“…To overcome this drawback, we can distinguish adaptive approaches based on a posterior error estimates in the natural W 1,p -setting of these equations (see, e.g., [7,8,5,29]); or approaches based on error estimates in fractional (and Hilbert) Sobolev norms H s (see, e.g., [31]); or approaches based on weighted Muckenhoupt norms (see, e.g., [2,3,4]). On another hand, we can also find methods based on mesh-grading techniques (see, e.g., [6,23]), and methods based on regularization techniques (see, e.g., [48,46,45,11,35,10,34]). Many of the former results also apply for line sources (see, e.g., [23,32]).…”