Considerable Uncertainties in Simulating Land Carbon Sinks Induced by Different Precipitation Products

Precipitation, a key determinant of soil moisture variations, plays an important role in regulating terrestrial carbon fluxes on multiple time scales. It is a critical meteorological forcing to drive terrestrial biosphere model (TBM), however, with a large uncertainty itself. We here investigated to what extent precipitation alone can cause uncertainties of model‐simulated carbon flux from terrestrial ecosystems to atmosphere (FTA), based on eight precipitation products and a TBM, VEGAS. We find that the pattern of uncertainties in simulated FTAobviously differs from the pattern of discrepancies in precipitation, owing to divergent water sensitivities of vegetation over different regions. Globally, the uncertainty in FTAcan be up to approximately 40.73% of the uncertainty in TRENDYv6 multi‐model simulated FTAwhich is caused by model structural and parameter uncertainty. A good linear relationship emerges between global area‐averaged land climatological annual precipitation and simulated total FTAwith the slope of −0.0040 PgC yr−1per mm yr−1(p= 0.03; negative for carbon sink), where 70% is explained by the sensitivity over extra‐tropical Northern Hemisphere (NH). For seasonal cycle, compared to nearly constant inter‐precipitation spreads over tropics plus extra‐tropical southern hemisphere (Trop + SH), uncertainties in corresponding simulated FTAshow obvious seasonal differences with the relatively larger uncertainties in March‐April‐May (MAM) and August‐September‐October (ASO). For interannual variability, uncertainties in simulated total FTAare, albeit smaller, nonnegligible, which are 40.61% (global), 38.17% (Trop + SH), and 29.63% (NH) of the TRENDYv6 inter‐model uncertainty, respectively. Therefore, generating better global precipitation product is important for reducing the uncertainty in simulating terrestrial carbon sinks. Precipitation, a key forcing in terrestrial biosphere model (TBM), has a large uncertainty itself. We here used eight land precipitation products to run TBM VEGAS and explored the resulting uncertainties in simulating carbon flux from terrestrial ecosystems to atmosphere (FTA) on climatology, seasonal cycle, and interannual variability. In climatology, the uncertainty in global FTAinduced by different precipitation can be approximately 40.73% of the uncertainty in TRENDYv6 multi‐model simulations induced by model structural and parameter uncertainty. Furthermore, we find a good linear relationship between global area‐averaged land climatological annual precipitation and simulated total FTA, implying more precipitation over land with stronger carbon sinks, especially over extra‐tropical Northern Hemisphere (NH). For seasonal cycle, uncertainties in simulated FTAover tropics plus extra‐tropical southern hemisphere (Trop + SH) show relatively larger values in March‐April‐May and August‐September‐October. For interannual variability, uncertainties in simulated total FTAare up to 40.61% (global), 38.17% (Trop + SH), and 29.63% (NH) of the TRENDYv6 inter‐model uncertainty, respectively. In general, uncertainty in land precipitation data sets can cause considerable uncertainties in simulating FTA, suggesting the importance of generating the better precipitation product for better simulating the terrestrial carbon sinks. Uncertainties in simulating land‐atmosphere carbon fluxes (FTA) on multiple time scales induced by different precipitation are investigatedUncertainties in climatology and interannual variability of global total FTAare about 41% of uncertainties in TRENDYv6 simulationsA good linear relationship emerges between global area‐averaged land climatological annual precipitation and simulated total FTA Uncertainties in simulating land‐atmosphere carbon fluxes (FTA) on multiple time scales induced by different precipitation are investigated Uncertainties in climatology and interannual variability of global total FTAare about 41% of uncertainties in TRENDYv6 simulations A good linear relationship emerges between global area‐averaged land climatological annual precipitation and simulated total FTA