COM.XINrNT RELATIONSHIPS 



growth, and for predation. Many of 

 these submodels, however, have only 

 been claimed to apply in greatly sim- 

 plified systems, and it remains to be 

 seen whether they are also relevant 

 in more complex natural systems. 



Uses for Models 



As for the use to which models can 

 be put, it is easier to indicate pos- 

 sibilities than to point to examples 

 of their actual use. We will leave 

 aside uses at the intermediate stages 

 of the model-building process, where 

 an imperfect model can itself, by the 

 development of internal inconsisten- 

 cies over a long computer run, or by 

 sensitivity analysis of various param- 

 eter estimates, point to ways in which 

 it can be improved. The process of 

 model building is indeed highly in- 

 structive, and aids greatly in the de- 

 velopment of insights into the func- 

 tioning of ecosystems. Once a model 

 has been built and validated, though, 

 it can be used for purposes extrinsic 

 to its construction. 



Experimentation — The model can, 

 for instance, be used for experimenta- 

 tion on scales that are impracticable 

 in real life, and many sources of 

 error inescapable in the field can 

 be eliminated. Questions can be 

 asked and answered, for instance, 

 on the effects of competition between 

 species under different meteorological 

 conditions. Such questions could be 

 included in a field experiment only by 

 extending it over different years or 

 different locations, where extraneous 

 and irrelevant sources of variation 

 would be introduced. 



Environmental Management — 

 When the treatments postulated for 

 the model are such as would be pos- 

 sible in practice, this use of the model 

 leads directly to its potential value 

 as a management tool. The effects of 

 any proposed manipulation may be 

 explored far more quickly and cheaply 

 than would be possible in the field, 

 and, either by trial and error or by 

 a formal optimization procedure, a 



choice can be made among a number 

 of possible management strategies, 

 once goals have been clearly defined. 



In the arid lands, for example, the 

 management goals that might be set 

 for particular areas could include 

 prevention of soil erosion by wind 

 and water; increased runoff of 

 groundwater recharge; increased (or 

 maintained) grazing capacity for 

 domestic livestock; increased num- 

 bers of wildlife (either for hunters or 

 as an amenity); and even increased 

 landscape values, insofar as they can 

 be defined (good strands of flowering 

 ephemerals following rain, or good 

 growth of the more spectacular 

 plants — Joshua tree, saguaro, palo 

 verde — might fill this bill). 



The practicable management treat- 

 ments would certainly include dif- 

 ferent grazing practices (livestock 

 type, density, and season, together 

 with methods of stock control); 

 shrub removal and/or seeding; wild- 

 life control — by hunting permits, 

 for example; introduction of exotics 

 (plants and animals); and perhaps 

 weather modification. The existence 

 of a reliable model of the system, 

 and a convenient computer imple- 

 mentation, would enable the effects 

 of any of these proposed treatments 

 to be evaluated in terms of the 

 selected goals (appropriately weighted 

 if multiple); the whole could then 

 be subjected to benefit/cost analysis. 



The arid lands of the United States 

 are under heavy developmental pres- 

 sure, which is likely to increase rather 

 than decrease. The multiple-use con- 

 cept often applied to them usually 

 means multiple stresses. Yet manage- 

 ment, except in limited fields, is per- 

 force largely intuitive at present. 

 Development of the management 

 tools outlined in the previous para- 

 graphs, accordingly, takes on the look 

 of urgency where our arid lands are 

 concerned. 



Needed Scientific Activity 



We should now examine what are 

 likely to be the roadblocks restrict- 



ing progress in this direction — 

 where and what sort of scientific 

 effort will need to be expended to 

 make these possibilities into realities. 



Monitoring — The range of eco- 

 systems currently being monitored 

 adequately to provide satisfactory 

 tests of alternative models is far too 

 small. It is of the greatest importance 

 that the ecosystem models produced 

 should be of high generality, even 

 though of limited precision; it is far 

 more valuable to be able to give ten- 

 tative predictions over a hundred 

 million acres than to predict accu- 

 rately the course of events on a 

 hundred acres. This means that ob- 

 servational areas against which model 

 results can be checked must be spread 

 widely enough, and be numerous 

 enough, to cover the variation over 

 which generalization is intended. 



Moreover, the establishment of 

 these monitored ecosystems for the 

 purpose of validating models under 

 development should be treated as a 

 matter of some urgency. Their value 

 largely depends on the period over 

 which observations have been made, 

 for long periods provide the most 

 exacting test of models. There are 

 a few sets of data already in exist- 

 ence — largely collected by the U.S. 

 Forest Service — extending back for 

 decades; these are of the greatest 

 value, even though only a limited 

 range of variables was monitored. 

 Field studies for the specific purpose 

 of validating ecosystem models are 

 also currently being set up under 

 the International Biological Program. 

 (See Figure IX-4) Many more such 

 sets of data will be needed for the 

 modeling work that lies ahead, and 

 in each of them a wide range of 

 variables should be recorded as a 

 routine. 



Ecosystem Modeling — It would be 

 premature to try to standardize ap- 

 proaches to ecosystem modeling. The 

 subject is not yet ten years old, and 

 it is far too early to try to put it 

 into a straitjacket. Several methods 

 of modeling are presently under test; 



283 



