Nature-based solutions in savanna-grasslands can be both uncertain and unstable

Altering management of disturbances in savanna-grasslands, a biome spanning >20 million km² and under extensive human management, can offset a substantial proportion of anthropogenic carbon emissions. Quantifying the long-term carbon accrual and storage can be challenging because much of the change occurs belowground and it requires an understanding of ecological processes in a diverse set of environmental conditions. Here, we focus on the role of altered fire regimes given ca. 3 million km² of savanna-grasslands burn annually. Combining repeated measurements of total ecosystem carbon stocks in a 58-yearlong fire-manipulation experiment with a global dataset, we demonstrate that while fire management leads to large changes in carbon sequestration, its magnitude and persistence over time is highly variable. In the experiment, fire exclusion resulted in large increases in carbon in soil organic matter via increased aboveground biomass inputs and faster decomposition. Increased burning frequency decreased carbon but this was partly offset by increased inputs from fine root turnover. However, repeated measurements illustrated both the magnitude of the sequestration within a plot and the differences across treatments changed—and even reversed—through time due to changes in tree inputs following a disease outbreak. The global dataset revealed that in wet sites, carbon sequestered in trees is most important but in drier sites the sequestration in soil organic matter is most important. Within soils, much of the variability in carbon accrual was due to variability in how much woody biomass inputs changed, with drier sites experiencing large changes. Consequently, much of the variability in carbon accrual is due to variability in the amount of woody biomass inputs and decomposition into soils. Because trees can be highly sensitive to changing disturbance regimes in drylands we propose that using fire management to sequester carbon in soils can be highly uncertain and unstable through time.