Design Concepts

Basic Principles

The model is based on human decision-making theory and the behavioural ecology theory of optimal foraging (Begossi, 1992; Lee et al., 2014; MacArthur and Pianka, 1966); individuals generally operate in a way that either directly, or indirectly, maximises their net money intake over time. The marginal value and central place foraging theories provide the basis for behaviours, such that vessels fish in areas that will indirectly maximise their profits (Charnov, 1976; MacArthur and Pianka, 1966; Orians and Pearson, 1979). These decisions are however bound by conditions that ensure their behaviour is realistic (such as a maximum distance they can travel from port), with values derived from

questionnaire surveys (Chapter 2).

181 Emergence

Vessel catches and profit, and the spatial distribution of fishing effort emerge from the model. The patterns emerge from the vessels’ behavioural rules used to indirectly maximise their profits.

Adaptation

Vessels select patches via behavioural rules that, to varying degrees, indirectly maximise their return rate. Vessels are not able to adapt to their experiences by changing the rules that they follow to make decisions; i.e. if the model is set so that vessels select a patch at random, they will follow this behaviour throughout the model simulation. Vessels can, however, adapt to the catch rate they have received in a patch, by deciding to remain on that patch, or moving to another patch. The number of trips a vessel makes is imposed to reproduce observed patterns of possible or available fishing days, and are not modelled as adaptive decisions.

Objectives

Vessels only have a true objective when using the patch choice rules that require them to select the cell with the greatest difference between expected catch and expected cost. Here, the objective of vessels is to maximise their money intake over time by catching as many scallops as possible using as little fuel/money/time in doing so, through minimising the ratio between expected catch rates and travel costs. In the majority of behavioural settings, vessels are indirectly trying to maximise their money intake, by selecting patches that are above a certain threshold, such as the minimum viable catch rate, or have the highest catch rate, but it is the model user who has defined this.

Learning

Vessels do not demonstrate learning in this model, rather they continue to use the same decision rules throughout.

Prediction

Vessels will predict the catch rate that they are going to receive in a cell, which assumes that vessels know the scallop biomass in each cell. Vessels do not necessarily receive the catch rate that they expect to receive from a cell (to simulate small scale variation in catch rates). As it is a small fishery with a relatively stationary resource, it is appropriate to allow vessels to sense catch rates over the whole model world. However, they only select a patch to fish in that is within a realistic distance threshold as identified in the questionnaire surveys (Chapter 2).

182 Sensing

Once fishing, vessels can sense the catch rates in the current and neighbouring patches to decide whether to move to another patch. They can also sense the catch rates across the whole model world when calculating if a patch has an above average catch rate.

Interaction

Vessels are indirectly competing for scallops in the model; whilst they do not directly compete, if one vessel removes scallops from a patch, these scallops are no longer available for the other vessels. There is no direct interaction between vessels, as during questionnaires fishermen stated they would tow very close to another vessel, so there is no interference competition. They also said they would discuss where there are good catch rates with other fishers, but rarely took this

information into account when deciding where to fish, therefore there are no interactions related to social communication in the model.

Stochasticity

In patch selection behaviour, if multiple cells offer equally good catch rates, or ‘qualify’ as potential patches, then one is chosen at random, introducing some spatial stochasticity. In addition, a vessel will not necessarily receive the expected catch rate; the received catch rate will be drawn from a random distribution with a mean of the expected catch rate, and a standard deviation of 10% of the expected catch rate. This is to simulate small scale variation in catch rates within a patch.

Nevertheless, this model has relatively little stochasticity.

Collectives

There are no collectives (e.g. a social group) in the model currently.

Observation

Each time a vessel completes a fishing trip, model logbook data are recorded (Table 5.5). Spatial information is also provided by the models animation, and in values recorded by the fishing patches (grid cells) themselves.

183 Table 5.5. Variables written to an output file at the end of a fishing trip, to mimic logbook records.

Variable Description

Date The day the trip took place on

Month The month in which the trip took place Who Which agent is recording the trip

VCU The size of the vessel

Strategy The strategy category of the vessel MyPort The port the vessel operates from TripNumber The trip number (i.e. trip ID per vessel) TimeSteaming The time the vessel spent steaming TimeTowing The time the vessel spent towing

SteamSpeed The vessels steaming speed, to calculate fuel costs TowSpeed The vessels towing speed, to calculate fuel costs FuelUsed The total amount of fuel used

TripDuration The total trip duration (i.e. time spent at sea) FishHours The total time spent towing

HoldStatus The hold status (i.e. kg of scallops on the vessel) at the end of the trip PatchDist The distance the vessel travelled from port

Patch-choice-decision The set of patch choice behavioural rules being followed

Between-patch-decision The set of between patch decision behavioural rules being followed Return-decision The set of return to port behavioural rules being followed

Visited-patches A list of the visited-patches

5.5. Initialisation

At the beginning of the simulation 26 fishing vessels are created. There are 3 different types of Manx vessel: quality maximisers (QLM), quantity maximisers (QTM), and efficient fishers (EFF) as

determined from a questionnaire survey (Shepperson et al., 2016).

The number of vessels of each type, in each port, was determined from logbook data. From the logbook records, the proportion of trips that took place from Douglas, Peel, Port St Mary, and Ramsey were 23%, 38%, 23%, and 17% respectively; the proportion of vessels in each strategy QLM, QTM, and EFF was 29%, 21% and 50% respectively, but the proportion of trips made by vessels in each strategy was 21%, 20%, and 60% respectively.

An average of 1634 trips were made by Manx vessels per season, equating to an average of 343 trips by QLM vessels, 327 vessels by QTM vessels, and 980 trips by EFF vessels per season. The model was initialised with 26 vessels, 8 QLM, 5 QTM, and 13 EFF. The model season is 210 days in length, so to make 343 trips the QLM vessels were attributed with a 20% probability of fishing each day, QTM a 31% probability of fishing, and EFF a 36% probability of fishing each day. The vessels in each strategy

184 were distributed between the ports according to the proportion of trips made at each port by each strategy in the logbook records (Table 5.6).

The model structure is also set at initialisation, determining the behavioural rules that vessels will use to select fishing patches and decide when to return to port throughout the simulation (see Submodels section). Each vessel is attributed with its state variables accordingly (Table 5.7).

Table 5.6. The proportion of trips made from each port by each strategy.

QLM QTM EFF

% No. in model % No. in model % No. in model

Douglas 0.16 1 0.33 2 0.10 1

Peel 0.67 5 0.35 2 0.52 7

Port St Mary 0.04 0 0.14 1 0.29 4

Ramsey 0.13 1 0.18 1 0.09 1

185 Table 5.7. Initialisation values for each type of vessel in the model

Category Variable Strategies Unit

QTM QLM EFF

NumberOfDredges = 0.019 (vcu) + 2.984 Number of dredges towed

DistancePerTick (18kmph / 60) * 4) = km per tick

HoldingCapacity 9006 4085 8512 tons

Financial

LikelihoodOfFishing 0.31 0.20 0.36 percent

Activity during

186