An indeclinable word given as {cíw >>} coe “earth” in The Etymologies of the 1930s as a derivative of the root ᴹ√KEM (Ety/KEM; EtyAC/KEM).
Possible Etymology: The primitive form of rejected cíw is given as ᴹ✶kēm and its derivation is clear: the long ē became ī and then the final m reduced to w after i as usual. The derivation of coe is more obscure, however. The likeliest explanation is that Tolkien imagined its ancient form with a slightly lowered vowel which he generally represented as ǣ in this period (in later writings as ę̄). According to the first version of the Tengwesta Qenderinwa and Comparative Vowel Tables from the 1930s (PE18/46; PE19/25), ǣ > ei > ai > ae, and in The Etymologies itself, it seems ai often became oe instead of ae.
Neo-Sindarin: Updating the derivation of hypothetical ✱kę̄m would produced ᴺS. cae in Sindarin phonology. But given the obscurity of its derivation, I recommend using 1950s S. ceven for “earth” instead.
A noun in The Etymologies of the 1930s glossed “soil” derived from the root ᴹ√KEM “soil, earth” (Ety/KEM). It had a rather unusual plural form ceif which shows i-intrusion but not the usual i-raising of e to i; under ordinary phonetic developments the expected plural would be ✱cif. I can think of no good reason for this abnormal plural, but it would be ^caif if adapted to Sindarin. It is mostly moot since this word is very unlikely to be used in the plural.