Title
Satellite Monitoring for REDD: Radar vs. Optical
Author(s)
Mitchard, E. T.;Saatchi, S. S.;Ryan, C.;Woollen, E.;Goodman, L. E.;Williams, M.;Gerard, F.;Starkey, M.;Meir, P.
Published
2010
Publisher
Geophysical Research Letters
Abstract
The proposed REDD (Reducing Emissions from Deforestation and Degradation) protocol will only succeed in reducing emissions if deforestation and degradation can be accurately monitored. Ground surveys are prohibitively expensive over large areas, so satellite monitoring will be essential for independently monitoring deforestation and degradation rates, and thus calculating payments. In addition, remote sensing will be needed for developing historical baselines. It is clear that different methodologies will be needed for different project areas, scales and threat types. In addition different methodologies are needed depending on the capabilities of the countries in question: in the context of Africa for example, in-country monitoring tends to be at a basic level, yet for REDD it is important that countries are able to do their own monitoring, so simple methods must be developed and tested. In this paper we present change detection results for different remote sensing methodologies for potential REDD projects in Cameroon and Mozambique. We have shown previously that a good relationship exists between aboveground biomass (AGB) and L-band radar backscatter (Mitchard et al. 2009). The errors in estimation of absolute biomass were still relatively high, in the region of ±25 %. However, it is suspected that some of these errors are intransient, being due to the structure of the landscape and vegetation within a site, and so it has been predicted that errors in change detection are smaller than those for absolute estimation. We present evidence from these sites that this is indeed correct, using ALOS PALSAR L-band radar data from 2007, 2008 and 2009. Optical satellite data is widely used for monitoring deforestation, for example the excellent system run by INPE in Brazil. However, while optical data is good at detecting deforestation occurring progressively in large clear-fell blocks, as in the Amazon, it is less good at detecting small-scale deforestation or degradation, unless at a very high resolution and revisit frequency (which will not be possible over large areas due to the limited capacity of current hyperspatial satellites). We have found in these study sites that optical data can be used, particularly to help set up historical baselines, with high correspondences between canopy cover and dry-season vegetation indices in the savanna and woodland dominated areas, using Landsat and ASTER data. However, at higher biomass levels the value of optical data may be limited to detecting clear-fell areas, with degradation and regrowth much harder to detect. Radar data also suffers from saturation in higher biomass areas, but in general appears to offer more chance of detecting deforestation and degradation than optical data alone; however a combination of both will always give the best results.
Access Full Text
A full-text copy of this article may be available. Please email the WCS Library to request.
![]()
Back
PUB14358