The simulator consists of (a) submodels of the decision making of several individuals and groups, and (b) the impacted ecosystem as follows.
The Chinese consumer submodel along with the wildlife protection agency submodel are aggregated, i.e., they simulate group-level decisions rather than the decisions of unique individuals. On the other hand, individual-level submodels are built for each identified trafficker be they a poacher or (potentially) a middleman. These latter, individual submodels are linked to the associated nodes of the investigation's social network model of all traffickers in their database.
All individual tiger poacher submodels share the same perceived causal structure across their cognitive variables as expressed in their respective .id files. Likewise, when there are multiple middlemen, they all share the same perceived causal structure across their own set of cognitive variables. Individuals differentiate via entries in their respective parameter files.
The files contained in intelinfiles.zip are:.
An actionable intelligence report is generated by running tigerpoleco.id with the command,
idalone tigerpoleco.id
at a Windows command prompt. This report contains social network analysis measures that support the report's Detain, Surveil, and Interdict lists. The Detain list identifies the tiger poacher who is doing the most damage to the ecosystem and hence is the most critical poacher to detain. This particular poacher is identified by running the simulator with each poacher temporarily removed from the model and noting the poacher whose removal results in the least ecosystem damage. For instance, if the removal of poacher #1 results in 10 tigers being poached by the remaining poachers, but the removal of poacher #2 results in 14 tigers being poached by the remaining poachers, then poacher #1 would be detained.
The actionable intelligence produced by this run produces the following output file, tigerintel.txt:
The criminal network just before arrests given in the Detain List are made:
The criminal network some weeks after these arrests:
The actions history plot generated with tiger poacher #2
temporarily removed from the model is:
There is a causal chain in the tiger ecosystem's influence diagram that is pertinent to tiger poaching. This chain flows from the chosen management option through to tiger abundance. This chain is as follows.
Management Option (deterministic decision node with values beginning state, poach for cash, translocate animals)
↓
Poaching Activity (deterministic discrete node with values reduced activity, no change, increased activity)
↓
Poaching Pressure (stochastic discrete node with values minor, moderate, severe)
↓
Tiger Death Rate (stochastic continuously-valued node with values death rate ∈ (0, 1))
↓
Tiger Abundance (stochastic continuously-valued node with values abundance ∈ (0, 3000)).
The last two nodes in this causal chain are part of a system of stochastic differential equations that model the population dynamics of tigers in Bandhavgarh National Park, India. The node representing the ecosystem's state that is readable by other simulator submodels is exclusively Tiger Abundance rather than the pair of nodes Number of Tigers Poached and Tiger Abundance. There is no node that is Number of Tigers Poached. Rather, the ecosystem submodel's Management Option has poach_for_cash as one of its possible values (see above). When this value is given as input to the ecosystem, poaching pressure is applied.
The carrying capacity, birth rate, and death rate parameters are all specified in the files tigereco-hyp.par and tigereco-ini.par. These parameters will be adjusted when the simulator is fitted to a new data set of tiger abundance observations. Wide differences in tiger birth and death rates can cause tiger abundance to blow up as time progresses. If this happens or the solution exhibits very high variance, try forcing this rate difference to be zero as a first try at fixing the problem.
As exhibited above, poaching's effect on tiger abundance is expressed by its effect on the death rate parameter. But if initial carrying capacity is close to initial tiger abundance, abundance will be almost completely a function of prey abundance (herbivore abundance). In this situation, tiger birth and death rates will have little effect on tiger abundance. To make abundance be sensitive to these rates, set the initial carrying capacity to be at least four times the initial abundance of tigers.
This Tool assumes that the confederation gathers evidence on the WTS at three different times. The first time is to find out the size, connectivity, and assets of the current, undisturbed WTS. The confederation quietly watches the network for several weeks and at the end of that period, observes its size and connectivity again. Then, the confederation recommends to law enforcement those WTS players to detain, surveil, and interdict. Finally, some weeks after these arrests, the confederation gathers information on the size and connectivity of the recovering WTS. Call these three time points, t1, t2, and t3.
The file tigerintel.txt contains predictions of those players in the WTS who are predicted to move into leadership roles (called Rising Stars); and the resiliency of the WTS (a measure of how fast the syndicate's functionality can recover from a series of player removals). All of these terms are discussed in Haas and Ferreira (2015).
Let EC(p,t) be player p's eigenvector centrality at time t. Player p is a Rising Star if (a) EC(p,t) is larger than the median eigenvector centrality of all network players at time t; and (b) EC(p,t2) > EC(p,t1).
Let CI(t) be a measure of a social network's connectedness at time t. Connectedness is one way to measure a social network's functionality. Let NRI be a measure of a social network's resiliency defined to be proportional to how quickly a social network recovers 90% of its functionality after removal of some of its players.
One quantitative definition of CI(t) is the largest eigenvalue of the social network's link weight matrix. And hence, one way to define NRI is to set it equal to 1/(t2 - t1) where CI(t2) = 0.9CI(t1) and t1 is the point in time just before the Detain list-recommended arrests are made. This definition is operationalized as follows. Let tobs be the time interval between t1 and a time point some number of weeks later. Then set NRI to CI(tobs) / (tobs0.9CI(t1)) when CI(tobs) < 0.9CI(t1), and declare that it is at least 1 / tobs otherwise.
The file polecotigers.dat contains observations on political-ecological actions. This file is as follows.
comment Political-Ecological Actions Data comment Date Actor Action begin 03-15-11 tigerpoacher1 poach_for_cash 11-05-12 tigerpoacher2 poach_for_cash 07-30-11 wccbureau arrest_poachers end
The simulator's parameters have been fitted to this data set. It is emphasized that it is assumed that the actions of identifiable tiger poachers can be acquired through the efforts of criminal investigators. This intelligence supports the construction of the above data set.
The file obstigers.dat contains observations on ecological nodes in the ecosystem influence diagram. This file is as follows.
comment Bengal tiger abundance estimates Region Time TigerAb begin Bandhavgarh 2010 1466 Bandhavgarh 2011 1621 Bandhavgarh 2012 1495 Bandhavgarh 2013 2968 Bandhavgarh 2014 2619 Bandhavgarh 2015 2875 Bandhavgarh 2016 3235 Bandhavgarh 2017 3235 Bandhavgarh 2018 3235 Bandhavgarh 2019 3235 Bandhavgarh 2020 3235 end
This data would be collected either by hired field ecologists or through technological methods such a camera traps and/or remote sensing techniques.