Digital Signal Processing Reference
In-Depth Information
From the comparison of the profiles for the image and the model for the two selected sections, the
above examples show in the most of the cases an acceptable trend of the model in comparison with the
image profiles for the three selected dates. This level of qualitative calibration indicates that the model
is predicting acceptable spatial profiles for the TSM concentrations, which can be observed as rough
plume patterns in the images. Analysing the spatial patterns from the 14 selected images it is clear that
there is a spatial and temporal variability of the patterns from one location to the other. Specific
variations are obvious at the entrance points of the main drains to the lake and also at the far south
western area. The 14 images representing seasonal variations of the flow patterns in the lake were
staked in ascending temporal sequence (January-July) using the image processing programme ENVI,
and a vertical common profile was drawn to represent the temporal variation at certain selected pixel
locations.
Pixel to Pixel Comparison
A group of five locations was selected inside the lake, representing some of the measuring points of
water quality samples namely (LO5, LO17, LO18, LO19, LO20), in order to test the variation. An
averaging window is selected around each location. These averaged values were compared to the
model data values at the same locations. The following section shows the compared averaged window
pixels and the corresponding model values for the five selected locations.
Figure (8-12)
shows the
selected measuring locations and the selected averaging pixels window.
LO20
LO5
Selected
locations
Average pixels window
(6x6)
LO19
LO18
LO17
Figure (8-12): Selected measuring locations for temporal comparison and average pixels window (6x6) applied
for each selected location
This qualitative calibration is based on a comparison of the time series of the model and the image
based on original pixel values of the image before applying any analytical algorithm to the images.
Figure (8-13-a,b,c,d)
shows the compared image pixel values and the corresponding model pixel
values; the trends are similar, e.g. LO5 and LO20, but there are variations within an acceptable range.
The reasons for the variations can be found mainly in the uncertainty in georeferencing coarse
resolution images that affects most of mixed pixels, the presence and dynamics of submerged and
floating vegetation, and the uncertainties inherent in the atmospheric corrections of the MODIS L2
products.
Search WWH ::
Custom Search