The　commonly　used　diffuse　fraction　models　are　generally　expressed　as　a　single-value　function　of　clearness　index　(Kt).　Whereas　in　fact,　the　same　Kt　occurring　at　different　times　of　the　day　may　correspond　to　different　diffuse　fraction　(Kf).　Using　a　10-year　dataset　from　2005　to　2014,　this　study　intended　to　model　hourly　solar　diffuse　fraction　based　on　sky　conditions.　Five　cities　in　China,　including　Golmud,　Kashgar,　Sanya,　Wuhan,　and　Wenjiang,　were　selected　as　representatives　of　five　solar　climatic　zones.　The　measured　radiation　data　on　a　horizontal　surface　were　used　for　k-means　clustering　of　the　sky　conditions　with　two　factors　of　Kf　and　Kt.　The　results　show　that　the　sky　conditions　could　be　divided　into　two　groups　according　to　time　dimension:　the　morning　(6:00–12:00)　and　the　afternoon　(13:00–18:00)　segments.　Subsequently,　a　morning-afternoon　segmented　Kf-Kt　model　for　each　zone　was　constructed　based　on　quartic　polynomial　regression,　with　the　smallest　error　rate　at　a　range　of　10.4%–28.5%.　Comparison　with　the　existing　models　further　validates　that　the　respective　models　for　the　morning　and　afternoon　periods　can　assist　in　improving　the　modeling　of　the　hourly　solar　diffuse　fraction　on　a　horizontal　surface　in　variable　solar　climatic　zones.　©　2023