And there's the additional complication of those ancient (usually late) authors who spell it (properly, they thought) with the movable nu, since the alpha privative occurs before a vowel (ἀνυδρία, "lack of water"). As for ἀυδρία vs. ἀϋδρία, Plato manuscripts show the former, Theophrastus manuscripts (Historia plantarum 8.6.5.2) the latter. Plato and Theophrastus had no diaeresis-sign, and no need for it, so it's really a question of taste for medieval scribes. My preference would be for the most simple: ἀυδρία.
Additional observation: the disappearance in late antiquity of the initial rough breathing (dasia) of ὗδωρ, "water," might have inspired the need for the insertion of nu-movable — so an-ydria instead of the former a-hydria. The same could perhaps be said of the medieval scribe(s) of Theophrastus, for whom the diaeresis was a necessary sign that the now-unaspirated upsilon did not make a diphthong with the preceding alpha.