A Proxy Consistency Loss for Grounded Fusion of Earth Observation and Location Encoders

TL;DR

This paper introduces a proxy consistency loss (PCL) that enhances Earth observation models by integrating proxy geographic data through a trainable location encoder, improving both in-sample and out-of-sample predictions.

Contribution

The paper proposes a novel proxy consistency loss for training a flexible location encoder that effectively incorporates proxy data, outperforming traditional fusion methods.

Findings

Outperforms existing fusion strategies in air quality prediction.

Enhances generalization to regions without training labels.

Improves robustness with limited labeled data.

Abstract

Supervised learning with Earth observation inputs is often limited by the sparsity of high-quality labeled or in-situ measured data to use as training labels. With the abundance of geographic data products, in many cases there are variables correlated with - but different from - the variable of interest that can be leveraged. We integrate such proxy variables within a geographic prior via a trainable location encoder and introduce a proxy consistency loss (PCL) formulation to imbue proxy data into the location encoder. The first key insight behind our approach is to use the location encoder as an agile and flexible way to learn from abundantly available proxy data which can be sampled independently of training label availability. Our second key insight is that we will need to regularize the location encoder appropriately to achieve performance and robustness with limited labeled data. Our experiments on air quality prediction and poverty mapping show that integrating proxy data implicitly through the location encoder outperforms using both as input to an observation encoder and fusion strategies that use frozen, pretrained location embeddings as a geographic prior. Superior performance for in-sample prediction shows that the PCL can incorporate rich information from the proxies, and superior out-of-sample prediction shows that the learned latent embeddings help generalize to areas without training labels.