Another part of the ground survey employs a detailed procedure based upon the 
recommendations of Congalton and Green (1999) to provide accuracy assessment data from 
randomly positioned stations within each stratum (Young et ah, in review). Results were 
obtained in spring 2004 from 51 randomly positioned stations within intertidal eelgrass meadows 
and 28 randomly positioned stations within bare substrate strata of Yaquina Estuary. Based upon 
a comparison of results from the image classification with those from the ground survey (taken 
as the reference), application of the classical error matrix analysis yielded an overall accuracy of 
97%, with a Kappa Index value of 0.9447 + 0.0024, indicating excellent agreement (Landis and 
Koch, 1977). The investigators attribute this very high accuracy level to the extensive training 
data provided via GPS mapping of the intertidal eelgrass meadow margins (Young et ah, in 
review). 
Depth Distribution ofZ. marina 
The amount of Z marina at a specific tidal height (Figures 11.2 and 11.3) was 
determined by overlaying the Z. marina maps with the results of an extensive bathymetric depth 
survey conducted by the U.S. Army Corps of Engineers in 2002. The bathymetric model and the 
seagrass classification are both Arclnfo format grids. The bathymetric depth data was 
interpolated onto a grid using TopoGrid in Arclnfo. A 2.0 m floating point bathymetric model 
grid was integerized and resampled to 0.25 cell size to match the binary seagrass classification 
grid. The bathymetric and seagrass grids were then overlaid using the map algebra function 
COMBINE, which produces a grid value attribute table with counts of cells for each unique 
combination of cell values from each grid. 
The aerial mapping method utilized is capable of classifying some submerged seagrass 
beyond the depth at which near-infrared radiation is absorbed by water; however, some of the 
data returned by the COMBINE function is undoubtedly a result of spatial misregistration 
between the aerial photo classification and the bathymetric model. A few outliers were trimmed 
from both tails of the distribution curve. It is also quite possible that a physical survey could 
return some higher percentages of overall seagrass distributions below -1 m MLLW. The result 
of this analysis is presented as distribution curves of total area of intertidal eelgrass for ocean and 
river dominated estuarine areas. 
143 
i f 
