Density methods may be applied to linear forms as well as point forms. Some statistics of interest include:

·         the number of line segments per unit area (sometimes called the line frequency, λ, rather than the density)

·         the length of line segments (by type) per unit area. The equation L_a=L₀e^‑r/b has been found to provide a good model of the distribution of urban road length with radial distance, r, from many city centres. L₀=the road density at the city centre, which for London is c.19.5 kms/km²; b is a parameter to be fitted from sample data. The inverse of this density measure is the average area associated with a unit of path length, sometimes called the constant of path maintenance

·         the number of line intersections per unit area

·         the number of line intersections per unit of length (the latter is known as the intersection density, k). This is a particular case of the density of points on lines, which may be subject to kernel density smoothing in the manner discussed earlier in this Section. It can be shown that if λ is the line density (frequency) then for a random network of straight lines there will be an average of k= λ ²/p intersections per unit length. K=k/λ² is sometimes called the crossings factor, and experimental evidence suggests that the random model provides an upper limit to this measure, so K has the range [0,1/p].

All these measures provide information that may be useful for analysis, including inter-regional comparisons. Applications include hydrology, ecology and road traffic analysis. For further discussion of applications in these areas and statistical measures see Haggett and Chorley (1969, Chapter 2.II) and Vaughan (1987, Chapter 3).

A number of GIS packages provide facilities for kernel density mapping of lines as well as points (e.g. ArcGIS Spatial Analyst, Density operations). Essentially these work in a similar manner to point operations, spreading the linear form symmetrically according to a specified kernel function. ArcGIS provides two sets of point density and line density functions (for SIMPLE or KERNEL density operations), one set that results in a grid being generated and the other for use in map algebra environments (i.e. where a grid already exists). SIMPLE density operations generate a set of grid values based on a neighbourhood shape (e.g. circular, annular, rectangular, or polygonal) and size (e.g. radius). Similar facilities do not appear to be available with most other mainstream GIS packages. The GRASS open source package include a vector processing function v.kernel (formerly s.kernel) that now includes linear kernel generation (i.e. a 1D kernel) in addition to its generation of 2D raster grids from vector point sets with a Normal (Gaussian) kernel.