Benchmarking ozone stress parameterizations in CLM5: a global mechanistic assessment of thresholds and memory effects

Tropospheric ozone remains a critical but uncertain driver of terrestrial productivity loss, and land surface models (LSMs) diverge markedly in how they represent vegetation ozone stress. We conduct a global, mechanistically consistent evaluation of three prominent ozone stress parameterization schemes, Sitch, Lombardozzi, and Li, within the Community Land Model version 5 (CLM5). Using unified meteorological and ozone forcing from CAM-chem and GSWP3.1, we designed five experiments to isolate the roles of ozone flux threshold selection and response function form. The mixed experiments using thresholds and response functions derived from the Sitch and Lombardozzi schemes were implemented without additional recalibration, allowing structural sensitivities to be evaluated consistently within the Li framework. Model output is benchmarked against MODIS and FLUXNET gross primary production (GPP) across spatial gradients, biomes, and among plant functional types (PFTs). All parameterizations capture the ozone–induced reduction in GPP relative to the ozone-free baseline, but their accuracy varies widely. The Li scheme, featuring PFT-specific thresholds and separate nonlinear responses for photosynthesis and stomatal conductance, best agrees with observed GPP patterns across scales. In contrast, the Lombardozzi scheme produces much larger reductions in high-flux regions. Analysis reveals that the structures of ozone response functions and memory-decay mechanisms primarily determine improvements in GPP simulation. Our results support a shift toward ozone parameterizations that couple stomatal flux with canopy phenology, dynamic water constraints, and regionally calibrated thresholds. These findings provide a transferable framework for quantifying ozone–carbon coupling in LSMs and highlight priorities for improving terrestrial biosphere models under atmospheric change.