Satellites are getting too good for forest carbon? / Feb 2025

There's a letter in Science today from a bunch of well known remote sensing researchers that make the unusual point that modern satellite resolution is getting too good to be accurate for forest carbon estimation.

Many new satellites can resolve fine features on the landscape, and even some individual trees outside forests, but this resolution (0.3-5m) is too high for mapping forest carbon. Forest carbon has a natural resolution constraint: the size of an individual tree. To create these maps, tree data from the ground are required because there is no direct measure of tree carbon nor any way to accurately divide trees into smaller components from space. [...] Because most carbon in a forest is stored in large trees, map resolutions should at minimum exceed the crown diameter of a typical large tree, which ranges from about 10m for temperate forests to about 20m for tropical forests --- Laura Duncanson et al, Spatial resolution for forest carbon maps, Science

The lead author Laura Duncanson is a remote sensing scientist at Maryland who works on the incredible GEDI instrument on the International Space Station. In her recent EEG seminar talk, she noted that their instrument is so sensitively calibrated that they can detect when astronauts on the space station are flushing the loo!

David A Coomes further notes that we shouldn't think of either field data or GEDI footprints as sole ground truths, but rather factor in the combined uncertainties in both ground and remote sensing data. This 2018 Geosciences paper goes through the details of how this error propagation works in Borneo rainforests:

By combining ALS imagery with data from 173 permanent forest plots spanning the lowland rainforests of Sabah on the island of Borneo, we develop a simple yet general model for estimating forest carbon stocks using ALS-derived canopy height and canopy cover as input metrics. An advanced feature of this new model is the propagation of uncertainty in both ALS- and ground-based data, allowing uncertainty in hectare-scale estimates of carbon stocks to be quantified robustly.

[...] Since the 1970s Borneo has lost more than 60% of its old-growth forests, the majority of which have been replaced by large-scale industrial palm oil plantations.

With the view of halting the further deforestation of carbon-dense old-growth forests and generating the necessary knowledge to better manage its forests into the future, in 2016 the Sabah state government commissioned CAO to deliver a high-resolution ALS-based carbon map of the entire state. The regional carbon model we develop here underpins this initiative [...] -- Tommaso Jucker, David Coomes et al, Estimating aboveground carbon density and its uncertainty in Borneo’s structurally complex tropical forests using airborne laser scanning

Michael Dales and Thomas Swinfield are just starting to refresh our PACT methodology spec, so this yet another timely warning to not race ahead with the latest satellite data without careful consideration of what it is we are actually measuring (in our case, forest carbon for carboncredits).