6 December 2011 | DURBAN | Daniel Tutu and Winston Asante barely noticed the morning glow creeping into the night sky outside their hotel on that second straight sunrise they’d worked through with hardly a break and no sleep at all.
“And no coffee,” says Asante. “It makes me hyper.”
The marathon session topped off more than two years’ work by scores of researchers at universities and government offices across Ghana, all cooperating on a project that could revolutionize the way developing countries measure and value their rainforests. It employed technology as basic as digging in the dirt and as advanced as mining the newest satellite data that the United States and Japan have to offer. If it worked, it would slash the cost of measuring forest cover and provide a tool that developing countries around the world could use to establish baselines and map out their forestry strategies.
And it was about to face its first big test.
Will it Work?
The two men packed up their computers and made their way down dusty streets to the Achimota Forest Reserve, a patch of woods that the city of Accra had saved for posterity in the 1930s.
It has been shrinking ever since, and today the modern campus of the Forestry Commission of Ghana occupies a clearing at the edge. Thirty men waited inside. They were the officials who, after months of debate, were charged with answering an apparently simple question: how should the country define a forest?
The question isn’t as peculiar as it seems, and the answer would determine the fate of the country’s forests and the flow of millions of dollars in carbon finance.
Cocoa, Carbon, and the Right Strategy
The Forestry Commission is charged with preserving forests and ensuring it is done in a way that doesn’t hurt the country’s legions of cocoa farmers – most of whom operate just above the subsistence level. It is this dual mandate to serve both the environment and the farmer that led the Commission to embrace forest carbon – but they first had to decide which type of forest carbon they wanted to pursue.
Most members of the commission favored a REDD (Reduced Emissions from Deforestation and Degradation) strategy, which would let farmers earn carbon credits by saving endangered forests. A growing minority, however, favored an A/R (Afforestation/Reforestation) strategy, which would let them earn carbon credits by planting trees on land that hadn’t been forest for a long time – if ever.
The choice was complicated by uncertainty over the future of both mechanisms. A/R is the only forest carbon vehicle recognized under the Kyoto Protocol’s Clean Development Mechanism (CDM), but it’s an unpopular vehicle and is not likely to see its prospects improve at talks underway this week in Durban, South Africa. REDD, on the other hand, is little more than an aspiration within the United Nations Framework Convention on Climate Change (UNFCCC), but it has been gaining wide acceptance in the voluntary markets and in newer regulatory regimes around the world, according to Ecosystem Marketplace’s State of the Forest Carbon Markets 2011 report.
What’s more, REDD is evolving into something called “REDD+”, which recognizes a broad range of activities aimed at reviving degraded forests, while pure REDD only recognizes programs that save endangered virgin forest.
With these factors in mind, most Commission members concluded that REDD+ provided the best vehicle for integrating conservation into the cocoa economy. Farmers, they reasoned, could accept lower yields, but only if “cocoa carbon” made up for the reduced income; and cocoa carbon is only feasible on a large scale under REDD+.
The Commission even wrote up a REDD Readiness Preparation Proposal (R-PP) that was accepted by the Forest Carbon Partnership Facility (FCPF), a global non-profit partnership that helps developing countries ready themselves for REDD.
What is a Forest?
But the Commission couldn’t chart a course until it knew for sure how its actions would be recognized under global agreements aimed at differentiating a cluster of trees from a forest. Such agreements had been hammered out through years of debate, and were designed to account for the environmental value that a cluster of trees delivers while taking into account regional differences and the limitations of carbon accounting. If Ghana embarked on a REDD+ course and found out later that the trees it saved didn’t fit the recognized legal definition of a forest, or that its baseline measurements weren’t up to international standards, then it would all be for naught.
And both of these were very real possibilities. For one, the forest inventories that Ghana did have were based on old estimates that the Food and Agriculture Organization (FAO) had made for general guidance. They weren’t rigorous enough for constructing a baseline or formulating a carbon strategy. For another, the definition that Ghana used in its R-PP was based on different criteria from those laid out in the Marrakesh Accords, an internationally-recognized set of agreements under the CDM that base their definition on a forest’s size, potential height, and the degree to which land is covered from above by branches and leaves (called “crown cover”).
Without a solid understanding of what type and number of forests it had, the Commission was paralyzed.
Finding an Answer
Yaw Kwakye, who runs the Forestry Commission’s Climate Change Unit, knew that Tutu and Asante had spent the previous two years working on a groundbreaking effort to map the country’s biomass, and he asked if they could use that biomass data to generate maps showing Ghana’s forest cover under different definitions.
