A primer on POMDPs
POMDPs (Partially observable Markov decision processes) are powerful decision models that allow users to make decisions under imperfect observations over time. This primer provides a much-needed entry point to ecologists.
POMDPs (Partially observable Markov decision processes) are powerful decision models that allow users to make decisions under imperfect observations over time. This primer provides a much-needed entry point to ecologists.
We’re excited to be able to offer a PhD top-up scholarship ($7k/year) to work with the Conservation Decisions Team! Dr Sam Nicol is partnering with the amazing Professor Richard Fuller from the Fuller Lab at UQ to find a student to work on Global cooperation for conserving migratory species.
Please note that successful candidates will need to apply for an obtain a scholarship from UQ to be eligible for the top-up. Application details for scholarships close 17 Jan 2020 for international students; 27 March 2020 for domestic students. See the post below for more details.
Many migratory species are in rapid decline and in need of effective conservation action. Interconnections between habitats used during migration mean that threats in any one part of an annual cycle can affect the entire population of a migratory species. Environmental management actions for migratory species therefore need to be coordinated across habitat types, seasons, and jurisdictions – yet this type of coordination has proven frustratingly elusive in practice. For example, the Convention on Migratory Species is signed by 125 countries, yet 91% of the world’s migratory birds are inadequately covered by protected areas across their annual cycle. There is a need to better understand where migrants move, how nations cooperate and the benefits of additional cooperation to protect migrants from extinction. This project will investigate these issues by measuring migratory connectivity and using the information gained to suggest improvements to the way migratory species are conserved. The methods that are developed will be tested on global datasets and used to provide guidance for international cooperation to better protect migratory species.
Top-up Scholarship Offer
We are looking for a motivated student to work with The Fuller Lab at UQ and the Conservation Decisions Team at CSIRO on an exciting new joint project. A top-up scholarship funded by CSIRO is available for this project for one PhD student contingent on the student receiving a UQ scholarship, an Australian Government Research Training Program (RTP) scholarship, or some similar domestic or international scholarship. We will work with the successful candidate to help put together their scholarship application. This top-up provides additional stipend ($7,000 per annum for 3.5 years) and generous support costs (including travel). UQ scholarship round dates can be found here: https://graduate-school.uq.edu.au/scholarships
The ideal candidate will need to have a strong interest in the conservation of migratory species. Excellent written and oral presentation skills are essential and the candidate will need to have or develop skills in data management and spatial analysis. Modelling and programming skills in at least one language (such as R), will be a distinct advantage. Prospective candidates with training in population ecology, conservation biology, spatial analysis or statistics are particularly desirable, but strong candidates with relevant skills from other quantitative disciplines are also encouraged to apply.
For more information contact Professor Richard Fuller (email@example.com) or Dr Sam Nicol (firstname.lastname@example.org)
For more details on research environment visit
Featured Image: Migrating Sandhill Cranes, Alaska USA. Image courtesy of: May-Le Ng
Invasive plant species threaten both agriculture and biodiversity globally and require ongoing management to minimise their impacts. However, the large number of invasive species means that prioritisation is required to ensure that limited resources are directed to managing species with both a high risk of impact as well as a good chance of management success. In our latest two collaborative papers, we used an expert-driven, risk-based approach to generate a State-wide priority list of invasive species for Queensland, Australia. Both papers were led by Dr Olusegun Osunkoya (Biosecurity QLD-DAF) and involved a number of other co-authors, including Dr Jens Froese (CSIRO Health and Biosecurity).
In our first study, we asked experts to assign risks to ~100 species considered by practitioners to pose significant risks across regions and likely to be management priorities. Risks were assessed using an impact metric that combined a measure of invasiveness (measured using the proportion of local government areas containing the species that also listed the species as a priority for management) and expert-generated severity scores. We found that the 5 invasive plants with highest impacts were: Parthenium, prickly acacia, rubber vine, Parkinsonia, and bellyache bush. We also carried out some regional analysis: while the invasive plant communities of most coastal regions were similar, the regions in western Queensland and the Torres Strait Island region had distinctly different invasive communities. This suggests that while there are many opportunities for common invasive species control plans, distinct approaches may be required in the west of the State and in the Torres Strait Island region.
