A　nonlinear　time　series　model　is　presented　to　describe　the　dynamics　of　ground　water　flow.　The　procedures　for　model-establishing　and　parameter　estimation　are　discussed.　The　proposed　nonlinear　model　uses　a　threshold　parameter　that　relates　different　precipitation　processes,　and　a　lag　parameter　that　relates　the　time　lag　between　precipitation　events　and　observed　increasing　spring　flow.　This　model　is　then　applied　to　the　prediction　of　time-dependent　flow　of　the　Longzichi　Spring,　Shanxi　Province,　northwest　China.　The　results　in　this　analysis　show　that　the　proposed　approach　can　accurately　describe　the　complicated　nonlinear　time　series　and　the　discharge　regime　of　the　Longzichi　Spring.A　nonlinear　time　series　model　is　presented　to　describe　the　dynamics　of　ground　water　flow.　The　procedures　for　model-establishing　and　parameter　estimation　are　discussed.　The　proposed　nonlinear　model　uses　a　threshold　parameter　that　relates　different　precipitation　processes,　and　a　lag　parameter　that　relates　the　time　lag　between　precipitation　events　and　observed　increasing　spring　flow.　This　model　is　then　applied　to　the　prediction　of　time-dependent　flow　of　the　Longzichi　Spring,　Shanxi　Province,　northwest　China.　The　results　in　this　analysis　show　that　the　proposed　approach　can　accurately　describe　the　complicated　nonlinear　time　series　and　the　discharge　regime　of　the　Longzichi　Spring.