Create your k zones with a gis and import data (shp) in feflow. You can create on the gis diffrent zones with a little gap between. With the gis, just generate points at equal distance between all nodes on the line of the k polygons. Import data in feflow and interpolate using inverse distance weithing. This will generate a kind of contouring of your k values and the gap between the k zones will create a smoothest k distribution, allowing feflow an easier convergence