Display of pixel loss and replication in re-projecting raster data from the sinusoidal projection.
They are the equal-area cylindrical, sinusoidal, Mollweide, Eckert IV, Hammer-Aitoff, interrupted Goode homolosine, integerized sinusoidal projections, and the equal area global gridding method with a fixed latitudinal metric distance (Seong 2005).
The longitudinal nature of data loss and duplication for resampling onto the sinusoidal projection can be examined by plotting the column tallies for loss and duplication cells in the DLDM quadrant illustrated in Figure 6.
The pattern is symmetric around the 45[degrees]E meridian and is nearly identical to the sinusoidal projection DLDM from around 60[degrees]N to the pole.
To begin, the plotted percentages do not form the narrow line seen with the sinusoidal projection, but rather a band of values at the equator that narrows to a line near the pole.
If the data loss curve for the sinusoidal projection (Figure 7) were overlaid on the data loss and no loss points in Figure 11, there would be an excellent fit from 45[degrees]N to the pole.
Similarly, all pseudo-cylindrical projections with horizontal parallels and smoothly curving meridians should have a data loss and duplication pattern on their DLDM that resembles the pattern for the sinusoidal projection.