Picture a globe divided into cells: near the poles they shrink into narrow wedges, while at the equator they stretch into wide bands. If you connect random points on such a globe with lines, the compass arrows will lie: connections will point to where there are more cells, not where the flows actually go. This is exactly what happens when scientists build networks from climate data or flight routes without accounting for the fact that nodes represent areas of different sizes.

The authors took a standard measure of directional bias, the mean direction and anisotropy (how much a node's connections point in one direction), and tested it on a sphere. They built a network of 18,432 nodes on a regular angular grid, where connections formed randomly but with a probability depending only on distance. Without correction, anisotropy grew from the equator toward the poles, peaking at about 60 degrees latitude, and the mean direction in the Northern Hemisphere pointed north. This is a pure illusion: random connections looked directional only because cells are smaller near the poles.

After correction, the zonal mean anisotropy dropped by a factor of one hundred, and the angle distribution became uniform, as expected for a random network. The method was then applied to real data. In a climate network based on surface temperature, zones of low anisotropy emerged in the centers of the Hadley and Ferrel cells, the giant atmospheric circulation patterns that transport heat from the equator toward the poles. In a precipitation network, the El Niño structure appeared: warm water off the coast of Peru and cold water near Australia create opposite directions of connections.

The aviation network of 3,425 airports showed that in Europe, departing flights on average head south, while in East Asia they converge toward central China. In the global trade network of 186 countries, imports and exports have different geometric patterns, with EU countries forming a zone where imports converge from all directions. The compass finally points true north, not grid north.

The method is ready for use in climatology to analyze teleconnections, which are long-distance links between atmospheric processes at different points on the planet. It will also be useful for transport logisticians and economists studying flows of goods and passengers. If there were no applications, it would be worth saying so plainly, but here they exist, and the data show it.