Page - 388 - in Book of Full Papers - Symposium Hydro Engineering

the above-mentioned goodness of fit measures (see Table 1). Mean absolute error as well as mean square deviation were significantly improved, the Pearson coefficient only slightly. Looking at the bias error, the perfect value of one is nearly achieved. Table 1 Error values before / after correction, example eastern Germany Goodness of fit measure before correction factor corrected qmap corrected qmap + factor corrected Pearson coefficient 0.51 0.62 0.54 0.55 Bias error 0.77 1.00 1.01 1.02 Nash Sutcliffe eff. coefficient 0.05 0.39 0.001 0.19 Mean absolute error 29.04 21.87 26.57 24.69 Mean square deviation 37.34 28.78 33.46 31.70 In addition, a qualitative statement can be made using a line graph (see Fig. 2), which maps the graphs for the forecast values before and after bias correction and the observed values. A direct comparison of the original NOAA monthly mean values with the values of the stations showed that amplitudes were not sufficiently reflected. After the bias correction, however, the range of observed values was enhanced and an improved representation of inner-annual patterns was achieved. Fig. 2 Comparison of NOAA forecast data before / after factor correction, example eastern Germany 388