Analyzing　the　uncertainty　of　flutter　propagated　from　different　sources　especially　flutter　derivatives　is　important　for　the　risk　assessment　of　long-span　bridges.　The　uncertainty　of　flutter　derivatives　is　often　characterized　based　on　the　experimental　data.　Dependence　and　sampling　error　would　therefore　exist　which　may　influence　the　onset　prediction　of　bridges.　This　paper　devotes　to　developing　a　framework　towards　the　uncertainty　analysis　of　bridge　flutter　considering　the　dependence　and　sampling　error　in　flutter　derivative　measurements.　In　this　framework,　the　dependence　among　different　flutter　derivatives　is　identified　with　Spearman＇s　rho　instead　of　conventional　Pearson＇s　product-moment　correlation　coefficient.　A　modified　procedure　based　on　bootstrap　is　proposed　to　characterize　the　sampling　error　by　representing　distribution　parameters　of　flutter　derivatives　as　interval　variables.　Regarding　the　probability-interval　hybrid　uncertainty,　a　double-layer　analysis　is　employed.　In　the　inner　layer,　probability　of　onset　velocity　is　estimated　using　the　point　estimate-based　probability　analysis.　In　the　outer　one,　the　distribution　envelope　is　obtained　by　the　genetic　algorithm-assisted　interval　analysis.　Application　of　the　framework　to　reliability　assessment　of　bridge　against　flutter　is　also　presented.　In　the　numerical　study,　the　influence　of　dependence　and　sampling　error　is　verified　in　detail.　By　this　framework,　the　distribution　envelope　of　onset　velocity　or　the　interval　of　reliability　index　against　flutter　are　expected　to　contain　the　true　estimates.　Consequently,　the　potential　risk　with　respect　to　bridge　flutter　can　be　assessed　in　a　conservative　way.