As one of the most widespread forest types, China's plantation plays an important role in global carbon balance. It is crucial to reduce the uncertainties in the estimation of carbon and water fluxes of plantation ecosystems, and model-data fusion technique provides an effective way. The purpose of this research is to improve the modeling accuracy of SIPNET model, the simplified Photosynthesis and Evapo-Transpiration(ET) model through two experiments, namely NEE alone and NEE & ET multi-constraints. The model-data fusion method used here is a combination of Metropolis-Hastings algorithm and Simulated Annealing algorithm. Based on eddy fluxes and meteorological observation data of Qianyanzhou subtropical coniferous plantation during 2004-2009 in ChinaFLUX (Chinese Terrestrial Ecosystem Flux Research Network), we estimated the key parameters of SIPNET model and simulated the corresponding carbon and water fluxes. Comparisons between the measured and modeled net ecosystem exchange of carbon dioxide (NEE) showed that the SIPNET model had approximately equivalent fits to the observed NEE under two optimization procedures (R² decreased from 0.934 to 0.929, and RMSE increased from 0.736 g C/m² to 0.763 g C/m²). In the case of ET, the NEE and ET parameterization produced a markedly better fit to the observed ET than the NEE parameterization (R² increased from 0.188 to 0.824, and RMSE decreased from 0.152 cm to 0.053 cm). As for transpiration, when optimized by observed NEE alone, SIPNET largely underestimated annual accumulated transpiration in 2004 compared with the measurements of sap flow technique. In comparison, while optimization based on NEE and ET, SIPNET led to a better fit of annual cumulative estimation of transpiration in 2004 to the sap flow measurement. These results indicated that the SIPNET model parameterized using NEE and ET observed fluxes could well reproduce the characteristics of carbon and water fluxes. In other words, more information can be extracted from simultaneous optimization, since there is additional process information in water flux observation data. Furthermore, we conducted a sensitivity test of precipitation on carbon fluxes through reduction of precipitation. We found that photosynthesis was more sensitive to precipitation reduction than respiration, and the model optimized using NEE and ET reproduced the response of NEE to precipitation reduction better than that optimized using NEE alone. In addition, we detected that the difference of NEE response to precipitation reduction in two optimizations of the SIPNET model was caused by gross ecosystem production rather than ecosystem respiration. Therefore, parameter estimation using NEE and ET altogether improved the performance of SIPNET model. And without optimization using both NEE and ET, the response of ecosystem carbon cycle to precipitation variation may be misrepresented.