To keep speed up, some techniques rely on fast and low resolution solutions which could result in wrong looking shadows which may be out of place in a scene.
Others require the CPU and GPU to calculate with algorithms the exact location and shape of a shadow with a high level of accuracy.
The method basically makes a 3D object occupy an enclosed volume in a scene where a shadow is being cast.
Chris Green of Valve, a video game maker, says that bump map data is derived from geometric descriptions of the objects surface significant lighting cues due to lighting occlusion by surface details are not calculated.
[3] A common fix is to use an additional texture channel to create an ambient occlusion field.