Division 2.7.26 Standard grid calculation and refinement

When noise contours are obtained by interpolation between index values at rectangularly spaced grid points, their accuracy depends on the choice of the grid spacing (or mesh size) Δ _G , especially within cells where large gradients in the spatial distribution of the index cause tight curvature of the contours (see Figure 2.7.s). Interpolation errors are reduced by narrowing the grid spacing, but as this increases the number of grid points, the computation time is increased. Optimising a regular grid mesh involves balancing modelling accuracy and run time.

A marked improvement in computing efficiency that delivers more accurate results is to use an irregular grid to refine the interpolation in critical cells. The technique, depicted in Figure 2.7.s, is to tighten the mesh locally, leaving the bulk of the grid unchanged. This is very straightforward and achieved by the following steps:

1. (1)

   Define a refinement threshold difference ΔL_R for the noise index.

2. (2)

   Calculate the basic grid for a spacing Δ _G .

3. (3)

   Check the differences ΔL of the index values between adjacent grid nodes.

4. (4)

   If there are any differences ΔL > DL_R , define a new grid with a spacing Δ _G /2 and estimate the levels for the new nodes in the following way:

   If	ΔL ≤ ΔLR	calculate the new value	by linear interpolation from the adjacent ones.
   	ΔL > ΔLR		completely anew from the basic input data.

5. (5)

   Repeat steps 1-4 until all differences are less than the threshold difference.

6. (6)

   Estimate the contours by linear interpolation.

If the array of index values is to be aggregated with others (e.g. when calculating weighted indices by summing separate day, evening and night contours) care is required to ensure that the separate grids are identical.