Climate change due to greenhouse gas emissions (mainly CO2) threatens ecosystems and human well-being. The land carbon sink through terrestrial vegetation is a key component to absorb atmospheric CO2 to offset global warming. A robust quantification of terrestrial vegetation responses to historical and future environmental changes is necessary to predict the strength of land carbon sink under climate change scenarios. Using a parsimonious model, the P model, I modelled changes in gross primary production (GPP) and demonstrated that this model reproduces historical variation of GPP against observations. Model experiments indicate that CO2 and vegetation greenness (fAPAR) are the major drivers. I further investigated the parameters that affect the sensitivity of modelled vegetation light use efficiency (LUE) to elevated CO2. The ratio of Jmax to Vcmax at 25°C and convexity of light response curve were found to be the most important variables. The effect of CO2 fertilization (CFE) on GPP was estimated to be 15.43% for a 100-ppm increase in CO2, consistent with previous studies. The modelled diminishing CFE during the past decades indicated a potential decreasing benefit from CO2 for vegetation growth under future climate change. An extended version of the P model was used to further quantify changes in the land carbon sink. Although a conceptually simpler model, it achieved similar performance compared to dynamic global vegetation models with reasonable bias. Finally, I predicted vegetation greenness based on the trade-off between vegetation carbon gain and water loss, both of which depend on the amount of leaf area quantified by fAPAR. Predicted seasonal maximum fAPAR explains 95% of the variation compared to observed remotely sensed MODIS fAPAR, with similar temporal trends. These studies provide a mechanistic understanding of vegetation responses to environmental changes and a first step towards more reliable land surface models for future climate projection.