Workshop Report on Hidden Markov Models for Animal Movement Data
Posted by Dr Enrico Gennari on March 27, 2016
The workshop was held by Prof. Roland Langrock and PhD student Théo Michelot (Universities of Bielefeld, Sheffield, and St Andrews) in Mossel Bay, 7-9 March 2016.
We first would like to thanks the National Research Foundation (NRF IRG - UK / South Africa Researcher Links Grants for Travel and Hosting of Scientific Events), and the South African Institute for Aquatic Biodiversity (SAIAB).
Keep up-to-date with all the latest Oceans Research marine science activities on our Facebook page
WORKSHOP PROGRAM:
Monday the 7th March
09:00-10:20 Theory: Motivation, preliminaries & basic model formulation
10:40-12:00 Practical session using R: loading and visualizing real data/generating artificial data
(Lunch)
13:45-14:45 Theory: Fitting an HMM to data
15:00-16:00 Practical session using R: fitting basic HMMs
Tuesday the 8th March
09:00-10:20 Theory: Model selection & model checking
10:40-12:00 Practical session using R: finding an appropriate model
(Lunch)
13:45-14:45 Theory: state decoding
15:00-16:00 Practical session using R: decoding the hidden states
Wednesday the 9th March
09:00-10:20 Theory: incorporating covariates into HMMs
10:40-12:00 Practical session: investigating covariate influence
(Lunch)
13:45-14:45 Theory: Extensions of the basic model formulation
15:00-16:00 Open discussion
In addition, as per grant requirement, Prof. Langrock gave a lecture on the ecological applications of Hidden Markov Models to an audience of postgraduates, post doctorates, as well as lecturers, at the NMMU Saasveld Campus in George the 10th March 2016.
Review from Dr XXXX:
Interests: conservation biologist, undertaking research on birds of the fynbos biome.
Positive aspects: I attended the workshop as I have an interest in statistical techniques, and I was wondering about applying the HMM methods to mark-recapture data. It was clear that another dataset we had on bird movements was more applicable, and it was fun applying the methods to our data through the moveHMM package. It is always a challenge for ecologists to transcend the gap between data and statistical techniques, which was recognised by Roland (likely because of his previous work with ecologists), and who tailored his explanations accordingly to the audience. He was patient and constantly monitoring the level of understandings of the audience. Theo was inspirational: amazing to see how someone so young (by my standards) has achieved so much. The moveHMM package is a good tool for implementing the techniques learnt in this course.
Improvements: Nothing comes to mind. It was clear that the organisers and presenters were flexible enough to deal with issues as they arose.
Review from Mr. XXXX:
Interests: My current research interests involve using acoustic telemetry to understand the habitat use and migratory patterns of fish and shark species in southern Angola, I am however involved in other studies that require the ageing of fish.
Positive aspects: Great workshop presented in a very basic manner. Roland and Theo both explained the modeling concepts so that every person who attended, irrespective of their previous modeling knowledge would understand the concepts brought forward.
Improvements: I do not think there is much that needs to be improved: an organised social event such as a braai during the event for networking purposes would have been a bonus.
Review from Dr XXXX:
Interests: N/A.
Positive aspects: The workshop was well taught with the lead lecturer clearly being a global leader in the field. It was well paced and mix of taught and practical was well balanced.
Improvements: There was perhaps a little too much focus on a single package and application of Hidden Markov models, but given the amount of background material that needed to be covered, this was probably a good thing.
Review from Dr XXXX:
Interests: understanding environmental and biological drivers of marine animal movement, with a specific focus on large sharks.
Positive aspects: Overall, I found the workshop helped progress my understanding of the application of Hidden Markov Models to my data. It gave me an overview of the possibilities, as well as the limitations. It also provided a much needed refresher on general statistical methodologies, and even helped clarify some overall concepts in statistics which previously I had not understood that well. I did feel that for me personally, some of the information was too technical and I could not follow, but I did make good contacts with statisticians in the group for future collaborations. The case studies were the most interesting to me.
Improvements: None
Review from Dr XXXX:
Interests: I am a statistician and am interested in working with others on analysing their data. I am fairly familiar with these techniques although have not done any publishing using them as yet.
Positive aspects: The workshop was really well presented and the theory was presented at the right level. The practical sessions were useful. Both Roland and Theo were very approachable, friendly and helpful.
Improvements: If there is a future session you might consider having a half day session where people real data could be analysed as well.
Review from Miss XXXX:
Interests: My research interests include the application of Independent Mixture models and Hidden Markov models to animal movement using GPS data to infer behavioural states of the animal and answer possible ecological questions. .
