The geometry-shader (GS) stage runs application-specified shader code with vertices as input and the ability to generate vertices on output. Unlike vertex shaders, which operate on a single vertex, the geometry shader's inputs are the vertices for a full primitive (two vertices for lines, three vertices for triangles, or single vertex for point). Geometry shaders can also bring in the vertex data for the edge-adjacent primitives as input (an additional two vertices for a line, an additional three for a triangle).
When trees are far away, a billboarding technique is used for efficiency. That is, instead of rendering the geometry for a fully 3D tree, a quad with a picture of a 3D tree is painted on it. From a distance, you cannot tell that a billboard is being used. However, the trick is to make sure that the billboard always faces the camera (otherwise the illusion would break).
A common CPU implementation of billboards would be to use four vertices per billboard in a dynamic vertex buffer. Then every time the camera moved, the vertices would be updated on the CPU, so that the billboard face the camera. This approach must submit four vertices per billboard to the IA Stage, and requires updating dynamic vertex buffers, which has overhead. With the geometry shader approach, we can use static vertex buffers because the geometry shader does the billboard expansion and makes the billboards face the camera. Moreover, the memory footprint of the billboards is quite small, as we only have to submit one vertex per billboard to the IA stage.
Introduction to 3D Game Programming with DirectX11, Frank Luna, 2012