Although forests store carbon, young forest canopies may take up less CO₂ than the underlying soil respires, making them a source of greenhouse gases . As a result, we are investigating the use of interferometric coherence from the ERS Tandem mission of 1995-2000 to estimate the age structure of British plantation forest stands (Fig. 1).

Figure 1. Coherence image of Kielder forest (left) and the resulting pixel classification map (right), showing forest compartment outlines in white. N.B. The very low coherence of Kielder Water (upper middle of the classified image) causes it be to classified as mature forest

Simple scattering models suggest that coherence is inversely related to the ratio of vegetation canopy backscatter and the underlying soil, and hence should be related to biomass (and age). Plots of coherence against age (Fig. 2) confirm this, and can be inverted to give estimates of forest age (Fig. 1). From Fig. 2 it is clear that coherence gives best discrimination of age for younger stands, but fortuitously it is the younger stands that exhibit the strongest variation of CO₂ exchange with age. Both coherence and CO₂ exchange become roughly independent of age for older trees. A more important problem for the inversion is that the shape of the plot varies as environmental conditions vary. We are investigating methods to invert coherence dynamically by modelling plant and soil water content in response to weather conditions (see Figure 3.).

The simulations in Fig. 3 underestimate the variations in observed coherence, but show the right trends, as long as we account for varying canopy moisture (Fig 3, top). This underlines the importance of looking at the dynamics of the whole canopy-soil system if we wish to understand the variations in coherence. However, these figures also illustrate that the temporal variability of coherence is both large and difficult to predict accurately. Possible weaknesses in the coherence modelling include incorrect description of the dielectric responses of living plants to changes in solar radiation, and inadequate treatment of the coherence of forest canopies as a function of penetration depth. The significance of such effects is currently under investigation, using this data set.

The difficulties in constructing predictive physical models of coherence have forced us to develop empirical techniques for practical forest age structure mapping. One approach, based on filtering, can reproduce much of the dynamic fluctuations of coherence in response to daily meteorology data, allowing some adjustment for meteorological affects. A more direct approach exploits the large resource of forestry data in the UK to deal with inter-acquisition variability. For large scale mapping purposes, five test sites in mainland Britain have been used to calibrate individual inversion functions (based on fitting to plots like Fig. 2) for the range of meteorological conditions under which UK SAR coverage was acquired. These functions have been used to extrapolate the inventory age data throughout the UK using coherence imagery (Fig. 4a). This method also allows the errors for specific plant ages and specific acquisitions to be estimated (Fig. 4b). Note that forest age information in the UK is only available for Forestry Commission stands, which make up less than a third of the forested area. Hence the plots shown in Fig. 4 provide information that is otherwise very difficult to obtain. These remote sensing techniques also provide the ability to avoid some of the systematic anomalies in inventory methods for deriving age in publicly owned forest.

       
Figure 4. (left) Satellite-derived age estimates (white = 0 or no forest, black is 70 year old forest) (right) Error SD estimates (white = 0, black =30 years) using a UK forest mask derived from Private Woodland and Forest Commission land-use data.

Satellite-based age structure estimates have been used to illustrate the potential uses of remote sensing for carbon accounting. Net Ecosystem Exchange (NEE) data from a chronosequence of sitka spruce stands , obtained using eddy covariance measurements in the Kielder area, allow tree age to be related to NEE. Combining this information with large scale age structure data allows detailed carbon accounting information to be inferred for entire forests and larger regions. Table 1 illustrates the application of both GIS and remotely sensed forest age structure information to derive NEE estimates for England, Scotland and Wales. For comparison, the table includes values from the national carbon emissions inventory (Milne et al. 2000). NEE estimates using both GIS and satellite data show much larger carbon sequestration than inventory.

On a country by country basis, inconsistencies between remotely sensed age structure and GIS-derived age structure have also been identified. These are evident by comparing the inventory and observed NEE estimates. Figure 5 illustrates the differences between GIS-based data for the Forest Commission (FC) woodland and the SAR-derived estimates. The GIS only includes planting events and is not updated when forest is felled, hence areas which are not replanted can be mistaken for mature forest. Typically UK forest is clear-felled between the ages of 40 to 60 years, implying that much of the planting in the 1940s and 50s would have been felled by the time of the satellite observations in 1995. The observations indeed show an age structure consistent with such a felling programme. Comparing private woodland with FC woodland shows a systematic difference in age structure, which accounts for the differences in NEE for these two categories of forest. (Note, however, that the histograms below only account for planting events and not planted area).

   
Figure 5 (a) Welsh Forest Commission GIS planting dates; felling times unavailable. (b) Welsh Forest Commission planting dates inferred from coherence-based age estimates. (c) Welsh privately owned forest planting dates inferred from coherence-based age estimates.

For comparison, the coherence-derived age structure of Scottish forest is shown in Fig. 6; here a much more uniform planting and feling programe appears to be in place. This age structure results in the higher estimates of carbon uptake than for Welsh forest (see Table 1).

Back to CTCD EO Overview page

HOME | MISSION STATEMENT | SCIENCE PROJECTS | PARTICIPATING ORGANISATIONS
PEOPLE | NEWS | CTCD ORGANISATION | CONTACTS