Author’s note: This is a reposting from my blog malloc(evol_ecol); the post introduces the GMSE R package as part of the ConFooBio project.

As part of the ConFooBio project at the University of Stirling, my colleagues and I have released the new R package GMSE, with v0.2.2.7 currently available on CRAN and GitHub. The GMSE package generalises management strategy evaluation (MSE), an adaptive management framework that incorporates social-ecological system dynamics, the process of observing and monitoring the social-ecological system, and the assessment and decision-making processes of managers (Bunnefeld, Hoshino, and Milner-Gulland 2011). Our generalisation includes a game-theoretic component in which both managers and stakeholders dynamically update their decision-making to maximise their own utilities. The GMSE package thereby simulates all aspects of MSE and uses genetic algorithms (Hamblin 2013) to find adaptive solutions for manager and stakeholder decision-making.

In keeping with the MSE approach (Bunnefeld, Hoshino, and Milner-Gulland 2011), GMSE does not attempt to find optimal strategies or solutions for management. Instead, genetic algorithms within the GMSE package find manager policies and stakeholder actions that are adaptive, meaning that policies and actions that are found reflect heuristic strategies of managers and stakeholders that are adopted within a set of constraints and potentially changing circumstances. A wide range of parameter values can be specified by the software user to simulate management (for a full list, see the available arguments to the gmse() function in the reference manual, or run help(gmse) after loading the GMSE package). The gmse() function simulates all aspects of management, including the natural resources (population) model, the process by which resources are sampled or observed, the process by which managers make decisions about policy, and the process by which stakeholders decide how to act in response to policy (see Figure 1 below). The latter two processes (upper left and right boxes in Figure 1, respectively) each call the genetic algorithm to find adaptive solutions in each time step.

Figure 1: General overview of GMSE

All sub-models (boxes in Figure 1) are individual-based (i.e., agent-based). The natural resources model simulates a single time step of resources (e.g., a managed population) that can undergo movement, reproduction, and death. The observation model simulates the monitoring of the resources (e.g., through capture-mark-recapture techniques). The manager model uses data from the observation model to set policy for stakeholders (e.g., how costly culling, scaring, or feeding resources should be), and the user model uses manager policy to determine actions to maximise their own utilities.

Installing the GMSE package

The GMSE package can be installed from CRAN or GitHub. The easiest way to install is through CRAN using the install.packages() function in R (Note, R version 3.3.3 or higher is required for GMSE).


To install this package from GitHub, the devtools library first needs to be installed.


Use install_github() to install using devtools.


A simple example illustrates the use of GMSE below.

A hunted population under management

Consider a population of managed resources that is hunted by four stakeholders (GMSE allows for any number of stakeholders; the default number is four to keep things simple). Assume that the population has a carrying capacity of res_death_K = 600 adults, and the manager wants to keep the population at manage_target = 400 individuals. The manager will use a capture-mark-recapture method of monitoring the population in each time step (observe_type = 1). Further assume that all other parameter values are set to default values (see the reference manual). We can run this scenario using the code below.

sim <- gmse(observe_type = 1, manage_target = 400, res_death_K = 600, plotting = FALSE);
## [1] 'Initialising simulations ... '
## [1] 'Generation 33 of 100'
## [1] 'Generation 62 of 100'
## [1] 'Generation 95 of 100'

To avoid automatic plotting, I have set plotting = FALSE. The output to sim is a very large data structure that includes output from each sub-model (natural resource, observation, manager, and user) in each of 100 (default) time steps. The results can be plotted using the plot_gmse_results function.

plot_gmse_results(res = sim$resource, obs = sim$observation, land = sim$land, agents = sim$agents, paras = sim$paras, ACTION = sim$action, COST = sim$cost);

Figure 2: Example of plotted GMSE simulation results.

Figure 2 above shows the simulation dynamics over time for the starting parameter values set in sim. The upper left panel of the plot shows the positions of resources on the simulated landscape in the last time step (had we set plotting = TRUE, the movement of resources over time would be observable). The upper right panel is entirely blue, representing public land that is not owned by stakeholders (to simulate stakeholders that own land and attempt to maximise land values — e.g., crop yield –, use the option land_ownership = TRUE. Land ownership will then be represented by different landscape colours).

The middle left plot shows how both actual and estimated resource abundance changes over time. The solid black line illustrates the actual abundance of resources (i.e., the number of individuals) alive in each time step; the solid blue line illustrates how many resources are estimated based on the observation model and the simulated mark-recapture technique of the manager (blue shading illustrates 95 percent confidence intervals, but these are for display only and are not used in manager decision-making). The black dotted line identifies the management target (manage_target = 400), and the red dotted line illustrates the population carrying capacity (res_death_K = 600); note
that carrying capacity is enacted on adult mortality (but see also res_birth_K), so abundance can increase over carrying capacity given a sufficient number of juvenile and adult resources (where juveniles are defined as individuals born in the same time step as the current time step). The orange line shows the mean percent yield (right axis) of landscape production, which may be decreased by resources if resources consume crops on the landscape (see res_consume). The middle right panel would show the percent yield of each individual stakeholder’s given land_ownership = TRUE, with solid line colours reflecting yield from identically coloured plots in the upper right panel.

