408 J. L. Tiwari et al. 



The next problem is to choose a modeler. The number of ecologists 

 who are really good modelers is quite small so we chose instead a geneticist 

 with a good mathematical background. It is important that a modeler be 

 familiar with the uncertainty and variability that are characteristic of 

 biological systems. Too many modelers with engineering backgrounds 

 appear to believe that a constant is inviolate and that the literature can be 

 trusted. Our choice of modeler meant that all of the biological insight had 

 to be provided by the ecologists and that modeling could not go on without 

 them. 



The model has to be detailed enough that it is biologically 

 sophisticated and satisfying to the ecologists. Thus, photosynthesis could 

 not be put into the model as a daily sine wave but, instead, had to be a 

 function of light, of nutrients, of temperature, etc. The actual functional 

 relationships, for example those between light and photosynthesis at 

 different temperatures, had to be incorporated into the equations. Other 

 details of the models are given in the next section. 



As a general philosophy, we tried to measure every parameter, 

 constant, and initial value that went into the model. This was about 90% 

 successful but some processes that we knew were important could not be 

 measured yet had to be incorporated into the model. One example is the 

 natural, or non-predation, death rate of microbes such as algae. Most 

 models of lakes or reservoirs contain this process but no one has measured 

 it directly. We asked one of the foremost engineering modelers about this 

 process. He replied that it had often been measured and that all one had to 

 do was follow the populations of plankton algae after enclosing them in a 

 bottle. This is absurd and naive. The situation becomes even more 

 upsetting when coefficients like that for algal death rate are used to "tune" 

 the model by adjusting them so that the output of the model agrees with 

 nature. 



A great deal of time, effort, and money was spent in developing a 

 computtr data bank for the Tundra Biome. Most of this was wasted as far 

 as the aquatic research was concerned. Only the most routine type of data 

 can be easily put into such a data bank and we stopped taking very much 

 of this type of data after the first year. From then on, most of the data 

 came from experiments which were run under conditions that changed 

 each time. Pages of explanation were necessary to put this type of 

 experimental data into the data bank. One extreme view of the situation 

 was that "if the result goes easily into the data bank it probably isn't worth 

 getting." This is really a reflection of the fact that we had lots of routine 

 data but needed more information about the relationships and controls. 



Once the modeling began, it became important to have some 

 interchange between the modelers and experimental scientists so that the 

 experiments could measure at least some of the things that were necessary 

 for the modeling. In our case, it did not prove profitable to have the 

 modeler in the field and so most of the interchange took place during the 