While our first study prioritised species according to impacts, our second study estimated the feasibility of management for 63 of the species prioritised in the first study. We used an expert elicitation approach to estimate the feasibility of three management approaches (biocontrol, mechanical and chemical control) in terms of cost, effectiveness and practicality. Across all species, chemical control was consistently rated highest on all three criteria, followed by biocontrol and mechanical control. Feasibility varied by life form: control of succulents and shrubs was generally rated more feasible than control of trees, vines and grasses. Experts were more confident about the feasibility of chemical and mechanical control than biocontrol.
To combine our two studies, we created a risk plot that mapped the 63 invasive plant species onto an impact/management feasibility plot, enabling managers to identify and target species with high impact and high management feasibility (good candidates for immediate site-specific control, e.g. Parthenium, Parkinsonia, Giant rat’s tail grass), as well as species with high impact but poor management feasibility (good candidates for research to improve management options, e.g. Prickly acacia, bellyache bush).
Read the papers:
Osunkoya, O. O., Froese, J. G., Nicol, S. , Perrett, C. , Moore, K. , Callander, J. and Campbell, S. (2019), A risk‐based inventory of invasive plant species of Queensland, Australia: Regional, ecological and floristic insights. Austral Ecology. doi:10.1111/aec.12776
Osunkoya, O.O., Froese, J.G. & Nicol, S. (2019) Management feasibility of established invasive plant species in Queensland, Australia: A stakeholders’ perspective. Journal of Environmental Management, 246, 484-495. https://doi.org/10.1016/j.jenvman.2019.05.052.
Our latest paper is out in Nature Communications! Read on for the details (TL;DR= we show that the value of information from resolving threat uncertainty is much bigger than we thought, and that priority threats for uncertainty reduction are high frequency fire, invasive predators and dieback).
*Note that the following text is reblogged from Nature Ecology and Evolution‘s “Behind the Paper” blog*
Managing threatened species requires making tough decisions. The lists keep getting longer, and management agencies globally are increasingly required to prioritise limited resources to save species. With limited resources, it’s important to use resources efficiently. Unfortunately, species management outcomes are often uncertain, so we rarely know where to target efforts to improve management efficiency, nor how many additional species could be saved by more efficient management.
To understand the potential gains from removing uncertainty about threatening processes, we undertook a value of information analysis. Value of information is a decision-science technique that measures the potential gains from the collection of new information. In our case, it allows decision-makers to prioritise the investment in improving knowledge about management effectiveness of threatened species. Working with the New South Wales government’s innovative Saving our Species program in Australia, we analysed almost 1,000 threatened species and threatened ecological communities affected by 20 key threatening processes including some of the big threats impacting species globally, such as fire, invasive species, and diseases. Calculating the value of information for such large numbers of species and threats is challenging and hasn’t been attempted before. One of our biggest challenges was to evaluate the gains in reducing two sources of uncertainty: the effectiveness in management of threats and the response of the species to that management. To collect the data for our analysis, we asked species experts to estimate the effectiveness of best practice management and the expected persistence of species with and without each threat.
Across all threats, we found that the average gain in species persistence if managed under current uncertainty would be 3% per species. If uncertainty could be removed, the gains would jump to 12% per species. The potential gains from removing uncertainty about threat management effectiveness could quadruple the gain in persistence achieved by managing under current uncertainty.
The implications of this finding are that uncertainty is far more influential than we thought for some threats and there are big opportunities to improve species management if we target research investment towards particular threats. We found that managers were confident about controlling invasive plant threats, but very uncertain about the benefits of managing high frequency fire, invasive predators and the plant dieback pathogen Phytophthora cinnamomi. Uncertainty about how species respond to threats had greater value than uncertainty about management’s ability to reduce the threat itself: i.e. it’s always better to attempt management than to do nothing even when the effectiveness is uncertain; but there are large potential benefits from better understanding how species will respond to threat reductions.
