INTERFEROMETRIC COHERENCE MEASUREMENTS OF TROPICAL, TEMPERATE AND BOREAL FORESTS

ABSTRACT

The relationship between forest biomass density and interferometric coherence was investigated using SAR image pairs acquired as part of the ERS tandem mission. Temperate, boreal and tropical forest test sites were imaged although so far results are available only from Corsican Pine stands at Thetford Forest in the UK.

Interferograms were generated for a variety of repeat-pass image pairs including four 1-day pairs, one 35-day pair and one 315 day pair. Corresponding coherence maps were produced and registered to a database containing information about the distribution of stands of Corsican Pine with a variety of biomass densities estimated from their age. The relationships between the estimated biomass density and both the backscattering coefficient and the coherence were derived from this database.

The 1-day repeat coherence shows a useful dependence on biomass density up to around 100 tonnes / hectare and this dependence is stronger than that between backscattering coefficient and biomass density. The coherence in the forest areas over longer timescales is uniformly low. However, it remains high in urban areas even after 315 days.

Keywords: Forest, interferometry, ERS-Tandem, biomass density

1. INTRODUCTION

ERS SAR remote sensing has the important advantages of radiometric stability, long-term support and relative independence from atmospheric conditions. However, for forest monitoring purposes, C-band data is not ideal because in general the backscattering coefficient of forests at this wavelength is poorly related to their biophysical characteristics. SAR interferometry is expected to provide improved capability for forest applications by making use of the additional information carried in the phase differences and in the coherence (g) between repeat-pass image pairs [Refs. 1 and 2]. The coherence is expected to decrease with forest density because there is more opportunity for small changes in scattering characteristics over time when more scatterers are present. Over time scattering characteristics might be affected by moisture, wind or growth. Also, the interferometric phase difference between clearings and adjacent forest canopies may allow tree height to be inferred.

This paper describes the preliminary findings of the ESA ERS Tandem project AOT.UK316 which aims to investigate the relationship between interferometric coherence and biomass density and to use the extra information derived from repeat-pass interferometry to assess forest characteristics at several established forest test sites. These include a temperate plantation at Thetford in the UK, a boreal forest test site at Siggefora in Sweden and tropical forest test sites at Tapajós and Manaus in Brazil.

Preliminary work suggests that 1-day interferometric coherence is very low over the tropical forest test sites but carries useful information at the temperate and boreal forests sites. As most data so far has been received and processed for the temperate forest site at Thetford, this paper describes the preliminary results from interferometric processing carried out at that site.

2. TEST SITE, FIELD AND IMAGE DATA

Thetford Forest is a predominantly coniferous plantation first developed in the 1930s on heath and low grade agricultural land in East Anglia. It consists mainly of Scots and Corsican Pine stands and includes a large range of ages of Corsican Pine from 3 years to 60 years.

Fieldwork had been carried out to measure the biomass density of Corsican Pine stands of a range of ages. This data was used to characterize the relationship between stand age and biomass density and hence to allow the biomass density to be estimated within a much larger set of forest stands. Plantation stocking maps provided by the UK Forestry Commission (now Forest Enterprise) were digitized to determine the location of forest stands of different ages of Corsican Pine within a geocoded database at a resolution of 10m.

Image data consists of four ERS Tandem image pairs from August 1995 and April, May and June 1996. This dataset allowed repeat-pass interferometry to be investigated over many time delays including 1, 35, 70, 244, 279 and 315 days.

3. IMAGE PROCESSING

Interferometric processing was carried out using the ESRIN/POLIMI ISAR software which proved to be effective though cumbersome.

All 8 images were co-registered to sub-pixel accuracy using the most recent image as a reference. Interferograms and optimized coherence maps were derived for the subset of the possible combinations of image pairs which were considered of most interest. The ones examined in this paper include all four possible 1-day repeat pairs, one of the possible 35-day repeat pairs and the maximum 315-day repeat pair.

The complex SAR images were detected and averaged to form 6-look amplitude products. These were calibrated (taking into account only the K-factor for this preliminary analysis) and registered to a geocoded database containing the forest stand information. The coherence maps were also registered to this same database.

The relationships between the estimated biomass density and both the backscattering coefficient (s⁰) and the coherence were derived from the database using the digitized forest stand map. In each case average values were calculated over image areas of at least 150 pixels representing homogeneous forest stands of at least 1.5 ha.

