Fig 1.
Default map (100 x 100 patches) with random values for resource quality and slope.
Resource quality map (left), slope map (middle) and source-target map (right). The source area is shown in brown and the target area in yellow. The water barrier is shown in blue with a size of 15 patches (default level).
Table 1.
Map settings.
Fig 2.
Behavioral decision tree for the scenarios with movement on land (left) and movement in water (right).
The movement on land is determined by the resource quality and randomness of the decisions, which is by default set at 10%, whereas the movement in water is determined by the selected scenario.
Table 2.
Default levels of model factors.
Table 3.
Factor levels of sensitivity experiments.
Table 4.
The four different behavioral scenarios tested in the simulation experiments and their respective levels of factors.
Fig 3.
Icons representing the water movement skills in the scenarios.
The black bar represents a drifting vessel or a raft. Note that in Scenario B the major part of the agent’s body is under water and is, therefore, affected by hypothermia and thermoregulation.
Fig 4.
Replication assessment with default levels of factors.
The precision is shown by the blue line and indicates the statistical validity of the simulation results at a given number of runs. The threshold for the accepted precision of 5% is shown in red. The number of runs we selected for the simulation experiments and where the precision is consistently below 5% is shown by the dotted line in red.
Fig 5.
One-factor sensitivity with respect to average CSRs.
The one-factor sensitivity experiments show how the systematic variation of levels of a single factor affect the average CSR response. Results are shown for water-barrier-size (a), perception-radius (b), hours-until-dehydration (c) and number-founder-agents (d). The behavioral scenarios are represented by the icons. Crossed out red circles indicate no successful crossing (CSR = 0).
Fig 6.
Two-factor sensitivity experiments for the radius of perception (perception-radius) and extent of the water barrier (water-barrier-size) for each scenario.
The two-factor sensitivity experiments show the combined effect of the systematic and simultaneous variation of two factors on the average CSR response. The default response is shown by the blue asterisk and the median response is shown by the yellow asterisk. The position of the yellow asterisk in relation to the diagonal line displays the weight of the contribution of each of the two factors.
Fig 7.
One-factor sensitivity with respect to average CSRs.
The one-factor sensitivity experiments show how the systematic variation of the levels of a single factor affect the average CSR response. Results are shown for current-speed (a), water-temperature-interval (b), decision-randomness (c) and population-density (d). The behavioral scenarios are represented by the icons. Crossed out red circles indicate no successful crossing (CSR = 0).
Fig 8.
Disappearance of agents during simulation runs.
An example of Scenario A, when varying the temperature of the water (left) and an example of Scenario D (right), when varying the hours until dehydration, are shown. Agents may die of dehydration, hypothermia or exhaustion, but may also leave the map at the border across land (left_map_land) or across water (left_map_water).