Temperature is a key climate indicator, whose distribution is expected to shift right in a warming world. However, the high temperature tolerance of trees is less widely understood than their drought tolerance, especially when it comes to sub-lethal impacts of temperature on tree growth. I use a large data set of annual tree ring widths, combined with a flexible degree-day model, to estimate the relationship between temperature and tree radial growth. I find that tree radial growth responds non-linearly to temperature across many ecoregions of the US: across temperate and/or dry ecoregions, spring-summer temperature increases are beneficial or mostly neutral for tree growth up to around 25-30°C in humid climates and 10-15°C in dry climates, beyond which temperature increases suppress growth. Thirty additional degree-days above the optimal temperature breakpoint lead to an average decrease in tree ring width of around 1-5%, depending on ecoregions, seasons, and inclusion or exclusion of temperature-mediated drought impacts. High temperatures have legacy effects across a 5-year horizon in dry ecoregions, but none in the temperate-humid South-East or among temperature-sensitive trees. I find limited evidence that trees acclimatize to high temperatures within their lifetime: local variation in exposure to high temperatures, which stems from local variation in the timing of tree birth, does not significantly impact the response to high temperatures, although temperature-sensitive trees acquire some heightened sensitivity from early exposure. I also find some evidence that trees adapt to high temperatures in the long-run: across humid ecoregions of the US, high temperatures are 40% less harmful to tree growth, where their average incidence is one standard deviation above average. Overall, these results highlight the strength of a new methodology which, applied to representative tree ring data, could contribute to predicting forest carbon uptake potential and composition under global change.

The data used in this paper have been compiled from freely and publicly available sources. Data processing is described in detail in the associated paper. Processing of tree ring data relies in part on Zhao et al. 2018 "The International Tree‐Ring Data Bank (ITRDB) revisited: Data availability and global ecological representativity".

• Tree ring width data and associated metadata come from the International Tree-Ring Data Bank (ITRDB; https://www.ncei.noaa.gov/access/paleo-search, accessed February 2019). I use a subset of these data, which covers the contiguous United States, and years 1903 to 2016.
• Temperature and precipitation data are based on the "Parameter-Elevation Regressions on Independent Slopes Model" (PRISM), which covers the contiguous US at a 4 km resolution. These data come from a version of PRISM, which is adapted to panel analyses (http://www.columbia.edu/~ws2162/links.html).
• Drought data come from the Standardised Precipitation-Evapotranspiration Index (SPEI) database (https://spei.csic.es/database.html).
• Ecoregion information come from the Köppen-Geiger climate classification (http://koeppen-geiger.vu-wien.ac.at/present.htm).