5 Take-Aways from London Climate Action Week 2024
Florian Reber, Head of Partnerships at Chloris Geospatial
Last week we convened an incredible group of people to discuss the question 'How can AI-supported monitoring technologies, such as direct biomass estimation, help bring trust to natural climate solutions?'
One part of the answer is: by enhancing the transparency, reliability, consistency, comparability, and scalability of the carbon estimates that underpin the development, MRV, and financing of natural climate solutions.
The other part of the answer is: by building trust in the ability of such novel technologies to deliver upon those promises.
The latter part of the answer is key, and deserves more elaboration. Here are 5 points that I heard last week, and that I believe are helpful to lay out to foster trust in AI-supported carbon monitoring technologies.
1 – Because every number is an estimate, uncertainty metrics matter. Generally speaking, every biomass and carbon number is an estimate, not a measurement of the truth. In fact, to know the true carbon number, destructive sampling would be needed. Therefore, when it comes to carbon accounting for nature-based solutions, every tonne is an estimate – whether it is calculated in the field or in the cloud. And at every step of every estimation process, there are specific sources of error. It means that uncertainty numbers, and transparency on uncertainty methodologies, matter.
2 – Accuracy matters, but compared to what. The accuracy of estimates matters a great deal, particularly at the project-level. But because every biomass number is an estimate, the first question to ask when confronted with accuracy assessments is: compared to what? To deliver robust and reliable accuracy insights, estimates should be compared at the site-level, not at the plot/pixel level.
3 Field data and remote-sensing data are both/and, not either/or. Every biomass estimate delivered today is possible thanks to previous scientific field work, conducted to establish specific allometric equations used to convert tree measurements into biomass estimates. Robust field data are also needed to validate (and sometimes calibrate) estimates generated by remote sensing based models. And while field data cannot deliver the same scalability, cost-effectiveness, frequency and consistency that Earth-Observation based remote-sensing technologies provide, field data is suitable for smaller, easier to access projects. And field campaigns are also relevant to local communities as a job-creation opportunity. So both remote-sensing data and field data have their role to play. By making smart use of both approaches, informed by their pros and cons, we can leverage their respective strengths and augment our ability to develop impactful nature-based solutions.
4 – Scientific integrity is everything. It goes without saying that for a technology that seeks to bring integrity to nature-based solutions, (scientific) integrity needs to be built into every step of the development and marketing of the product. Even more so as understanding whether a specific remote-sensing technology is fit-for-purpose constitutes a challenging process, especially for decision-makers who may not have a background in remote-sensing, forestry, or carbon science. To support real progress and build confidence in new tools, integrity-led marketing is key.
5 – The market needs a data benchmarking system. Because every number is an estimate, and because every algorithm works differently, different algorithms can predict different carbon estimates for the same site. How to know which one to trust? To infuse trust and foster comparability of the “safety” and quality of algorithmic carbon data, there is a need for a high-quality, well-designed data validation and benchmarking infrastructure. It would equip the market with a common, standardized measurement stick, increase transparency and foster trust in the numbers. Efforts such as the benchmarking exercise conducted by ERS in October 2023 can be an excellent example to inform how a system for more industry-wide benchmarking could be developed.
At Chloris, we look forward to continuing the dialogue on this important topic and hearing your thoughts on what else is needed to infuse trust in AI-supported carbon monitoring.
Get in touch to learn more: info@chloris.earth