Skip to contents

02: Advanced setup

Fallert, S. and Cabral, J.S.

Source: vignettes/advanced_setup.Rmd
advanced_setup.Rmd

More realistic and complicated simulations may require a more advanced setup. This vignette will cover some of the more advanced features of the metaRange package.

Time layer mapping

The setup and processing of the environmental SDS can take quite some time, especially when the environmental data is large. In order to save the user time, the method set_time_layer_mapping() allows the user to define a custom mapping between the time steps and the layers of the environmental raster. Use cases may be:

  1. Using a “burn-in” period in the simulation, before the main simulation begins. This is usually done to let the simulation state reach a (semi) equilibrium state, from which the simulation objectives can be studied.
  2. If the environment is static and doesn’t undergo any change, the user may want to use the same environmental raster layer for all time steps.
  3. Shortening or lengthening the number of time steps of the simulation.

Default mapping

The default configuration is that each layer in the environment represents one time step of the simulation. I.e:

library(metaRange)
#> metaRange version: 0.0.0.9000
library(terra)
#> terra 1.7.55
sim <- create_simulation(create_example_landscape())
sim$number_time_steps
#> [1] 10
sim$time_step_layer
#>  [1]  1  2  3  4  5  6  7  8  9 10

Burn-in period

To add a burn-in period of 10 time steps, where the environmental data is not updated, one could use the function as follows:

sim$set_time_layer_mapping(c(rep_len(1, 10), 2:10))
sim$number_time_steps
#> [1] 19
sim$time_step_layer
#>  [1]  1  1  1  1  1  1  1  1  1  1  2  3  4  5  6  7  8  9 10

Static environment

To use the same environmental raster (i.e. the first one) for all time steps, one can use:

sim$set_time_layer_mapping(rep_len(1, 10))
sim$number_time_steps
#> [1] 10
sim$time_step_layer
#>  [1] 1 1 1 1 1 1 1 1 1 1

Changing the number of time steps

In the same way described above, one can also pick the specific layer that are used to use and shorten the simulation length.

sim$set_time_layer_mapping(c(2, 4, 6))
sim$number_time_steps
#> [1] 3
sim$time_step_layer
#> [1] 2 4 6

Global variables and processes

metaRange uses an object orient programming paradigm, this means each species is described by its own biologically relevant processes and traits. However, there may be cases where the user wants to define a global variable or process that is shared between all species, be it to store intermediate results that don’t belong to one species or processes to visualize or store output. This can by done by using the add_globals() function for adding global variables and the add_process() function with no species argument specified, for adding global processes. The global variables and processes are accessible through in the globalsand processes field of the simulation object itself, respectively. The benefit of using a global process is that the self keyword refers to the simulation object itself, which allows for easier indexing across multiple species.

sim$add_globals(
    global_var = 1
    # ... more global variables
)
sim$globals$global_var
#> [1] 1
sim$add_process(
    # Note the missing species argument
    process_name = "global_process",
    process_fun = function() {
        # self = simulation object
        self$globals
        # easy access to simulation functions
        self$species_names() # no species yet
    },
    execution_priority = 1
)
sim$processes$global_process
#> Process name:  global_process 
#> PID:  PID-127d751e71-global_process 
#> execution_priority:  1 
#> execution_environment_label:  simulation_192951c9 
#> $fun: function() {
#>         # self = simulation object
#>         self$globals
#>         # easy access to simulation functions
#>         self$species_names() # no species yet
#>     }
#> <environment: 0x000002557ed31180>

Interaction with the priority queue

With some specific study questions, it may not be desired to simulate all species from the first time step. As an example, during the simulation of invasion dynamics, one may want to have a burn-in period without the invasive species present and then introduce it after this point. On the other hand, there may be a need to simulate a species for a specific time period and then remove it from the simulation (e.g. there is no point in calculating the reproduction of a species that has gone extinct). To accommodate this, metaRange allows the user to manually add and remove processes from the priority queue during the simulation.

Queuing a process

The default behavior of add_process() is to immediately add the process to the priority queue. Setting the argument queue = FALSE will add the process to the simulation, but not to the priority queue. In that case, the user has at any point during the simulation the option to add the process to the priority queue using the enqueue() method of the priority queue.

sim <- create_simulation(create_example_landscape())
sim$set_time_layer_mapping(c(1:6))
sim$add_species(name = "species_1")
sim$add_process(
    species = "species_1",
    process_name = "invasion",
    process_fun = function() {
        message("Species invades!")
    },
    execution_priority = 1,
    # Note the queue = FALSE argument
    queue = FALSE
)
sim$add_process(
    process_name = "activate_species_1",
    process_fun = function() {
        message(paste0("time step: ", self$get_current_time_step()))
        # Note that when manually changing the queue,
        # the changes will take place in the
        # _next_ time step
        # e.g. the following will lead to the process
        # being first executed in time step 4)
        if (self$get_current_time_step() == 3) {
            message("Activating species 1")
            for (pr in self$species_1$processes) {
                self$queue$enqueue(pr)
            }
        }
    },
    execution_priority = 1
)
sim$begin()
#> time step: 1
#> time step: 2
#> time step: 3
#> Activating species 1
#> time step: 4
#> Species invades!
#> time step: 5
#> Species invades!
#> time step: 6
#> Species invades!

Dequeue a process

The dequeue() function of the priority queue allows the user to remove a process from the priority queue.

sim <- create_simulation(create_example_landscape())
sim$set_time_layer_mapping(c(1:6))
sim$add_species(name = "species_1")
sim$add_process(
    species = "species_1",
    process_name = "invasion",
    process_fun = function() {
        message("Species invades!")
    },
    execution_priority = 1,
)
sim$add_process(
    process_name = "stop_invasion",
    process_fun = function() {
        message(paste0("time step: ", self$get_current_time_step()))
        if (self$get_current_time_step() == 3) {
            message("Extiction species 1")
            for (pr in self$species_1$processes) {
                # Here we are querying the process ID,
                # which is a unique identifier for each process
                # so that the priority queue knows what to remove
                self$queue$dequeue(pr$get_PID())
            }
        }
    },
    execution_priority = 1
)
sim$begin()
#> Species invades!
#> time step: 1
#> Species invades!
#> time step: 2
#> Species invades!
#> time step: 3
#> Extiction species 1
#> time step: 4
#> time step: 5
#> time step: 6

Ending the simulation early

To end the simulation safely, before the last time step, the user can use the exit() method of the simulation. This will end the simulation at the end of the process it is called inside of. A possibly use case would be to conditionally end the simulation if all species are extinct.

sim <- create_simulation(create_example_landscape())
sim$set_time_layer_mapping(c(1:6))
sim$add_species(name = "species_1")
sim$add_process(
    species = "species_1",
    process_name = "invasion",
    process_fun = function() {
        message("Species invades!")
    },
    execution_priority = 1,
)
sim$add_process(
    process_name = "end_simualtion",
    process_fun = function() {
        message(paste0("time step: ", self$get_current_time_step()))
        if (self$get_current_time_step() == 4) {
            message("Ending simulation early")
            self$exit()
        }
    },
    execution_priority = 1
)
sim$begin()
#> Species invades!
#> time step: 1
#> Species invades!
#> time step: 2
#> Species invades!
#> time step: 3
#> Species invades!
#> time step: 4
#> Ending simulation early