Multiple tools in ENVI 4.5 were used to complete the following
processing steps: Data conversion from digital number to radiance,
correction for atmosphere and water surface reflection,
linearization of spectral decay as function of depth, masking of
non-applicable values, and bathymetric derivation using variables
from multiple linear regression analysis. Multiple tools in
ArcGIS 9.3 were used for dataset integration and to extract values
for the multiple linear regression analysis and subsequent error
analyses. The statisitcs program S-Plus was used for the multiple
linear regression analysis to provide original variables for depth
derivation. The multivariate slope intercept formula used to
derive depth was Depth=Yint+(Mblue)(Xblue)+(Mgreen)(Xgreen). Due to
drastically different water qualities between interior and exterior area of
the atoll, two images were split into regions representing areas inside and
outside of the atoll. Effectively creating four images that were analyzed
seperately. Forty two derived products, nineteen from the image
opo_86356_msi_0000000 (Pal35600in: fourteen products, Pal35600out: six
products) and Twenty-three from the image opo_86356_msi_0010000 (Pal35601in:
fifteen products, Pal35601out: eight products) were integrated to expand spatial
coverage.
The original MLR variables for Pal35600in were Yint=7.5076,
Mblue=-6.0909, and Mgreen=18.8197. The product derived from these
variables (Pal35600in_DB) was mosaiced with thirteen additional products derived
using increasingly reduced Y-intercept values as follows:
Yint=6.5076 (Pal35600in_DB1), Yint=5.5076 (Pal35600in_DB2), Yint=4.5076
(Pal35600in_DB3), Yint=3.5076 (Pal35600in_DB4), Yint=2.5076 (Pal35600in_DB5),
Yint=1.5076 (Pal35600in_DB6), Yint=0.5076 (Pal35600in_DB7), Yint=-1.5076
(Pal35600in_DB8), Yint=-2.5076 (Pal35600in_DB9), Yint=-3.5076 (Pal35600in_DB10),
Yint=-4.5076 (Pal35600in_DB11), Yint=-5.5076 (Pal35600in_DB12), Yint=-6.5076
(Pal35600in_DB13) and Yint=-7.5076 (Pal35600in_DB14). This was done to increase
coverage in shallow areas. More information on this integration process is
provided in the product error analysis, availble from PIBHMC upon request.
The original MLR variables for Pal35600out were Yint=23.5804,
Mblue=-1.5047, and Mgreen=16.1530. The product derived from these
variables (Pal35600out_DB) was mosaiced with five additional products derived
using increasingly reduced Y-intercept values as follows:
Yint=19.5804 (Pal35600out_DB4), Yint=15.5804 (Pal35600out_DB8), Yint=11.5804
(Pal35600out_DB12), Yint=7.5804 (Pal35600out_DB16) and Yint=3.5804
(Pal35600out_DB20). This was done to increase coverage in shallow areas. More
information on this integration process is provided in the product error analysis,
availble from PIBHMC upon request.
The original MLR variables for Pal35601in were Yint=9.1819,
Mblue=-7.8389, and Mgreen=23.5717. The product derived from these
variables (Pal35601in_DB) was mosaiced with fourteen additional products derived
using increasingly reduced Y-intercept values as follows:
Yint=8.1819 (Pal35601in_DB1), Yint=7.1819 (Pal35601in_DB2), Yint=6.1819
(Pal35601in_DB3), Yint=5.1819 (Pal35601in_DB4), Yint=4.1819 (Pal35601in_DB5),
Yint=3.1819 (Pal35601in_DB6), Yint=2.1819 (Pal35601in_DB7), Yint=1.1819
(Pal35601in_DB8), Yint=0.1819 (Pal35601in_DB9), Yint=-1.1819 (Pal35601in_DB10),
Yint=-2.1819 (Pal35601in_DB11), Yint=-3.1819 (Pal35601in_DB12), Yint=-4.1819
(Pal35601in_DB13), Yint=-5.1819 (Pal35601in_DB14) and Yint=-6.1819 (Pal35601in_DB15).
This was done to increase coverage in shallow areas. More
information on this integration process is provided in the product error analysis,
availble from PIBHMC upon request.
The original MLR variables for Pal35601out were Yint=30.4352,
Mblue=-7.2710, and Mgreen=27.6316. The product derived from these
variables (Pal35601out_DB) was mosaiced with seven additional products derived
using increasingly reduced Y-intercept values as follows:
Yint=26.4352 (Pal35601out_DB4), Yint=22.4352 (Pal35601out_DB8), Yint=18.4352
(Pal35601out_DB12), Yint=14.4352 (Pal35601out_DB16), Yint=10.4352 (Pal35601out_DB20),
Yint=6.4352 (Pal35601out_DB24) and Yint=2.4352 (Pal35601out_DB28).
The derived bathymetry from each image were in turn mosaiced blending the
the concurrent data from image Pal35600out and Pal35601out to minimize seams in the
area of overlap. Data from this combined grid was prioritized of that from Pal35600in
and data from Pal35601in received the lowest priority. Data with better statistical
accuracy was given higher "priority" so that they replaced concurrent data with lowere
accuracy during the mosaic process.
This mosaiced derived bathymetry product (Pal_DBall) was then resampled to a 5 meter
grid and integrated with the multibeam sonar data, with sonar data prioritized over
derived data, to create the final product (Pal_DBMB).
A detailed description of all processing steps is available at:
http://oregonstate.edu/~hogrefek/Cookbook/Cookbook_042108.pdf
An error analysis of each derived bathymetry grid used is available
upon request.
