Shah et al. (2025) Constrained negentropy optimisation (CoNE-opt): Using independent components to merge satellite data products

Identification

• Journal: Remote Sensing of Environment
• Year: 2025
• Date: 2025-12-05
• Authors: Suraj Shah, Yi Liu, Seokhyeon Kim, Ashish Sharma
• DOI: 10.1016/j.rse.2025.115170

Research Groups

• School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia
• Department of Civil Engineering, College of Engineering, Kyung Hee University, Yongin, Republic of Korea

Short Summary

This paper introduces Constrained Negentropy Optimisation (CoNE-opt), a novel method for merging uncertain geophysical datasets by maximizing non-Gaussianity. CoNE-opt outperforms traditional merging techniques, particularly in the presence of high error cross-correlation and outliers, as demonstrated by superior performance in merging global satellite-derived surface soil moisture products.

Objective

• To propose and validate Constrained Negentropy Optimisation (CoNE-opt), a novel data merging technique that assumes the true dataset exhibits the greatest departure from Gaussianity, for combining multiple uncertain geophysical datasets.

Study Configuration

• Spatial Scale: Global
• Temporal Scale: Not explicitly stated for the study period, but applied to global satellite-derived products, implying continuous or long-term observation capabilities.

Methodology and Data

• Models used: Constrained Negentropy Optimisation (CoNE-opt) based on negentropy maximization and a linear error model constraint; principles of Independent Component Analysis (ICA). Compared against Ordinary Least Squares (OLS)-based merging methods.
• Data sources: Synthetic experiments; three global satellite-derived surface soil moisture products (SMAP, SMOS, and SMOS-IC); reference datasets for validation.

Main Results

• CoNE-opt successfully maximizes negentropy as an objective function while incorporating a linear error model constraint, enabling joint estimation of the full error spectrum and merging weights.
• The method significantly outperforms existing merging alternatives that rely on second-order error statistics.
• CoNE-opt demonstrates improved error magnitudes, especially in scenarios with high error cross-correlation (ECC) and outliers, where OLS-based methods typically struggle.
• In surface soil moisture merging, CoNE-opt achieved an overall normalised Root Mean Square Error (RMSE) of 0.15, which is substantially better than the 0.34 observed for the best OLS-based alternative.
• The merged product consistently surpassed the performance of individual input products and existing merging methods.

Contributions

• Introduction of a novel data merging framework, CoNE-opt, which leverages negentropy maximization and Independent Component Analysis (ICA) principles to address limitations of traditional methods.
• Offers a robust solution for merging geophysical data by jointly estimating the full error spectrum and merging weights without restrictive assumptions like zero error cross-correlation.
• Provides a significant advancement for generating high-quality global soil moisture datasets, thereby enhancing land surface and hydrological modeling, particularly in data-sparse regions.
• Demonstrates superior performance over existing methods, especially under challenging conditions characterized by high error cross-correlation and outliers.

Funding

Not specified in the provided text.

Citation

@article{Shah2025Constrained,
  author = {Shah, Suraj and Liu, Yi and Kim, Seokhyeon and Sharma, Ashish},
  title = {Constrained negentropy optimisation (CoNE-opt): Using independent components to merge satellite data products},
  journal = {Remote Sensing of Environment},
  year = {2025},
  doi = {10.1016/j.rse.2025.115170},
  url = {https://doi.org/10.1016/j.rse.2025.115170}
}