Uncertainty Quantification for EO Regression Tasks: Building Height, Tree Canopy Height and Above-ground Biomass Estimation

2026-07-13

AI Summary

This study addresses the widespread lack of reliable uncertainty quantification in Earth observation regression tasks—such as building height, canopy height, and aboveground biomass estimation—which hinders their application in domains like urban planning and climate policy that demand both accuracy and trustworthiness. To this end, the authors integrate year-round Sentinel-1 SAR and Sentinel-2 MSI time series and propose two complementary approaches: Gaussian uncertainty modeling for heteroscedastic errors and quantile regression for asymmetric error structures. This work delivers, for the first time, well-calibrated and interpretable pixel-level uncertainty estimates at 10-meter resolution across multiple regression tasks. Experimental results demonstrate that the proposed models match or exceed the performance of existing deterministic baselines and global products across all three tasks, and notably outperform current state-of-the-art uncertainty-aware methods in canopy height estimation.

Abstract

Earth Observation regression tasks such as building height, canopy height, and above-ground biomass estimation underpin critical applications in urban planning, forest monitoring, and climate policy, where both accuracy and reliability are critical. Yet most deep learning models yield only deterministic predictions, providing no indication of per-pixel reliability. These regression tasks are inherently challenging due to heterogeneous land surfaces, skewed target distributions, sensor noise, and signal saturation at high target values, making uncertainty (UC) estimation essential for reliable inference. We address this gap by modeling aleatoric uncertainty using year-long Sentinel-1 SAR and Sentinel-2 MSI time series, proposing two complementary approaches: (i) Gaussian UC, which jointly predicts mean and standard deviation under a Gaussian assumption, and (ii) Quantile UC, which estimates the 10th, 50th, and 90th quantiles to capture asymmetric and heteroscedastic error distributions. Both models are evaluated on three representative EO regression tasks at 10 m spatial resolution. Results show that both approaches match or surpass deterministic benchmarks and existing global products, while delivering well-calibrated, interpretable, and operationally useful confidence estimates. Notably, both models outperform the current 10 m state-of-the-art uncertainty-aware model for canopy height estimation. Our implementation will be available at: https://github.com/RituYadav92/EO-Regression-Uncertainty-Estimation

Problem

• Uncertainty Quantification
• Earth Observation
• Regression Tasks
• Aleatoric Uncertainty
• Remote Sensing

Innovation

• Uncertainty Quantification
• Earth Observation
• Aleatoric Uncertainty
• Quantile Regression
• Deep Learning