the development of an individual -based tilapia farming simulation

14
THE DEVELOPMENT OF AN INDIVIDUAL-BASED TILAPIA FARMING SIMULATION MODEL Gertjan de Graaf & Pieter Dekker Amsterdam, the Netherlands May 2003

Upload: others

Post on 12-Sep-2021

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

THE DEVELOPMENT OF AN INDIVIDUAL-BASED TILAPIA FARMING SIMULATION MODEL

Gertjan de Graaf & Pieter Dekker

Amsterdam, the Netherlands May 2003

Page 2: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

TABLE OF CONTENTS

BACKGROUND OF THE MODEL .................................................................. 1

THE SIMULATION MODEL............................................................................ 2

THE SET UP .................................................................................................... 2 CALIBRATION OF THE MODEL............................................................................ 4

WORKING WITH THE MODEL ...................................................................... 4

RESULTS OF THE MODEL.................................................................................. 6

EXTENSIONS OF THE MODEL ................................................................... 10

OPTIMISATION MODULE ................................................................................. 10

REFERENCES.............................................................................................. 12

TABLE OF FIGURES Figure 1: The basic pathway of the Tilapia model. Red lines indicate flow of

information for key state parameters ......................................................... 3 Figure 2: Comparison of observed and simulated net yields of Tilapia farming4 Figure 3: The four menus of the Tilapia model ................................................ 5 Figure 4: Saving a run as an ASCI text file...................................................... 6 Figure 5: Basic set up of a “Run” in the Tilapia model..................................... 7 Figure 6: Results of a completed run: yields and average length of the

fingerlings at stocking................................................................................ 8 Figure 7: Average length of fingerlings at stocking and net profits (U$/ha) for

10 consecutive rearings............................................................................. 9 Figure 8: Simulated growth during one rearing................................................ 9

Page 3: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

1

BACKGROUND OF THE MODEL Gertjan de Graaf (NEFISCO1, Amsterdam, the Netherlands) and Pieter Dekker (XI2, Delft, the Netherlands) developed an Individual-Based Tilapia Farming Simulation Model. The development of this model is the result of experiences gained by G.J. de Graaf, with the large-scale farming of Nile Tilapia (Oreochromis niloticus) in Congo Brazzaville during a FAO/UNDP aquaculture project. In Congo Brazzaville, using proper feeding and stocking regimes, net productions of 6-8 t/ha/yr. were obtained and the famous “stunting of Tilapia” was not a problem. These results were published in a number of scientific articles (de Graaf et al. 1996, 1999). During the analysis of the data of Congo Brazzaville and the writing of the articles it was realised that there could be some basic “biological driving forces” behind the failure or success of Tilapia farming, especially in rural Africa. Testing the validity of the ideas with field experiments was too expensive and could not be realised. However, the availability of a large data set obtained during five years of Tilapia farming in Congo Brazzaville and the development of the computer industry, with the availability of cheap, powerful computers and software, provided the idea: “Why not build a simulation model and try to find out what are the most

important driving forces within the system?”

The Tilapia model was built in MATLAB and calibrated with the data set of Congo Brazzaville. The first results of the model provided good results and we are convinced that it gives realistic outputs. The present model can simulate a mixed culture of Nile Tilapia and the integrated farming with predators (African catfish and Snakehead) and gives biological and financial results of the different systems.

1 www.Nefisco.org 2 www.Xi-advies.nl

Page 4: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

2

THE SIMULATION MODEL

The set up The Tilapia model uses the principles of Individual-Based modelling; in other words, each individual fish is followed during the simulation process. Most aquaculture simulation models are based on nutrient flows and operate with a kind of empirical black box method. Our model differs from this approach and the mathematics are based on commonly known population dynamics as used in fish stock assessment. Growth and survival are described with: Von Bertalanffy Growth Function

( )( )t

K t

L L e t= −

− −*

*1 0

Where

Lt Length at time t Loo L infinitive or asymptotic length K growth parameter to T zero, or time when the fish are born or entered in the system The growth rate at any point in the life span of the fish is can be calculated as follows:

( )tLLKdt

dL−= ∞

Exponential Decay model

( )[ ]eNN tPM

tt

+−

+= .