Our study quantifies the opportunities from better understanding how threat management influences species persistence but also creates new questions about how to best capitalise on these opportunities. When we looked at the interactions between threats, we found that the threats with the greatest potential gains from management (fire, invasive predators, dieback) were also those that interacted most with other threats. How to disentangle these interactions to reduce uncertainty is an ongoing challenge. Our next project will use adaptive management algorithms from artificial intelligence to strategically learn how to improve threat management over time while managing the system.
Our study is the first to quantify the value of information about threat management at such a broad scale. While many previous value of information studies have focused on single species and found minor benefits from removing uncertainty, our approach showed that when small benefits are aggregated across the many species impacted by threats, there can be large potential gains to management from reducing those uncertainties.
The paper citation:
Nicol, S., Brazill-Boast, J., Gorrod, E., McSorley, A., Peyrard, N. & Chadès, I. (2019) Quantifying the impact of uncertainty on threat management for biodiversity. Nature Communications, 10, 3570.
When: 28-30th August 2019, EcoSciences precinct, Brisbane, Australia
Reducing farmed animal protein consumption by introducing edible insects into the Australian diet promises significant health, cultural and environmental benefits. The global insect industry is expected to grow 23% by, to $1 billion, by 20231, yet few Australian companies produce and sell insects2 (mostly non-native crickets and mealworms) for human consumption. This symposium will bring together leading scientific and traditional knowledge of native Australian edible insects, promote their inclusion in our diet, and drive commercialisation opportunities. We will produce a much needed roadmap for an Australian edible insect industry, providing innovative edible insect solutions.
The symposium will run over three days, and presentations will be followed up by group discussions. The symposium will be facilitated by the CSIRO ON team, with the support of a facilitation team (Rocio, Bryan and the CSIRO Conservation Decisions Team). We will include the following topics:
Participating enterprises will be able to set up a stand and promote/sell their products during the event. Please note that selling will only be permitted after 2:30 PM due to building restrictions.
We are looking for donations and sponsors!
Rocio Ponce-Reyes (L&W), Bryan Lessard (CSIRO- NCMI), Emma Woodward (L&W-IF), Rene Cerritos (National Autonomous University of Mexico); Iadine Chades (L&W), Josie Carwardine (L&W), and Sam Nicol (L&W)
#CSIROediblebugs #ConservationDecisions #CSIRO #edibleinsects #innovation
1 http://www.fao.org/edible-insects/en/ “Global Edible Insects Market Will Reach USD 1,181.6 Million by 2023: Exclusive Market Research Report”
(Photo by Bryan Lessard, Grilo Spread)
In this paper we look at biodiversity conservation from a different perspective: we analysed how our current eating habits drive agricultural production and affect not only our health but the environment.
Using global databases, we analysed the links between health, agricultural production and environmental data together. We found that the Western diet—dominated by processed foods, refined sugar, fats and flours—has negative implications for all three.
Martin was awarded an outstanding doctoral thesis award (top 5% of PhD graduate).
Martin first joined the team as an internship student from the French engineering school “Ecole des Ponts”. After 6 months discovering our research environment, Martin decided to embark onto a PhD.
During his PhD, Martin published 4 papers. I am particularly fond of two papers he led, because they exemplify the type of AI research needed to tackle challenging problems in applied ecology: Read more
Nations, regions, governments, organisations and communities around the world are increasingly concerned about declines and losses of biodiversity. For many, there are important moral and legal requirements to improve the outlook for biodiversity and avoid species extinctions. However, understanding how to best invest resources in management to improve biodiversity is far from simple. Unless we are considering the right information in a logical way, we won’t be able to make the best use of the resources we have. Read more
We are seeking a highly motivated and dynamic postdoctoral research fellow to work on developing AI specifically designed to deliver interpretable and explainable solutions to environmental decision problems.
Together with Dr Kate Helmstedt and Dr Mike Bode from the Queensland University of Technology (QUT) School of mathematical sciences, Dr Sam Nicol from our team is offering a PhD scholarship to study optimal sampling and control of spatial populations in real time. It’s a great opportunity for someone with an interest in applied mathematics and ecology to come and work with our team. On top of the QUT PhD Scholarship, we’re offering a CSIRO top-up of A$7000/year to the successful candidate. If you’re interested to discuss more, get in touch! Applications close September 30, 2018.
