Caveats.

The reader should be warned that this is an exercise in modeling and all that that implies. Errors can accumulate from a large number of sources...

• The river flow inputs are themselves the output of a model. The River flow model was calibrated from sparse and sometimes unreliable flow gauges. Some rivers were not gauged and inflow records were not calibrated at all.
• Data from two Raingauges, which are mere centimeters in diameter, were used as representative of a lake with surface area 3000000000000 .
• Evaporation records were insufficient and the record was artificially patched.
• The rule base model relies on the expert testimony of biological experts who...
  • have the awesome task of attempting to understand the immensely complex interacting behaviours of thousands of species swimming in of very murky water.
  • did not necessarily understand what we asked.
  • their thoughts were not necessarily correctly interpreted by us.
  • their rules were not necessarily correctly implemented in the model.
• Species interact. We have ignored interactions. Example. With increasing salinity massive die offs of vast aquatic weed beds occur sending a pulse of nutrients back into the system. This is not accounted for in the models.

Having said the above, then as Gavin Quibell argued at the inception of this project, ``This model is the best there is available, and the system must be managed. Even doing nothing is a management choice.''