Positive aspects: I thoroughly enjoyed the workshop and would like to be a part of something similar again in the future. I felt that the material was presented and explained in a clear understandable manner that not only the statisticians could follow and understand. The learning environment was nice and relaxed and the presenters were really friendly and approachable. I enjoyed the practical sessions and felt like we got a good feel of how to implement the package to real data. The structure of the workshop worked well in the sense that the theory didn't get too overwhelming and the practical sessions kept our interest. I would like to thank the organiser again for a really well organised workshop and thank the presenters for giving such a high quality presentation.
Improvements: If there is a future session you might consider having a half day session where people real data could be analysed as well. The only area of improvement would be to include more practical examples especially when fitting the more complex models covered on the final day.
Review from Mr. XXXX:
Interests: My interests lie with the ecological interactions surrounding the dynamics between herbivory and fire, this makes up my M-tech thesis at NMMU. I am looking into this dynamic by using movement data on a number of species in combination with fire history records to determine population growth, habitat preference and predator avoidance.
Positive aspects: Thank you for organizing this workshop. I really enjoyed the workshop, in that it was highly informative and very well presented, furthermore the topic was interesting and I feel that it could be greatly beneficial to my project.
Improvements: My criticisms are that I often felt lost due to the fact that I was not familiar with either R or HMM’s, however this may be my fault. My other comment is that I thrive on practical experience and although the example supplied for the tutorials worked well, when I tried to substitute it for my own data, there were a number of real world problems that had to be addressed. I feel that some of these issues could have been addressed in the course, for example working with data that contained zero.
Review from Mr. XXXX:
Interests: quantitative ecology with a specific interest in animal movement.
Positive aspects: extremely knowledgeable instructors. The first lectures provided a great theoretic understanding of the methods and made the ideas and motivations behind the more complex extensions easy to grasp. Overall, I think the best workshop we've hosted yet.
Improvements: Not everyone might agree but I would have liked to see more custom programming of HMM from scratch, and I understand this was an introductory course.
Review from Miss XXXX:
Interests: my research interests are still very introductory but are centered around modelling animal behaviour using different statistical methods (such as Independent Mixture models, Hidden Markov Models and Bayesian State-space Models). I have been exposed to some capture-recapture methods recently and would like to include GPS tracking data for more accurate results.
Positive aspects: I can only say positive things regarding the workshop that was organised. It was structured well by including a practical component after each theory section completed with enough time to apply what was learnt to our own data. The learning experience was of a high standard as Roland went into detail to explain any content in a manner understandable to all and not only statisticians, while Theo was always available for assistance when something was unclear. All material covered was openly given to those that attended to always refer back to in the future (which is already becoming of assistance). Lastly, the environment that we worked in was a relaxed and open one, where questions and concerns could be shared to the group.
Improvements: I cannot recall anything that needed to change. It was an all-round great experience and I must thank all that contributed to ensure the success of this workshop. I look forward to hearing of any more that are organised.
Review from Mr. XXXX:
Interests: My research interests are monitoring and modelling animal movements in estuaries especially where there has been an increase of pollutants. Basically the triggers of movement of a specific population and the ones that stay in the area.
Positive aspects: the workshop was an introduction to R software and HMM. Also Roland and Theo were amazing. They know their work very well. Group participation was amazing and we had people who are specialists in their fields so examples that were brought up were relevant to the workshop.
Improvements: there was nothing negative.
Review from Miss XXXX:
Interests: My interest was on the models that I can use for my project on latent movements of predators.
Positive aspects: The workshop was very helpful in understanding how I can apply the model to my data, also all the codes that I should need. The workshop was very inspiring, brought about many more ideas to work on and looking forward to. I would be more than keen to attend those workshops again, meeting people from different field of work and learning from them as well was very interesting.
Also am very thankful to Roland and Theo, it was an amazing workshop! WELL DONE !
Improvements: None
Review from Dr XXXX:
Interests: Animal movement modelling, state-space models and hidden Markov models for ecological time series, habitat use of large marine vertebrates, animal telemetry, statistical ecology.
Positive aspects: Being taught by experts in the field of animal movement modelling, meeting the community of scientists who use these methods in South Africa.
Improvements: None
Review from Mr. XXXX:
Interests: Movement ecology, dietary analysis, reef fish community structure.
Positive aspects: The workshop provided an excellent introduction into Hidden Markov Models, and explained very well the principles underlying the statistical model very well.
Improvements: While the main focus of the workshop was limited to gps tracking data, the course did briefly go into other areas where HMMs are useful. I would have liked to see an expansion of this side, where more information is provided on other useful aspects of HMMs. Additionally, the practicals were quite limited. One could copy-paste the entire code from the example into R, and be provided with all the answers. A more diverse set of questions during the practical would have been more interesting, and would allow for practitioners to play with various aspects of the R package.
Review from Miss XXXX:
Interests: Personality within the predator-prey context
Positive aspects: Great explanation of the statistical methods, of the mathematics and theory supporting the processes and formulas. It was very useful to be guided through some examples where this statistical approach was used to target ecological questions, and through different areas where this methods can be applied other than the direct analysis of animal movement. I now have a fairly well understanding of how I could analyze a data set in R using this package.
