Global evaluation of terrestrial biogeochemistry in the Energy Exascale Earth System Model (E3SM) and the role of the phosphorus cycle in the historical terrestrial carbon balance

. The importance of carbon (C)–nutrient interactions to the prediction of future C uptake has long been recognized. The Energy Exascale Earth System Model (E3SM) land model (ELM) version 1 is one of the few land surface models that include both N and P cycling and limitation (ELMv1-CNP). Here we provide a global-scale evaluation of ELMv1-CNP using the International Land Model Bench-marking (ILAMB) system. We show that ELMv1-CNP produces realistic estimates of present-day carbon pools and ﬂuxes. Compared to simulations with optimal P availability, simulations with ELMv1-CNP produce better performance, particularly for simulated biomass, leaf area index (LAI), and global net C balance. We also show ELMv1-CNP-simulated N and P cycling is in good agreement with data-driven estimates. We compared the ELMv1-CNP-simulated response to CO 2 enrichment with meta-analysis of observations from similar manipulation experiments. We show that ELMv1-CNP is able to capture the ﬁeld-observed responses for photosynthesis, growth, and LAI. We investigated the role of P limitation in the historical balance and show that global C sources and sinks are signiﬁcantly affected by P limitation, as the historical CO 2 fertilization effect was reduced by 20 % and C emission due to land use and land cover change was 11 % lower when P limitation


Text S1: Additional description of the representation of Non-structural (NSC) pool in ELMv1-CNP
In the current model configuration, there are no direct feedbacks of the non-structural carbon pool on plant activities.When soil nutrient supply is high, allocation to new NSC decreases and allocation to biomass construction increases, and the size of the NSC pool declines; however, existing NSC turns over due to respiration and releases CO2 into the atmosphere and is not utilized by plants.
The respiration from the NSC pool RNSC is calculated using the following equation: Where brcpool is the respiration base rate at 25 o C, which is set to 10 -9 gC m -2 s -1 .This is equivalent to roughly a 3-year turnover time for the NSC pool CNSC, which is broadly consistent with observations that indicate a range of values across plant components and NSC forms with a weighted mean of 2-3 years (Richardson et al., 2015).The Q10 parameter, which has a value of 1.5 in these simulations, controls the strength of the temperature response.Tair is the 2-meter air temperature (K).
Nutrients are allocated from the non-structural pools using the following equation: Where Ndemand is the amount of nutrient (nitrogen or phosphorus) required to allocate all NPP during a model timestep to plant structural pools given the stoichiometry.NNS is the size of the non-structural nutrient pool, and NNS,max is the size of the pool above which there is no nutrient limitation.
!",9,< =  *  !,,::=,' NPPN,annual is the previous year's total of annual net primary productivity for the nutrient of interest.The parameter Nstor is set to 3 for these simulations.Therefore, nutrient limitation only affects allocation when there are less than 3 years of non-structural nutrient storage.This is consistent with the average age of NSC found by Richardson et al. (2015), though that analysis found a 2-pool NSC model may have more predictive skill, which we will consider implementing in future work.
These computations of Nalloc are done for both nitrogen and phosphorus.The actual allocation of carbon, nitrogen and phosphorus is set by whichever nutrient is more limiting at that timestep given the lower of the ratios Nalloc:Ndemand.

Figure S1 :
Figure S1: Time series of globally-integrated P pools for the control run, which is the continuation of the normal spinup.

Figure S7 :Figure S8 :
Figure S7: The comparison of atmospheric CO2 concentrations inferred from ELM v1-CNP landcarbon fluxes (green lines) with in situ flask measurements from NOAA's global Cooperative Air Sampling Network (gray lines) in ILAMB: (a) the mean seasonal amplitude over flask sites [ppm], (b) the range of interannual variability [ppm], (c) the average month-of-year when the peak CO2 concentration occurs, and (d) the average month-of-year when the lowest CO2 concentration occurs.Observations and measurements are binned within 30• latitude increments; small gray dots indicate individual flask sampling locations.

Figure S9 :
Figure S9: Simulated change in land carbon storage in response to changes in CO2, land use and land cover change, N deposition, climate during 1850-2010.Unit: Pg C. CN is for ELMv1-CN and CNP is for ELMv1-CNP

Table S1
Yang et al. (2016) parameters are provided in detail in the supporting information ofYang et al. (2016) *

Table S2 :
Yang et al. (2016)nt parameters* The sources of the parameters are provided in detail in the supporting information ofYang et al. (2016) *

Table S3 :
List of Input Data

Table S4 :
Observational Dataset Used for Carbon Cycle Evaluation in ILAMB