1

During initialisation of the model, all individuals at start are given, at random, a set of parameters such as: sex, length and growth rate. Variation among the mean values is normally distributed and controlled with a variation level, which is one of the inputs for the model. The model follows each individual over time in a matrix. Each day they grow a little, they can die and are removed from the matrix, females reach the length of first reproduction and produce offspring, the offspring enters the matrix and are treated the same way and followed individually over time. At any day during the simulation, the model can provide information on average length, biomass, numbers, net yield, etc. The basic pathway of the model is presented in Figure 2.

Page 5: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

3

Figure 1: The basic pathway of the Tilapia model. Red lines indicate flow of information for key state parameters

Page 6: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

4

Calibration of the model Key state parameters such as: growth and mortality were estimated from the dataset of Congo Brazzaville (de Graaf et al, 1996) and the same dataset was used to calibrate the model. An example of a comparison of observed and simulated yields is presented in Figure 2 an it is concluded that the model is accurate up to net yields of 8000 kg/ha/year. Beyond these yields, the simulation model underestimates the yields. The main reason is that the high yields were obtained from small ponds, which are easy to manage and where non-biological factors, which cannot be simulated, were the driving force behind the high yields.

Figure 2: Comparison of observed and simulated net yields of Tilapia farming

WORKING WITH THE MODEL The major input variables for the model are: • Pond size • Stocking density • Number of rearing days • Sex distribution • Feeding level (high, medium, low) • Predators, no predators or integrated culture with Catfish or Snakehead • Predator density • The length range and number of fingerlings to be selected at a simulated

harvest, to be used in a second iteration of the model • The number of rearing periods3 and simulation per rearing period

3 Also called iteration

0

2000

4000

6000

8000

10000

12000

14000

0 5000 10000 15000

Observed Yield

Sim

ula

ted

yie

ld

Page 7: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

5

• Financial or economic parameters • Growth and mortality parameters for the different species Setting of the basic parameters is done with four menu screens (Figure 3):

• Main menu: contains the basics of the simulation such as: pond size, stocking density, number of rearing days, feeding level, stocking with a predator and the size range of the fingerlings to be selected at harvest for stocking of the next rearing period.

The main menu has three toggle buttons to go to the Tilapia, predator and economics menu.

• Tilapia menu: contains the basic calibrated settings of Tilapia. In the stand-alone version you can change only the length of the stocked fingerlings.

• Predator menu: contains the basic calibrated settings of the used predator. Again, here you can only change the length of the stocked predator.

• Economic menu: here the basic financial parameters such as investments, labour price, feed costs, farm-gate price of Tilapia, etc, can be entered. Financial parameters of Tilapia differ greatly all over the world, and all parameters can be changed in the stand-alone version.

Once all settings are entered, clicking on the “start button” will start the simulation. Depending on the pond size, the number of simulations and the number of iterations to be carried out, it takes 3-35 minutes to run a simulation.

Figure 3: The four menus of the Tilapia model

Main menu Tilapia menu

Predator menu

Economic menu

Page 8: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

6

Results of the model Once the simulation is carried, out you can export the results as an ASCI text file through “File” -à “Save run as ASCI table” (Figure 4) or you can view the different graphic plots with the plot selection menu in the main menu.

Figure 4: Saving a run as an ASCI text file In order to understand the different types of plots that can be selected, you have to understand the unique “stochastic character” of the model. As discussed before, the model works with probabilities, at random each fish gets its growth parameters, the individuals to die are selected at random, etc. The variation in the model is set with a selected coefficient of variation (CV) for each key state variable. This process allows studying the variation in the results after simulating the same process a number of times. The model can simulate an unlimited number or consecutive rearings: i.e., a pond is stocked with an initial number of fingerlings at rearing no. 1 (start/rearing 1). After a certain number of days, the pond is harvested and fingerlings with the indicated size are selected at random and used to stock the pond for the second time (rearing 2). The pond is again harvested after the indicated number of days and again fingerlings are selected to stock the pond for the third time (rearing 3), etc. A simulation is the result of a number of consecutive rearings and the total of all simulations and rearings is combined in a run (Figure 5):

Page 9: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

7

Figure 5: Basic set up of a “Run” in the Tilapia model

Simulation 1 start

start

start

start

start

start

Rearing 2

Rearing 2

Rearing 2

Rearing 2

Rearing 2

Rearing 2

Rearing 3

Rearing 3

Rearing 3

Rearing 3

Rearing 3

Rearing 3

Rearing 4

Rearing 4

Rearing 4

Rearing 4

Rearing 4

Rearing 4

Rearing 5

Rearing 5

Rearing 5

Rearing 5

Rearing 5

Rearing 5

Simulation 2

Simulation 3

Simulation 4

Simulation 5

Simulation 6

Page 10: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

8

The model allows plotting the results with a variable degree of details.

