Dam construction significantly alters the hydrodynamic characteristics of rivers, rendering reservoirs important sources of global greenhouse gas emissions; however, current emission estimates remain highly uncertain. This study conducted a summer field survey in August 2025 at 11 typical cascade reservoirs along the mainstem of the Lancang River, systematically investigating the spatial heterogeneity of diffusive fluxes of methane (CH4) and carbon dioxide (CO2) and their gas transfer velocities (k), and individually dissecting potential driving factors including gas species, elevation gradients, and relative positions within the reservoirs. The results revealed that the two gases exhibited significantly distinct emission patterns across the 11 cascade reservoirs. The CH? diffusive flux averaged 1.23 ± 2.53 mmol m?² d?¹, with an emission hotspot identified in the middle reservoir reaches. CO? fluxes ranged from ?287.25 to 143.46 mmol m?² d?¹; the middle and downstream reservoir reaches generally acted as carbon sources (median: 23.10 mmol m?² d?¹), whereas the dam-proximal zone functioned as a carbon sink due to enhanced photosynthesis (median: ?11.69 mmol m?² d?¹). Measured k values showed a statistically significant trend of k(CH4) > k(CO2) (means of 8.02 and 5.46 cm h?¹, respectively; p < 0.05). In the high-flow, air-entrainment reaches near the reservoir tailwaters, the order-of-magnitude difference in solubility between CH? and CO2 was the primary reason why the air-entrainment process induced a physical enhancement effect on CH4. In addition, altitudinal gradients may exert potential physical controls on the gas transfer coefficient by altering atmospheric pressure and gas solubility. k(CH4) exhibited pronounced spatial heterogeneity (8.02 ± 5.81 cm h?¹); the uncertainty in CH4 flux estimation introduced by spatial variability in k (coefficient of variation: 72.4%) was comparable in magnitude to that arising from spatial heterogeneity in dissolved CH4 concentrations (71.4%-73.1%). Empirical k models relying solely on wind speed systematically underestimated CH4 diffusive fluxes from Lancang River reservoirs by approximately one order of magnitude, whereas models incorporating flow velocity markedly improved k prediction accuracy. This study confirms that spatial heterogeneity in gas transfer velocity constitutes one of the core sources of uncertainty in reservoir carbon emission estimates, and future assessments should integratively account for the synergistic effects of hydrodynamic gradients and gas solubility differences.