Vegetation distribution and terrestrial carbon cycle in a carbon cycle configuration of JULES4.6 with new plant functional types

Dynamic global vegetation models (DGVMs) are used for studying historical and future changes to vegetation and the terrestrial carbon cycle. JULES (the Joint UK Land Environment Simulator) represents the land surface in the Hadley Centre climate models and in the UK Earth System Model. Recently the number of plant functional types (PFTs) in JULES was expanded from five to nine to better represent functional diversity in global ecosystems. Here we introduce a more mechanistic representation of vegetation dynamics in TRIFFID, the dynamic vegetation component of JULES, which allows for any number of PFTs to compete based solely on their height; therefore, the previous hard-wired dominance hierarchy is removed. With the new set of nine PFTs, JULES is able to more accurately reproduce global vegetation distribution compared to the former five PFT version. Improvements include the coverage of trees within tropical and boreal forests and a reduction in shrubs, the latter of which dominated at high latitudes. We show that JULES is able to realistically represent several aspects of the global carbon (C) cycle. The simulated gross primary productivity (GPP) is within the range of observations, but simulated net primary productivity (NPP) is slightly too high. GPP in JULES from 1982 to 2011 is 133 PgC yr(-1), compared to observation-based estimates (over the same time period) between 123 +/- 8 and 150-175 PgC yr(-1). NPP from 2000 to 2013 is 72 Pg C yr(-1), compared to satellite-derived NPP of 55 Pg Cyr(-1) over the same period and independent estimates of 56.2 +/- 14.3 PgC yr(-1). The simulated carbon stored in vegetation is 542 PgC, compared to an observation-based range of 400-600 PgC. Soil carbon is much lower (1422 Pg C) than estimates from measurements ( > 2400 Pg C), with large underestimations of soil carbon in the tropical and boreal forests. We also examined some aspects of the historical terrestrial carbon sink as simulated by JULES. Between the 1900s and 2000s, increased atmospheric carbon dioxide levels enhanced vegetation productivity and litter inputs into the soils, while land use change removed vegetation and reduced soil carbon. The result is a simulated increase in soil carbon of 57 Pg C but a decrease in vegetation carbon of 98 Pg C. The total simulated loss of soil and vegetation carbon due to land use change is 138 PgC from 1900 to 2009, compared to a recent observationally constrained estimate of 155 +/- 50 PgC from 1901 to 2012. The simulated land carbon sink is 2.0 +/- 1.0 PgC yr(-1) from 2000 to 2009, in close agreement with estimates from the IPCC and Global Carbon Project.