4. PRELIMINARY RESULTS

Results are shown quantitatively in Figure 1 and qualitatively in Figures 2 to 11.

Although much of the Thetford area is covered by forest, strong phase fringes were developed for each of the 1-day repeat image pairs over the whole area. In the interferograms generated from image pairs spanning greater than 1-day, fringes were visible only outside the forest area.

The fringes, before and after the removal of orbital effects, for one of the 1-day repeat image pairs are shown in Figures 2 and 3. These have not been normalised but the appearance of only one fringe over the area in Figure 3 confirms the flat nature of this region. Ongoing work will determine if the phase difference between clearings and adjacent forest canopies may allow tree height to be inferred.

Figure 1 shows that there is a strong relationship between interferometric coherence and biomass density for each of the 1-day repeat pairs but that this relationship is not significant over time scales of 35 days or more. In each of the 1-day repeat cases, this relationship is stronger than that between s⁰ and biomass density. This is partly due to the soil scattering component which becomes significant at low biomass density values in some images, presumable because of higher soil moisture.

There is some variability in the relationship between biomass density and 1-day repeat coherence and future work will investigate whether this variability is related to meteorological or atmospheric conditions. This variability is also apparent in Figures 4 and 5 which show the coherence maps for two of the 1-day repeat pairs. Figure 10 shows a colour composite of three of the 1-day repeat coherence maps showing that a large amount of extra information is available using this technique. Here the area with lowest coherence indicates the mature stands of Thetford Forest.

Figures 6 and 7 show coherence maps for one of the 35-day repeat and the 315-day repeat pairs. Some coherence remains in the towns and over the Feltwell agricultural area after a month in the early growing season but little remains elsewhere. Over a time scale of nearly a year, there still remains high coherence in the urban areas and over the airfields that are characteristic of this region. Figure 11 combines coherence maps from 1, 35 and 315 days showing the limits of coherence over 1-day in red, 35 days in yellow and long term in white.

Another source of information in the ERS data lies in the difference in intensity between images from different dates. Figures 8 and 9 show composites of coherence, s⁰ and Ds⁰ for 1-day and 35-day repeat image pairs [Refs. 3 and 4]. This demonstrates that for 1-day repeat pairs, most information is carried in the coherence while for longer time periods more information is carried in the intensity difference than in the coherence.

5. CONCLUSIONS AND FUTURE WORK

This study has investigated the relationship between interferometric coherence and biomass density of Corsican Pine stands at Thetford Forest in the UK for repeat-pass ERS SAR image pairs over a variety of time scales.

The 1-day repeat coherence is strongly related to biomass density although this relationship is variable with the date of acquisition, probably because of environmental factors. Preliminary results show the dependence to be monotonic although noisy up to biomass densities of around 100 tonnes per hectare. Coherence over longer time scales than 1-day is uniformly low over forest areas. The relationship between biomass density and 1-day repeat coherence is stronger than that between biomass density and s⁰ suggesting that the extra information provided by SAR interferometry may be useful for forest applications.

Well developed phase fringes from 1-day repeat data may allow phase difference information to be used in the estimation of tree height.

Future work will concentrate on determining the sources of variability in the relationships and using the phase information to estimate tree height. The analysis will be repeated at the boreal forest test site at Siggefora in Sweden and further attempts will be made to perform similar analyses of the tropical forest test sites.

6. REFERENCES

[1] J. O. Hagberg, L. M. H. Ulander, and J. Askne, ìRepeat-Pass SAR Interferometry Over Forested Terrain,î IEEE Transactions on Geoscience and Remote Sensing, vol. 33, pp. 331-339, 1995.

[2] J. I. H. Askne, P. B. G. Dammert, L. M. H. Ulander, and G. Smith, ìC-band Repeat-Pass Interferometric SAR Observations of the Forest,î IEEE Transactions on Geoscience and Remote Sensing, vol. 35, pp. 25-35, 1997.

[3] U. Wegmuller and C. Werner, ìSAR Interferometric Signatures of Forest,î IEEE Transactions on Geoscience and Remote Sensing, vol. 33, pp. 1153-1161, 1995.

[4] U. Wegmuller and C. Werner, ìRetrieval of Vegetation Parameters with SAR Interferometry,î IEEE Transactions on Geoscience and Remote Sensing, vol. 35, pp. 18-24, 1997.

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry