摘要：Increasing use of ecological models for management and policy requires robust evaluation of
model precision, accuracy, and sensitivity to ecosystem change. We conducted such an
evaluation of hypoxia models for the northern Gulf of Mexico and Chesapeake Bay using
hindcasts of historical data, comparing several approaches to model calibration. For both
systems we find that model sensitivity and precision can be optimized and model
accuracy maintained within reasonable bounds by calibrating the model to relatively
short, recent 3 year datasets. Model accuracy was higher for Chesapeake Bay
than for the Gulf of Mexico, potentially indicating the greater importance of
unmodeled processes in the latter system. Retrospective analyses demonstrate
both directional and variable changes in sensitivity of hypoxia to nutrient loads.
