Towards weather-based forecasting of annual seed production in six European forest tree species

Abstract

Ecological forecasting is essential for climate change adaptation and mitigation. In reforestation and restoration, seed production forecasting has the potential to support planning and resource allocation, while providing benefits for wildlife management and public health. We hind- and forecast seed production using statistical models based on weekly weather and high-resolution seed data of six tree species recorded in two Austrian old-growth forest sites. Using a sliding-window approach and model selection, we model annual reproduction for three coniferous (Silver fir, European larch, Norway spruce) and three broad-leaved species (Sycamore maple, European beech, European ash). We investigate changes in explained variance with decreasing time before seed rain and evaluate hindcasting proficiency and the potential forecast horizon using quantitative and categorical measures useful to practitioners. Furthermore, we compare the performance of local models with a general forecasting model for seed production in European beech. Most local models show unbiased but partly imprecise predictions with a broad range in explained variance (0.15–0.93) in the year prior to seed rain. Nonetheless, within this timeframe, hindcasting seed rain above 10% of the long-term maximum, a threshold relevant to practitioners, works well for all species. Previous seed rain explains a considerable proportion of the variation in seed rain of fir, ash, and maple. We forecast seed rain for 2022 to 2025, with mixed results for 2022 and 2023. Overall, categorical year out predictions seem feasible for most studied species, but model refinement is required. Local models for beech outperformed the general model in predicting crop failures and bumper crops. Synthesis and applications: Seed production can be predictable with a promising degree of accuracy for six European tree species in the year prior to seed rain, if combined with on-site monitoring of seeds, phenology, and weather. This holds value for seed harvesters, nurseries and forest managers, and may inform orchard management and public health risk anticipation. Seed forecasts will help address seed scarcity and support climate change adaptation and mitigation. Future efforts should prioritise species based on rarity and seed storability, as well as forming stronger partnerships with potential forecast users.