Project Details

Description

ASU-GDCS Forest Carbon Mapping and Analysis Program Undisclosed Objectives Worldwide Tropical Forest Carbon Monitoring This project aims to map and monitor aboveground carbon stocks and emissions from the world's tropical forests with unprecedented detail by combining high-resolution daily satellite imagery and advanced artificial intelligence analytics. Creating low-cost, automated, spatially explicit indicators of forest carbon stocks and emissions will greatly contribute to sustainable pathways in economic development for tropical countries while benefiting the entire world. It will serve as a transformative tool that quantifies and helps to value the climate change mitigation services forests provide, as part of the MRV (Measurement, Reporting, and Verification) systems for REDD+ mechanisms developed by UNFCCC (United Nations Framework Convention on Climate Change). Approach Worldwide Tropical Forest Carbon Monitoring A proof-of-concept approach for carbon monitoring was recently developed and is curreully ueiug 1efi11eu for the country of Peru, where Planet Dove's high-resolution daily satellite imagery was combined with the Asner Lab's Carnegie Airborne Observatory LiDAR (light detection and ranging) data using machine learning techniques to create a high-resolution carbon map for the entire country. Technological developments in Earth observation have reduced costs, increased the quality of images, and decreased revisit times of the satellites, making it possible to map forest carbon stocks at highresolution, every day. We aim to measure aboveground carbon stocks in all tropical forests. We previously showed that a correlation between aboveground forest carbon stock and tree height is maximized in grid cells (pixels) of I-hectare in size. Tree heights were measured by the carbon-sensitive CAO LiDAR, and in the case of Peru, cover approximately 5% of its area. The Asner Lab used this information in 2014 to generate the I-hectare high-resolution carbon samples throughout Peru. The remaining 95% of the country's carbon map were generated from a model relating the airborne LiDAR carbon samples to full-coverage, Landsat multispectral data, along with environmental data. Landsat provides imaging of Peru (and other tropical regions) once every 16 days, but with cloud cover, this equates to a clear mapping of the region about once per six months. This limits the use of Landsat and all other freely available sources of government data to satellite-based mapping estimates of changes in forest carbon stocks. Coarse resolution data, such as from Landsat, are not adequate to extract detailed patterns indicating the structure of trees, including their heights. In contrast, Planet imagery has a high spatial resolution (3 .7 meter), which makes possible the extraction of textural features from images, which can capture characteristics of trees canopies including height, shape, and roughness. As a result, we have demonstrated in Peru that Planet imagery can be used to predict top of canopy heights and, ultimately, to estimate aboveground carbon stocks using machine learning techniques. Creating a carbon estimation model using spectral and textural information from Planet is changing how we approach a low-cost, automated and timely indicator of carbon stocks and emissions over the world's tropical forests. Our current model developed for Peru and ready to be scaled across the world's tropical forests is based on spectral and textural information extracted from Planet Dove satellites, intersected with CAO's LiDAR top of canopy heights data using gradient boosting algorithms and random forest machine learning regression. Spatial texture analysis approaches comprised of Fourier Transformed Textural Ordination (FOTO) are able to detect patterns in the data at multiple scales and to reconstruct the characteristics of tree canopies. Our new modeling approach was evaluated against samples of CAO's top of canopy LiDAR samples not used in the training step, achieving an agreement of more than 80% with the airborne reference carbon samples. Our proof-of-concept was focused on the mega-diverse prototyping landscapes of Peru for multiple reasons. To adapt to a global range of tropical forest conditions, we will make use of CAO LiDAR campaigns from the past made for multiple countries including Panama, Malaysia, Madagascar, Costa Rica, Ecuador and Colombia. Additionally, we will make use of NASA's newest mission, GEDI (Global Ecosystem Dynamics Investigation), which was launched on December 6, 2018, and will provide complete LiDAR coverage of world's tropical forests, measuring the height and 3D structure of the forest at a resolution of 25m. Planned Outcomes We believe that It is now technically feasible Lu generale high-1esululiu11111as, updated frequently through time, of aboveground carbon stocks and emissions for global tropical forests. Such a dynamic monitoring system could be served to the public via an online interactive platform for exploration and extraction of changing forest carbon composition, which will greatly advance multi-sector efforts in forest conservation, management, and policy decision-making. Such efforts include the specific objectives for tropical forest conservation in Indonesia and Malaysia laid out in this proposal.
StatusFinished
Effective start/end date4/1/204/2/20

Funding

  • Morgan Family Foundation: $1.00

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.