GEDI and Sentinel data integration for quantifying agroforestry tree height and stocks.

Area of Science:

• Forestry science
• Remote sensing
• Climate change mitigation

Background:

• Agroforestry, particularly poplar plantations, enhances landscape resilience and sequesters carbon.
• Conventional forest inventories are insufficient for monitoring fast-growing poplar plantations.
• Remote sensing offers an effective solution for tracking plantation structural variables.

Purpose of the Study:

• To estimate carbon stocks in poplar plantations in Italy's Padan Plain using remote sensing.
• To develop a high-resolution canopy height model (CHM) for monitoring poplar plantations.
• To quantify key forestry variables like growing stock volume and aboveground biomass.

Main Methods:

• Developed a 10m resolution CHM using a deep learning U-Net model with Sentinel-1 and Sentinel-2 data.
• Utilized GEDI waveforms for tree height estimation.
• Integrated annual plantation data and a poplar-specific yield table for variable prediction.

Main Results:

• Achieved a 2.6m mean absolute error for the U-Net CHM against NFI data.
• Estimated average growing stock volume (GSV) at 70 m³ ha⁻¹ and carbon stock (CS) at 12 MgC ha⁻¹.
• Quantified 370,000 m³ of harvested GSV (66,000 MgC) between 2021 and 2022.

Conclusions:

• Integration of remote sensing and machine learning enables effective monitoring of dynamic poplar plantations.
• This methodology accurately quantifies forest variables essential for climate change mitigation strategies.
• The study demonstrates the value of remote sensing for precise and frequent updates in forestry management.