456 J. L. Tiwariet al. 



From our experiences, we conclude that the construction of an 

 ecological model should be a long-term effort measured in decades, not in 

 months or a few years. It is too bad that the tundra pond models could not 

 be used as such a long-term guide for research but the questions that were 

 being asked during construction of the model were so basic that they could 

 best be studied in temperate regions, not in the Arctic. 



SUMMARY 



Several small models of processes such as phosphorus cycling and 

 primary production of benthic algae were constructed, and gave valuable 

 information about the importance of such things as the mixing of algae 

 into the sediments or the contribution of protozoans to the phosphorus 

 flux. Two large models were also constructed. These models incorporated 

 relationships of the physiology of populations to factors of the environ- 

 ment. Thus, algal photosynthesis was related to temperature, light, and 

 algal biomass. Almost all of the parameters and constants used in the 

 model were measured in the ponds; a few could not be measured but were 

 incorporated because they are known to be important. 



The output of the models successfully simulated the seasonal cycles of 

 biomass and production in the ponds. However, some constants and 

 parameters had to be tuned and some were very sensitive in that slight 

 changes caused drastic changes in output. Because of this, and because we 

 felt we did not know all of the controls operating in nature, we did not use 

 the models for prediction of future states of the ecosystem. 



The main value of the modeling came from the construction itself. We 

 were obliged to attempt to answer the most difficult questions of controls 

 and carbon flow. All the scientists had to look beyond the borders of their 

 own specialties and to think about the whole system. This stimulated 

 fruitful discussion and joint experimentation. It also caused us to realize 

 how important it was to concentrate on one or two ponds rather than 

 attempting to study a large number of ponds. 



The large models were deterministic but we also constructed several 

 stochastic models that incorporated variability into the initial conditions, 

 the input variables (e.g., light and temperature), and various parameters. 

 The introduction of the variability caused the models to become more 

 stable and also allowed us to calculate a mean of output values along with 

 a standard deviation. This approach appears promising because all of the 

 information about the natural system may be incorporated and because all 

 parts of the natural system are variable to some degree or other. 



