Trade-off between spring phenological sensitivities to temperature and precipitation across species and space in alpine grasslands over the Qinghai-Tibetan Plateau

Elucidating climatic drivers of spring phenology in alpine grasslands is critical. However, current statistical estimates of spring phenological sensitivities to temperature and precipitation (beta T and beta P) might be biased and their variability across sites and species are not fully explained. We benchmarked species-level beta T and beta P statistically inferred from historical records with observations from a field manipulative experiment. We then analyzed landscape scale beta T and beta P estimated from the best statistical approach in the benchmark analysis across 57 alpine grassland sites in the Qinghai-Tibetan Plateau. Compared with manipulative experiment results, process-agnostic regression-based approaches underestimate beta T by 2.36-3.87 d degrees C-1 (54-88%) while process-based phenology model fitting predicts comparable beta T and beta P. Process-based estimates of beta T and beta P are negatively correlated across species (R = -0.94, P < 0.01) and across sites (R = -0.45, P < 0.01). beta T is positively correlated with mean annual temperature, and beta P is negatively correlated with elevation at the regional scale. Using process-based model fitting can better estimate spring phenological sensitivities to climate. The trade-off between beta T and beta P contributes to species-level and site-level variabilities in phenological sensitivities in alpine grasslands, which needs to be incorporated in predicting future phenological changes.