The best time to plant a tree was 20 years ago. The second best time is now. Chinese Proverb
Idea Transcript
APPENDIX SENSITIVITY ANALYSIS The robustness of model outcomes (total returns and total mean fish catch) to changes in parameter values was assessed with a sensitivity analysis. Table A1 gives the parameters tested in the analysis and the ranges within which each parameter was systematically varied. The sensitivity of the centralized regime to changes in the maximum total number of fields has already been analyzed in the main part of the paper. Other parameters are not relevant for model outcomes in the centralized regime (see also below). The following analysis thus only concerns the decentralized regime. The presentation of the sensitivity analysis is given with respect to changes in delta to assess the impact of changes in the respective parameter under different delta scenarios. Table A1: Parameters and their ranges used in the sensitivity analysis Parameter
Definition
Default
Range
Increment
Namaxnumfields
maximum total number of fields (centralized)
180
90-900
50
Maxnumfields
maximum number of fields/farmer (decentralized)
20
10-100
5
minincome
minimum yield requirement (decentralized)
80
40-200
10
alpha
birth rate of fish population
2
0.1-4.1
0.4
phi
scaling factor for inflow of larvae (runoff/phi)
1
1-10
1
lambda
scaling factor for fish income
10
0-30
2.5
setharvest
target catch
10
1-28.5
2.5
The parameters maximum yield, consumption and cropcosts affect the magnitude of the results in a linear way. They are thus not further explored. Some of the parameters of the decentralized regime such as the maximum number of fields per farmer or the minimum yield requirement are expected to have significant impact on model results because they influence the structure of the model (maximum number of fields per farmers) or the decision making process of the agent (minimum yield requirement). Their impact on the performance of the regime will be subject of more in depth investigations in the future, which will include the effects of heterogeneity between agents with respect to those attributes. The implications of changing the maximum number of fields per farmer have been discussed in the main part of the paper. Minimum yield requirement (decentralized - no fishing) The minimum yield requirement influences the decision of the farmer with respect to when to risk increasing the number of fields despite lack of information whether there
will be enough water to irrigate them. The requirement thus reflects the level of risk a farmer is willing to take, because an unsuccessful attempt to irrigate will result in the loss of the irrigation investment. The default value for the minimum yield requirement was set at 80 which is twice the consumption of one year. With a very high minimum yield requirement (minyield = 200) the farmers will always try to increase the number of fields as long as their financial reserve permits (if the maximum number of fields is 20, 200 is the maximum amount of yield a farmer can receive). They thus do not any longer take their estimation of water availability into account (delta has no effect). This strategy performs better than the satisfycing strategy of lower minimum yield requirements. The implications of the decision strategy of individual farmers will be further explored in future work. An increase in the minimum yield requirement above 80 to a value of 140 leads to positive total returns at lower delta values but only little increase in total returns (Figure A2) up. Beyond this at higher delta values the predictions become better and the first farmers use more water, making the risk strategy too costly. The system eventually can not sustain itself any longer because too many farmers go out of business. The final accumulated returns are sensitive to the number of farmers that remain in business, because of the upper limit on the amount of land a single farmer can irrigate. With low delta values the predictions of the upstream farmers are not very accurate, leaving more water to go to the downstream farmers which by increasing their number of fields more frequently at higher minyield values can benefit more from the additional water. For values of minyield below 80 the system breaks down except for high delta values (delta = 0.7 – 1.0), because too little fields are irrigated.
8000 7000 6000 s Return Total
5000 4000 3000 2000 1000 0 200 180
1.0 0.9 0.8
160 140
0.7
M
ld yie in
120
0.6 0.5
100
0.4 80
0.3
lta De
0.2
60 0.1 40
0.0
Figure A1: Accumulated total returns at end of simulation with change in the parameter minyield (lower income limit of farmer) for different delta scenarios.
Parameters related to the fishpopulation model and fishing activities (decentralized – with fishing) Fish population parameters such as the birthrate or the immigration of larvae with inflow of water to the lake affect the outcome of the model mainly at low parameter and low delta values. With low delta values agricultural performance is low, thus fish catch can have a significant effect on the performance of individual farmers and decide whether they stay in business and contribute to the overall irrigation returns or not. Thus at low delta values (0