Characterizing peatland carbon balance estimates using freely available Landsat ETM+ data
Résumé
We demonstrate the potential of using freely available satellite data from the Landsat ETM+ sensor for generating carbon balance estimates for lowland peatlands. We used a lowland ombrotrophic peatland in the UK as our test site representing a range of peatland conditions. A literature survey was undertaken to identify the simplest classification schema that could be used to distinguish ecohydrological classes for carbon sequestration on the peatland surface. These were defined as: active raised bog, Eriophorum-dominated bog, milled unvegetated peat and drained or degraded bog, with bracken and Carr woodland to define the bog edges. A maximum likelihood classifier (MLC) was used to map the spatial distribution of the six classes on the peatland surface. A Landsat ETM+ band-5 derived brightness-texture layer created using geostatistical methods greatly improved classification accuracies. The results showed the best accuracy of the MLC, when compared to finer scale methods, with Landsat ETM+ bands alone was 74%, which increased to 93% when including the brightness-texture layer. An estimate of carbon sequestration status of the site was performed that showed good agreement with the results of a finer-scale-based estimate. The coarse-scale map estimating −12 000 kg carbon and fine scale map estimating +23 000 kg carbon per annum. We conclude that with further development of our tool, if textural measures are used alongside optical data in MLC, it is possible to achieve good quality estimates of carbon balance status for peatland landscapes. This represents a potentially powerful operational toolkit for land managers and policy makers who require spatially distributed information on carbon storage and release for carbon pricing and effective land management. Copyright © 2014 John Wiley & Sons, Ltd.