Unlike our previous analysis, now that we do not have input points or areas to choose from, we have to delimit those areas based on different criteria. That alone raises the idea of using rasters. Additionally, this time we not only have Boolean criteria (inside or outside), but also have some continuous preferences (closer, or farther, the better). This factor calls for raster analysis. In raster analysis, we can consider almost the same classification as in vector analysis:

• Overlay analysis: Masking a raster layer with a binary mask layer. Where the binary mask layer has a zero value, we drop the value of the other raster layer, or set it to zero.
• Proximity analysis: Analyzing the distance between features or cells, and creating a raster map from the results. The raster map can contain real-world distances (Appendix 1.12) or raster distances (number of cells) from features or non-null cells in the input vector or raster map.
• Neighborhood analysis: Analyzing the neighborhood of the input raster. It usually involves convolution, which calculates some kind of statistics from the neighboring rasters of every cell, and writes the result in the appropriate cell of the output raster. The search radius can be circular or rectangular, and take an arbitrary size.

As you can see, the definitions have changed, as we cannot talk about geometries and attributes separately in case of raster data. Rasters offer full coverage of a rectangular area, therefore, if we use two perfectly aligned raster layers with coincident cells, the result will have the same cell number and cell size, and only the values matter. If not, a sophisticated GIS will resample one of the raster layers by simply aligning it with the other one, or interpolating its values during the process.