• Plotting the results of a run: plots the overall results of all simulations and rearings. In Figure 6, an example of a graphical plot of the simulated yield and the average length of the fingerlings at stocking of a run is presented. Plotting the results of a run gives you at a quick glance an idea of the variability of the results

Figure 6: Results of a completed run: yields and average length of the fingerlings at stocking

• Plotting the results of a simulation: plots the results of a number of simulated consecutive rearings. In Figure 7, an example of a plot for the average length of the stocked fingerlings of one simulation with 10 consecutive rearings is presented. Viewing the individual simulations allows you to see changes in time. In the example, the first rearing has a wide range of fingerlings, while this is not the case in the next rearings. This is caused by the settings of the model: in the first rearing the average length at stocking and its CV is used, while in the next rearings the range is limited by the size range setting for selection of fingerlings taken for stocking from the previous harvest.

Page 11: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

9

Figure 7: Average length of fingerlings at stocking and net profits (U$/ha) for

10 consecutive rearings

• Plotting the results of an individual simulated rearing or one iteration

provides all the details. In Figure 8 an example of the simulated growth of males, females and recruits during one grow-out period is presented, and we see how they grow over time.

Figure 8: Simulated growth during one rearing

Page 12: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

10

EXTENSIONS OF THE MODEL The first goal for the development of the model was scientific research: simulation of different management scenarios in order to get a better insight into different biological processes of Tilapia farming. However, we realised that the model could be easily extended with economic, marketing, product values, etc. With extension of such modules, it became a more practical decision model, which can give first indication on questions such as: • Considering certain economic conditions, should I go for mono-sex culture

or should I use poly-culture with predator? • With a certain farm-gate price for Tilapia, what is the optimal farming

system? • If a predator is used, what are the optimal stocking densities?

Optimisation module In Tilapia farming, a number of different systems are used. In order to avoid overpopulation by fingerlings, an “all male culture” or predators are used. If the market accepts all different sizes, a traditional mixed culture can be used and the stocking rate can be adapted to the requirements. All systems will result in a production rate with certain sizes of fish and with a net profit depending on the local economic conditions. The model can provide the results at present. However, running all the different options is a time-consuming business. The model would really become a practical decision-making tool, if considering the economic conditions in a country an optimal Tilapia farming system would be automatically detected by the model. The proposed optimisation module will be a kind of “goal seeking” module. It will consider the major biological and economic parameters and seek automatically the most optimum management system. The development of the model has been funded until now by Nefisco and XI. However, we do not have the financial capacity anymore to continue with further development. Therefore, Nefisco and Xi are looking for donor support to finalise the model and make it available through the Internet. The cost to finalize the present model is estimated at 15.000 euro.

Page 13: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

11

An additional note In general the model can be used for simulating the farming of any fish species as long as the basic parameters such as growth, etc. are known. As the model is based on general principles of fish stock assessment, the model can easily be extended to simulation of “culture based fisheries” or even to ” fisheries.” Nefisco and Xi have the capality to make such extension, but this is not covered by the present proposal.

Page 14: THE DEVELOPMENT OF AN INDIVIDUAL -BASED TILAPIA FARMING SIMULATION

12

REFERENCES de Graaf, G.J., Galemoni, F. and Banzoussi, B., 1996, Successful recruitment

control of Nile Tilapia, Oreochromis niloticus by the African catfish, Clarias gariepinus (Burchell 1822) and the African snakehead, Ophiocephalus obscuris, I A biological analyses. Aquaculture 146: 85-100.

de Graaf, G.J., Galemoni, F. and Huisman, E.A., 1999. The reproductive

biology of pond reared Nile Tilapia (Oreochromis niloticus), Aquaculture Research, 30, 25-33.