The map was the first major project undertaken by the West Africa Katoomba Incubator, which is a joint project of the Katoomba Group Ecosystem Services Incubator (KI) and the Nature Conservation Research Center (NCRC).
The idea for the biomass map came in November, 2008, when NCRC CEO John Mason invited Oxford Professor Yadvinder Malhi to a two-day workshop that NCRC was hosting in Ghana.
“NCRC wanted to help Ghana get ready for REDD,” says Malhi, who’d spent more than a decade examining the impacts of climate change on forests around the world – in the process becoming an expert in measuring the amount of carbon in trees, grass, and everything in between.
“Measuring biomass is core to understanding the impacts of climate change,” he says. “And there are two primary tools for doing it: one involves going out in person and hugging trees with measurement tapes, and the other is remote sensing.”
Measuring trees is effective but requires lots of people and sophisticated logistics to get enough data, while remote sensing is cheap but generally considered less precise at the local scale. Mahli wanted to test new methods of incorporating ground data into the tools that analyze forests from the sky.
The Limits of Technology
One of those improvements was being developed by Sassan Saatchi, a physicist at the US National Aeronautics and Space Administration (NASA) who had worked with Malhi at Oxford while mapping biomass in the Amazon and Peruvian Andes. Malhi and Saatchi both knew that the technology for doing large-scale mapping would not be deployed for years, but they wanted to see what they could accomplish with the technology that exists today.
“New remote sensing approaches using light detection and ranging (Lidar) and radio detection and ranging (radar) from airborne sensors have been successful in providing high-resolution estimates of forest carbon density for small areas,” he wrote in a paper published in May by the Proceedings of the National Academy of Sciences of the United States.
“The space-borne sensors needed to use these approaches for large-scale mapping and monitoring efforts will not be available before the end of this decade,” he continued. “Until then, cost-effective mapping of carbon stocks for project- and national-scale assessments will rely on a combination of satellite imagery and ground-based inventory samples of forest carbon density.”
His solution was conceptually simple: gather data from Lidar and radar, mix it with photographic images and compare it to hard data from the ground, then come up with an algorithm that could take the best of all readily-available inputs from the sky and distill it into a reliable presentation of what was on the ground.
“On the flight back to London, I kept thinking of all this, and how useful a carbon map could be for Ghana,” he says. “I was also keen to try these ideas out on a larger scale.”
So, after landing, he e-mailed Mason and arranged for one of his Cambridge students, David Aitken, to spend some time with NCRC. Mason in turn called Jacob Olander, head of the Katoomba Incubator and his partner in the West Africa Katoomba Incubator. Together they applied for a grant from the Moore foundation to get the map off the ground.
When the funding came in, they hired Asante and Tutu – old schoolmates who had each pursued different careers in forestry. As the project began, Asante was working as a free-lance consultant, and Tutu was working for the Ghanaian Environmental Protection Agency (EPA). Tutu joined the project part-time and headed off to NASA’s Jet Propulsion Laboratory at the California Institute of Technology, while Asante became a full-time employee of NCRC and oversaw the gathering of ground data.
“They were critical to the success of this project,” says Malhi. “They not only had a good scientific skill set, but they had the talent and connections to build on networks and relationships of trust. If I’d hired two Oxford researchers to come over and do this, it would have been a much more difficult process.”
Looking for Help
Malhi knew that the Forestry Commission and universities had already taken many of the measurements they needed, but he also knew from past experience that researchers are protective of their data.
“People have put in a lot of work into gathering this data,” he says. “They’re always afraid someone will run away with it and not involve the people who did the heavy lifting.”
He wasn’t sure people would share, and he could understand why.
Malhi asked Asante to invite all the key players to a workshop in Accra to make sure they all understood the nature of the project, and then to ask them to share information. The workshop took place in December, 2009, and to his surprise, every participant willingly shared their data with the Incubator mapping project.
“People understood that this was part of a national effort to make REDD+ a reality,” says Malhi. “We weren’t writing a paper, but were creating a resource that would be freely available for Ghana to use.”
As the data came in, he and Asante realized it was extensive in some places – mostly in the forest reserves the wooded south – and scant in others – mostly in the northern savanna woodlands.
“We weren’t expecting much data in the north, because it’s mostly sparse woody vegetation and other stuff that most Ghanaians don’t think of it as forest,” says Malhi. “Plus, the research capacity is mostly in the south – clustered around Kumasi and Accra.”
Filling the Holes
Asante assembled a team of one botanist and five assistants, and spent the months of July, August, and September of 2010 canvassing the nation with them to document the dominant land use in each region – first by asking locals to simply tell them what was there, and then by visiting some plots themselves.
