Multiple linear regression (MLR) is widely used to attribute causes of the interannual variability (IAV) of land carbon uptake, yet, parameter estimation in MLR can be problematic if the predictors are strongly inter-correlated. Recently, Humphrey et al., (2021) used MLR method to conclude that the indirect effect of soil moisture (SM) via land-atmosphere coupling, rather than the direct effect of SM on photosynthesis and respiration, controls the IAV of NBP. Here we assess the validity of MLR as used by Humphrey et al. (2021) by comparing the true contribution of SM in a terrestrial biosphere model, derived from the difference between a control run (CTRL) and an experiment with prescribed climatological SM (EXP), with the MLR method applied to the CTRL outputs.

We ran two experiments (CTRL and EXP) with the ORCHIDEE-MICT terrestrial biosphere model at 2º spatial resolution. The control (CTRL) run followed the protocol of “S3” experiment of TRENDY-v6, forced by CRUNCEP-v8 climate forcing, increasing atmospheric CO₂ concentration, and varying land use maps. Monthly outputs for the period 1960-2005 were used for analysis. For the EXP run, to remove the interannual variability (IAV) of soil moisture (SM) while keeping its seasonal cycle, a climatological monthly SM averaged for the years 1960-2005 simulated by the CTRL run was prescribed in the model. Note that the intrinsic time-step of hydrology and photosynthesis in ORCHIDEE-MICT is half-hourly, thus SM within the same month took the same monthly mean value. Other configurations in the EXP run were identical to the CTRL run.