Contribution of land use to the interannual variability of the land carbon cycle

Abstract

Understanding the driving mechanisms of the interannual variability (IAV) of the net land carbon balance (S_net) is important to predict future climate-carbon cycle feedbacks. Past studies showed that the IAV of S_net was correlated with tropical climate variation and controlled by semiarid vegetation. But today's land ecosystems are also under extensive human land use and management. Here, we report a previously hidden role of land use in driving the IAV of S_net by using an improved biosphere model. We found that managed land accounted for 30-45% of the IAV of S_net over 1959-2015, while the contribution of intact land is reduced by more than half compared with previous assessments of the global carbon budget. Given the importance of land use in modulating future land climate-carbon cycle feedbacks, climate mitigation efforts should strive to reduce land-use emissions and enhance the climate resilience of carbon sinks over managed land.

Fig. 1. LUC processes and associated carbon fluxes.

LUC processes considered in this study are: conversion of intact land (exemplified as intact forest (a)) into agricultural land (c), forest wood harvest for fuel wood (d), and industrial wood (e), and regeneration of secondary forest following harvest or agricultural abandonment (f). An old forest was used as an example for intact land in this figure, but similar land transitions involving natural grassland were also included. Likewise, pasture was also included as a form of agricultural land. The individual carbon fluxes comprising E_LUC are: b, d E_fire, immediate emissions following forest clearing through burning of aboveground biomass and other on-site disturbance, plus emissions from harvested fuel wood assumed to be burned at the year of harvest; e E_legacy, emissions from recently established agricultural land that is dominated by the decomposition of legacy slash and soil carbon inherited from former intact land; b E_wood, long-term, gradual carbon release from industrial wood product degradation; and f S_recov, carbon sink in recovering secondary forest and grassland. The net land-use change emissions (E_LUC) is defined as: E_LUC = E_fire + E_wood + E_legacy − S_recov, with a positive sign indicating a carbon source to the atmosphere. The dashed arrows indicate conversion of secondary forest (or grassland) into agricultural land in shifting cultivation, or reharvest of wood in case of forest management.

Fig. 2. Land–atmosphere carbon fluxes over managed and intact lands.

The annual mean values for 1990–2015 were shown. a E_LUC over managed land. b S_Intact. c S_net. The further disaggregation of E_LUC into its component emission fluxes of E_fire, E_wood, and E_legacy, and the sink flux of S_recov are shown in d–g, respectively.

Fig. 3. Forest carbon sinks by ORCHIDEE and an inventory-based study.

Data for 2000-2007 were shown. a Temperate and boreal regions, with sinks of intact (below horizontal white lines) and secondary (above horizontal white lines) forests being separated for ORCHIDEE simulation. b Forest carbon sinks over intact forests for tropical regions. Refer to Supplementary Fig. 1 for the global distribution of these eight regions.

Fig. 4. Comparison of E_LUC by ORCHIDEE and the bookkeeping model.

a ORCHIDEE model. b The HN2017 bookkeeping model. Refer to Fig. 1 for the meanings of the four component fluxes of E_LUC.

Fig. 5. Overview of the global carbon cycle for 1959–2015.

a E_LUC was derived from the HN2017 bookkeeping model with the intact land sink, as a residual of the global carbon budget. b All E_LUC component fluxes and the intact land sink were simulated by ORCHIDEE. Note that all land fluxes by ORCHIDEE include full impacts of environmental changes and climate variations. The difference between the bottom edge of the atmosphere sink and the pink line in b indicates the mismatch between the modeled and observation-based global carbon budgets.

Fig. 6. Attribution of S_net and its variance into E_LUC and S_Intact.

a S_net. b E_LUC. c S_Intact. Data were shown for 1959–2015. The red line indicates ORCHIDEE result; the black line shows the result, following the bookkeeping and residual budget approach (i.e., IPCC AR5 and ref. ), with shaded gray areas indicating the uncertainty of one standard deviation. Pearson correlation coefficients (r) between the two estimates are shown, with the period of 1991–1993 under strong Pinatubo influence being excluded when calculating r (n = 54, p-value calculated as a two-sided p-value). d shows the decomposition of temporal variances (Var) of S_net into the variances of E_LUC, S_Intact, and their covariance (n = 57, Eq. (5) in “Methods” section).

Fig. 7. Temperature sensitivity (γ) of S_net, S_Intact, and E_LUC.

Red color indicates results from ORCHIDEE. Black color indicates results for the bookkeeping and residual budget approach (i.e., IPCC AR5 and ref. ). a The temperature sensitivities of S_net (${\gamma }_{\mathrm{LAND}}^{\mathrm{T}}$). b The decomposition of ${\gamma }_{\mathrm{LAND}}^{\mathrm{T}}$ into ${\gamma }_{\mathrm{Intact}}^{\mathrm{T}}$ and $-{\gamma }_{\mathrm{ELUC}}^{\mathrm{T}}$, following the equation “S_net = S_Intact − E_LUC”. Negative values of ${\gamma }_{\mathrm{LAND}}^{\mathrm{T}}$ and ${\gamma }_{\mathrm{Intact}}^{\mathrm{T}}$ mean that elevated tropical land warming leads to less land carbon uptake, while positive values of ${\gamma }_{\mathrm{ELUC}}^{\mathrm{T}}$ mean that warming leads to enhanced carbon emissions from managed land (note that $-{\gamma }_{\mathrm{ELUC}}^{\mathrm{T}}$ is shown in the figure). All linear regressions were significant with a p < 0.05 (n = 57, two-sided p-value). Shaded area in subplot a indicates the 95% confidence interval of fitted values. Error bars in subplot b indicate the standard error of fitted γ values.