Low soil moisture during hot periods drives apparent negative temperature sensitivity of soil respiration in a dryland ecosystem: a multi-model comparison

Arid and semiarid ecosystems (drylands) may dominate the trajectory of biosphere-to-atmosphere carbon (C) exchange, and understanding dryland CO2 efflux is important for C cycling at the global-scale. However, unknowns remain regarding how temperature and moisture interact to regulate dryland soil respiration (R (s) ), while 'islands of fertility' in drylands create spatially heterogeneous R (s) . At a site in southeastern Utah, USA we added or removed litter (0-650 % of control) in plots associated with either shrubs or biological soil crust-dominated interspaces between vascular plants. We measured R (s) , soil temperature (T-s), and water content (theta) repeatedly from October 2013 to November 2014. R (s) was highest following rain in late summer at T-s similar to 30 A degrees C, and lowest mid-summer at T-s > 40 A degrees C, resulting in apparent negative temperature sensitivity of R (s) at high temperatures, and positive temperature sensitivity at low-moderate temperatures. We used Bayesian statistical methods to compare models capturing a range of hypothesized relationships between T-s, theta, and R (s) . The best model indicates that apparent negative temperature sensitivity of R (s) at high T-s reflects the control of water content, not high temperatures. Modeled Q(10) ranged from 2.7 to 1.4 between 5 and 45 A degrees C. Litter addition had no effect on Q(10) or reference respiration (R (ref) = R (s) at 20 A degrees C and optimum theta) beneath shrubs, and little effect on R (ref) in interspaces, yet R (ref) was 1.5 times higher beneath shrubs than in interspaces. Altogether, these results suggest reduced R (s) often observed at high T-s in drylands is dominated by the control of theta, and, on shorter-timescales, variable litter inputs exert minimal control over R (s) .