Flood　and　landslides　in　mountainous　cities　triggered　by　rainstorm　can　severely　impact　people＇s　lives,　property　and　socioeconomic　development.　The　pre-hazard　early　warning　system　are　crucial　to　the　disaster　prevention,　and　would　be　an　important　part　of　smart　city　planning.　This　paper　introduced　a　way　to　support　the　pre-hazard　identification　based　on　ground　water　level　change　fast　prediction,　which　is　the　key　factor　for　occurrence　of　the　rainstorm-induced　hazard.　Firstly,　the　remote　monitoring　stations　supplied　by　solar　power　are　established,　the　data　about　the　water　content　of　surface　soil　and　rainfall　were　real-time　collected　from　different　sensors.　By　introducing　the　sliding　windows　of　historical　data　and　for　prediction,　the　early　warning　system　are　effective　in　pre-hazard　identification　as　considering　the　vulnerable　environmental　factors　such　as　rainfall,　surface　runoff,　temperature,　sunshine,　vegetations,　soil　properties　and　structures.　Based　on　the　model　analysis　on　the　4982　time-series　samples,　taking　the　sliding　window　T=7d　of　historical　data　and　sliding　window　G=7d　for　prediction　as　an　example,　the　root-mean-square-error　(RMSE)　of　the　predicted　result　of　the　model　reached　0.0812,　with　R2　up　to　0.9776.　Thus,　Our　study　aims　to　provide　a　strategic　way　in　quick　response　for　the　connection　of　pre-disaster　planning　and　urban　planning,　and　the　improvement　of　disaster　prevention,　mitigation　capacities,　increasing　the　resilience　of　mountainous　cities　and　their　inhabitants.　©　2021　Elsevier　Ltd