Sign in

April 2010 - Adjusted standard deviation water level predictions

2018-04-25 10:01:16
The standard deviations for the water level predictions are reduced to the absolute average prediction error plus the historical standard deviation of predictions.

PROTIDE uses a standard deviation around its data settings to model the uncertainty of the predictions. This is done in a monte carlo simulation where water levels are drawn from a normal distribution with an average of the prediction and a given standard deviation of the data setting. The standard deviations were taken on the safe side (larger then the expected error of the prediction) while the actual prediction errors had to be investigated. The actual prediction errors have now been calculated for most water level data settings. The result can be found under in the Settings - Calculation Data - Data settings menu. All data settings now have a lower standard deviation, which means that tidal windows will become larger.

The historical errors in the predictions have been analysed to make these improvements. In this analysis there (evidently) is an average error and a standard deviation. The average error suggests a structural problem in the predictions and the standard deviation represents the actual variance between prediction and observation. To be on the safe side we added the two to get the standard deviation that PROTIDE uses in its calculations.