Gap Filling of Biophysical Parameter Time Series with Multi-Output Gaussian Processes

TL;DR

This paper demonstrates that multi-output Gaussian Process models effectively fill gaps in biophysical parameter time series, such as LAI and fAPAR over rice, outperforming single-output models especially with high missing data.

Contribution

The study introduces and evaluates multi-output Gaussian Process models based on the linear model of coregionalization for biophysical parameter gap filling, highlighting their ability to transfer information across domains.

Findings

MO-GP models outperform SO-GP in high missing data scenarios.

MO-GP models successfully predict LAI and fAPAR variables.

Implicit information transfer improves gap filling accuracy.

Abstract

In this work we evaluate multi-output (MO) Gaussian Process (GP) models based on the linear model of coregionalization (LMC) for estimation of biophysical parameter variables under a gap filling setup. In particular, we focus on LAI and fAPAR over rice areas. We show how this problem cannot be solved with standard single-output (SO) GP models, and how the proposed MO-GP models are able to successfully predict these variables even in high missing data regimes, by implicitly performing an across-domain information transfer.