A time geography approach for data fusion of movement and environmental data: accounting for uncertainty of movement in resource selection models


Animal movement behaviour is influenced by a set of complex external and internal factors interacting at diverse temporal and spatial scales [1]. Changes in environmental conditions, such as wind, temperature and precipitation often lead to different movement behaviours in wildlife, reflected as specific spatio-temporal patterns in movement data. Identifying and understanding these patterns is one of the challenges for movement ecology [2].

Studying the relationship between environmental conditions and wildlife movement requires the use of trajectory annotation, a set of data fusion techniques combining movement locations with environmental data [3]. Environmental data used in trajectory annotation can originate from diverse sources, such as meteorological stations [4] and remote sensing satellites [5]. Current trajectory annotation methods use simple interpolation based on the nearest value in space and/or time [3]. This works well when movement trajectories are sampled at high temporal frequencies, but is problematic when the sampling rate is coarse. For example, weather radar data are typically collected at 5 min intervals, which is a finer frequency than most GPS tracking data for larger animals whose movement is potentially affected by precipitation – for these animals, tracking points are often collected at sparse temporal resolution of hours or days [6]. This introduces a considerable uncertainty in terms of how well the environmental data matches the patterns observed in wildlife tracking data. Therefore, in order to understand how environmental conditions shape wildlife movement, it is crucial that we account for this uncertainty when investigating use of resources and their effect on movement.

This is an interdisciplinary project between geographic information science and movement ecology, which will look at the potential of time geographic methods [7,8] to explicitly address uncertainty in sparsely-sampled animal movement data within the data fusion process.

Click on an image to expand

Image Captions

Deer_Weather.jpg – A camera trap collage of movement of deer in Scottish Highlands in varied weather conditions. Photos by Solene Marion, current PhD student at the University of St Andrews, supervised by UD and JL, used with her permission.


We will develop new methods for movement data fusion of wildlife tracking data with environmental data, using principles from time geography. We will explore the potential of modelling environmental conditions experienced between two recorded tracking locations within the accessible space, that is, the space within which the animal had had to move given the physics of its movement. The accessible space is represented using what is termed a space-time prism [9] and captures the space where movement was possible in two [8] or three physical dimensions [7] plus time. Because space-time prisms focus on the physics of movement, we will develop specific interpolation methods that model values of environmental data across these accessible spaces. This has the potential for new insights in wildlife tracking analyses, because it captures the range of environmental conditions experienced between recorded tracking data locations in a more realistic way.

Additionally, we will re-define the accessible space using the physical characteristics of the environmental source field. For example, while potential path volumes [7] (derived from space-time prisms in 4D) model the accessible space between two 3D observations of a moving object (e.g. a bird in flight) as an ellipsoid, this space can be stretched, twisted and distorted according to the forces exerted by the environmental conditions (such as the wind). We will use information on the wind field to model this distorted space and create a more realistic representation of accessible space, which takes into account the constraints posed to movement by the environment.

Our methods will be tested on simulated and real tracking data. We will have two case studies:
• CS1 – deer movement. We will link 2D GPS tracking data from white-tailed deer [10] to meteorological data (e.g. rainfall radar and other open weather data) to investigate deer’s responses to periodic and extreme weather events.
• CS2 – bird migration. We will link 3D GPS tracking data [7] to three-dimensional wind data (e.g. 3D wind profiles from ESA’s Aeolus satellite and other openly available data on atmospheric dynamics). This will contribute to the understanding how birds respond to external atmospheric factors to navigate across large-scale distances.

The final step will be validation of our new methods through a comparative analysis of resource selection processes: we will fit step-selection functions (SSF [11]) to fused movement-environment data from our methods and compare these to results of SSF models where data were integrated using the traditional nearest-in-space-time interpolation.

Studentship will be located at the University of St Andrews with co-supervision at the University of Stirling.

Anticipated project outcomes:
1. A set of new methods for data fusion of sparsely sampled movement trajectories and environmental data.
2. A completed PhD thesis.
3. Publications: 3 papers in GIScience/Movement Ecology journals and 3 conference papers at international conferences (GIScience, Geocomputation, Bio-logging).
4. Impact outcomes: implementation and publication of new methods as FOSS software

Project Timeline

Year 1

Student will perform an in-depth interdisciplinary literature review of trajectory annotation methods. He/she will start method development for case study 1, undertake training and present the work at an international conference.

Year 2

Student will develop new movement data fusion methods and write the first journal paper. He/she will continue analysis and method development for case study 1, undertake further training and present at one international conference.

Year 3

Student will finalise case study 1 with a further journal paper. He/she will work on case study 2 (analysis and method development). Training courses continued as well as one presentation at an international conference.

