I assume you're referring to D2D render target clips, and swap chain presentation dirty regions? The two aren't mutually exclusive, so you can use both.
At a high level, a clip rect restricts subsequent drawing operations to the specified area, which allows the system to optimize what actually gets drawn. Clip rects can be pushed and popped as needed while drawing each frame. There
is some further info here:
http://msdn.microsoft.com/en-us/library/windows/desktop/dd372260(v=vs.85).aspx#clipping_an_arbitrary_shape
http://msdn.microsoft.com/en-us/library/windows/desktop/dd316860(v=vs.85).aspx
Invalid rects (assuming you mean dirty rects) and scroll rects on the other hand relate to the content after it is finished rendering. Dirty rects specify the area(s) of the backbuffer which have actually been updated with new content
for the frame you're presenting. The system can then use that info to only update the regions that have changed since the previous frame - this is particularly beneficial for scenarios like viewing an app over Remote Desktop, since only
the modified regions need to be transmitted and updated. There's a lengthy overview of the details available here:
http://msdn.microsoft.com/en-us/library/windows/desktop/hh706345(v=vs.85).aspx
