ArcGIS Tutorial: Raster cell size and resampling

Different raster datasets do not need to be stored with the same cell resolution. However, when working with multiple datasets, it is best to use the same cell resolution. When you enter multiple raster datasets of different resolutions into any ArcGIS Spatial Analyst extension tool, one or more input datasets are automatically resampled to the most coarse resolution of the input dataset.

By default, the nearest neighbor allocation resampling technique is used. This is because it applies to both discrete and continuous data, while other resampling types (bilinear interpolation and three convolution) apply only to continuous data. It is necessary to use resampling techniques because the center of the input cell is rarely aligned with the center of the cell after the transformation to the desired resolution. Before merging rasters of different resolutions, you can use the Resampling tool as a preprocessing step using bilinear and three convolution techniques.

You can use the cell size environment parameter to control the default resampling resolution, in which you can specify whether the tool uses the minimum resolution of the input raster or the specific cell size that is defined.

As shown, the cell size set in the analysis environment is coarser than the cell size of the input raster in the tool. When executed, the input raster is first resampled to a coarser resolution, and then the tool is applied.

　　

When performing the analysis, determine if the cell size you are setting is appropriate. For example, when the cell size is 5-kilometer, it is unlikely that you will be studying mouse movement. 5-kilometer of cells may be more suitable for studying the effects of global warming on the earth.

Resampling

To find the value that each cell should take on the resampled output raster, you must map the center of each cell in the output raster to the original input coordinate system. The center coordinate of each cell is reversed to determine the position of the point on the original input raster. When the input location is determined, a value is assigned to the output location based on the neighboring cells in the input raster. The output cell Center is rarely aligned exactly with the input raster cell center. Therefore, several techniques have been developed to determine the output value based on the location of the point relative to the input raster cell center and the values associated with those cells. The three techniques used to determine the output value are nearest neighbor allocation, bilinear interpolation, and three convolution interpolation methods. Each technology assigns output values in different ways. Therefore, the values assigned to the cells of the output raster may vary depending on the technology being used.

　　Nearest Neighbor allocation method

Because the nearest neighbor assignment does not change the value of the input cell, it is a resampling technique for discrete (categorical) data. After locating the location of the cell center in the output raster dataset to the input raster, the nearest neighbor assignment determines the closest cell center location on the input raster and assigns the value of that cell to the cells on the output raster.

The nearest neighbor assignment does not change any of the values of the cells in the input raster dataset. The value 2 in the input raster will still be 2 in the output raster, never 2.2 or 2.3. Because the output cell values remain constant, the nearest neighbor allocation method should be used for nominal or ordinal data, where each value represents a class, a member, or a taxonomy (categorical data, such as land use, soil, or forest type).

Considering that the output raster created from the input raster will rotate 45 ° in the operation, resampling will occur. For each output cell, you get the value from the input raster. In, the cell center of the input raster is a gray point. The output cell is shaded in green. The cells being processed are shaded in yellow. In the nearest neighbor allocation method, the center of the input raster (orange dot) closest to the center of the cell being processed (the red dot) is determined and specified as the output value of the cell (yellow shadow) being processed. Repeat this procedure for each cell in the output raster.

　　

　　Bilinear interpolation method

bilinear interpolation uses values from the center of the four nearest input cells to determine the value on the output raster. The new value of the output cell is a weighted average of these four values, adjusted according to their distance from the center of the output cell. This interpolation method produces a smoother surface than the nearest neighbor allocation method.

As with the legend of nearest neighbor interpolation, the cell center of the input raster is a gray point, the output cell is a green shadow, and the cell to be processed is a yellow shadow. For bilinear interpolation, first determine the four input cell centers (orange dots) nearest to the center of the cell being processed (the red dots), then calculate their weighted average, and then specify the resulting value as the output value of the cell (yellow shadow) being processed.

　　

Because the output cell value is calculated based on the relative position and value of the input cell, bilinear interpolation is the preferred method for determining the data (that is, continuous surfaces) that is assigned to the cell value by the location of a known point or phenomenon. The elevation, slope, noise intensity of the airport and the salinity of the groundwater near the estuary are all phenomena that are represented as continuous surfaces and are best suited for resampling using bilinear interpolation.

　　Three convolution interpolation method

The three-time convolution interpolation method is similar to bilinear interpolation, which calculates the weighted average value by the 16 nearest input cell centers and their values.

Shows how to calculate the output value of the three convolution interpolation method. First determine the 16 cell centers (orange dots) closest to the center of the cell being processed (the red dots), then calculate their weighted average, and then specify the resulting value as the output value of the cell (yellow shadow) to be processed.

　　

Compared with bilinear interpolation, the three-time convolution method tends to sharpen the edges of the data because more cells are involved in calculating the output value.

　　Resampling and data types

You should not use bilinear interpolation or three convolution interpolation for categorical data, because the output raster dataset does not retain categories. However, these three technologies are available for continuous data, where nearest neighbor generates blocky output, bilinear interpolation produces smoother results, and three convolution methods produce the clearest data.

ArcGIS Tutorial: Raster cell size and resampling