Proximate drivers of population inter-annual variation in seed output for a masting conifer species

•One-year lagged summer mean temperature was the strongest predictor of seed output in Picea abies.•Standard deviation of one-year lagged summer mean temperature negatively affected population inter-annual variation in reproductive output (CVp).•We found a significant non-linear, hump-shaped relationship between one-year lagged autocorrelation between past and current reproduction and CVp. Assessing the drivers of population inter-annual variation in reproductive output (CVp) is key for conservation and management of forest resources, as these drivers determine variation in seed crops which are closely related to seedling recruitment (i.e. forest regeneration) and understanding the mechanisms by which they affect reproductive output sheds light into population-level resilience to changing abiotic conditions. Proximate drivers of CVp include weather conditions which act as cues for reproduction and resource-driven lagged negative autocorrelations between past and current reproduction. Increased temporal variability in weather cues and strong negative autocorrelations are both expected to increase CVp, but very few studies have robustly tested these predictions at the intra-specific level using long-term, multi-population datasets. Based on a published dataset, we used approximately 2,000 yearly observations spanning 130 years (1886–2015) and 61 populations to test for effects of temporal variability in weather cues and lagged autocorrelations on CVp in seed output for a masting conifer species (Picea abies). We found that lagged (lag –1) summer (June–July) mean temperature was the best predictor of population-level annual seed output. Contrary to expectations, however, we observed a significant negative (not positive) effect of the standard deviation of lag –1 summer mean temperature on CVp. In addition, we found a non-linear, hump-shaped relationship between lag –1 reproductive autocorrelation and CVp, suggesting a qualitative change in the effects of resource constraints on reproductive variability whereby expected positive effects change to negative when the strength of negative autocorrelations exceeds a certain level. These patterns point at unexpected mechanisms whereby temporal variability in weather cues dampens variability in reproductive output, whereas the non-linear association with the lagged autocorrelations suggest thresholds associated with resource availability leading to qualitative changes in temporal patterns of reproduction.