Conference Agenda

The second year of the China ESA Forest Observation project (CEFO-2) has strengthened the joint objective of developing a dynamic forest inventory framework for China and the United Kingdom through the integration of near-surface, airborne and satellite remote sensing. During this phase, work in both countries has moved beyond the initial establishment of baseline inventories and has concentrated more directly on the observation of change through time, the improvement of biomass estimation, and the development of methods capable of supporting future forecasting and management scenarios. Taken together, the activities undertaken in China and the United Kingdom demonstrate the value of combining detailed local measurements with scalable Earth observation approaches to deliver forest information that is both scientifically robust and operationally useful.

The Chinese research team conducted investigations across three domains. In individual tree species identification, a cognition-inspired structure-visual fusion framework was developed using multi-platform point cloud datasets, achieving species-level and genus-level classification accuracies of 86.12% and 87.49%, respectively. A Kalman filtering-based polar coordinate cross-section reconstruction method was further proposed to address point cloud thickness bias in handheld laser scanning (HLS), reducing DBH estimation RMSE from 2.48 cm to 1.96 cm. In biomass modeling, individual tree LBI–AGB models were constructed and validated across 17 major tree species in China, Oceania, and Europe, yielding R² values of 0.79–0.95 and RMSE of 16.41–51.93 kg. To support large-area estimation, an individual tree-to-footprint scale inference (L-IFS) framework was developed for AGB estimation using spaceborne full-waveform LiDAR systems (TECIS and GEDI), achieving a maximum R² of 0.82 and R² ≥ 0.62 at complex terrain sites. By further integrating spaceborne LiDAR footprints with Sentinel-2 imagery through deep neural network modeling, continuous spatial mapping of forest carbon stocks was achieved across China and the United Kingdom, with cross-validation R² of 0.70–0.91. In dynamic forest monitoring, a Random Forest model was developed for pine bark beetle infestation prediction in Yunnan Province using monthly cloud-free Sentinel-2 imagery, achieving prediction accuracies of 0.88–0.90 from January to March 2023.

The UK research team focused on three principal study areas: the Forest of Dean, Aberfoyle, and Thetford Forest. Tree lists derived from high-resolution airborne LiDAR enabled more granular spatial mapping of dominant species within sub-compartments, supporting detailed forest inventories including tariffing, height and diameter distributions, gap fractions, and natural regeneration. These datasets were also applied to wind risk assessment using the ForestGALES model. Recognizing that allometric relationships vary considerably across the UK due to latitudinal and longitudinal gradients, extensive mobile laser scanning was undertaken to capture site-specific DBH and stem form, which were subsequently incorporated into locally adjusted allometric models for biomass and volume estimation. An area-based analysis using a 20×20 m grid was further applied to estimate site productivity expressed as Yield Class, by combining species data from Planet satellite time series, age data from the Sub-Compartment Database, and LiDAR-derived heights. This hybrid approach enables chronosequence construction for forecasting future forest increments, with planned refinements to incorporate competition and disturbance dynamics in collaboration with Oxford partners. Additionally, a Digital Twin platform is being developed to convert remote sensing outputs into a web-based application capable of displaying spatial data, generating reports, and running alternative management scenarios tailored to diverse end-user needs.

The Chinese and UK research strands are highly complementary. China has advanced scalable LiDAR processing, forest carbon mapping, and disturbance detection, while the UK has refined structural inventories, allometric calibration, productivity analysis, and risk modelling. Together, they demonstrate that combining LiDAR-derived structural data with spaceborne systems enables high-resolution biomass estimation, and that time-series observations can reveal growth trajectories, phenological variation, and disturbance signals critical for forest management decisions. In its second year, CEFO-2 has successfully transitioned from baseline inventory development towards dynamic monitoring, model integration, and practical application, laying robust foundations for precision forestry across diverse ecological and management settings.

This project aims to develop approaches to characterize forests and vegetated environments and to analyze their temporal evolution using multidimensional SAR data and techniques.

The project also seeks to exploit the recently launched ESA BIOMASS mission and to develop synergies with other sources of Earth Observation (EO) data. It proposes innovative multivariate P-band radar signal processing techniques capable of estimating key 3D forest features, as well as advanced approaches for monitoring vegetated areas at large scale using asynchronous time series of multi-physics data. These techniques are tailored to the specific characteristics of the BIOMASS mission and are applied to global mapping applications. A synergistic use of multi-source EO sensors will be developed to accurately monitor the temporal evolution of forests and to quantitatively assess their above-ground biomass.

Specific contributions related to the BIOMASS mission, as well as to other SAR-based applications, include advanced processing schemes for estimating forest height, above-ground biomass density, forest disturbance, and ground topography beneath dense forest cover. The consortium has also actively contributed to the calibration and validation (Cal/Val) of the mission. A rigorous theoretical framework, combined with a lightweight operational process, enables continuous forest monitoring over time, facilitating the detection of deforestation and disturbances related to forest fires.