The PhD advertisement is here:
Optimal monitoring and control of spatial populations in real time
In a new paper, available online now in Biological Conservations, Rocio together with researchers from the Wildlife Conservation Society and the University of Queensland assessed the potential impact of climate change on the distributions of ecosystems.The distribution effects that we investigated here that may result from climate change are six: contraction, expansion or shifting of existing ecosystems, creation of new ecosystems, complete destruction to ecosystems, or no change to ecosystem distribution (Picture: Queen Elizabeth National Park, Acacia brachystegia grasslands (Northern Savannah) by Andy Plumptre)
Sam, Josie, Iadine, Nicole and Stephanie will test and strengthen their Conservation Technology idea through ON Prime 2!
We are very excited and can’t wait to start! We can’t say much at this stage, however we will use our social media platforms to share the different stage of our adventure.
Urgent decision making can’t wait for big data!
In this week Nature’s correspondence, Iadine and Sam raise that the shift of private and public funding towards big data problems could impact our ability to solve some of our most urgent decision problems – for which we have no or very little amount of data available: biodiversity, health and biosecurity issues to cite a few. Sam and Iadine also provide some solutions and call for big ideas. It’s free for all to read.
Chades, I. & Nicol, S. (2016) Information: Small data call for big ideas. Nature, 539, 31-31.
Iadine and Sam wrote a bigger piece on the topic. We welcome your comments: Chades, I. & Nicol, S. (2016, November 2). Small data, big ideas. Zenodo. http://doi.org/10.5281/zenodo.164443
Our new paper came out in Ecological Indicators yesterday. In the paper we review the literature on metrics for managing spatially structured populations (i.e. metapopulations, migratory species populations). We ask the question: “what is the best way to measure the contribution of a habitat to the total population?” It’s a complicated question, partly because different theoretical approaches all have their preferred metrics. In the paper, we take a quick tour of graph theory, metapopulation theory and matrix models/perturbation analysis, and try to link the metrics developed in these literatures into a usable framework. We argue that the best metric depends on the management objective and the data available, and provide a simple framework for metric selection using these ideas.
The work was completed in collaboration with scientists from the US, Austria and Canada as a product of the NIMBioS Working Group on Habitat for Migratory Species. You can see their press release here.
Mobile species with complex spatial dynamics can be difficult to manage because their population distributions vary across space and time, and because the consequences of managing particular habitats are uncertain when evaluated at the level of the entire population. Metrics to assess the importance of habitats and pathways connecting habitats in a network are necessary to guide a variety of management decisions. Given the many metrics developed for spatially structured models, it can be challenging to select the most appropriate one for a particular decision. To guide the management of spatially structured populations, we define three classes of metrics describing habitat and pathway quality based on their data requirements (graph-based, occupancy-based, and demographic-based metrics) and synopsize the ecological literature relating to these classes. Applying the first steps of a formal decision-making approach (problem framing, objectives, and management actions), we assess the utility of metrics for particular types of management decisions. Our framework can help managers with problem framing, choosing metrics of habitat and pathway quality, and to elucidate the data needs for a particular metric. Our goal is to help managers to narrow the range of suitable metrics for a management project, and aid in decision-making to make the best use of limited resources.
Nicol S, Weiderholt, Diffendorfer J, Mattsson B, Thogmartin W, Semmens D, Lopez-Hoffman L, Norris DR (2016) . A management-oriented framework for selecting metrics used to assess habitat- and path-specific quality in spatially structured populations. Ecological Indicators 69: 792-802. DOI:10.1016/j.ecolind.2016.05.027
We are very excited to to announce our new report ‘Priority Threat Management for Imperilled Species of the Queensland Brigalow Belt’ (PDF) (see The Conversation article, CSIRO website).
Ponce Reyes, R., Firn, J., Nicol, S., Chadès, I., Stratford, D.S., Martin, T.G., Whitten, S., Carwardine, J. (2016) Priority Threat Management for Imperilled Species of the Queensland Brigalow Belt CSIRO, Brisbane.
