Frequent burning causes large losses of carbon from deep soil layers in a temperate savanna

Fire activity is changing dramatically across the globe, with uncertain effects on ecosystem processes, especially below-ground. Fire-driven losses of soil carbon (C) are often assumed to occur primarily in the upper soil layers because the repeated combustion of above-ground biomass limits organic matter inputs into surface soil. However, C losses from deeper soil may occur if frequent burning reduces root biomass inputs of C into deep soil layers or stimulates losses of C via leaching and priming. To assess the effects of fire on soil C, we sampled 12 plots in a 51-year-long fire frequency manipulation experiment in a temperate oak savanna, where variation in prescribed burning frequency has created a gradient in vegetation structure from closed-canopy forest in unburned plots to open-canopy savanna in frequently burned plots. Soil C stocks were nonlinearly related to fire frequency, with soil C peaking in savanna plots burned at an intermediate fire frequency and declining in the most frequently burned plots. Losses from deep soil pools were significant, with the absolute difference between intermediately burned plots versus most frequently burned plots more than doubling when the full 1 m sample was considered rather than the top 0–20 cm alone (losses of 98.5 Mg C/ha [−76%] and 42.3 Mg C/ha [−68%] in the full 1 m and 0–20 cm layers respectively). Compared to unburned forested plots, the most frequently burned plots had 65.8 Mg C/ha (−58%) less C in the full 1 m sample. Root biomass below the top 20 cm also declined by 39% with more frequent burning. Concurrent fire-driven losses of nitrogen and gains in calcium and phosphorus suggest that burning may increase nitrogen limitation and play a key role in the calcium and phosphorus cycles in temperate savannas. Synthesis. Our results illustrate that fire-driven losses in soil C and root biomass in deep soil layers may be critical factors regulating the net effect of shifting fire regimes on ecosystem C in forest-savanna transitions. Projected changes in soil C with shifting fire frequencies in savannas may be 50% too low if they only consider changes in the topsoil.