|Paper title||Blockchain Applications for Biomass Measurement and Deforestation Mitigation|
|Form of presentation||Poster|
Blockchain Applications for Biomass Measurement and Deforestation Mitigation
Accurate estimation of forest above-ground biomass and its change over time is critical to forest conservation efforts and the consequence of the current voluntary carbon market. A positive change in biomass through planting more trees is important, but it must be noted that afforestation is merely the first step in producing an increase in global carbon sequestration. The true variable influencing the outcome of these efforts is the resilience of the biomass, and tree growth that resilience results in. Typically, forest densities range from 1000 to 2500 trees per hectare and have a carbon sequestration rate of up to about 10 tonnes per hectare per year in the tropics (0.8 to 2.4 tonnes in boreal forests, 0.7 to 7.5 tonnes in temperate regions and 3.2 to 10 tonnes in the tropics). By the age of 100, one broadleaf tree (commonly found in the tropical rainforests) could have sequestered up to one tonne of carbon. In comparison, chopping and burning just about 5 to 10 average-sized pine trees (~450 Kg dry weight each) instantly releases back all of the carbon that one hectare of trees captured in one year. This points to an extreme negation effect that logging can have, and can be a powerful point of change for policy makers.
The new forests that have been planted in the past two decades represent merely 5% of the net global carbon sink. While these numbers will grow with time, it is currently far more important to monitor and prevent deforestation of existing mature forests. It is also potentially one of the most difficult interventions to implement. Most deforestation and forest degradation are concentrated in the tropics because of both illegal logging activities that can go undetected due to the small spatial scales they occur at, and because of general forest clearing for cattle pasture and agricultural expansion. The most drastic effects of this rampant deforestation have been witnessed in the Amazon rainforest, which, at the time of writing this, has already / is strongly tending towards becoming a net emitter of carbon instead of a net sink, if the given rates of deforestation continue. Since the turn of the century, Brazil alone has released 32.5Gt of CO2 from deforestation and for reference, the average amount of CO2 annually emitted across the globe varies around 40 Gt. The rate of carbon emission also varies according to forest type. Old-growth primary forests, unlike their secondary and fast-rotation counterparts, can release carbon that has taken centuries to get stored. Hence, illegal logging, especially deep within primary forests needs early detection through regularly updated AGB change estimation.
At a policy level, accurate and timely detection of changes in biomass can be of value to countries that are attempting to recognize indigenous peoples and local communities as owners of their lands. Enforcing the rights of indigenous communities is a proven strategy to protect standing forests and enhance the carbon stored in them.
Governments across the globe have been actively incorporating forest landscape restoration measures in their policies. However, the effectiveness of these interventions towards carbon removal and climate change mitigation is difficult to quantify, especially in regions where there is insufficient biomass data. To fill these gaps in knowledge, various programs have provided open-source access to sensor-fused datasets at resolutions varying between 100m (ESA Climate Change Initiative) to 500m (NASA Pantropical AGB dataset).
Our work aims to utilize machine learning techniques such as Random Forest and XGBoost to train our algorithm on recently developed AGB datasets and Vegetation Indices extracted from satellite image radiances. The correlation between these inputs is then used to predict AGB in a different location and year. Since above-ground biomass accounts for about 27% of the entire carbon sequestered by a tree, the mathematical difference between the pixel values of two independent AGB predictions over simultaneous years allows us to estimate the total carbon sequestration at that pixel, in that year. Moreover, biomass change detection can help clearly identify logging activity, storm damage, restoration after forest fires and reforestation efforts being undertaken by forest managers. This allows for monitoring of policy implementations as well. Through sensor fusion with multiple satellite datastreams including SAR data, we can monitor large scale regions (including remote and inaccessible ones) at night, and through clouds (a major issue in imaging the tropics), with a rapid revisit rate.
Predictions of biomass change and carbon sequestration both occur at the pixel resolution of the training dataset, although we are also working on methods to increase pixel resolutions of the final products through the use of deep neural networks. The ability to monitor biomass change at 100m resolution and lower, will also assist with forest boundary change measurement. Forest boundaries can be substantial areas of change in forest expanse due to easy access, and estimating those changes can be very helpful to forest managers.. Moreover, we test several vegetation index combinations to better understand and potentially provide insight towards standardizing best practices for AGB prediction models globally.
Our work further contributes to solving the issue of the severe lack of multi-temporal AGB datasets. Projects such as the NASA GEDI mission, while providing very high-resolution AGB estimates, are currently only a one-time estimate. A similar problem occurs with the ESA CCI biomass datasets which are only valid for 2010, 2017, and 2018. Internally consistent AGB change datasets were only made available in December of 2021. Other datasets are also only available for one year randomly placed throughout the past two decades. This points to the fact that while work is being consistently done to acquire and formulate these datasets, there is a need for predictive software to estimate continuous time series of AGB change across several years, which can then be validated by ground truth and reference datasets whenever they become available.
Environmental protection does, however, come at the cost of economic growth; and this is a major hurdle, especially in developing nations. Therefore, a highly effective way of incentivizing countries to strictly control deforestation is to provide them monetary compensation through the use of carbon credits.
The carbon credit market has a key role to play in the solution to the problem of climate change and dClimate is entering this field using machine learning mechanisms and advanced AI algorithms. The current voluntary carbon market does not put enough emphasis on preventing deforestation. Only 32% of carbon offsets deal with preventing deforestation, and the IPCC believes the number of carbon removal projects must increase in order to limit warming to 1.5°C.
One of the primary issues with carbon offsets is the lack of transparency in the market. dClimate is revolutionizing the industry by verifying our own offsets using blockchain technology, which will allow buyers and project creators to view a transparent immutable ledger with all needed information available. To do this, we are creating an above-ground biomass estimation and monitoring system, which will allow us to price tokens based on the amount of carbon sequestered by existing biomass, and stored in the form of AGB change. Currently, the VCM registries are plagued with double or triple counting wherein the same parcel of land is sold multiple times. This not only prevents adequate market dynamics for the price of carbon offsets, but also limits the growth of the industry. In addition, the measurement of deforestation and carbon output is traditionally non-standard as it entails very bespoke methodologies which are not equipped to handle the problem at scale.
Leveraging the intersection of the simultaneous advances in many decentralized technologies such as Chainlink, IPFS, and distributed ledgers (Ethereum) we are able to create new financial ReFi (regenerative finance) primitives which facilitate carbon price discovery. Additionally, through the use of decentralized execution environments, all computations are transparent and can be inspected by anyone providing a platform for trustless interoperability without having to rely on centralized failure points. By creating this infrastructure we not only create the tools to mitigate deforestation, but also accurately measure other parts of the collective climate economy.
As a result of bringing together the new multi-resolution (spatial and temporal) datasets from multiple global organizations, increasing end-to-end transparency, and creating pressure on countries and stakeholders through a penalty system for anthropogenic biomass reduction, our work will develop detailed maps of Above Ground Biomass and its spatio-temporal variation. This will not only provide financial impetus to all nations who choose to use our services (especially to low-income countries with high biomass reserves), but will also assist the global scientific community by providing a rapidly updated database through an easily accessible API, aimed at creating a standard system of carbon emission control and sequestration measurement.