Run to the hills: Forest growth responsiveness to drought increased at higher elevation during the late 20th century.

Climate warming is expected to enhance forest growth in cold-limited biomes while triggering reductions in drought-limited biomes. However, as temperature raises, it is unclear how temperature- and drought-growth couplings shift across elevation gradients in different biomes. We still lack comprehensive analyses on how altitude modulates the influence of temperature and drought on tree growth during the second half of the 20th century when climate warming accelerated. We compared the worldwide responses of tree growth (RWI, ring-width indices) to two of its major climatic constraints, growing-season minimum temperatures and drought (SPEI index), across biomes and elevation gradients during two periods with different warming rates (1960-1980 vs. 1980-2000). We found a decrease in the correlations of minimum temperatures with growth, but a strengthening of drought-growth relationships. However, these patterns varied across biomes because correlations between growth and temperature decreased in temperate forests and woodland shrubland, while correlations between growth and SPEI increased in boreal forests and decreased in temperate forests. Differences in growth responsiveness to climate between the two periods were more marked for mid-latitude forests situated between 1200 and 1600 m. The slopes of the relationships between growth-temperature correlations and elevation decreased in late spring and midsummer. The slopes of the relationships between growth-drought correlations and elevation increased in temperate forests and woodland shrubland suggesting that drought impacts are "climbing" in these biomes. Temperature controls on forest growth are relaxing as the climate warms, while drought is becoming a more significant constraint for tree growth, particularly for mid-elevation forests and in drought-prone woodland and shrubland. The strengthening of drought-growth coupling should be considered in vegetation models to reduce the uncertainty on forest climate mitigation.