[Paper Review] Estimation of boreal forest biomass from ICESat-2 data using hierarchical hybrid inference

Petri Varvia, Svetlana Saarela+6 more|arXiv (Cornell University)|Jul 10, 2023

Remote Sensing and LiDAR ApplicationsEnvironmental Science3 citations

TL;DR

This study proposes a hierarchical hybrid inference approach to estimate boreal forest above-ground biomass density (AGBD) from ICESat-2 lidar data, integrating uncertainty from multiple modeling stages including allometric models and proxy airborne lidar data. The method achieves a mean AGBD estimate of 63.2 ± 1.9 Mg/ha in the Nurmes validation area, with a relative standard error of 2.9%, closely matching the reference wall-to-wall ALS-based estimate of 63.9 ± 0.6 Mg/ha (1.0% relative standard error), confirming the method’s accuracy and robust uncertainty quantification.

ABSTRACT

The ICESat-2, launched in 2018, carries the ATLAS instrument, which is a photon-counting spaceborne lidar that provides strip samples over the terrain. While primarily designed for snow and ice monitoring, there has been a great interest in using ICESat-2 to predict forest above-ground biomass density (AGBD). As ICESat-2 is on a polar orbit, it provides good spatial coverage of boreal forests. The aim of this study is to evaluate the estimation of mean AGBD from ICESat-2 data using a hierarchical modeling approach combined with rigorous statistical inference. We propose a hierarchical hybrid inference approach for uncertainty quantification of the AGBD estimated from ICESat-2 lidar strips. Our approach models the errors coming from the multiple modeling steps, including the allometric models used for predicting tree-level AGB. For testing the procedure, we have data from two adjacent study sites, denoted Valtimo and Nurmes, of which Valtimo site is used for model training and Nurmes for validation. The ICESat-2 estimated mean AGBD in the Nurmes validation area was 63.2$\pm$1.9 Mg/ha (relative standard error of 2.9%). The local reference hierarchical model-based estimate obtained from wall-to-wall airborne lidar data was 63.9$\pm$0.6 Mg/ha (relative standard error of 1.0%). The reference estimate was within the 95% confidence interval of the ICESat-2 hierarchical hybrid estimate. The small standard errors indicate that the proposed method is useful for AGBD assessment. However, some sources of error were not accounted for in the study and thus the real uncertainties are probably slightly larger than those reported.

Motivation & Objective

• To develop a statistically rigorous method for estimating boreal forest above-ground biomass density (AGBD) from ICESat-2 lidar data.
• To account for uncertainty arising from multiple modeling steps, including allometric models and proxy airborne lidar data, using hierarchical inference.
• To validate the ICESat-2 AGBD estimate against a high-resolution wall-to-wall airborne lidar reference in a boreal forest setting.
• To quantify the uncertainty of the ICESat-2 AGBD estimate using a hybrid inference framework that combines design-based and model-based variance components.
• To demonstrate the feasibility and accuracy of ICESat-2 for large-scale AGBD monitoring in boreal forests with quantified uncertainty.

Proposed method

• The study employs a hierarchical hybrid inference framework that combines design-based sampling variance with model-based propagated uncertainty from proxy and satellite lidar models.
• A two-stage modeling process is used: first, an allometric model is trained on field plots in Valtimo to predict AGBD from tree-level measurements, then applied to 15×15 m subcells of ICESat-2 segments to produce proxy AGBD estimates.
• The covariance of the proxy AGBD model parameters is derived using the law of total variance and Taylor approximation, accounting for both sampling and prediction uncertainty.
• The ICESat-2 AGBD model is fitted using the proxy AGBD values as response variables, with uncertainty propagated through the Jacobian matrix of the nonlinear model and the covariance of the proxy estimates.
• The final AGBD estimate and its variance for the Nurmes validation area are computed using the Jacobian of the ICESat-2 model and the full covariance structure of the model parameters.
• A reference hierarchical model-based estimate is computed using wall-to-wall ALS and Sentinel-2 data, with variance estimated via the law of total variance applied to the model parameters and spatial averaging.

Experimental results

Research questions

• RQ1 Can a hierarchical hybrid inference approach accurately estimate boreal forest above-ground biomass density (AGBD) from ICESat-2 lidar data while quantifying uncertainty across multiple modeling stages?
• RQ2 How does the uncertainty in the ICESat-2 AGBD estimate compare to a high-resolution wall-to-wall airborne lidar reference estimate in terms of precision and coverage?
• RQ3 To what extent does the inclusion of propagated uncertainty from proxy models and allometric equations improve the reliability of satellite-based AGBD estimation?
• RQ4 Is the ICESat-2 AGBD estimate within the 95% confidence interval of the reference estimate, indicating statistical consistency?
• RQ5 What is the relative standard error of the ICESat-2 AGBD estimate, and how does it compare to the reference estimate’s precision?

Key findings

• The ICESat-2 estimated mean AGBD in the Nurmes validation area was 63.2 ± 1.9 Mg/ha, with a relative standard error of 2.9%.
• The reference wall-to-wall airborne lidar-based estimate was 63.9 ± 0.6 Mg/ha, with a much lower relative standard error of 1.0%.
• The ICESat-2 estimate fell within the 95% confidence interval of the reference estimate, indicating statistical consistency.
• The small standard errors suggest the proposed hierarchical hybrid inference method is effective for AGBD assessment with reliable uncertainty quantification.
• Despite the low reported uncertainty, the study acknowledges unaccounted error sources, implying the true uncertainty may be slightly higher than reported.
• The method successfully integrates uncertainty from multiple modeling stages—field data, allometric models, proxy ALS data, and ICESat-2 data—into a coherent statistical framework.