19 Forecasting the Coastal Optical Properties Using Satellite Ocean Color
343
there will be a high standard deviation compared to the low mean error. Significant
standard deviation errors (darker red) are observed in coastal areas associated with
strong tidal regimes and the dissipation of the Mississippi River plume.
However, the total number of points used to determine the mean and standard
deviation is critical to determine statistical validity. The number of matchup’s pairs
used to compute the mean and standard deviation for the month is represented
in Fig. 19.5c at each location. This number represents the number of times during the month that a difference between the forecast field and a satellite retrieved
backscattering product was computed. The greater number of matchup pairs, i.e.
>10+ (color coded in white), represents valid statistical relationships,compared to
the low numbers such as 4 and 5 (color coded in light and dark blue).
Notice that in areas where there are high number of data pairs (white in
Fig. 19.5c), the monthly mean forecast errors are low (white in Fig. 19.5a) especially in the western region. The Mississippi River plume statistics are mixed.
Representative errors occur where there are high numbers (white in Fig. 19.5c) in
addition to a high mean error (red or blue in Fig. 19.5a). In the areas where there are
lower number of match ups (light or dark blue in Fig. 19.5c) and the forecast error is
high (red in Fig. 19.5a) the forecast errors are not representative. At these locations
where the numbers are low, the statistical forecast error is unreliable.
In order to determine how the forecast of the backscattering compares with
monthly “climatology”, we examined the monthly mean and standard deviation of
the backscattering coefficient for October 2009, which was computed based only
on satellite derived backscattering (Fig. 19.6). As expected, the mean backscattering distributions do not show the small scale plumes and eddies along the coast as
observed in October 19, and in the optical forecast. The monthly mean distribution
is much different from the individual day’s imagery and the forecast. The monthly
standard deviation of backscattering (Fig. 19.6b) represents substantial changes in
coastal backscattering which we believe is primarily resulting from the monthly
Fig 19.6 Monthly October 2009 backscattering (551) coefficient derived from MODIS –Aqua:
(a) monthly mean and (b) standard deviation from the mean
343
there will be a high standard deviation compared to the low mean error. Significant
standard deviation errors (darker red) are observed in coastal areas associated with
strong tidal regimes and the dissipation of the Mississippi River plume.
However, the total number of points used to determine the mean and standard
deviation is critical to determine statistical validity. The number of matchup’s pairs
used to compute the mean and standard deviation for the month is represented
in Fig. 19.5c at each location. This number represents the number of times during the month that a difference between the forecast field and a satellite retrieved
backscattering product was computed. The greater number of matchup pairs, i.e.
>10+ (color coded in white), represents valid statistical relationships,compared to
the low numbers such as 4 and 5 (color coded in light and dark blue).
Notice that in areas where there are high number of data pairs (white in
Fig. 19.5c), the monthly mean forecast errors are low (white in Fig. 19.5a) especially in the western region. The Mississippi River plume statistics are mixed.
Representative errors occur where there are high numbers (white in Fig. 19.5c) in
addition to a high mean error (red or blue in Fig. 19.5a). In the areas where there are
lower number of match ups (light or dark blue in Fig. 19.5c) and the forecast error is
high (red in Fig. 19.5a) the forecast errors are not representative. At these locations
where the numbers are low, the statistical forecast error is unreliable.
In order to determine how the forecast of the backscattering compares with
monthly “climatology”, we examined the monthly mean and standard deviation of
the backscattering coefficient for October 2009, which was computed based only
on satellite derived backscattering (Fig. 19.6). As expected, the mean backscattering distributions do not show the small scale plumes and eddies along the coast as
observed in October 19, and in the optical forecast. The monthly mean distribution
is much different from the individual day’s imagery and the forecast. The monthly
standard deviation of backscattering (Fig. 19.6b) represents substantial changes in
coastal backscattering which we believe is primarily resulting from the monthly
Fig 19.6 Monthly October 2009 backscattering (551) coefficient derived from MODIS –Aqua:
(a) monthly mean and (b) standard deviation from the mean