Nrdly
Get Nrdly Free Trial Built with Nrdly

Modeling Forest Canopy Height using GEDI and Sentinel Data in Earth Engine

Introduction

Forest conservation is one of the most critical solutions to manage the ecosystem and mitigate climate change. Therefore, spatially explicit monitoring of forests is vital to assess forest conservation activities and transparent reporting of forest carbon. Over the past decades, International and national initiatives have implemented activities to reduce and stop tropical forest loss. In addition, global programs such as REDD+ are being implemented in some countries to mitigate climate change. While forest cover mapping has improved, forest carbon monitoring programs have been hampered by a lack of forest structure information.

To date, forest carbon is poorly quantified in the tropical regions due to a lack of timely spatial explicit forest structure information, such as forest canopy height. Forest canopy height is essential for assessing forest degradation and deforestation, estimating above-ground biomass (AGB), and monitoring ecosystem changes. Traditionally, forest researchers and ecologists used field measurements to collect forest canopy features. However, collecting forest canopy height information through ground measurements is time-consuming, costly, and often limited in spatial extent. Currently, remote sensing researchers use terrestrial laser scanning (TLS) and airborne laser scanning (ALS) to map forest canopy height. In general, TLS and ALS provide more accurate forest canopy height data. However, TLS and ALS are costly and have limited spatial and temporal coverage in large forest areas.

Current and Upcoming Earth Observation (EO) Systems

To date, remote sensing experts use spaceborne LiDAR systems such as NASA’s Ice, Cloud, and Land Elevation Satellite (ICESat) and the Global Ecosystem Dynamics Investigation (GEDI) to produce global forest canopy height maps. ICESat-2 (launched in September 2018) measures the elevation of ice sheets and glaciers (cryosphere). While NASA designed ICESat-2 to focus on the cryosphere, the satellite also globally measures forest heights. GEDI measures ecosystem structure and provides global coverage from 51.6° south to 51.6° north. The LiDAR system acquires forest vertical structure measurements in temperate and tropical forests. GEDI comprises three lasers that produce eight parallel observation tracks. Each laser illuminates a 25-meter spot on the surface, where the 3D structure is measured. Every 25-meter site is separated by 60 meters along the track. The GEDI system acquires data along eight tracks 600 meters apart.

Several upcoming Earth Observation (EO) missions will collect satellite data that will help to map forest structure and AGB. For example, the NASA Indian Space Research Organization (ISRO) Synthetic Aperture Radar (NISAR) mission will provide high-resolution data to map and monitor forest structures. The NISAR mission is scheduled to be launched in 2022 and will collect wall-to-wall L-band data (and S-band in some regions) globally. NISAR will repeat its orbit every 12 days during its three-year mission. In addition, other spaceborne missions such as the European Space Agency (ESA)’s BIOMASS, and the German Aerospace Center (DLR)’s TanDEM-L will also contribute to mapping and monitoring forest structure.

The Need for National Forest Canopy Height Maps

Currently, global forest height maps are available. While these global forest height maps are essential, their accuracy varies regionally. Scientists have reported lower accuracy due to variability in land cover types, topography, geological structure, soil types, and climatic conditions. However, some studies show that the accuracy of forest height maps improves when restricted to local areas or specific forest types. For example, scientists report high accuracies for project-level or national-level forest height maps. Therefore, national forest height maps are required to design effective climate change solutions at the local and national levels.

National forestry agencies need forest height maps calibrated to the local environments (e.g., national land cover types) to monitor forest structure and produce reliable AGB estimates. However, the production of national forest height maps requires technical, financial, and human resources (e.g., experts and high-performance computing resources), which are barriers for most national forestry agencies in tropical countries. Therefore, these agencies can use EO data such as GEDI and cloud computing platforms such as Google Earth Engine to produce their maps. In addition, governments and international development partners need to provide long-term funding instead of ad-hoc one-time financing for forest conservation programs. Properly executed long-term funding will provide a solid foundation for progressively producing reliable national forest canopy height and AGB maps.

Modeling Forest Canopy Height

This blog tutorial will model forest canopy height using GEDI’s Level 2A Geolocated Elevation and Height Metrics Product (GEDI02_A). The “GEDI02_A” dataset comprises 100 Relative Height (RH) metrics. We will use the dataset “LARSE/GEDI/GEDI02_A_002_MONTHLY” available in Earth Engine. This dataset is a raster version of the original “GEDI02_A” product, organized as monthly composites of individual orbits in the corresponding month. For more detail about the dataset, please refer to the Earth Engine catalog. We will also use seasonal Sentinel-1 and Sentinel-2 data and elevation data as predictor variables in the RF model.

Next Steps

We will use Mafungautsi Forest Reserve in Zimbabwe as a test site in this post. Readers can access the tutorial and GEE script at Ai.Geolabs.