Name

CRIT — Function to calculate numerical quality criteria

Calling Sequence

[Crit]=CRIT(OBS,CALC)

Parameters

OBS

Reference dataset (Matrix [Nx1] where N is the number of data).

CALC

Simulated dataset (Matrix [Nx1] where N is the number of data).

Crit

Calculated criteria.

Crit(1) : Root Mean Square Error

Crit(2) : Nash-Sutcliffe Criteria (%)

Crit(3) : C2M Criteria (%, bounded version of the Nash-Sutcliffe criteria: C2M = 100 x Nash / [200 - Nash])

Crit(4) : Mean error CALC-OBS

Crit(5) : Absolute error exceeded 20% of time

Description

  • This function permits to calculate different numerical criteria to evaluate simulation quality.

Examples

   // Fictious observed data
   t=(0:0.1:10)';
   OBS=sin(t);
   
   // Fictious calculated data (random error + autocorrelation)
   e=convol(1/5*ones(1,5),3*(rand(size(t,1),1)-0.5));e=e(5:$)';
   SIM=OBS+e;
   
   // Comparison
   Crit=CRIT(OBS,SIM);
   plot(t,OBS,'k:',t,SIM);legend(['Observed' 'Calculated'],a=3,%f);
   
  

Authors

Julien Lerat

CEMAGREF Antony, HBAN Unit, julien.lerat@cemagref.fr

Bibliography

Nash, J. E. and J. V. Sutcliffe (1970). "RIVER FLOW FORECASTING THROUGH CONCEPTUAL MODELS-PT 1." Journal of Hydrology 10(3): 282-290.