Abdelmounaime Safia put it well - It depends on how you are going to derive the map.

a) Are you using spectral classification for your landcover map If not, probably no correction unless the difference is strong.

b) Are you using pre-determined spectral class definitions? If so, yes.

c) Are you using adjacent images from the same sweep? If so, and the classes at the common edges match, probably no need for att. corr. but sun position still relevant.

d) Are you using images from multiple days, probably yes, but one can check if there is a significant atmospheric difference by comparing the same classification settings on land cover patches one has reason to expect will be the same.

Atmospheric correction is not necessarily benign. Over correction, or the wrong sort of correction, will introduce errors. Ultimately one needs to check that areas for which one knows the landcover class, e.g. water, are consistently classified in both cases.

It dependent on how are you going to derive the land use map. The idea is that if all your processes will be conducted using data from a single scene, so no atmospheric correction is needed. However, if your are using more than one scene (mosaic) or change detection study, you need to remove the atmospheric effect.

Relative measures need atmospheric correction, while in absolute measures, atmospheric correction can be omitted.

Also, I've just noticed that you are in the field of geology. So if you are using a spectral library, you need to perform atmospheric correction, because in this case it is about relative measures.

Atmospheric correction is a standard operation in image processing. I guess as suggested by Safia you should do it. However if in your landcover map you are looking at urban sprawl, infrastructure etc. you may probably go ahead without, provided you are using shape recognition. If your change detection is based on spectral characteristics alone, then stmospheric correction is a must.

Yes necessary when you are more concerned about the accuracy.

Abdelmounaime Safia put it well - It depends on how you are going to derive the map.

a) Are you using spectral classification for your landcover map If not, probably no correction unless the difference is strong.

b) Are you using pre-determined spectral class definitions? If so, yes.

c) Are you using adjacent images from the same sweep? If so, and the classes at the common edges match, probably no need for att. corr. but sun position still relevant.

d) Are you using images from multiple days, probably yes, but one can check if there is a significant atmospheric difference by comparing the same classification settings on land cover patches one has reason to expect will be the same.

Atmospheric correction is not necessarily benign. Over correction, or the wrong sort of correction, will introduce errors. Ultimately one needs to check that areas for which one knows the landcover class, e.g. water, are consistently classified in both cases.

Atmospheric correction is superfluous also for mosaics of satellite imagery for your purpose; please have a look on my paper with Westinga on Rwanda.

Thank you for your answers but the neccessary factor to caried out atmospheric corection for Landsat 8 are not avaliable till now so can I use the same factor of Landsat TM7 for Landsat 8 or the same equations.

Changing a little subject and according to its comments, could be said that the atmospheric correction is necessary to generate maps by means of digital cartography of soils? That is, to interpolate georeferenced soil data, using a mosaic of Landsat bands 8 among other inputs? For example organic carbon map of the soil, where the variable NDVI is very necessary.