To　accurately　estimate　freeway　traffic　carbon　dioxide　(CO2)　emissions,　this　paper　proposes　a　spatiotemporal　cellbased　model　by　taking　traffic　dynamics　into　account.　High-fidelity　vehicle　trajectory　data　is　used　to　construct　a　spatiotemporal　traffic　(ST)　diagram　and　to　calculate　the　exact　CO2　emissions　of　the　traffic　in　the　ST　diagram.　The　factors　impacting　the　CO2　emissions　in　the　ST　diagram　are　selected　and　taken　as　model　inputs.　First-　and　second-order　regression　models　are　employed　to　fit　the　exact　CO2　emissions.　It　is　found　that　the　relationship　between　complicated　traffic　dynamics　and　CO2　emissions　can　be　simply　described　by　using　a　linear　or　nearly　linear　function;　i.e.,　for　larger　cells　(such　as　90　center　dot　150　sec　center　dot　m)　that　are　used　to　construct　an　ST　diagram,　a　first-order　regression　model　is　able　to　well　reflect　the　relationship,　while　for　small　cells　(such　as　30　center　dot　50　sec　center　dot　m)　a　second-order　model　is　more　accurate.　To　validate　the　proposed　model,　another　trajectory　dataset　that　was　collected　in　a　different　freeway　segment　is　introduced,　and　the　transferability　and　predictability　of　the　model　are　demonstrated.　The　proposed　spatiotemporal　cell-based　model　allows　us　to　accurately　estimate　CO2　emissions　by　inputting　the　prevailing　ST　diagram.　It　opens　a　gate　for　estimating　CO2　emissions　from　widely　available　low-fidelity　traffic　data,　since　the　ST　diagram　can　be　constructed　by　using　various　traffic　flow　data,　such　as　loop　detector　data　and　floating　car　data.