Quantifying　the　decoupling　states　of　carbon　emissions　from　a　multi-sectoral　and　dual-perspective　can　guide　more　detailed　emission　reduction　strategies.　Based　on　the　single-regional　input-output　(SRIO),　Tapio　decoupling　analysis　(TDA),　and　structural　decomposition　analysis　(SDA),　this　study　investigated　the　dynamic　variation　feature　and　decoupling　state　of　multi-sectoral　carbon　emissions,　and　revealed　their　driving　factors　of　consumption-based　emissions　in　Guangdong　province　from　2002　to　2017.　The　main　discovery　can　be　summarized　as　follows　from　results　analysis.　Firstly,　electricity　production　sector　and　construction　sector　were　the　largest　direct　and　embodied　carbon　emission　sources,　and　capital　formation　was　the　most　important　factor　with　the　contribution　of　approximately　100　%　that　led　to　embodied　carbon　emissions　of　construction.　For　most　of　the　manufacturing　and　service　sectors,　the　embodied　carbon　emissions　caused　by　international　export　exceed　50　%.　Secondly,　the　consumption　structure,　consumption　per　capita,　and　population　effect　promoted　the　embodied　emissions　during　2002-2012,　while　the　emission　intensity　effect　was　the　greatest　offsetting　factor　for　all　sectors.　Consumption　structure　effect　was　becoming　a　major　driver　to　the　increase　of　embodied　carbon　emissions　for　construction.　Thirdly,　agriculture,　mining,　energy　transformation,　and　service　sector　showed　the　unsatisfactory　decoupling　relationship　between　direct　carbon　emissions　and　economic　output.　According　to　the　decoupling　states,　the　decoupling　relationships　in　some　secondary　industries　were　overestimated　under　the　situation　of　only　considering　direct　carbon　emissions.　The　obtained　results　and　policy　implications　are　expected　to　provide　holistic　reference　for　policymakers　to　promote　the　short-term　carbon　peak　and　long-term　carbon　neutrality　of　Guangdong　province　from　the　sectoral　perspective.