Perhaps the most impactful decision support tool we have developed to date, Priority Threat Management (PTM) is a team adventure where a dedicated bunch of interdisciplinary scientists and kick ass facilitator work alongside stakeholders to establish the best strategies to manage threats to biodiversity. Stay tuned as we have grand plans to improve PTM and make it available to the entire world.
I was asked to provide a list of reference on priority threat management, sharing it with you here:
Freely accessible material:
Journal paper (PDF available upon request):
Did you know that there are fish living in the Australian desert? Obviously, fish need water, and that’s usually scarce in the desert. However large parts of arid Australia sit over the Great Artesian Basin, which is an area where groundwater is under such pressure that if you dig a well, it’ll come to the surface all by itself… no pumping required! This is good news for fish, because weak points in the aquifer mean that water seeps to the surface and forms permanent springs in the desert. Because these springs are so isolated from other sources of fish, unique species have evolved that live only in spring complexes. One of the best looking of these fish is the red-finned blue eye, Scaturiginichthys vermeilipinnis, which lives only in springs on a single property in central western Queensland, Edgbaston Reserve. I was lucky enough to visit Edgbaston at Easter Read more
Can climate change and biodiversity loss be tackled together by restoring forests? Our recent published papers investigate how and where carbon farming in Australia can be targeted to sequester carbon and benefit wildlife and threatened ecosystems. This work is summarised by an article in The Conversation – here it is!
Megan Evans, Australian National University; Anna Renwick, The University of Queensland; Josie Carwardine, CSIRO, and Tara Martin, CSIRO
Climate change and the loss of biodiversity are two of the greatest environmental issues of our time. Is it possible to address both of those problems at once?
In Australia, farmers and landholders will this week be able to apply for payments through the Federal government’s A$2.55 billion Emissions Reduction Fund. Bidders can request funding for projects that reduce emissions using agreed methods, which include approaches relevant to the transport, waste and mining sectors, as well as the land sector: for example, by managing or restoring forests.
Forests hold carbon in vegetation and soils and provide important habitat for native wildlife. Restoring forests in areas where they have been cleared in the past could be good for the climate, good for biodiversity, and generate additional income for landholders.
How well the Emissions Reduction Fund can achieve these benefits will depend on three things: the right approach, the right price, and the right location.
… improve the protein uptake and therefore the health of millions of people globally;
…reduce the land clearing and use of pesticides while obtaining economic profit;
…reduce the carbon dioxide an methane emission
In our new paper:Exploiting a pest insect species Sphenarium purpurascens for human consumption: ecological, social, and economic repercussions published in the first issue of the Journal of Insects as Food and Feed we developed a model of sustainable exploitation for pest insect species and explore the potential benefits to humans in different areas, like health/nutrition, conservation, economy and even climate change!
Please contact me if you would like a copy of the paper!
The picture below is by Rene Cerritos the lead author of this paper.
I was fortunate to be a co-author on the paper recently published in Frontiers in Ecology and the Environment led by Viv and Ayesha Tulloch, “Why do we map threats? Linking threat mapping with actions to make better conservation decisions.” (see media release PDF).
This is a great thinking piece about why we should probably not use threat maps “as is” to inform conservation decisions and how threat maps should be included as part of the decision process. I would recommend reading this paper if you are new to structured decision making and would like an easy read. The authors have done a fantastic job at explaining the risk of not following a transparent decision making process.
As a side note, our priority threat management work in the Pilbara was highlighted as a good example of making transparent decisions (No doubts they are other good examples in the literature!). It’s always a privilege to be acknowledged by our peers, it’s even better when these are esteemed close collaborators. Thanks for your support!
Vivitskaia JD Tulloch, Ayesha IT Tulloch, Piero Visconti, Benjamin S Halpern, James EM Watson, Megan C Evans, Nancy A Auerbach, Megan Barnes, Maria Beger, Iadine Chadès, Sylvaine Giakoumi, Eve McDonald-Madden, Nicholas J Murray, Jeremy Ringma, and Hugh P Possingham 2015. Why do we map threats? Linking threat mapping with actions to make better conservation decisions. Frontiers in Ecology and the Environment 13: 91–99. http://dx.doi.org/10.1890/140022