ABSTRACT FINAL ID: B31D-0346;

TITLE: Assessing Variability in Belowground Carbon for CMIP-5 Models

SESSION TYPE: Poster

SESSION TITLE: B31D. Improving Predictions of the Global Carbon Cycle and Climate in Earth System Models: New Mechanisms, Feedback Sensitivities, and Approaches for Model Benchmarking II Posters

AUTHORS: Katherine E Todd-Brown1, Forrest M Hoffman4,1, James Tremper Randerson1, Wilfred M Post2, Steven D Allison3,1

INSTITUTIONS:
1Earth System Science, Univ. of California, Irvine, Irvine, CA, United States.
2Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN, United States.
3Ecology and Evolutionary Biology, Univ. of California, Irvine, Irvine, CA, United States.
4Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, TN, United States.

ABSTRACT BODY: Soil carbon stocks and heterotrophic respiration are critical components of the global carbon cycle. This study compared soil carbon survey data with modeled stocks and assessed the magnitude of change in modeled soil carbon between 1850 and 2005. Model means (1995–2005) from the CMIP-5 historical experiment were compared to soil carbon estimates from the Harmonized World Soil Database (HWSD) across biomes derived from the MODIS/Terra+Aqua land cover type. Changes in soil carbon stocks were calculated as the difference between the 1850–1860 and 1995–2005 model means.

In general, the models overestimated modern soil carbon stocks in areas of relatively low soil carbon (<100 kg m-2) by 310 Pg (28%) and underestimated modern soil carbon stocks in areas of relatively high soil carbon (>100 kg m-2) by 1012 Pg (85%) (n=7). Total modern soil carbon ranged from 1120–2523 Pg among the models versus 2280 Pg from the HWSD. These results imply that the models do not represent the high soil carbon of peatlands well. Models underestimated soil carbon in the boreal forest by 417 Pg (61%) (sd=47, cv=0.11, n=7). Soil carbon predictions for temperate forests were closest to HWSD with an overestimate of 13 Pg (21%) (sd=13, cv=0.96, n=7). Over the course of the simulations, average model soil carbon stocks increased 6% gobally (sd=4, cv=0.65, n=7) from 1563 Pg to 1664 Pg. This change was accompanied by a 22% (sd=5, cv=0.26, n=4) increase in modeled NPP from 73 Pg-C/yr to 89 Pg-C/yr for models that report NPP.

In summary, the CMIP5 models tend to underestimate total soil carbon stocks, particularly in areas with high soil carbon concentration. Despite rising temperatures over the past 150 years, these models also predicted a substantial increase in soil carbon stocks that is likely due to increased NPP. Whether this rate of increase can be verified with empirical data or sustained over the 21st century is unclear. Given the discrepancies between modeled and observed distributions of global soil carbon, this study provides a strong rationale for improving soil carbon representation in global climate models.

KEYWORDS:
[0414] BIOGEOSCIENCES / Biogeochemical cycles, processes, and modeling,
[0428] BIOGEOSCIENCES / Carbon cycling,
[0550] COMPUTATIONAL GEOPHYSICS / Model verification and validation.

SPONSOR NAME: Katherine Todd-Brown

CONTACT: Katherine Todd-Brown <ktoddbro at uci dot edu>