Improvements: On the first day, I found it a little more challenging to follow the workshop as the explanation of the subject was very technical. Such problem was readily improved on the following days, as Roland adjusted the way he presented the material to suit an audience of ecologists with different statistical backgrounds.
Review from Mr. XXXX:
Interests: fisheries management
Positive aspects:
- Roland has deep understanding of statistics in general and in particular Markov which forms basis of the method HMM.
- Demonstrating the derivation of the HMM method from first principles was the most illuminating thing for me. It helped my understanding much better.
- Testing alternative assumptions of the model such as assumed distributions, underlying properties of the method(such as discrete or continuous state), he demonstrated to us the flexibility of the method
- The clarity with which he presented relative complex ideas into understandable concepts was greatest point of the workshop.
- Demonstrations of the method through applied examples from different ecological spheres was another major highlight of the workshop
- The ability to implement the method using R either using the package Theo and Roland developed or showing how to write the implementation of the method from scratch was the highlight (although we did not have enough time to have a look at R code for fully implemented procedure, it was interesting to see how the likelihood is implemented in R, this demonstrate the level of understanding of the concepts by these guys)
Improvements:
- I did not have any negative comments
- What I would like going forward is maybe a greater platform for collaboration going forward where we can try and apply these methods on our data and perhaps we can have mini-symposium next year to see how the method performed on our various data bases.
Review from Dr XXXX:
Interests: animal movement ecology.
Positive aspects: I have attended several movement ecology / statistics workshops, and this was one of the best ones. The focus on a single analysis (with increasing layers of complexity) was good because it allowed delegates to fully understand the method. The course instructors were of exceptional caliber, and it was a privilege to learn from such skilled scientists. The length of course was appropriate, materials received were excellent, and the size of the audience was also well suited to the course. Finally I thank the workshop organizers, hosts, and the funders of this workshop. I learnt a lot, and now feel entirely capable of undertaking some of the learnt analyses on my own data.
Improvements: I have very few criticisms of this successful event. Perhaps the only difficulty I personally had was following the very complicated statistical explanations, particularly on the first day. I recognize this was unavoidable given the mixed audience of statisticians and ecologists (I being an ecologist). There is no easy resolve given that this is an interdisciplinary field, and perhaps the more statisticians and ecologists work together (in workshops such as these), the closer the divide will become in time.
Lastly, I, as organizer of this workshop, and on behalf of all the attendees, would like to thank NRF and SAIAB for having allowed this workshop to happen. It brought people from different parts of South Africa to meet, discuss and start conversations about collaborations.
It allowed South African Master and PhD students, as well as Post-Docs to fast track their understanding of ecological modeling, and therefore speeding up their competitiveness in the scientific realm.
A big thanks to the two presenters as they did not hesitate to come to South Africa and use their vast knowledge and skills as fantastic lecturers, to open a new research avenue for South African on modeling animal behavior using movement data, as well as powerful statistical tools.
Kindest regards
Dr Enrico Gennari
Keep up-to-date with all the latest Oceans Research marine science activities on our Facebook page
The 21 attendees were:
NAME | ORGANISATION | COUNTRY |
Mr Dylan Irion | Oceans Research | US / South Africa |
Mr Ralph Watson | Rhodes University | South Africa |
Dr Enrico Gennari | Oceans Research | Italy / South Africa |
Valentina Scarponi | Sussex University | UK / South Africa |
Dr Theoni Photopoulou | University of Cape Town / SEEC | South Africa |
Alexander Winkler | Rhodes University | South Africa |
Dr Alan Lee | University of Cape Town | South Africa |
Dr Simon Elwen | University of Pretoria / SeaSearch | South Africa |
Miss Monique Lambscher | University of Pretoria | South Africa |
Mr Christopher Brooke | Nelson Mandela Metropolitan University | South Africa |
Miss Francis Martens | Nelson Mandela Metropolitan University | South Africa |
Mr Lungisani Mbhele | Bayworld Aquarium | South Africa |
Dr Allan Clark | University of Cape Town | South Africa |
Dr Alison Kock | University of Cape Town / Shark Spotters | South Africa |
Tamlyn Engelbrecht | University of Cape Town | South Africa |
Miss Bracken van Niekerk | Nelson Mandela Metropolitan University | South Africa |
Miss Lee Watchurst | Nelson Mandela Metropolitan University | South Africa |
Miss Jenicca Poongavanan | Nelson Mandela Metropolitan University | South Africa |
Dr Linda Harris | Nelson Mandela Metropolitan University | South Africa |
Mr Sobahle Somhlaba | Department of Agriculture, Forestry and Fisheries | South Africa |
Dr Trevor McIntyre | University of Pretoria | South Africa |