Spin-up processes in the Community Land Model version 4 with explicit carbon and nitrogen components

Given arbitrary initial conditions, terrestrial biogeochemistry models typically require hundreds to thousands of years for carbon and nitrogen in various pools to reach steady-state solutions. Such long spin-up processes not only pose a significant burden to computational resources, but also are against observational evidence. The objectives of this study are to: (1) compare four spin-up methods and their steady-state solutions using the Community Land Model version 4 with explicit carbon and nitrogen processes (CLM4CN. ²2SOM, soil organic matter; ND, native dynamics; CLM4CN, the Community Land Model version 4 with explicit carbon and nitrogen; PFT, plant functional types; LPJ, Lund-Potsdam-Jena Dynamic Global Vegetation Model; AD, accelerated decomposition; DDL, decelerated bulk denitrification and leaching; SI, initialization of soil carbon and nitrogen pools; CESM, the Community Earth System Model; LBA-DMIP, Large Scale Biosphere-Atmosphere Data Model Intercomparison Project; GSDT, Global Soil Data Task; ORCHIDEE, ORganizing Carbon and Hydrology in Dynamic EcosystEms.); (2) elucidate the potential weaknesses of the model that are responsible for long spin-ups. The four methods can be classified into two groups: (1) the model spins up from arbitrary initial conditions (e.g., the traditional native dynamics or ND method); (2) the model is initialized with observed soil organic matter (SOM) pools. Our results show that: (1) compared to ND, accelerating SOM decomposition rates during spin-up reduces the spin-up timescales in tropical forests, grasslands, temperate forests, and boreal forests; (2) in some temperate forests, decelerating the denitrification and leaching rates and accelerating the decomposition rates during spin-up saves more computational time than the method only with decomposition rates accelerated; (3) a reasonable SOM initialization helps the model reach its steady state quickly. We also find that in some ecosystems the vegetation seasonality described by methods with decomposition or denitrification and leaching rates changed is inconsistent with that from the ND method. CLM4CN has the potential of improving the simulations and reducing the long spin-up timescales if the model structure and ways in representing decomposition and immobilization are improved.