The lower left and right plots show manager policy and stakeholder actions over time, respectively. Under default simulation parameter values, only culling is available as a management policy and a stakeholder action, but (as indicated by the legend) other policy options can be set, including any combination of the following: scaring (moving resources), culling (killing resources), castration (removing resource’s ability to reproduce), feeding (increasing a resource’s survival probability), helping (increasing a resource’s reproduction), tend crop (increasing the yield from landscape cells), and kill crop (destroying yield of a landscape cell). Managers create policy by setting the cost of stakeholders performing available actions on resources (i.e., all actions except tending and killing crops), as constrained by the manger’s budget (manager_budget) and a minimum cost of performing actions (minimum_cost). In the above plot, there are time steps in which the manager has set culling costs to be high or low, based on the degree to which the estimated abundance of the population (blue line in middle left panel) is above or below the manager’s target. In response, stakeholders (in this case hunters) cull a lot when the manager sets culling to be cheap, but very little when culling costs are set to be high (‘costs’ in this case might be interpreted as quotas, or as strongly enforced prohibitions if the cost of culling is higher than the stakeholders budget, user_budget).

The gmse() function thereby provides a method of simulating multiple aspects of population management. This simple example is only one of many possible scenarios that can be simulated.

Development of GMSE

The code underlying GMSE has been flexibly developed for future expansion and new features. If there is something that would be useful to add that does not appear to be available by setting parameter values in gmse(), then chances are it can already be done by a few tweaks to the source code, or would require only a bit of additional coding. For example, the code already allows for future expansions to any number of resource types, allowing for multiple populations and structured populations — this is also true for agents.

A wiki can be found in the GMSE GitHub repository, and there is a place to post issues, including suggestions for new features.


Bunnefeld, Nils, Eriko Hoshino, and Eleanor J Milner-Gulland. 2011. “Management strategy evaluation: A powerful tool for conservation?” Trends in Ecology and Evolution 26 (9): 441–47. doi:10.1016/j.tree.2011.05.003.

Hamblin, Steven. 2013. “On the practical usage of genetic algorithms in ecology and evolution.” Methods in Ecology and Evolution 4 (2): 184–94. doi:10.1111/2041-210X.12000.

April’s Conservation Conversation was inspired by the Future of Conservation Project. This joint project hosted by UN WCMC and including researchers from universities Cambridge, Leeds & Edinburgh looks to explore the views of conservationists across a range of issues. The rearchers wonder if current conservation science paradigms are actually monopolised by a few loud voices and the project aims to find out how the values of conservation scientists en masse, align with those voices or otherwise.

The researchers’ previous work (Holmes, Sandbrook & Fisher, 2016) grouped conservation scientists around subjective positions. This then informed the building of a web-based questionnaire which scores respondents on two parameters: Conservation & Capitalism and People & Nature. The website then helpfully plots these two scores (along with those of a bunch of other recent respondents) and places the user in one of four positions: New Conservationist, Traditional Conservationist, Market Biocentrist and, Critical Social Scientist. Full methods & detailed descriptions of each of these positions are available on the project website.


Figure 1 – Screen grab of Future of Conservation output plot. We’ve added in the four positions in each corner of the plot

Here at StiCS we asked our Conservation Conversationists to complete the questionnaire then come along with their lunch and discuss the aims, positions used and potential opportunities the resulting data could provide. We also had Tunnock’s Tea Cakes.

Here’s where our group ended up:


Figure 2 – beautiful MS paint version plotting the approx position of the Conservation Conversationists (we are professionally available for figure production)

Rather than concentrating on where we fell individually and collectively on the plot, our discussion turned in two other directions: How we felt we answered the questions; and the relationship between values and interventions.

How we answered the questions

We discussed how it was often difficult to decide whether to answer more aspirationally (this is how it should be) or more realistically (this is how it should be, given the circumstances). The questions seemed designed to be context neutral, but many of us found it hard to decouple our general views from real world examples. Bold ethical questions were not apparent in the questionnaire: would we prefer to keep a person alive for 1 year or save a thousand hectares of rainforest? This may well be an unrealistic choice but such a question could encourage deeper thinking about how we value nature and people on such a linear scale.

Values & interventions

Can certain interventions be mapped onto the four positions of conservationist? A carbon trading scheme such as EU ETS may clearly fall onto the top half of the plot, using capitalist tools to achieve a conservation objective. However, rewilding which may at first glance seem like a Traditional Conservationist movement, could equally attract those from other positions (Table 1).

