Better to light a candle than to curse the darkness

The Guardian published an article last week claiming that a very large proportion of carbon offsets certified by the Verra Verified Carbon Standard are worthless.

There are a series of potential issues with the scientific studies behind that article, and the way they have been reported (see statement from Verra here and here , from Everland here , and from our Chief Scientist Prof Ed Mitchard here ).

As forest scientists, we welcome assessment of the impact and quality of forest projects.

At best, rigorous assessment of forest conservation project impacts will drive scarce conservation finance into the highest-performing projects.

At worst, blanket criticisms of the sector risk undermining one of the best means that we have to mitigate climate change and biodiversity loss, now.?

In that spirit we will also highlight one very significant flaw with all three of the scientific studies behind that article.

In order to understand how much forest existed at the beginning of the study period, and to measure forest clearances over time, all of the studies cited by The Guardian used global forest maps that were created using only one source of satellite data: NASA’s Landsat satellite series. These maps are useful in that they cover the entire world, and allow a broad understanding of regional trends in forest cover. Yet this also means that the maps need to work broadly across a very wide range of vegetation types and conditions. This involves compromises in analytical methods used to map forests and changes within them.

Consider for example the differences between classic wet tropical forests of the Amazon, and the dry seasonal forests of Cambodia. These are very different ecosystems with different ecological features and properties that change over space and time. It is not possible to capture these ecological differences in one analytical process. Consequently, the global forest maps which the studies in The Guardian use feature significant forest mapping errors.

Such global maps, quoted in The Guardian, are therefore not sufficiently accurate to be used for forest carbon project development, nor to assess their performance.

Indeed, the production of forest maps of sufficient accuracy for forest conservation project Monitoring, Reporting and Verification requires deep expertise at the intersection of satellite data analytics; AI and machine learning (ML); and tropical forest ecology.

Fortunately, there is a large body of work on how to do this in practice, in order to produce maps which are sufficiently accurate, and thereby establish high quality forest conservation projects that the market trusts. (see here )

Specifically, the highest quality forest maps are produced using satellites with a 10x higher spatial resolution than the broad scale maps used by the studies cited by The Guardian, (e.g. incorporating ESA’s Sentinel-1 Synthetic Aperture Radar and Sentinel-2 optical data?in addition to Landsat and other datasets).?

The Synthetic Aperture Radar data in particular warrants special attention since this technology allows users to ‘see’ through clouds which cover large areas of the world’s forests, and moreover to discover information about a forest’s structure. Our Chief Science Officer co-authored this paper on the subject. Further, LiDAR data directly measures the height of trees.?

Such detailed information allows not only forest clearance to be mapped, but also forest degradation. This is the more subtle removal of around 30% of trees from an area of forest.

Hence these data give a far more detailed and accurate picture of forests than the Guardian source and are the gold standard to be used when undertaking a robust analysis of forest conservation projects, including identifying areas of forest against which to compare the performance of forest conservation projects producing carbon credits.

Until such accurate data is widely employed by projects, governments, ratings agencies, and academics, we urge caution in suggesting that the entire system of forest conservation and carbon crediting is broken.

By doing so we will undermine the rapid investment in forest carbon credits which are funnelling new precious finance into forest conservation.

We must continue supporting forest conservation projects, both under the various voluntary-sector carbon schemes, and under the emerging national schemes. Further, we must continue to improve the integrity of the systems used to Monitor, Report and Verify these projects, including through digitisation and uptake of satellite technologies.

Our forests depend on it:? we are still destroying ~1.5% of tropical forest every year, with all the carbon and biodiversity that is lost with it.?

Better to light a candle than to curse the darkness: let next generation satellite earth observation analytics show the way.?

At Space Intelligence we make forest maps, building on over 60 years of accumulated experience of our science team, embodied in our analytical processes bringing together huge volumes of satellite data in a machine learning framework.

We work directly with project developers and their teams on the ground to iterate and maximise the accuracy of these products, and monitor forest changes over time.

Our team’s work and commitment to forests is documented in over 100 publicly available scientific articles on satellite mapping technology and forest policy.?