Buffering

<< Click to Display Table of Contents >>

Navigation:  Building Blocks of Spatial Analysis > Distance Operations >

Buffering

Buffering is the process of creating one or more zones around selected features, within a pre-specified distance (almost always Euclidean distance) from these features. The resulting buffer is an area object or series of area objects (typically polygons) that may be automatically merged (dissolved) where they overlap. In many cases the buffer regions are created from a combination of straight lines and circular arcs centered on the vertices of the convex hull. In Figure 4‑72A simple symmetric buffers generated using the GRASS v.buffer operation are shown, whilst in Figure 4‑72B the result of applying a 100metre buffer to sections of road network is shown, without buffer segment merging (selected segment highlighted). Buffers may be applied to:

point, line or polygon objects, or to cells in a grid dataset

they may be symmetric, i.e. spreading outwards from lines and polygon boundaries in both directions (left and right or inside and out), or

they may be asymmetric, spreading inwards or outwards only (see for example the buffering used in polygon skeletonization, Figure 4‑14)

buffers may also be applied differentially, for example larger buffers applied to major roads and smaller buffers applied to minor roads, as part of a pollution analysis exercise (air-borne, noise, vibration etc.)

buffers may be applied incrementally, progressively increasing and/or reducing in size

Although the idea of buffering appears very straightforward, and has many applications, different GIS packages implement the facility in a variety of ways. In ArcGIS extensive support is provide in ArcToolbox for buffer creation of selected features, with facilities to select buffering: to the left, right or both sides of polylines and polygons; to select whether linear feature buffers are terminated with rounded or squared-off ends; and to select features which may have their buffered regions combined (dissolved together) and optionally a selected attribute (field) to be used for determining whether dissolve operations should be performed. An additional buffer tool is provided to generate multiple buffer rings in a single operation. MapInfo provides a similar facility, which it calls “Concentric ring buffers”.

Figure 4‑72 Simple buffering

A. v.buffer in GRASS

B. Buffer tool, line element, ArcToolbox

clip0126.zoom50

clip0127.zoom50

Manifold’s principal buffering facility is invoked via the Transform toolbar, and the standard Buffers operation when applied to a polygonal area provides a buffer that extends outside the polygon plus the entire polygon internal area, effectively growing the entire polygon. To limit the buffer zone to the boundary of a polygon the Border buffer zone operation must be used (Figure 4‑73B). An additional operation, Inner buffer zone is provided which selects the inner part of a polygon, comprising all areas at least the buffer distance from the boundary (Figure 4‑73C). Repeated buffering operations may be applied, either to the original feature or to the resulting buffered entity itself. Buffer features may also by modified, e.g. by clipping operations or intersection, to generate new or modified zones, as required. The Buffers operation in Manifold creates distinct buffer zones (Figure 4‑74A), whilst an alternative function, ‘Common buffers’ creates merged zones (Figure 4‑74B).

Figure 4‑73 Manifold: Buffer operations

A. Selected polygon

B. Border buffer

C. Inner buffer

clip0128

clip0129

clip0130

 

Figure 4‑74 Manifold: Buffering options

A. Buffer

B. Common buffer

clip0131

clip0132

Once a buffer has been created, or as part of the creation process, a new map layer containing the buffer polygons will be generated. This map layer will have an associated attribute table, and if not already generated, attributes such as polygon areas can be computed. The new layer may, in turn be combined with other layers (e.g. via an Intersect overlay) to analyze attribute data within the buffer zone.

Raster buffering

Buffering may also be applied in raster GIS, whereby cells are classified according to their distance from one or more target cells (grid distance, Euclidean distance or a more complex, cost-related distance). For example, in Idrisi by default target cells are coded as 1, buffer cells as 2 and other cells as 0, and the buffer width to apply is specified in reference system units (based on true Euclidean distances or grid-based cost distances). The process of generating a buffer in this environment is very similar to performing a standard image processing task, that of generating a distance transform of the (binary) input image, and then selecting from the resulting output image all those cells that have entries (distance values) less than or equal to the buffer width specified. These selected cells are then re-classified (RECLASS in Idrisi) as required (e.g. to a value of 2; see also, Section 4.4.2, Cost distance).

Hybrid buffering

Buffer operations within vector GIS generate regions, and distances within those regions meet the overall buffering criterion set. The use of multiple buffering does enable a degree of gradation within such buffer regions, but for more detail some form of hybrid buffering may be more appropriate. In this approach either hybrid buffering is directly supported by the GIS tool or the relevant vector objects will require conversion to grid format and then grid buffering applied. The latter may be achieved using cost distance (ACS) methods or distance transforms (DTs), as described in Section 4.4.2, Cost distance.

Network buffering

A simple buffer operation applied to a vector map of roads will create a set of merged or unmerged regions either side of the selected roads. An alternative form of buffering for such networks is a time or distance buffer from a point source or sources with respect to the network. In the case of roads, this form of buffer generates travel-distance or drive-time zones. The Manifold GIS Business Tools option provides for such buffering. In their model drive times are computed to all network points in one or more time zones, plus a distance buffer (speed related) for off-road travel (e.g. walking). This is a very processor intensive operation. TransCAD and ArcGIS Network Analyst provide similar functionality via their “drive time” analysis facilities (see, for example, Figure 7‑18). Basemap provide more advanced toolsets that incorporate multi-modal travel and timetables.