Table 1 – How rewilding could fit in each of the conservationist position
New Conservationist

A rewilding enterprise financing facility, promoting the business case for investing in natural capital (European Commission, 2017)

Market Biocentrist

Rewilding paid for by a business as part of a biodiversity offset scheme managed to maximise returns for biodiversity (Maron et al., 2016)

Critical Social Scientist

An area for rewilding selected to maximise the ecosystem services it can provide (Cerqueira, et al, 2015)

Traditional Conservationist

An area for rewilding is selected to maximise returns for biodiversity

(Rewilding Europe)

Devisive / Inclusive

One can imagine that categorising individuals into defined positions could have both divisive and inclusive outcomes. Divisive, from building potentially opposing camps into which people feel they should choose one position over another. Inclusive, from a clarification of values which can be used to gain a meaningful appreciation of alternate viewpoints.

The Future of Conservation survey was a fascinating spark for our group to explore why we do conservation and subsequently, what we do for conservation. Those behind the project have thus achieved the goal of stimulating discussion! We at StiCS look forward to seeing the results which come out of the project, particularly how demographic characteristics (age, gender, geography, education) impact conservationist values.

We suspect this won’t be the last time that the Future of Conservation Project is discussed at the Conservation Conversation!


We weren’t the first group to blog about this project! Check out how the Conservation Science group at RMIT in Melbourne got on in their discussion, here.

Perspectives from Stirling Biological & Environmental Sciences on how Brexit may affect our research, and what we can do about it…


EU flag. Source: Wikimedia Commons

Biological and Environmental Sciences (BES) here at Stirling is full of international researchers and collaborations. Since the Brexit vote, there has been uncertainty over how we, as UK- and non-UK, EU- and non-EU nationals, will continue to work on EU-funded projects, begin new international collaborations, and remain as residents in the UK. Read More

Kirsty Park, University of Stirling

The grand finale of the BBC’s Planet Earth II showcased the ingenious strategies that some animals use to thrive in urban environments. Though impressive, these species are in the minority. As the number of people living in cities around the world continues to rise, we should really be turning our attention to those animals that find city living too hard to handle. Read More

Mosquito nets are often used for fishing. A smart response is needed

Emma Bush, University of Stirling and Rebecca Short, Zoological Society of London

mnf_man2The human race is extremely resourceful, particularly when resources are limited. Inevitably, when poor rural communities are given access to a new asset they will find a number of uses for them. Anti-malarial bednets – the fine-mesh nets used to protect people from mosquito bites while they sleep – are a good case in point.

Read More

By Penelope Whitehorn

“I arrived confused about this topic and I will leave as confused as ever.” This was the parting comment from the only MP in the room and not the outcome we were hoping for! The event was a scientific briefing about neonicotinoid pesticides and pollinators, organised by the Soil Association in a classy venue in Westminster. Unfortunately, such confusion seems typical of the political response to an issue that has generated passionate controversy in many other sections of society. Read More


Resolving conflicts between food security and biodiversity conservation under uncertainty (ConFooBio)

Funded by an ERC Starting Grant to Nils Bunnefeld for 5 years from 1st September 2016

Find out more about the project and how to apply for the postdocs here


Applications now being received for the 2016 Interdisciplinary Conservation Network (ICN) workshop!

ICN 2016 flyer_updated

Date of workshop: 26-28 June 2016
Application deadline: 31 March 2016

The Interdisciplinary Centre for Conservation Science (ICCS), Stirling Conservation Science (STI-CS) and the Centre for Biodiversity and Conservation Science (CBCS) are pleased to invite PhD students and early-career researchers in the field of conservation science to apply to participate in a three-day workshop to be held at the University of Oxford, UK.

The aim of this workshop is to provide early-career researchers with an opportunity to collaborate with other researchers from around the world, including leading figures in their field, and to learn key skills for the development of their careers.

More information about the workshop and how to apply is available here.

Luc Bussiere, University of Stirling

This article was originally published on The Conversation. Read the original article.

In a post-apocalyptic future, what might happen to life if humans left the scene? After all, humans are very likely to disappear long before the sun expands into a red giant and exterminates all living things from the Earth. Read More

Katie Murray and Zarah Pattison


We recently held our departmental lunchtime “Conservation Conversation”, discussing whether or not invasive non-native species (INNS) are really that bad after all. This is an interesting concept to think about, especially for Zarah Pattison and myself who both work on different groups of invasive species in Stirling University’s Natural Sciences department. This is particularly in light of the flurry of books, namely Fred Pearce’s “The New Wild” and Ken Thompson’s “Where do camels belong?” which are promoting INNS. There has been a storm of surrounding media attention and outrage of invasion biologists worldwide. But who is right? And if they are “Nature’s Salvation” (Pearce, 2015), then are we wasting money on biological control of these organisms? Read More