Since soil CO2 flux is a key component of ecosystem carbon balance, quantifying its contribution to the ecosystem carbon flux and understanding the factors that underlie its temporal variation is crucial for a better comprehension of ecosystem carbon dynamics under climate change and for optimal ecosystem use and management. Our objectives were to quantify the contributions of total soil CO2 efflux (FS) to ecosystem respiration (RE) and heterotrophic soil CO2 efflux (FH) to FS in two chinampa ecosystems with different natural grass covers. We also aimed to identify the main environmental drivers of seasonal variability of these contributions. The CO2 fluxes were measured on each site about every 14 days from September 2008 to August 2009 in the Xochimilco Ecological Park in Mexico City using dark chamber techniques. For two studied sites, RE, FS and FH were estimated on average as 94.1 ± 8.5, 34.7 ± 3.5 and 16.5 ± 1.7 (± S.E.) mg C-CO2 m-2 h-1, respectively. On average over the study period and sites, the annual cumulative RE, FS and FH fluxes were 824 ± 74, 304 ± 31 and 145 ± 15 g C m-2 year, respectively. The RE, FS and FH varied between the winter and summer seasons; this variation was explained mostly by seasonal variations of soil temperature, soil water content and shoot plant biomass. Temperature sensitivity of CO2 fluxes depended on vegetation type and plant growth differences among the sites and decreased in the following order: RE > Rs > RH. The contribution of FS to RE and FH to FS for the two studied sites and period averaged about 38% and 50%, respectively regardless of the site vegetation type, but the degree of FS/RE and FH/FS variability depended on the differences in seasonal dynamics of plant cover. The contribution of FH to FS varied from 37% in summer to 73% in winter at the site without a seasonal shift in dominant plant species, but FH/FS was close to constant during the year at the site with a seasonal change in dominant plant species. During the cold period, the contribution of FH to FS increased following plant growth decrease. The linear regression analysis showed that plant biomass was the dominant factor controlling the seasonal variation of FH/FS ratios, whereas the plant biomass dynamic followed the dynamics of soil water content, water table depth, and soil temperature. Our results suggest that seasonal variation of soil contribution to total fluxes from the chinampa ecosystem is locally differentiated. These differences were related to differences in seasonal dynamics of cover productivity which has been associated with localization of soil water content. This finding has important implications for assessing the contribution of the chinampa ecosystem to the global carbon budget.