“Once we got there, we would take as many types of measurements as we could – water, soil, etc.,” says Asante. “We didn’t really know if we’d have a chance to come back, and wanted to use each trip as best we could.”
Off to NASA
While Asante was scouring the Ghanaian countryside, Tutu was at NASA working with Saatchi. They began working with three sets of data – two provided by NASA and one by the Japan Aerospace Exploration Agency (JAXA).
From NASA came visual data generated by the agency’s workhorse Landsat and MODIS (Moderate Resolution Imaging Spectroradiometer) projects and Lidar (Light Detection And Ranging) data from its GLAS (Geoscience Laser Altimeter System) project. From JAXA came radar data from its ALOS-Palsar satellite.
Lidar uses lasers the way radar uses radio-waves, and would essentially estimate a forest’s height, with the laser’s first bounce hitting the top of the forest and the second one hitting the bottom. By combining that data with the images, Saatchi and Tutu were able to get a fairly good idea of what was on the ground.
The process was long and tedious, and involved first filtering and then crunching data that had come from three different sources in three different forms. The filtering involved removing obvious errors, such as outlier data caused by light or sound bouncing at bad angles. The crunching was more of a challenge.
Crunching the Data
The satellite technology had proven adequate for determining the amount of biomass when the land in question had one clearly dominant type of land-use, such as farms, forests, and savannah, but it was less effective on land where the uses were more mixed, so Tutu and Saatchi created three maps.
The first was a “probability map” that depicted the likelihood of a given patch of land having a given biomass. The next was an “error map” that aimed to depict the biomass from secondary land-use types. The third was a map based on regression analysis that aimed to distill a single, reliable figure for each pixel.
In the end, the probability map proved best on larger scales, even though it was weak on a pixel-by-pixel basis. As Saatchi wrote in his paper, Benchmark Map of Forest Carbon Stocks in Tropical Regions Across Three Continents, the probability map was often just 50/50 at the level of individual pixels – each of which represented one hectare – but it was 95% accurate on plots the size of a typical forest carbon project (10,000 hectares or more), and to 99% when scaled up to the size of a country (1 million hectares or more).
“The regression-based map worked well when we were looking at flat land, but it didn’t really work when we got into something like the shade of a mountain,” says Saatchi. “In the end, the probability map was the most accurate representation of what was on the ground.”
Fixing the Algorithms
Tutu returned from California in September, 2010, just as Asante was wrapping up his field work. Then they began fine-tuning the maps by comparing data from the sky with data from the ground and adjusting the algorithms when necessary. Then they would test the new algorithms by seeing how well they predicted what they knew was already there.
As the process continued, the maps became more and more accurate.
By the time of the meeting at the Forestry Commission, the maps were almost ready to be made public, and Kwakye called Tutu and Asante with a request: could they, he asked, use the maps to project the shape of Ghana’s forests under the different forest definitions allowed under the Marrakesh Accords?
That’s because the Accords don’t just give countries an absolute definition of what constitutes a forest, but rather proscribe a range within which countries can set their own minimum thresholds. This means that one country can define a forest as being any cluster of trees larger than a half-hectare with crown cover above 10% and populated by trees that can grow at least two meters high is considered a forest. Another country could set its minimum thresholds at one hectare in size with 30% crown cover and a minimum height of five meters.
So, for 48 hours, the two combed through the digital maps pixel-by-pixel, punching in four different recognized definitions of a forest and generating maps showing how much forest the country had under each definition. The results were astonishing.
When they set the canopy threshold at 10%, forests covered large swathes of the country; but when they the threshold at 30%, forest cover plunged, and the forests that disappeared were among the most degraded – and thus endangered. These were exactly the forests that could benefit the most from REDD+ financing.
Setting the threshold higher wouldn’t give them extra land on which to plant trees, but rather would swap land that was eligible for the up-and-coming REDD+ mechanism for land that might be eligible for the outgoing and cumbersome A/R mechanism.
In the end, they settled on a threshold of 15%, setting the stage for REDD+.
“Without that map, we wouldn’t have known what we had,” says Kwakye. “It made it possible for us to formulate a strategy based on a realistic understanding of our resources.”
The mapping process isn’t yet powerful enough to support projects – even under current voluntary standards, which require higher accuracy on a hectare-by-hectare basis. It is, however, proving to be a powerful decision-making tool.
Malhi believes it can also help guide the development of new sensor techniques, such as the European Space Agency’s upcoming BIOMASS sensor which will allow for direct forest biomass assessments.
“We’re not trying to replace ground inventories, because if you actually undertake a project, you will still need that,” says Malhi. “But we are trying to see if, by identifying a few local stars, engaging the wider community, and having a few international experts involved you can provide enough information to help inform decision-making.”