Geodesic grid

A geodesic grid is a technique used to model the surface of a sphere (such as the Earth) with a subdivided polyhedron, usually an icosahedron.

Introduction

A geodesic grid is a global Earth reference that uses cells or tiles to statistically represent data encoded to the area covered by the cell location. The focus of the discrete cells in a geodesic grid reference is different from that of a conventional lattice-based Earth reference where the focus is on a continuity of points used for addressing location and navigation.

In biodiversity science, geodesic grids are a global extension of local discrete grids that are staked out in field studies to ensure appropriate statistical sampling and larger multi-use grids deployed at regional and national levels to develop an aggregated understanding of biodiversity. These grids translate environmental and ecological monitoring data from multiple spatial and temporal scales into assessments of current ecological condition and forecasts of risks to our natural resources. A geodesic grid allows local to global assimilation of ecologically significant information at its own level of granularity.

When modeling the weather, ocean circulation, or the climate, partial differential equations are used to describe the evolution of these systems over time. Because computer programs are used to build and work with these complex models, approximations need to be formulated into easily computable forms. Some of these numerical analysis techniques (such as finite differences) require the area of interest to be subdivided into a grid — in this case, over the shape of the Earth.

Geodesic grids have been developed by subdividing a sphere to developing a global tiling (tessellation) based on a geographic coordinates (longitude/latitude) where a rectilinear cell is defined as the intersection of a longitude and latitude line. This approach is easily understood in terms of accepted Earth reference, accessible using the longitude and latitude as an ordered pair, and implemented in a computer coding as a rectangular grid. However, such a pattern does not conform to many of the main criteria for a statistically valid discrete global grid, primarily that the cells' area and shape are not generally similar; this is especially noticeable as the cells are developed towards the poles.

Another approach gaining favour uses geodesic sphere grids generated by the subdivision of a platonic solid into cells or by iteratively bisecting the edges of the polyhedron and projecting the new cells onto a sphere. In this geodesic grid, each of the vertices in the resulting geodesic sphere corresponds to a cell. One implementation uses an icosahedron as the base polyhedron, hexagonal cells, and the Snyder equal area projection is known as the Icosahedron Snyder Equal Area (ISEA) grid. Another method, using the intersection of a tetrahedron into triangular quadtrees, is known as the Quaternary Triangular Mesh (QTM). A triangular mesh conforms well to representation in a graphics pipeline, and its dual cells are hexagons, convenient for encoding data. The hexagonal geodesic grid inherits many of the virtues of 2D hexagonal grids, and specifically avoids problems with singularities and oversampling near the poles. Along the same line, different Platonic solids could also be used as a starting point instead of an icosahedron or tetrahedron — e.g. cubes that are common in video games can be used to represent the Earth with a small aperture and an efficient equal area projection.

The quadrilateralized spherical cube is a kind of geodesic grid based on subdividing a cube into equal-area cells that are approximately square.

Positive traits

• Largely isotropic.
• Resolution can be easily increased by binary division.
• Does not suffer from over sampling near the poles like more traditional rectangular longitude/latitude square grids.
• Does not result in dense linear systems like spectral methods do (see also Gaussian grid).
• No single points of contact between neighboring grid cells. Square grids and isometric grids suffer from the ambiguous problem of how to handle neighbors that only touch at a single point.
• Cells can be both minimally distorted and near-equal-area. In contrast, square grids are not equal area, while equal-area rectangular grids vary in shape from equator to poles.

Negative traits

• More complicated to implement than rectangular longitude/latitude grids in computers

Parameters

The simplest tessellation of the icosahedron is characterized by the following parameters.

Let r_u be the radius of the circumscribed sphere and a₁ be the edge length of the icosahedron. From a view point at the center of the sphere, each edge appears under an angle γ₁,

$\sin\frac{\gamma_1}{2} = \frac{a_1}{2r_u}.$

By splitting each edge into s line segments of length a₁/s, and by projection of the intermediate points back onto the sphere, each triangle is split into s² smaller triangles, with associated viewing angles

$\gamma_s \equiv \gamma_1/s$

and edge lengths a_s

$\sin\frac{\gamma_s}{2} = \frac{a_s}{2r_u},$

$a_s=2r_u\sqrt{\frac{1-\cos(\gamma_1/s)}{2}}.$

(The out-projection lets a_s become larger than a₁/s.) The edges that are not generated by out-projection of one of the 30 edges of the icosahedron but from points inside the triangular mesh of any of the 20 surfaces have different lengths. That is, the triangles are generally not isosceles if s>1.

The solid contains 20s² tetrahedra with apexes at the sphere center, with base surface areas A_s,t and orthogonal distance r_i,s,t to the center of the sphere, t=1,...,s².

The total volume in all tetrahedra is

$V_s=20 \sum_{t=1,\ldots,s^2}\frac{r_{i,s,t}A_{s,t}}{3}.$

The volume filling factor

$f_s\equiv \frac{V_s}{4\pi r_u^3/3}$

relative to the volume of the circumscribed sphere approaches 1 as s grows to infinity.

The following table collects numerical values of these parameters, assuming that the (s+1)(s+2)/2 points of a regular flat triangular grid on each of the 20 surfaces are projected radially onto the circum-sphere:

s	as/ru	fs	image
1	1.051462224238267	0.6054613829125257	Simple icosahedron with 20 faces
2	0.5465330578253433	0.87345315725681	Subdivision with 4*20 faces
3	0.3669588160039673	0.9409379804627	Subdivision with 9*20 faces
4	0.2759044842552674	0.9661513525328	Subdivision with 16*20 faces
5	0.2209776477808279	0.97814560438654	Subdivision with 25*20 faces

History

The earliest use of the (icosahedral) geodesic grid in geophysical modeling dates back to 1968 and the work by Sadourny, Arakawa, and Mintz and Williamson. Later work expanded on this base.

A generalization to any number of dimensions, spherical design, was developed in 1984.

External links

• BUGS climate model page on geodesic grids
• Discrete Global Grids page at the Computer Science department at Southern Oregon University
• the PYXIS innovation Digital Earth Reference Model .
• Interpolation on spherical geodesic grids: A comparative study

• Williamson, 1969
• Cullen and Hall, 1979.

• Randall et al., 2000; Randall et al., 2002.