4 ULANOWICZ 



change in dimensionality of the problem and/or the necessity for a 

 new functional descriptor, f, of the system dynamics. Deviations in 

 the neighborhood of Scrit are signs of a stressed ecosystem in the 

 physiological sense of the word. 



Thus two classes of stress analyses are readily identifiable. In the 

 first, there is no explicit mention of a critical deviation, a change in 

 dimensionality, or a switch in function (topological form). The 

 second class is identified by the prominence of at least one of these 

 characteristics. The first class of models will be referred to as local 

 and the second as far-field. 



Unfortunately, not all ecosystem stress analyses fall neatly into 

 these two classes. There are critics (e.g., Mann, 1975) who find 

 compartmental modeling reductionistic. They claim that time would 

 be better spent searching for emergent properties of the ecosystem as 

 a whole and that these properties would serve as more reliable 

 indicators of the response of the community to stress. To press the 

 earlier analogy (perhaps a little far!), this is akin to observing the 

 strain response of an assemblage of metal rods (such as a bridge truss) 

 to various imposed stresses without being concerned with the 

 properties of the individual members. Certainly there are character- 

 istic dimensions or dimensionless ratios of the total structure at 

 which strain responses are indicative of impending collapse. Such 

 approaches to ecosystem stress will be termed macroscopic in nature. 



Finally, I indulge in speculation on a principle that I believe 

 would crystallize research on ecosystem response and, more impor- 

 tantly, might provide a theoretical basis to bridge the gap between 

 ecological systems research and evolutionary theory. 



LOCAL ANALYSES OF STRESS 



Most of the ecosystem models in existence today were con- 

 structed to elucidate the response of the community to a stress. As I 

 see it, most of these efforts have been local in nature. This is not to 

 imply that local analyses are necessarily uninteresting or uninforma- 

 tive. In fact, some appear to be nonlinear and have been quite 

 successful in portraying the response to exogenous stresses. For 

 example, the highly reeilistic aquatic ecosystems models of Nixon and 

 Kremer (1977), Di Toro et al. (1975), and Park, Scavia, and Clesceri 

 (1975) are capable of predicting significant changes in response to 

 exogenous stresses (e.g., temperature, nutrient input, and light 

 availability). 



As an analysis of stress, however, these simulations are basically 

 local. To see this, we should appreciate that most of the exquisite 



