The identification of a word for "soil" with that for "the whole planet," or at least the extent of its non-water regions, is quite old in many societies. It goes back as far as we can tell in all three branches of Germanic, and the Norse have a jötunn (titan) Jörð cognate with this word, who is often said to be the mother of Thor and is said by Snorri Sturluson to be biologically related in some way to night, prosperity, and day.

Beyond this, we are presented with a tantalizing number of leads, each of which is uncertain.

Epic and Doric have an adverb ἔραζε, (spelled ἔρασδε in Doric) "toward the ground" from which it can be inferred that there was an archaic word ἔρα "ground," but it never appears as such in documents we have, and even though it probably existed, that doesn't mean it's necessarily cognate with our word.

The Welsh word erw "acre" (formerly 11664 sq. ft.) is probably not cognate with "earth," but from the PIE root for "to plough."

Douglas Harper claims that there is a Middle Irish suffix -ert "ground," but it does not show up as a standalone word for earth, and even if it does exist, it could have been borrowed from the English.

Lithuanian erdvė̃ or érdvė meaning "space, expanse" is of unknown etymology. Likewise, Tocharian B yare "gravel," Armenian erkir "earth" are probably unrelated.

So our strongest lead for a Proto-Indo-European root is one Greek adverb; beyond that, each purported connection is unlikely. Nostraticists, of course, go even further to allege that not only does a PIE root exist, but that it is cognate with Proto-Semitic *ʾarṣ́ and some Proto-Dravidian root. This is exceedingly unlikely.