or wave staff data, the columns headed "lower and upper limits" are the wave 

 height values for the 5 and 95% confidence levels shown graphically as 

 vertical lines through the values on Figures 2 through 9. The "long range" 

 columns under the predicted wave heights are the predicted values for the 

 30 or 36-hourly forecasts at the times of 0600Z and 1200Z, respectively. 

 Table V gives the forecast intervals for each of the 6-hourly times. 



Figures 10 and 11 show two scatter diagrams for wave height, one 

 for the USWB input and the other for the FNWF input. These figures disclose 

 the bias for the observed to predicted wave heights to be -1.6 feet and the 

 RMS error ±6.2 feet for 124 observations using the USWB input; for the FNWF 

 input the bias and RMS error are slightly less, +0.4 feet and ±5.1 feet, 

 respectively, for 66 observations. Bias as used in this report is the average 

 of the differences between the observed and predicted values. RMS values 

 were computed by taking the square root of the average of the square of the 

 differences between the observed and predicted values. Note that on Figures 

 10 and 11 the central diagonal line represents perfect correspondence between 

 predicted and observed wave heights. The vertical distance from this line 

 represents the plus or minus bias for each observation depending on whether 

 the plot lies above or below the central diagonal. The diagonals labeled 

 RMS show the positions on the graphs of the computed RMS errors for each 

 figure. 



In addition to evaluating the wave height data, a similar statistical 

 analysis was made of the wind directions and speeds. Table VI shows the 

 observed wind directions and speeds together with the machine predicted 

 values for the same times at the same various forecast intervals as shown in 

 Table V. It should be realized that there is a discrepancy between the locations 

 of the observation stations and the gridpoint locations used in the evaluation as 

 indicated on Figures 2 through 9. The effect of this discrepancy on both the 

 wave heights and the winds was not determined in this evaluation. 



Scatter diagrams for the wind directions, using the two different inputs, 

 are shown in Figures 12 and 13. Figures 14 and 15 are scatter diagrams for the 

 wind speeds, using the two different inputs. For the wind directions, the 

 USWB input produces slightly better results with a bias of +46 degrees and an 

 RMS error of ±75 degrees as against a bias of +52 degrees and an RMS error 

 of ±78 degrees for the FNWF input. A positive bias indicates that the predicted 

 winds veered from the observed directions. The analysis of the wind speeds 

 discloses very little difference between the two inputs. There is a bias of -5 

 knots and an RMS error of ±10. 1 knots for the USWB input as against a bias of 

 -4 knots and an RMS error of ±10.5 knots for the FNWF input. 



