before jumping too hard on the USACOE for missing the weather, one of the problems on all predictive models (like salmon) is that they look backwards to predict forwards. If the (climate, survival, etc) is changing you are always predicting from behind. So, even in a reasonably stable situation you are playing from behind. Throw in some sort of anomaly and you get way off.

But, if you are managing for flood protection just how much hell are you going to catch if you go out on a limb, predict a dry spring, and have no flood pocket? You'd get crucified for all the resulting flood damage.