Maps of tree cover that were used for developing schemes to reduce
emissions from deforestation and degradation have errors. It’s all about scale
and pixels, say
Betha Lusiana and colleagues
By Robert Finlayson
The ability of any scheme to meet its
national target to reduce greenhouse gas emissions from deforestation and
forest degradation plus conservation (REDD+) requires understanding how its
processes are linked across scales, from local through provincial to national
and international levels. A single approach to reduce deforestation that is
effective for a project in several villages might not be as effective at an
aggregated level, such as a district.
Accordingly, scale must be addressed in REDD+
schemes, including highly technical activities such as satellite mapping of
vegetation cover. This is a critical feature, since knowing how the amount of
carbon stock in the form of vegetation, of what type, and how it changes over
time determines payments to local people for preserving, adding to, or
depleting the stock.
Having a good carbon map is important for
being able to monitor carbon being sequestered or emitted over time. For
incentive schemes, having a map that fits closely to the reality on the ground is
also important. Developing emission maps requires information in the form of
land-cover maps and aboveground carbon stocks for every land-use type in the
landscape. However, both types of information have errors and uncertainty. For example, when looking at a satellite
image, rubber agroforests can be visually mistaken for natural forests (even in
the field it can be difficult for untrained eyes to tell them apart) and the
amount of carbon stock in each type of tree cover can vary substantially, which
means that when changes to the stock are monitored and aligned with payments
for preservation, enhancement or reduction of said stock, there could be large
errors and hence incorrect payments.
To address this, we set out to identify an
appropriate resolution for mapping carbon stock in a REDD+ scheme. This work
was part of a study we conducted—discussed more fully in Mitigation and Adaptation Strategies for
Global Change—to design effective emission-reduction activities in
Tanjung Jabung Barat (a high-emission district in Jambi province, Indonesia)
that can be implemented by the district government.
The study of resolution accuracy involved
two steps. First, we developed emission maps for the district that identified changes
in aboveground carbon stocks between 2000 and 2009. The maps included calculations
that allowed for uncertainty caused by errors in land-cover-map classifications
and the variation of carbon, representing the many possible carbons stored in a
similar-size plot of a given type of vegetation. Second, we calculated
estimates of emissions based on various resolutions from the maps developed in
From this process, we were able to propose
an appropriate scale for monitoring emissions from land-use changes: for anThe effect of scale on hot spots of carbon emissions in Tanjung Jabung Barat, Jambi, Indonesia, between 2000 and 2009. Pixel resolution of 100 m equals pixel area of 1 ha and pixel resolution of 1000 m equals pixel area of 1 km2. Source: World Agroforestacceptable error of 5% (to put it another way: 95% accuracy), planners should
use an emissions map with pixel resolution of 1000 m, equal to a pixel
size of 1 km2.
We compared this with a map developed by planners in
Tanjung Jabung Barat, who had been involved in a participatory planning process
with communities, businesses and government agencies to come up with ways of
reducing greenhouse-gas emissions and found that the schemes they had in mind
would be served well by a map with resolution of 1 km2.
B, van Noordwijk M, Johana F, Galudra G, Suyanto, Cadisch G. 2014.Implications
of uncertainty and scale in carbon emission estimates on locally appropriate
designs to reduce emissions from deforestation and degradation (REDD+). Mitigation and Adaptation Strategies for
Global Change 19(6).
This work is linked to the CGIAR Research Program on Forests, Trees and Agroforestry