Aarhus University Seal / Aarhus Universitets segl

Hunter decision model

Overview

1. Purpose

The hunter decision model was created to model the distribution of hunting areas to predict hunting pressure at a regional level. This predicative simulation model was created within the existing the ‘Animal, Landscape and Man Simulation System’, ALMaSS. Development of this model is in two phases. The first stage, documented here, was to predict the distribution of hunting locations within a modelled landscape, representing a region in north-west Denmark.

This first stage required modelling potential decision processes that hunters might use in deciding ‘where to go goose hunting’, without geese being in the model but accounting for possible future goose distributions. Model development was designed to systematically test possible decision rules and explore the degree of complexity needed to model the selection of hunting locations by hunters, and to predict the distribution of hunting locations.

2. Entities, state variables, and scales

The model is comprised of individual hunters and a farming landscape as a bio-physical entity.

Hunters

Individual hunters are given home locations in the model, from where hunters can travel to their selected hunting locations. Home locations were based on known home addresses for those hunters who had shot geese in the model area, wherever they resided in Denmark. Initially, a fixed number of hunters (343) was included in the model. However, as part of the model’s development it was used to predict potential goose hunter numbers. Support information on determining the number of hunters can be found here [link]. Hunters use decision rules to select hunting locations as described in 3. Process overview and scheduling.

Landscape

The model represents an 864 square kilometer area (36km x 24km) named Vejlerne, which straddles Thisted and Jammerbugt municipalities in north-west Denmark. The Vejlerne area is a rural agricultural landscape predominately composed of farm properties. Rights to hunt are given by ownership of property or by securing hunting rights.

Farms were chosen to represent hunting locations to be selected by hunters. The model represents several hundred farm units (558) of varying sizes but with similar agricultural usage, predominantly cattle, pig or mixed livestock (45%) and mixed ‘hobby’ farms (34%). Each farm property encompasses fields with known boundaries and sizes (hectares). Farms are the entity for hunting locations, determining distances for hunters to travel (farm centre points to home locations) and density of hunters per hectare. Field boundaries and sizes are used to determine ‘openness’, as describe in 3. Process overview and scheduling.

Roost sites are locations where geese disperse from to foraging areas (farm fields), determining distances from roost to farm properties.

3. Process overview and scheduling

For the first stage of model development the key process was for hunters to select hunting locations, represented by farm properties. Modelling this decision-making process started by testing very simplistic rules (e.g. hunters’ randomly selecting hunting locations) and then adding greater complexity by including a variety of utility-based decision rules, analogous to a bounded rationality approach. Six decision scenarios were created using combinations of the six decision rules, with increasing complexity from 1-6 (below). Hunters would select a farm based on these rules and allocated a hunting location. When combining decision rules all conditions had to be fulfilled e.g. if the closest farm was not suitable because of other decision rules (e.g. minimum field openness) then hunters would select the next closest farm, and so on. The first four decision rules were combined to create an initial set of decision scenarios (1-4) that were run and tested to allocate hunters to farms. An additional decision rule was hard-coded within the model enabling each hunter to select between 1-5 hunting locations. The last two decision rules were implemented after an initial development phase creating a further two decision scenarios (5 & 6). Supporting information on decision rules can be found here [link].


The decision process for each hunter to select a hunting location was a one-time decision, because of the known tendency for hunting leases, in Denmark, to be long-term agreements. An additional decision rule was hard-coded within the model enabling each hunter to select between 1-5 hunting locations.

 Decision scenarios

  1. RandomFarm
  2. ClosestFarm
  3. RandomFarm_MinOpenness_MaxDensity
  4. ClosestFarm_MinOpenness_MaxDensity
  5. ClosestFarmProbability_MinOpenness_MaxDensity
  6. DistanceRoostProbability_ ClosestFarmProbability_MinOpenness_MaxDensity
  • No.HuntingLocations (hard coded)

4. Design Concepts

4.1. Basic principles

To systematically develop and test a sequence of possible decision rules and explore the degree of complexity needed to model this decision-process. Starting with individual simplistic rules (e.g. hunters’ randomly selecting hunting locations) and then add greater complexity by including and combining a variety of utility-based decision rules, analogous to a bounded rationality approach. A few published studies indicated factors that might influence how hunters choose hunting locations and informed our choice of decision rules [link decision rules ID 14555].

4.2. Emergence

The patterns that emerge at the landscape level are the number of hunters distributed to individual farms within the Vejlerne model area. These emerge because of interactions between decision rules for selecting hunting locations (farms) by hunters.

4.3. Adaptation

The selection of hunting locations in this current model is a one-time decision because of the known tendency for hunting leases, in Denmark, to be long-term agreements.

4.4. Objectives

The objective for hunters is to select hunting locations that provide opportunities to shoot geese based on hypothesized decision rules, where there are no geese included in the model.

4.5. Learning

In this decision-model hunters’ do not learn or change hunting locations (one-time decision).

4.6. Prediction

Hunters select hunting locations based on predictions that they will have the potential to offer goose hunting opportunities.

4.7. Sensing

Hunters use the information inherent within the decision rules to select hunting locations e.g. ClosestFarm rule requires hunters to assess the distance from home locations to potential farm properties (centroid of nearest farm).

4.8. Interaction

The decision rule Maximum density simulated interaction between hunters, whereby they selected farms based on the number of goose hunters already using it, to avoid ‘crowding’. Expressed in terms of hunters per hectare, with farms selected below a maximum density of hunters per hectare of ‘open fields’.

4.9. Stochasticity

Stochastic events are included in the model as part of the following decision-rules:

  • Random farm
  • Number hunting locations
  • Closest farm probability
  • Distance to roost probability

4.10. Collectives

The model does not use collectives.

4.11. Observation

The outputs of the model used to assess the spatial distribution of hunting locations included:

  • Distance travelled: Model travel distances (Euclidean) calculated from hunter home locations to the centroid of allocated farm.
  • Hunters per hectare: Density - number hunters per hectare of open fields for each farm.
  • Number of hunters per farm: Number of modelled hunters allocated to each of 558 farms in the model.

Details on data used to test and validate model outputs can be found in the Background data section.