PLUTONIUM DYNAMICS IN A DECIDUOUS FOREST ECOSYSTEM 521 



a regional assessment involves the impact of a fire that burns litter, ground vegetation, 

 tree wood, and leaves. From field experiments transfer coefficients for plutonium have 

 been measured in forest stands over the region of interest with a precision such that CV 

 values are nearly 0.2 for all model parameters. Given that the soil is contaminated with 

 1.7 X 10^ pCi Pu/m^, the predicted amount of plutonium in the forest, at steady state, at 

 risk of release by fire from an average forest is 26.7 /iCi/ha. This amount could range 

 from 12 to 87 /.tCi/ha, however, depending on which forest was contaminated. 



Attempts to model plutonium dynamics in ecosystems are hampered by uncertainties 

 in predictions arising from problems with system identification, quality of data, and lack 

 of validation. System identification, which involves determining transfers and model 

 structure in a way that model performance fits real-world data, is a problematic area in 

 ecosystem modeling because of the variability in ecological data and our lack of control 

 over the natural environment (Halfon, 1975). The recommendation of O'Neill (1973) and* 

 of Shelley (1976) to "build the simplest model appropriate to achieving the objective" 

 was followed in designing the present plutonium forest model. In simple models the 

 effects of measurement error on predictions are reduced, but inaccuracies arising from 

 systematic bias are increased (O'Neill, 1973). 



Even when an optimal model structure can be found (i.e., one that simultaneously 

 minimizes inaccuracies due to systematic bias and measurement error), problems remain 

 because of natural variation in plutonium data from ecosystems. By simultaneously 

 varying all transfers, COMEX is a statistically based simulation technique that permits an 

 assessment of the variation in predicted amounts of plutonium in the forest, recognizing 

 that variance in transfers exists. Even with small amounts of variation about the 

 model parameters (CV = 0.2), there are typically eightfold differences in model 

 predictions. As Shelley (1976) points out, we cannot expect the variability in predicted 

 values from ecosystem models to be less than the variation in the data used to calibrate 

 the model. The COMEX simulations of plutonium in forest biota support this argument. 

 Given that CV values on plutonium data from field studies range from 0.5 to 3.0, we 

 conclude that the ability to adequately model plutonium transport in ecosystems is 

 strongly dependent on better data from natural environments and on an understanding of 

 the causes of variation in the data. 



Data on plutonium in the White Oak Creek floodplain forest were used to calibrate 

 this model. Calibration data cannot be used for model validation (Shelley, 1976), and 

 criteria for validity (Mankin et al., 1977) are difficult to define. The model cannot be 

 judged valid, but it has been useful for the identification of areas where research is needed 

 to better our understanding of plutonium dynamics in forests and thereby develop more 

 precise models. Future field studies should provide the data necessary for systems analysis 

 and comparison of plutonium dynamics in forests and other ecosystems. It is unlikely, 

 however, that these models will be validated over the full time frame of some simulations 

 (e.g., >100yr) before advances in ecosystem analysis make the models obsolete. 

 Nevertheless, the long radioactive half-life of ^^^Pu (24,400 yr) and its potential for 

 accumulation in the biosphere, necessitate some predictions in lieu of none at all. The 

 model reported here represents an hypothesis that presents testable predictions about the 

 dynamics and distribution of plutonium in deciduous forests. 



Acknowledgments 



We thank S. I. Auerbach, W. F. Harris, C. A. Little, D. E. Reichle, H. H. Shugart, Jr., and 

 anonymous technical reviewers for their helpful comments and suggestions. The research 



