The　uncertain　mechanical　parameters　of　clay　layer　under　torrential　rain　are　the　key　to　the　dynamic　evolution　process　and　stability　assessment　of　landslide　geological　hazards.　Due　to　the　complex　environment,　engineering　geology　and　physical　chemistry　process,　the　mechanical　parameters　of　clay　layer　show　significant　spatial　variability　and　correlation.　In　addition,　due　to　technical　and　economic　conditions　constraints,　the　actual　investigation　and　test　data　of　soft　cohesive　soil　are　very　limited,　which　seriously　restricts　the　stability　evaluation　of　clay　slope　and　the　prevention　of　instability　disaster.　To　characterize　anisotropic　spatial　variations　of　uncertain　mechanical　parameters　for　clay　layer　using　incomplete　probability　data,　the　elastic　modulus,　Poisson　ratio　and　shear　strength　under　saturated　conditions　were　measured,　and　statistical　data　and　variation　properties　of　uncertain　mechanical　parameters　were　analyzed.　A　modeling　approach　was　proposed　for　characterizing　incomplete　probability　data　of　clay　layer.　The　accuracy　of　the　proposed　approach　is　verified　by　comparison　of　the　statistical　characteristic　for　measured　data　and　simulated　data.　A　novel　linear　fitting　method　was　proposed　for　assessing　scale　of　fluctuation　and　autocorrelation　distances.　The　variability　and　correlation　of　uncertain　mechanical　properties　for　soft　cohesive　soil　layer　are　discussed.　The　results　show　that　the　mechanical　properties　of　the　clay　layer　are　uncertain　in　spatial　position.　Both　the　original　observation　data　and　the　simulated　data　of　three　mechanical　parameters　have　symmetrical　correlation　structure.　The　clay　layer　display　the　horizontal　layered　structure　on　the　soil　profile,　and　the　vertical　autocorrelation　distances　are　shorter　than　the　horizontal　distances.　This　paper　clearly　illustrates　the　anisotropic　spatial　variations　of　uncertain　mechanical　parameters　for　clay　layer　using　incomplete　probability　data　and　it　can　provide　scientific　data　for　the　uncertainty　analysis　and　risk　assessment　of　clay　slope　under　torrential　rain　conditions.　©　2024　Elsevier　Ltd