different positions A and B either in real or synthetic mode. The distance between A and B is called baseline. In the real mode two antennae on the same platform a used simultaneously. In the synthetic mode the second image is generated by the same SAR-antenna during the second overpass of the area. With the Earth Resources Satellites (ERS1 and ERS2) data only the synthetic interferometry is possible.

The interferometry technique is described in detail by Hartl and Thiel(1993). Here it is sufficient to state that in the interferometry not the ranges but only the phase differences in range are measured. The phase difference between the two images contains the information on the ground elevation. The system parameters which affect the accuracy of the topographic mapping using INSAR techniques are errors in determination of the baseline, thermal noise, surface change and decorrelation caused by the baseline length. Because of the fact that SAR's echo is registered at two different positions (In synthetic mode the base line length is in order of 30 - 600 meters) the structure of the surface is the driving power of the decorrelation. In the result the phase difference caused by the structure of the objects is interpreted as difference in elevation. But this serious problem of the elevation mapping with radar interferometry is the advantage of the roughness length mapping. However, if there is a change in the objects characteristics during the time of the acquisition of the two images, there will be additional decorrelation.

For an area of approximately 5299 ha near Bonn, Germany, two ERS-1 SAR interferometric image pairs have been processed. The images of the first image pair were acquired at 26.03.92 and 29.03.92 and the images of the second pair at overpasses of the area at 14.04.92 and 17.03.92. The baseline length is 107 meters in the first case and 405 in the second.

In principle the roughness length information can be derived from both the phase difference image and the interferometric Digital Elevation Model (DEM). The proposed measure is the standard deviation of the phase difference or the elevation model calculated within a image window of certain number of pixels (i.g. 3 x 3, 5 x 5, 7 x 7 etc.). Thus, the result is only a index which has to be transformed into the roughness length.

Figure 1 shows the roughness length derived from the CORINE land use data. Here, to each major land use category roughness length values have been assigned as shown in Table 1.
 
 
 
 

Table1 Roughness length (z0)values for major land use categories (Deursen et al., 1993)
 
 
 
Land Use Category
Roughness Length in cm
Open water large
1
Open water small
8
Agriculture
15
Low vegetation area
15
Urban
100
Deciduous forest (closed canopy)
94
Coniferous forest (closed canopy)
134

 

Figure 2 shows the standard deviation of the relative phase in a window of 5 by 5 pixels. Highest values occur within forest and urban areas. Also more details can be seen related to scattered trees and scrub. However, there is one exception. Waters show a very high standard deviation values. The reason is the decorrelation caused by change of the objects.

5. Main conclusions

The interferometric SAR-Data is a very promising source of information for mapping roughness length required by environmental models. The first results show major advantages over the method which employs land use data. The proposed method is independent of the land use data. In addition, the seasonal variation of the vegetation can easily be taken into account by images pairs acquired in different months. Also other parameters, such as forest edges, can be extracted from the image. Further research is needed on the processing of the interferometry data for the roughness length mapping. The problems are the time
 
 


 



 

Fig. 1 Roughness Length derived from the CORINE land use data



Fig. 2 Standard deviation of the relative phase (white - low, dark - high)



dependence in the synthetic interferometry and the influence of the base line length and the window size, for which the standard deviation is calculated, on the roughness length estimates.

There is also research needed on the transfer procedure from the standard deviation image to the roughness length.

6. Aim for coming year

The main aim for the coming year is the acquisition of real-time interferometry data from air borne SAR systems. These data will be used to study the influence of the decorrelation in synthetic interferometry.

7. References

Deursen, W. van , G. W. Heil, and A, Boxtel, van; Using remote sensing data to compile roughness length maps for atmospheric deposition models. International Symposium "Operationalisation of Remote Sensing" 19-23 1993, ITC Enschede, The Netherlands ,(1993).

Hartl, P. and K.-H. Thiel; Bestimmung von topographischen Feinstrukturen mit interferometrischen ERS-1-SAR. ZPF, Zeitschrift für Photogrammetrie und Fernerkundung 61(3) (1993) 108-114.