Extinction debt and delayed colonization have had comparable but unique effects on plant community–climate lags since the Last Glacial Maximum

AIM: Plant communities typically exhibit lagged responses to climate change due to poorly understood effects of colonization and local extinction. Here, we quantify rates of change in mean cold tolerances, and contributions of colonization and local extinction to those rates, recorded in plant macrofossil assemblages from North American hot deserts over the last 30,000 years. LOCATION: Mojave, Sonoran and Chihuahuan Deserts. TIME PERIOD: 30–0 thousand years before present (kybp). MAJOR TAXA STUDIED: Vascular plants. METHODS: Colonization and local extinction dates for 269 plant species were approximated from macrofossils in 15 packrat (Neotoma) midden series. Cold tolerances estimated from contemporary climate were used to quantify assemblage‐mean cold tolerances through time. Rates of colonization and local extinction, and their effects on rates of change in assemblage‐mean cold tolerances, were estimated for 30–20 kybp (Late Pleistocene, no directional warming), 20–10 kybp (deglaciation, rapid warming) and 10–0 kybp (Holocene, no directional warming). RESULTS: Rates of change in all metrics were negligible during the Late Pleistocene. Rates of change in assemblage‐mean cold tolerances (mean 1.0°C × 10⁻⁴/yr) lagged behind warming during deglaciation, and continued at similar rates (1.2°C × 10⁻⁴/yr) throughout the Holocene. Colonization and local extinction contributed equally to delayed responses to warming, but their dynamics differed through time: Colonization by warm‐adapted species predominated during deglaciation, while the most heat‐adapted species exhibited long delays in colonization. Only the most cold‐adapted species went locally extinct during deglaciation, followed by slow repayment of the extinction debt of cool‐adapted species during the Holocene. MAIN CONCLUSIONS: Responses to rapid warming can persist for millennia, even after cessation of warming. Consistent patterns from different midden series across the region support a metacommunity model in which dispersal interacts with environmental filters and buffers against local extinction to drive community–climate disequilibrium during and after periods of warming.