Land　use　patterns　are　governed　by　a　broad　variety　of　potential　driving　forces　and　constraints　which　act　over　a　large　range　of　scales　and　multi-scale　investigation　of　land　use　patterns　is　essential　for　full　understanding　of　its　complexity.　The　main　purpose　of　this　paper　was　to　perform　a　multi-scale　analysis　of　arable　land　distribution　pattern　of　Fujian　province　by　means　of　statistical　analysis　through　overall　study　and　each　agro-zone　respectively.　27　variables　were　selected　as　the　candidate　land　use　drivers　representing　bio-geophysical,　socio-economic　and　infrastructural　conditions.　The　basic　spatial　organization　in　the　analysis　was　a　1km×1km　geographical　grid.　Through　aggregations　of　these　cells,　a　total　of　10　artificial　aggregation　levels　were　obtained.　The　independent　models　of　the　whole　study　area　and　each　6　agro-zones　of　arable　land　distribution　patterns　were　constructed　at　multiple　scales　respectively.　The　results　showed　that　Land　use　models　varied　with　aggregation　level　and　also　between　agro-zones.　Independent　variables　explained　more　of　the　variance　for　the　explanation　of　land　use　type　at　higher　aggregation　levels.　Except　slope,　the　highest　ranking　variable,　other　variables　of　the　arable　land　use　model　vary　between　agro-zone　I　to　VI.　But　the　general　rule　is　that　arable　land　in　all　6　agro-zones　is　strictly　restricted　by　topographic　factors　which　changes　little　along　with　time.　It　is　argued　that　these　types　of　analyses　can　support　quantitative　multi-scale　understanding　of　land　use,　needed　for　the　spatially　explicit　land　use　change　models.