Mean period statistics in Table 4 for NDBC 42020
show negative bias for all the models indicating mean
periods are lower than measurements. The average
RMS errors for all models are 0.6-0.7 sec indicating
all the models have similar results. WAM shows a
slightly lower correlation mean over the year with 0.66
compared to WIS's 0.76. WIS scatter index mean and
skill score indicate slightly better comparisons for
WIS. Monthly mean period statistics for 42020 are
shown in Tables 20-22 in the Appendix. Peak period
results for 42020 in Table 5 again show negative
period biases for all models. WIS and WW3 show the
best mean correlations. Other statistics are similar for
all models. Monthly 42020 peak period statistics are
shown in Tables 23-25 in the Appendix.
6.2. WAVE COMPARISONS AT 42035
Monthly wave height comparison statistics for NDBC
42035 near Galveston, TX.,are shown in Tables 26-28
in the Appendix. Table 3 shows the mean statistics.
The 3G models show negative bias and WIS shows
positive bias indicating WIS is slightly over-predicting
and the 3G models under-predicting. RMS errors are
approximately 0.2 m for all models. WIS shows a
slightly higher correlation and WAM and WW3 show
slightly better skill scores and scatter indexes.
Figure 4. WW3 comparison results for October
Mean period statistics for 42035 are shown in Table 4.
1995 at NDBC 42001. Measured results in red and
Monthly statistics are shown in Tables 29-31 in the
WAM results in blue. Plots from top to bottom
Appendix. WIS shows a positive bias and WAM and
include significant wave height, wave period, wave
WW3 show a negative bias at this location. WW3
direction, wind speed, and wind direction.
shows the lowest RMS with 0.55 sec (WW3 RMS is
for the first three months of 1995). WW3 shows the
6.1 COMPARISONS AT 42020
best correlations and scatter indexes but all models
have similar skill scores. All model mean bias
Tables 17-19 in the Appendix show the monthly
statistics for peak period in Table 5 are negative
significant wave height statistics for the three
indicating peak period is slightly under-predicted.
hindcasts at NDBC 42020 located near Corpus Christi,
WAM and WIS have biases of 0.01-0.02 sec
TX. Table 19 only includes the first three months of
indicating very good agreement. Mean RMS error for
results for WW3 since the last Level 3 hindcast runs
peak period is 1.54 sec for WAM and WIS with WW3
are currently in the computer queue. Level 3 WW3
giving a lower mean of 1.04 sec over the Jan-Mar
for the full year will be shown at the conference.
period. WW3 shows the best values for scatter index,
Looking at the yearly mean statistics for WAM and
skill score and correlation for Jan.-Mar. Tables 32-34
WIS in Table 3 for 42020, WIS has a small positive
in the Appendix show monthly peak period statistics.
bias indicating over-prediction and WAM has a
negative bias indicating under-prediction. RMS
6.3. WAVE COMPARISONS AT 42036
values are very close with 0.38 m for WAM and
0.35 m for WIS. WW3 shows a negative bias
Significant wave height comparison statistics for
indicating it is lower than the measurements. WW3
NDBC 42036 located west of Tampa, FL, are shown
RMS error for January through March is similar to
in Tables 35-37 in the Appendix. Using the summary
both WIS and WAM. WIS shows slightly higher
Table 3, WAM and WW3 show negative bias
correlation than the other two models for the January
indicating slight under-prediction with WIS showing a
through March period, and the WIS yearly correlation
positive bias of 0.1m. WW3 shows the lowest RMS
mean is slightly higher than WAM.
statistic for Jan-Mar with 0.31m but WIS has the
lowest mean RMS error with 0.3 m. The scatter index
Previous Page
