A Generative Trajectory Interpolation Method for Imputing Gaps in Wildlife Movement Data

Date:

Zijian Wan1, Somayeh Dodge1 1Department of Geography, University of California Santa Barbara, USA ## Abstract Advances in tracking technologies have resulted in growing repositories of large and long-term movement data of wildlife at an unprecedented rate. Nevertheless, many of these movement datasets come with missing records, termed gaps in this paper, which need to be imputed before further movement analysis. However, existing trajectory interpolation methods have certain limitations. Their effectiveness might be restrained by users’ domain knowledge of the moving entity or by the properties of the trajectories, to name a few. Moreover, the uncertainty of movement data has not received enough attention and is often neglected in the interpolation process. A review of existing literature suggests a need for designing more robust and broadly applicable data-driven interpolation methods that can self-adapt to the subject tracking data, and meanwhile, can take movement uncertainty into consideration. This study proposes a new trajectory interpolation model that leverages a generative adversarial network (GAN) architecture supported by long short-term memory (LSTM) layers to interpolate missing trajectory points. The model uses a latent code in addition to the noise input to deal with the uncertainty in movement behaviors. We apply and evaluate the proposed model against a real-world GPS trajectory dataset of migratory white storks to assess its effectiveness for imputing migration paths. ## Keywords Trajectory interpolation, movement modeling, uncertainty, generative adversarial network (GAN), long short-term memory (LSTM), GPS trajectories, wildlife tracking ## Session info 1st ACM SIGSPATIAL International Workshop on AI-driven Spatio-temporal Data Analysis for Wildlife Conservation (GeoWildLife ’23), ### About the Workshop In collaboration with ACM SIGSPATIAL, we are pleased to announce the call for papers for GeoWildLife 2023, a workshop dedicated to bridging the gap between AI-enabled spatio-temporal data analytics and wildlife conservation. The primary objective of this workshop is to advance the state-of-the-art in AI-driven spatio-temporal data analysis for wildlife conservation. By connecting computer scientists, geospatial scientists, ecologists, and conservation practitioners, the workshop seeks to promote interdisciplinary collaboration and drive real-world impact. Through a series of keynote presentations, panel discussions, and interactive sessions, participants will explore various topics including remote sensing technologies, predictive modeling, movement ecology, species distribution modeling, habitat quality assessment, and mitigating human-wildlife conflict. Special focus will be given to ethically and responsibly harnessing AI to ensure the sustainability of conservation efforts. Accepted papers will be included in the workshop proceedings, which will be published in the ACM Digital Library. At least one author of each accepted paper must register for the workshop and present the paper. We also offer authors the option to opt out of the proceedings. Such papers will be published on the workshop’s website and will not be considered archival for resubmission purposes. ### Call for Papers Topics of interest include but are not limited to: * **Remote Sensing, UAV Imagery, and Wildlife Monitoring:** AI interpretation of remote sensing data for tracking animal populations, identifying habitats, and monitoring ecosystem health. This includes techniques such as deep learning-based spatial data analysis. * **Predictive Modeling and Mobility Simulations for Conservation:** AI-driven models to predict species movements, population changes, and the effects of environmental changes on wildlife, leveraging high-performance spatio-temporal mobility simulation systems. * **Movement Ecology, Urban Mobility, and AI:** Using AI to understand species behavior and movement patterns based on spatial and temporal data in both natural and urban environments. * **Species Distribution and Semantic Trajectory Modeling:** Applications of AI in predicting and mapping the spatial distribution of different species under changing climate scenarios and leveraging generative models for semantic trajectory analysis. * **AI in Habitat Quality Assessment and Urban Landscape Analysis:** Employing AI algorithms to analyze, interpret spatial data, and evaluate urban habitats using augmented street-level imagery and points of interest. * **Mitigating Human-Wildlife and Human-Urban Conflicts:** Developing AI tools to predict and prevent conflicts between humans and wildlife and understanding spatio-temporal patterns in urban habitats. * **Spatio-temporal Data Analytics for Movement Ecology, Wildlife Monitoring, and Natural Disaster Prediction:** Integrating multi-output neural networks to predict natural phenomena like earthquakes, drought that could impact wildlife habitats. * **Conservation Planning, Decision Support Systems, and Urban Infrastructure Mapping:** Using AI to aid in spatial planning for conservation, identifying priority areas for protection, and mapping urban structures that might impact wildlife movement. * **AI, Ethics, Data Privacy, and Wildlife Conservation:** Delving into the ethical implications of using AI in wildlife conservation, including concerns related to privacy, surveillance, data ownership, and location-based biases. * **Socio-Political Implications of AI in Conservation and Urban Mobility:** Understanding the socio-political dimensions of AI-based conservation strategies and their impact on urban landscapes and communities. * **AI-driven Citizen Science, Crowd-sourced Data:** Leveraging AI-enhanced citizen science approaches in wildlife monitoring, conservation efforts, and understanding spatio-temporal patterns through narrative techniques. * **Urban Representation Learning and Urban Mobility Analysis in Conservation:** Applying graph neural networks and deep learning techniques to understand urban economic statuses and their implications for wildlife movement. * **Distributed Data Warehousing and Indexing for Wildlife Conservation:** Techniques like efficient storage and indexing for handling massive datasets related to wildlife monitoring and urban landscapes.