aiqbench analysis
MoveBench establishes a critical new standard for evaluating probabilistic trajectory forecasting by addressing the unique stochasticity and environmental constraints of wildlife movement data. The study reveals that deep learning models do not universally outperform simpler baselines, highlighting the importance of rigorous covariate selection and temporal generalization over individual adaptation. This benchmark provides an essential foundation for advancing machine learning methods in ecological conservation and global-scale movement prediction.
MoveBench: A Benchmark for Global-Scale Wildlife Movement Forecasting
arXiv:2609.15780v1 Announce Type: new Abstract: Understanding and predicting wildlife movement is critical for ecology and conservation. While trajectory forecasting has advanced for human and vehicle movement, wildlife trajectories present distinct challenges: they are unconstrained in space, highly stochastic, and influenced by environmental conditions. We introduce MoveBench, the first large-scale benchmark for probabilistic wildlife movement forecasting, containing 2.6M GPS locations from 800+ individuals across 110 species in 127 countries, paired with 1.6B environmental raster tiles capturing 160 covariates known or hypothesized to influence movement. We propose a probabilistic evaluation protocol for movement trajectory forecasts, addressing limitations of point-prediction metrics for inherently stochastic phenomena. Through comprehensive empirical evaluation of four method families across multiple temporal and spatial scales, we reveal that: (1) existing predictive methods generalize better to future timepoints than to unseen individuals, (2) deep learning approaches do not consistently outperform simpler baselines, and (3) environmental covariate selection significantly impacts performance. MoveBench enables standardized evaluation of movement forecasting methods and provides a foundation for methodological advances on this ecologically important task.
Authors:Justin Kay, Shir Bar, Ellen O. Aikens, Martin Becker, Francesca Cagnacci, Juliet Cohen, Scott W. Forrest, Jessica Kendall-Bar, Madeleine Lucas, Macon Overcast, Meredith S. Palmer, Will Rogers, Nicholas J. Russo, Christian Rutz, Larissa T. Beumer, Michael Brown, Ying-Chi Chan, Sarah C. Davidson, Diego Ellis Soto, Anne G. Hertel, Roland Kays, Benjamin Koger, Guram Mikaberidze, Thomas Mueller, Ruth Oliver, Thorsten Papenbrock, Robert Patchett, Jared A. Stabach, Dane Taylor, Scott W. Yanco, Sara Beery
Abstract:Understanding and predicting wildlife movement is critical for ecology and conservation. While trajectory forecasting has advanced for human and vehicle movement, wildlife trajectories present distinct challenges: they are unconstrained in space, highly stochastic, and influenced by environmental conditions. We introduce MoveBench, the first large-scale benchmark for probabilistic wildlife movement forecasting, containing 2.6M GPS locations from 800+ individuals across 110 species in 127 countries, paired with 1.6B environmental raster tiles capturing 160 covariates known or hypothesized to influence movement. We propose a probabilistic evaluation protocol for movement trajectory forecasts, addressing limitations of point-prediction metrics for inherently stochastic phenomena. Through comprehensive empirical evaluation of four method families across multiple temporal and spatial scales, we reveal that: (1) existing predictive methods generalize better to future timepoints than to unseen individuals, (2) deep learning approaches do not consistently outperform simpler baselines, and (3) environmental covariate selection significantly impacts performance. MoveBench enables standardized evaluation of movement forecasting methods and provides a foundation for methodological advances on this ecologically important task.
Machine Learning (cs.LG)
arXiv:2609.15780 [cs.LG]
(or arXiv:2609.15780v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.15780 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 255 words
Related coverage
Vertical Advantage: Transforming Industries with Lakebase and Agentic AI
Databricks · Aug 27, 2026
A New Era of Discovery: Google Research at I/O 2026
Google Research · May 28, 2026
DST-Net: Decomposed Spatio-Temporal Network for Time Series Forecasting
arXiv · Oct 17, 2026
Who’s afraid of the big, bad GPU?
The Verge · Jul 21, 2026