CLM5-PFT-PPE

The CLM5-PFT-PPE is a perturbed-parameter ensemble (PPE) of Community Land Model (CLM) simulations in which plant functional type (PFT) specific parameters are varied independently. It provides CLM-BGC transient simulations for 1850–2014 on a representative subset of global land grid cells and is intended for analyzing PFT interactions and parameter uncertainty in carbon, water and energy cycling. Example scripts for working with the dataset are available at: https://github.com/linniahawkins/clm5-pft-ppe

CLM is the land component of the Community Earth System Model, representing carbon, water and energy cycling by the land surface. The CLM version 5 is described in depth in Lawrence et al., (2019); the model represents processes related to hydrology, vegetation, biogeochemistry, agriculture, fire, land-use and land-cover change and others at varying degrees of complexity. CLM is open source; code and documentation are available online (https://github.com/ESCOMP/CTSM). This experiment used CLM version 5.1 which includes biomass heat storage (Swenson et al., 2019) and parameter adjustments (Birch et al., 2021), with additional code modifications that facilitate parameter perturbations as described in Kennedy et al., (2025). The model code for this experiment is contained in a development tag of the CLM (https://github.com/ESCOMP/CTSM/tree/branch_tags/PPE.n11_ctsm5.1.dev030). The CLM component set is: 2000_DATM%GSWP3v1_CLM51%BGC_SICE_SOCN_SROF_SGLC_SWAV_SIAC_SESP. We used the CLM-BGC configuration with prognostic vegetation and active biogeochemical cycling (Lawrence et al., 2019). Dynamic crop management was turned off and land-use land cover change was prescribed. All simulations were performed with prescribed meteorology using the Global Soil Wetness Project reanalysis product (GSWP3v2; http://hydro.iis.u‐tokyo.ac.jp/ GSWP3/) which is a 3 hourly 0.5° global product for 1901 through 2014.

The simulation protocol follows Kennedy et al., (2025). Briefly, the protocol starts from a pre-existing spun-up state and runs the model for 20 years in accelerated decomposition mode (Thornton and Rosenbloom, 2005) then for 80 years in semi-analytic spin-up mode (Lu et la., 2020; Liao et al., 2023), and finally for 40 years in native dynamic spin-up mode. We used the final state as initial conditions for transient simulations of 1850 through 2014. We recycled 1901-1920 GSWP3v2 meteorology for all spin-up stages as well as 1850-1901. This process is repeated for each perturbed parameter ensemble member. To further reduce the computational expense, we ran a subset of representative land gridcells. See Kennedy et al., (2025) for full detail, but briefly the global land gridcells at 2° resolution were clustered based on meteorological variables, ecosystem state variables, and ecosystem flux variables. One gridcell was selected to represent each of the 400 unique clusters.

We perturbed 32 CLM-BGC parameters identified from prior one-at-a-time sensitivity analysis, using the same ranges as in Kennedy et al. (2025): 20 global parameters and 12 PFT-specific parameters (the latter varied independently across PFTs). We first generated a large random sample of parameter sets and used Gaussian process emulators for PFT leaf area index (LAI) to estimate the CLM simulated LAI for each parameter set. We compared the emulated LAI to PFT-level LAI targets from MODIS (Lawrence & Chase, 2007) used in CLM’s prescribed satellite phenology mode and retained the parameter sets that yielded LAI within the observational uncertainty. Of the retained parameter sets we formed a 500-member stratified ensemble by first selecting 100 global sets and pairing each with five distinct PFT-parameter sets that assign different values to each PFT. This design preserves observationally consistent LAI while sampling broadly from the parameter space for analyses of PFT interactions and parametric uncertainty.

References:

Birch, L., Schwalm, C. R., Natali, S., Lombardozzi, D., Keppel-Aleks, G., Watts, J., et al. (2021). Addressing biases in Arctic–boreal carbon cycling in the community land model version 5. Geoscientific Model Development, 14(6), 3361–3382. https://doi.org/10.5194/gmd-14-3361-2021

Kennedy, D., Dagon, K., Lawrence, D. M., Fisher, R. A., Sanderson, B. M., Collier, N., ... & Wood, A. W. (2025). One‐at‐a‐time parameter perturbation ensemble of the Community Land Model, version 5.1. Journal of Advances in Modeling Earth Systems, 17(8), e2024MS004715.

Liao, C., Lu, X., Huang, Y., Tao, F., Lawrence, D. M., Koven, C. D., et al. (2023). Matrix approach to accelerate spin-up of clm5. Journal of Advances in Modeling Earth Systems, 15(8), e2023MS003625. https://doi.org/10.1029/2023MS003625

 Lawrence, P. J., & Chase, T. N. (2007). Representing a new MODIS consistent land surface in the Community Land Model (CLM 3.0). Journal of Geophysical Research: Biogeosciences, 112(G1). https://doi.org/10.1029/2006JG000168

 Lawrence, D. M., Fisher, R. A., Koven, C. D., Oleson, K. W., Swenson, S. C., Bonan, G., et al. (2019). The community land model version 5: Description of new features, benchmarking, and impact of forcing uncertainty. Journal of Advances in Modeling Earth Systems, 11(12), 4245–4287. in press. https://doi.org/10.1029/2018MS001583

 Lu, X., Du, Z., Huang, Y., Lawrence, D., Kluzek, E., Collier, N., et al. (2020). Full implementation of matrix approach to biogeochemistry module of CLM5. Journal of Advances in Modeling Earth Systems, 12(11), e2020MS002105. https://doi.org/10.1029/2020MS002105

 Swenson, S. C., Burns, S. P., & Lawrence, D. M. (2019). The impact of biomass heat storage on the canopy energy balance and atmospheric stability in the community land model. Journal of Advances in Modeling Earth Systems, 11(1), 83–98. https://doi.org/10.1029/2018MS001476

 Thornton, P. E., & Rosenbloom, N. A. (2005). Ecosystem model spin-up: Estimating steady state conditions in a coupled terrestrial carbon and nitrogen cycle model. Ecological Modelling, 189(1), 25–48. https://doi.org/10.1016/j.ecolmodel.2005.04.008