Aims: Our aims were to identify the primary factors involved in soil respiration (Rs) variability and the role that spectral vegetation indices played in Rs estimation in irrigated and rainfed agroecosystems during the growing season. Methods: We employed three vegetation indices (i.e., normalized difference vegetation index (NDVI), green edge chlorophyll index (CI green edge) and enhanced vegetation index (EVI)) derived from the moderate resolution imaging Spectroradiometer (MODIS) surface reflectance product as approximations of crop gross primary production (GPP) for Rs estimation. Different statistical models were used to analyze the dependencies of Rs on soil temperature soil water content and plant photosynthesis, and accuracy of these models were compared in the irrigated and rainfed agroecosystems. Results: The results demonstrated that a model based only on abiotic factors (e.g., soil temperature and soil water content) failed to describe part of growing season variability in Rs. Residual analysis indicated that Rs was influenced by a short-term gross primary production (GPP) and a longer-term (>_3 days) accumulated GPP in the irrigated and rainfed agroecosystems. Therefore, photosynthesis dependency of Rs should be included in the Rs model to descibe the growing-season dynamics of Rs. Among the three VIs, CIgreen edge showed generally better correlations with GPP at different cumulative times and canopy green leaf area index than EVI and NDVI. Adding the CI green edge into the model considering only soil temperature and soil water content significantly improved the simulation accuracy of Rs. Conclusions: Our results suggests that spectral vegetation index from remote sensing could be used to estimate Rs, which will be helpful for the development of a future Rs model over a large spatial scale.