ire ction of arriva l ( DOA) est imation algori thm in non-G aussi an im pul se noise envir onment, Alpha s tabl e distri-but ion was ta ken as imp ulse noise mode l. Firs tly, two no nlin-ear transfor m methods , fra ctiona l l ow order t rans form and compr ession tra nsform, we re used t o prepro cess th e array re-ceived signal to effectiv ely suppr ess impulse noise. Then, two robust dual-size unitary ESPRIT ( U-DS-ESPRIT) algorithms were p ropose d based  on  the cova rianc e matr ix of the  p repro-cessed dat a. Finally, the Crame r R ao boun d (CR B) of DOA estimati on for distributed array s in Alpha stable dis tributed impu lse n ois e enviro nment was analyzed . Si mulation r esults show tha t i n im puls e nois e environment, the estim ation per-fo rman ce of the two robust U-DS-E SPRI T al gor ithm s proposed in thi s paper is signifi cant ly better th an that of t he U-DS -ES-PR IT algorith m based on t rad ition al covariance m atrix, and the e stim ation perfor mance of the U -DS-E SPRIT al gorithm base d on compression tra nsformation is much clo ser to CRB.