Corrosion-induced　deterioration　of　the　in-service　marine　reinforced　concrete　(RC)　structures　may　result　in　unsatisfactory　serviceability　or　insufficient　safety.　Surface　deterioration　analysis　based　on　random　fields　can　provide　information　regarding　the　future　development　of　the　surface　damage　of　the　in-service　RC　members,　but　its　accuracy　needs　to　be　verified　in　order　to　broaden　its　applications　in　durability　assessment.　This　paper　performs　an　empirical　study　to　verify　the　accuracy　of　the　surface　deterioration　analysis　based　on　random　fields.　The　batch-casting　effect　is　considered　to　establish　the　＂step-shaped＂　random　fields　for　stochastic　parameters　in　order　to　better　coordinate　their　actual　spatial　distributions.　Inspection　data　from　a　23-year-old　high-pile　wharf　is　obtained　and　analyzed　in　this　study.　The　simulation　results　of　the　RC　panel　members＇　surface　deterioration　are　compared　with　the　in-situ　inspection　results　with　respect　to　the　steel　cross-section　loss,　cracking　proportion,　maximum　crack　width,　and　surface　damage　grades.　It　shows　that　the　simulation　results　coordinate　well　with　the　inspection　results.　On　this　basis,　four　maintenance　options　are　established　and　compared　in　terms　of　the　total　amounts　of　RC　panel　members　needing　restoration　and　the　total　economic　costs.　It　provides　a　comparative　tool　to　aid　the　owners　in　selecting　the　optimal　maintenance　action　given　the　inspection　results,　to　minimize　the　lifecycle　cost　and　guarantee　the　sufficient　serviceability　and　safety　of　the　structures.