Rainfall-triggered carbon losses from drylands are widely underestimated in current ecosystem models, according to a study published in Nature by Trevor F. Keenan and colleagues. Using a dataset of 1,857 rainfall "pulse events" from 34 dryland sites, the researchers show that these short-term CO₂ emissions contribute about 17% of annual ecosystem respiration and nearly 10% of net ecosystem productivity.
Existing carbon flux partitioning methods, both parametric and machine learning, failed to capture these events, underestimating CO₂ emissions by up to 27%. The study introduces FluxPulse, a new correction method that incorporates rain-induced microbial respiration surges, known as the Birch effect.
The team also found a consistent exponential decay in carbon pulses after rain, influenced by vegetation productivity, aridity, and soil pH. These findings suggest that more intense and less frequent rainfall-projected under climate change-could increase carbon losses from drylands, underscoring the need to revise global carbon models.