Year 3.5

The student will finalise case study 2 with a journal paper. He/she will write and submit the PhD thesis based on results of case studies and published papers.

& Skills

The student will, in collaboration with the supervisors, prepare a Training Needs Analysis
and identify where he/she needs to add skills. These will be acquired in GradSkills courses given by the Centre for Academic and Professional Development (CAPOD) at St Andrews, in the IAPETUS2 training events and in external courses, such as the AniMove science school (http://animove.org/) and the Association of the Geographic Information Laboratories of Europe (AGILE) PhD school.

An additional component of training will be through active supervision, consisting of regular meetings (live and virtual) of the supervisory team with the student. Meetings will be weekly in year 1, while later on the student will be expected to initiate meetings every 2 weeks (or as needed) in order to train to work increasingly independently. Student will further attend research meetings of the Bell-Edwards Geographic Data Institute (https://begin.wp.st-andrews.ac.uk/), of which he/she will be a member. These meetings run monthly on topics relevant to career progression (publishing, time management, proposal writing, etc.). Student’s progress will be evaluated annually in the PhD review process of the University of St Andrews.

References & further reading

1. Nathan R. et al, 2008. A movement ecology paradigm for unifying organismal movement research. PNAS, 105(49):19052-19059.
2. McClintock BT et al, 2014. When to be discrete: the importance of time formulation in understanding animal movement. Movement Ecology, 2:21.
3. Dodge S et al, 2013. The environmental-data automated track annotation (Env-DATA) system: linking animal tracks with environmental data. Movement Ecology, 1:3.
4. Shamoun-Baranes et al, 2010. Integrating Meteorology into Research on Migration. Integrative and Comparative Biology, 50(3):280-292.
5. Brum-Bastos VS et al, 2019, A systematic literature review of context-aware movement analysis in movement ecology. Under review.
6. De Groeve et al, 2016. Extracting spatio‐temporal patterns in animal trajectories: an ecological application of sequence analysis methods. Methods in Ecology and Evolution, 7(3):369-379.
7. Demšar U and Long JA, 2019. Potential path volume (PPV): a geometric estimator for space use in 3D. Movement Ecology, 7:14.
8. Long JA and Nelson T, 2012. Time geography and wildlife home range delineation. Journal of Wildlife Management, 76(2):407-413.
9. Miller H, 2005. A Measurement Theory for Time Geography. Geographical Analysis, 37(1):17-45.
10. Marantz S et al, 2016. Impacts of human hunting on spatial behavior of white-tailed deer (Odocoileus virginianus). Canadian Journal of Zoology, 94(12):853-861.
11. Thurfjell H et al. 2014. Applications of step-selection functions in ecology and conservation. Movement Ecology, 2:4.

Further Information

Supervision and fit with project. The two primary supervisors are Dr Urska Demsar (UD) at the School of Geography and Sustainable Development (SGSD) in St Andrews and Dr Thiago Silva (TS) in Biological and Environmental Sciences at the University of Stirling. There will be further two supervisors: Dr Vanessa Brum Bastos (VBB) at SGSD and Dr Jed Long (JL) at the Western University, Canada. The roles of each supervisor are:

UD (https://udemsar.com/) is Senior Lecturer in Geoinformatics. Her research interests are in spatio-temporal visual analytics and in particular in analysis of movement – a topic on which she is collaborating with movement researchers from other disciplines (movement ecologists, human-computer interaction specialists). To date she has supervised ten PhD students (six to completion). She will be the main supervisor and will support the student in research, career planning, publishing and all other aspects. She is available for further information at urska.demsar@st-andrews.ac.uk.

TS is Lecturer in Ecosystem Change and Environmental Informatics and runs the international Ecosystem Dynamics Observatory research group (https://tscanada.wixsite.com/ecodyn). His research program bridges ecology, geosciences and data science to understand how ecosystem spatial and termporal dynamics respond to natural environmental variability and anthropogenic-induced change. He is also keenly interested in developing new methods and technologies for ecosystem observation and monitoring. He will support the student in method development as well as in publishing and other career development aspects.

VBB (https://sgsup.asu.edu/vanessa-da-silva-brum-bastos) is currently postdoctoral researcher at the Arizona State University, but will start a new postdoctoral position at SGSD in Jan 2020, to work on a project on data science for geomagnetic bird navigation. She has a background in remote sensing and will support the student in case study 2, with experience in obtaining and modelling data from satellite sources.

JL (https://jedalong.github.io/) is Assistant Professor at the Department of Geography, Western University. He is interested in developing and applying novel methods for spatial and space-time analysis in the study of movement and has extensive experience working with spatial ecologists. He will participate in case study 1, where he will provide white-tailed deer data through his existing collaboration with the Noble Research Institute and support the student in method development for data fusion of deer data.

Apply Now