Abstract. Annually resolved tree-ring records extending back to pre-industrial conditions have the potential to constrain the responses of global land surface models at interannual to centennial timescales. Here, we demonstrate a framework to simultaneously constrain the representation of tree growth and physiology in the ORCHIDEE global land surface model using the simulated variability of tree-ring width and carbon (Δ13C) and oxygen (δ18O) stable isotopes in six sites in boreal and temperate Europe. We exploit the resulting tree-ring triplet to derive integrative constraints for leaf physiology and growth from well-known mechanistic relationships among the variables. ORCHIDEE simulates Δ13C (r=0.31–0.80) and δ18O (r=0.36–0.74) better than tree-ring width (r<0.55), with an overall skill similar to that of a tree-ring model (MAIDENiso) and another isotope-enabled global vegetation model (LPX-Bern). The comparison with tree-ring data showed that growth variability is not well represented in ORCHIDEE and that the parameterization of leaf-level physiological responses (stomatal control) to drought stress in the temperate region can be constrained using the interannual variability of tree-ring stable isotopes. The representation of carbon storage and remobilization dynamics emerged as a critical process to improve the realism of simulated growth variability, temporal carryover, and recovery of forest ecosystems after climate extremes. Simulated forest gross primary productivity (GPP) correlates with simulated tree-ring Δ13C and δ18O variability, but the origin of the correlations with tree-ring δ18O is not entirely physiological. The integration of tree-ring data and land surface models as demonstrated here should guide model improvements and contribute towards reducing current uncertainties in forest carbon and water cycling.

Abstract. The search for a long-term benchmark for land-surface models (LSM) has brought tree-ring data to the attention of the land-surface community as they record growth well before human-induced environmental changes became important. The most comprehensive archive of publicly shared tree-ring data is the International Tree-ring Data Bank (ITRDB). Many records in the ITRDB have, however, been collected almost exclusively with a view on maximizing an environmental target signal (e.g. climate), which has resulted in a biased representation of forested sites and landscapes and thus limits its use as a data source for benchmarking. The aim of this study is to propose advances in land-surface modelling and data processing to enable the land-surface community to re-use the ITRDB data as a much-needed century-long benchmark. Given that tree-ring width is largely explained by phenology, tree size, and climate sensitivity, LSMs that intend to use it as a benchmark should at least simulate tree phenology, size-dependent growth, differently-sized trees within a stand, and responses to changes in temperature, precipitation and atmospheric CO2 con¬cen¬tra¬tions. Yet, even if LSMs were capable of accurately simulating tree-ring width, sampling biases in the ITRDB need to be accounted for. This study proposes two solutions: exploiting the observation that the variation due to size-related growth by far exceeds the variation due to environmental changes; and simulating a size-structured population of trees. Combining the proposed advances in modelling and data processing resulted in four complementary benchmarks - reflecting different usage of the information contained in the ITRDB - each described by two metrics rooted in statistics that quantify the performance of the benchmark. Although the proposed benchmarks are unlikely to be precise, they advance the field by providing a much-needed large-scale constraint on changes in the simulated maximum tree diameter and annual growth increment for the transition from pre-industrial to present-day environmental conditions over the past century. Hence, the proposed benchmarks open up new ways of exploring the ITRDB archive, stimulate the dendrochronological community to refine its sampling protocols to produce new and spatially unbiased tree-ring networks, and help the modelling community to move beyond the short-term benchmarking of LSM.

Abstract. The search for a long-term benchmark for land-surface models (LSMs) has brought tree-ring data to the attention of the land-surface modelling community, as tree-ring data have recorded growth well before human-induced environmental changes became important. We propose and evaluate an improved conceptual framework of when and how tree-ring data may, despite their sampling biases, be used as century-long hindcasting targets for evaluating LSMs. Four complementary benchmarks – size-related diameter growth, diameter increment of mature trees, diameter increment of young trees, and the response of tree growth to extreme events – were simulated using the ORCHIDEE version r5698 LSM and were verified against observations from 11 sites in the independent, unbiased European biomass network datasets. The potential for big-tree selection bias in the International Tree-Ring Data Bank (ITRDB) was investigated by subsampling the 11 sites from European biomass network. We find that in about 95 % of the test cases, using ITRDB data would result in the same conclusions as using the European biomass network when the LSM is benchmarked against the annual radial growth during extreme climate years. The ITRDB data can be used with 70 % confidence when benchmarked against the annual radial growth of mature trees or the size-related trend in annual radial growth. Care should be taken when using the ITRDB data to benchmark the annual radial growth of young trees, as only 50 % of the test cases were consistent with the results from the European biomass network. The proposed maximum tree diameter and annual growth increment benchmarks may enable the use of ITRDB data for large-scale validation of the LSM-simulated response of forest ecosystems to the transition from pre-industrial to present-day environmental conditions over the past century. The results also suggest ways in which tree-ring width observations may be collected and/or reprocessed to provide long-term validation tests for land-surface models.

No abstract

Supplementary information 1. Description of ORCHIDEE in function of the aggregation model 1.1. Dependency of tree growth on climate For deciduous trees, when the phenological thresholds are exceeded in ORCHIDEE r5698, leaves emerge using carbon from the reserve pool and as such an essential condition for carbon assimilation is fulfilled. Carbon assimilation is calculated following the analytical solution of the Farquhar and Ball and berry model, defined by Yin and Struik (Yin and Struik, 2009):