EarthShift: a benchmark for measuring robustness to real-world distribution shifts in Earth observation

• URL: http://arxiv.org/abs/2605.29330v1
• Date: Thu, 28 May 2026 04:06:30 GMT
• Title: EarthShift: a benchmark for measuring robustness to real-world distribution shifts in Earth observation
• Abstract summary: We introduce EarthShift: the first public testbed for benchmarking robustness across realistic distribution shifts in remote sensing.<n>Our experiments show GFMs consistently perform 15-20% worse out-of-distribution on average regardless of model architecture, size, pre-training or fine-tuning strategy.<n>This highlights a need for future research to strive for improvements in distributional robustness, not just performance, which can be benchmarked using EarthShift.
• Score: 10.357172510221021
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Current Earth observation benchmarks focus on measuring performance on diverse tasks and applications, typically measuring generalization in-distribution. But when models are deployed, they must generalize to myriad out-of-distribution scenarios, such as new time periods, geographies, scales, and sensors. We introduce EarthShift: the first public testbed for benchmarking robustness across multiple realistic distribution shifts encountered in remote sensing. EarthShift enables users to measure distributional robustness by comparing performance in- and out-of-distribution using datasets from paired datasets from different sources, temporal windows, geographic locations, and sensors. Our experiments on 8 geospatial foundation models (GFMs) and 11 tasks covering 5 shift types show that GFMs consistently perform 15-20% worse out-of-distribution on average regardless of model architecture, size, pre-training or fine-tuning strategy. We show that GFM robustness is similar to that of generic vision foundation models, and even fully-supervised models. This highlights a need for future research to strive for improvements in distributional robustness, not just performance, which can be benchmarked using EarthShift. We release our code and datasets to provide a testbed to guide future work to create foundation models that are robust and reliable in real-world applications. Code and data for EarthShift are available at: https://earthshift.github.io

Related papers

• Earth System Foundation Model (ESFM): A unified framework for heterogeneous data integration and forecasting [52.53742727060265]
  Foundation models (FMs) for the Earth system learn statistical relationships between physical variables across massive datasets.<n>Here, we introduce Earth System Foundation Model (ESFM), a fully open model building on the 3Dwin UNet backbone of the pioneering model.<n>We present findings using dense gridded data (ERA5, CMIP6), regionally masked dense data, sparse gridded MODIS satellite data, and station data.<n>Results demonstrate competitive or superior performance relative to state-of-the-art benchmarks.
  arXiv  Detail & Related papers  (2026-04-20T11:40:39Z)
• TerraFM: A Scalable Foundation Model for Unified Multisensor Earth Observation [65.74990259650984]
  We introduce TerraFM, a scalable self-supervised learning model that leverages globally distributed Sentinel-1 and Sentinel-2 imagery.<n>Our training strategy integrates local-global contrastive learning and introduces a dual-centering mechanism.<n>TerraFM achieves strong generalization on both classification and segmentation tasks, outperforming prior models on GEO-Bench and Copernicus-Bench.
  arXiv  Detail & Related papers  (2025-06-06T17:59:50Z)
• Benchmarking Object Detectors under Real-World Distribution Shifts in Satellite Imagery [9.723791642707738]
  Domain Generalisation (DG) seeks to bridge the gap by enabling models to generalise to Out-Of-Distribution (OOD) data without access to target distributions during training.<n>We examine the generalisability and robustness of state-of-the-art object detectors under real-world distribution shifts.<n>We introduce Real-World Distribution Shifts (RWDS), a suite of three novel DG benchmarking datasets.
  arXiv  Detail & Related papers  (2025-03-24T23:04:06Z)
• Towards Understanding Link Predictor Generalizability Under Distribution Shifts [34.58496513149175]
  We introduce a novel splitting strategy, known as LPShift, to induce a controlled distribution shift.<n>We verify LPShift's effect through empirical evaluation of SOTA LP models on 16 LPShift variants of original dataset splits.
  arXiv  Detail & Related papers  (2024-06-13T03:47:12Z)
• Wild-Time: A Benchmark of in-the-Wild Distribution Shift over Time [69.77704012415845]
  Temporal shifts can considerably degrade performance of machine learning models deployed in the real world. We benchmark 13 prior approaches, including methods in domain generalization, continual learning, self-supervised learning, and ensemble learning. Under both evaluation strategies, we observe an average performance drop of 20% from in-distribution to out-of-distribution data.
  arXiv  Detail & Related papers  (2022-11-25T17:07:53Z)
• AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection [7.829710051617368]
  We introduce an unsupervised anomaly detection benchmark with data that shifts over time, built over Kyoto-2006+, a traffic dataset for network intrusion detection. We first highlight the non-stationary nature of the data, using a basic per-feature analysis, t-SNE, and an Optimal Transport approach for measuring the overall distribution distances between years. We validate the performance degradation over time with diverse models, ranging from classical approaches to deep learning.
  arXiv  Detail & Related papers  (2022-06-30T17:59:22Z)
• Extending the WILDS Benchmark for Unsupervised Adaptation [186.90399201508953]
  We present the WILDS 2.0 update, which extends 8 of the 10 datasets in the WILDS benchmark of distribution shifts to include curated unlabeled data. These datasets span a wide range of applications (from histology to wildlife conservation), tasks (classification, regression, and detection), and modalities. We systematically benchmark state-of-the-art methods that leverage unlabeled data, including domain-invariant, self-training, and self-supervised methods.
  arXiv  Detail & Related papers  (2021-12-09T18:32:38Z)
• Evaluating Predictive Uncertainty and Robustness to Distributional Shift Using Real World Data [0.0]
  We propose metrics for general regression tasks using the Shifts Weather Prediction dataset. We also present an evaluation of the baseline methods using these metrics.
  arXiv  Detail & Related papers  (2021-11-08T17:32:10Z)
• WILDS: A Benchmark of in-the-Wild Distribution Shifts [157.53410583509924]
  Distribution shifts can substantially degrade the accuracy of machine learning systems deployed in the wild. We present WILDS, a curated collection of 8 benchmark datasets that reflect a diverse range of distribution shifts. We show that standard training results in substantially lower out-of-distribution than in-distribution performance.
  arXiv  Detail & Related papers  (2020-12-14T11:14:56Z)
• BREEDS: Benchmarks for Subpopulation Shift [98.90314444545204]
  We develop a methodology for assessing the robustness of models to subpopulation shift. We leverage the class structure underlying existing datasets to control the data subpopulations that comprise the training and test distributions. Applying this methodology to the ImageNet dataset, we create a suite of subpopulation shift benchmarks of varying granularity.
  arXiv  Detail & Related papers  (2020-08-11T17:04:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.