Abstract
Data from a paired-catchment manipulation experiment are used to evaluate a model of the biogeochemical responses of catchments to acidic deposition. The data are from the Watershed Manipulation Project at Bear Brook, Maine, USA. The model is MAGIC, (Model of Acidification of Groundwater In Catchments) which has been widely used in assessments of acidification effects. The catchment manipulation consisted of 3 years of application of dry ammonium sulfate to the catchment on a bimonthly schedule. The principal stream responses to treatment included increased concentrations of SO 4, NO 3, Ca, Mg, Na, K, Al and H, decreased alkalinity and dissolved organic carbon and essentially no change in the concentrations of Cl and Si. The model was calibrated to pre-treatment data on the manipulated catchment as well as to 4 years of data from the paired control catchment. In general, the calibration to the pre-treatment data from the manipulated catchment gave smaller residual differences (simulated minus observed values for the treatment period) than did the calibration to the data from the control catchment. The larger differences resulting from calibration to the control catchment were decomposed into components arising from biases in the calibration and manipulation data and from biases in the model structure and/or parameterization. The former were of the same order as the latter. Comparisons of the dynamic responses of systems and models (like rates of change variables) also provided important information concerning the adequacy of the model. The trends in variables simulated by the model paralleled the trends in variables observed in the manipulated system. While there are problems related to interannual variability, spatial aggregation and incompletely determined parameters, the basic structure of the model appears to be consistent with the observed responses to the manipulations. It is concluded that paired-catchment manipulation experiments are extremely useful for evaluating models of catchment biogeochemistry. However, rather than seeking to verify or validate (or invalidate) models, model evaluations should strive to provide increased awareness of the weakest elements in the models which hopefully will lead to improvements not only in the model structure but also in understanding of the system being modelled.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have