Process-oriented models of autumn leaf phenology: ways to sound calibration and implications of uncertain projections

Abstract

Abstract. Autumn leaf phenology marks the end of the growing season, during which trees assimilate atmospheric CO2. The length of the growing season is affected by climate change because autumn phenology responds to climatic conditions. Thus, the timing of autumn phenology is often modeled to assess possible climate change effects on future CO2-mitigating capacities and species compositions of forests. Projected trends have been mainly discussed with regards to model performance and climate change scenarios. However, there has been no systematic and thorough evaluation of how performance and projections are affected by the calibration approach. Here, we analyzed >2.3 million performances and 39 million projections across 21 process-oriented models of autumn leaf phenology, 5 optimization algorithms, ≥7 sampling procedures, and 26 climate model chains from two representative concentration pathways. Calibration and validation were based on >45 000 observations for beech, oak, and larch from 500 central European sites each. Phenology models had the largest influence on model performance. The best-performing models were (1) driven by daily temperature, day length, and partly by seasonal temperature or spring leaf phenology; (2) calibrated with the generalized simulated annealing algorithm; and (3) based on systematically balanced or stratified samples. Autumn phenology was projected to shift between −13 and +20 d by 2080–2099 compared to 1980–1999. Climate scenarios and sites explained more than 80 % of the variance in these shifts and thus had an influence 8 to 22 times greater than the phenology models. Warmer climate scenarios and better-performing models predominantly projected larger backward shifts than cooler scenarios and poorer models. Our results justify inferences from comparisons of process-oriented phenology models to phenology-driving processes, and we advocate for species-specific models for such analyses and subsequent projections. For sound calibration, we recommend a combination of cross-validations and independent tests, using randomly selected sites from stratified bins based on mean annual temperature and average autumn phenology, respectively. Poor performance and little influence of phenology models on autumn phenology projections suggest that current models are overlooking relevant drivers. While the uncertain projections indicate an extension of the growing season, further studies are needed to develop models that adequately consider the relevant processes for autumn phenology.

Process‐oriented models of leaf senescence are biased towards the mean: Impacts on model performance and future projections

“…While the leaf senescence process can arguably be captured more accurately by improved models, the BTM may still be mitigated by the performance metric rather than by the sample used for calibration. Meier and Bigler (2023b) or NSE (e.g., Delpierre et al, 2009;Keenan & Richardson, 2015;Zani et al, 2020), which tend to favor smaller simulated versus observed variability (Gupta et al, 2009). This tendency is lessened with the Kling-Gupta Efficiency, for example (Gupta et al, 2009;Kling et al, 2012).…”

“…Meier and Bigler (2023b) used different sampling procedures for selecting the observations and sites to calibrate the site-specific F I G U R E 1 Scheme of analyzed data. This study analyzed a subset of the data produced by Meier and Bigler (2023b), which contains the simulated timing of leaf senescence for the past and future. Based on observations for Fagus sylvatica, Quercus robur, and Larix decidua, process-oriented leaf senescence models were calibrated with different calibration samples (NA, not available).…”

“…The CDD model (Dufrêne et al, 2005) served as reference for all LMMs. In addition, the first climate scenario (CMC 1, i.e., the first combination of a regional and global climate model under RCP 4.5; supplement S1: table S5 in Meier & Bigler, 2023b) served as the reference for the LMMs regarding Δ 100 . The Δ 100 data were too numerous to be processed by one site-specific and one species-specific LMM.…”

“…For Δ 100 as response variable, the fixed effects were the factorized leaf senescence models and climate scenarios, δ RC , and δ SO as well as the interaction terms of the models with the climate scenarios and with δ RC . Because the Δ 100 from the site-specific models were based on calibrations with all data per site (Meier & Bigler, 2023b), the sampling procedures were only relevant for Δ 100 from the species-specific models. The residuals for all LMMs generally showed no heteroscedasticity (i.e., the assumption of constant variance is fulfilled), but were heavy tailed (see Section S2.4), indicating that they featured some deviation from the normal distribution.…”

“…In addition, models with a BTM are likely to lead to distorted shifts of future leaf senescence. While simulated trends that are overly flat (see above) distort future shifts in general, they are often calculated using the projected versus observed leaf senescence for the future and reference periods, respectively (Meier & Bigler, 2023b;Zani et al, 2020). Yet, the observations for the reference period may differ from the simulations for the same period, which results in a reference bias.…”