UC Irvine
UC Irvine Previously Published Works
Title
Global soil carbon projections are improved by modelling microbial processes
Permalink
https://escholarship.org/uc/item/3hd31556
Journal
Nature Climate Change, 3(10)
ISSN
1758-678X
Authors
Wieder, WR
Bonan, GB
Allison, SD
Publication Date
2013-10-01
DOI
10.1038/nclimate1951
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
Global soil carbon projections are improved by modeling microbial processes
William R. Wieder
1
Gordon B. Bonan
1
Steven D. Allison
2
1
National Center for Atmospheric Research, Boulder, CO 80307, USA
2
Department of Ecology and Evolutionary Biology & Department of Earth System Science,
University of California, Irvine, CA 92697, USA
Corresponding Author:
William R. Wieder
Phone: 303.497.1352
Fax: 303.497.1348
email: wwieder@ucar.edu
address: TSS, CGD/NCAR
PO Box 3000
Boulder, CO 80307-3000
2
Society relies on Earth system models (ESMs) to predict future climate and carbon (C) 1
cycle feedbacks. However, the soil C response to climate change is highly uncertain in these 2
models
1,2
, and they omit key biogeochemical mechanisms
3-5
. Specifically, the traditional 3
approach in ESMs lack direct microbial control over soil C dynamics
6-8
. Thus, we tested a new 4
model that explicitly represents microbial mechanisms of soil C cycling at the global scale. 5
Compared to traditional models, the microbial model simulates soil C pools that more closely 6
match contemporary observations. It also predicts a much wider range of soil C responses to 7
climate change over the twenty-first century. Global soils accumulate C if microbial growth 8
efficiency declines with warming in the microbial model. If growth efficiency adapts to warming, 9
the microbial model predicts large soil C losses. By comparison, traditional models predict 10
modest soil C losses with global warming. Microbes also change the soil response to increased C 11
inputs, as might occur with CO
2
or nutrient fertilization. In the microbial model, microbes 12
consume these additional inputs; whereas in traditional models, additional inputs lead to C 13
storage. Our results indicate that ESMs should simulate microbial physiology in order to more 14
accurately project climate change feedbacks. 15
Contemporary ESMs use traditional soil C models, which implicitly simulate microbial 16
decomposition via first-order kinetics that determine turnover rates of soil C pools
1,2
. Although 17
such models can replicate extant soil C pools at various scales
9,10
, their ability to predict soil C 18
response in a changing environment remains unresolved
11,12
. In the past 30 years, researchers 19
have identified key processes and feedbacks that could be important for accurately simulating 20
future C cycle—climate feedbacks. For example, traditional models neglect microbial 21
physiological processes that transform and stabilize soil C inputs
3-5
. In contrast, recent microbial 22
models explicitly simulate microbial biomass pools that catalyze soil C mineralization
6,8
and 23
3
produce notably different results in transient simulations
6
. By representing microbial 24
physiological responses, such models may provide a better fit to observations, especially in a 25
changing environment
13,14
. Yet to date, no modeling studies have tested the relevance of 26
microbial mechanisms for soil C responses to climate change at the global scale. 27
We created a new soil biogeochemistry module for use in the Community Land Model 28
that explicitly simulates microbial biomass pools (hereafter referred to as the CLM microbial 29
model; Fig. 1; modified from ref.
6
). The CLM microbial model represents aboveground and 30
belowground processes and separates belowground pools into surface (0-30 cm) and subsurface 31
(30-100 cm) horizons. Microbes in this model directly catalyze the mineralization of litter and 32
soil C pools according to Michaelis-Menten kinetics. In this formulation, decomposition losses 33
can be limited by both substrate availability (the organic C pools) and the microbial biomass, 34
which is assumed to be the source of enzymatic activity. This structure differs from traditional 35
models in which decomposition losses depend only on first-order decay of substrate (soil C) 36
pools
6
. 37
Temperature affects three key microbial parameters in our model. The Michaelis-Menten 38
relationship requires two parameters: K
m
, the substrate half-saturation constant, and V
max
, the 39
maximal reaction velocity (Fig. 1). We used observational data to constrain these parameters 40
and their temperature sensitivities, which generally follow an exponential form
15
. The third key 41
parameter is microbial growth efficiency (MGE), which determines how much microbial 42
biomass is produced per unit of substrate consumed
16
. MGE probably declines with increasing 43
temperature, although the magnitude of the response is uncertain
17
. Consequently, C 44
decomposition depends on temperature, substrate availability, and the size of the microbial 45
biomass pool. 46
4
After running to steady-state, we compared soil C pools from the CLM microbial model 47
to soil C pools from two traditional models (illustrated with model parameterizations from 48
CLM4cn
18
and DAYCENT
10
). We also compared model outputs to observations from the 49
globally gridded Harmonized World Soils Database
19
. Global simulations were forced with 50
observationally-derived litter inputs (see methods) and with soil temperature and moisture from a 51
20
th
century simulation
18
. Overall, the CLM microbial model explained 50% of the spatial 52
variation in the soil C observations, whereas the traditional models explained 28-30% of the 53
variation and showed greater average deviations from soil C observations (Fig. 2). 54
Other traditional models perform even worse than the two reported here. For example, a 55
prior version of CLM4cn, using modeled litter inputs, explained only ~2% of the spatial 56
variation in observed soil C stocks at the 1º grid scale, and no other ESM explained more than 57
16% of the variation
2
. Some of this poor performance may be due to ESM errors in simulating 58
litter inputs. We avoided these errors by using litterfall observations for our current analysis. 59
Still, the CLM microbial model explained 20% more soil C variation than traditional CLM4cn 60
with observed litterfall, an improvement rivaling the entire explanatory power of previous 61
models. Moreover, the CLM microbial model accurately simulates observed soil C pools in both 62
surface soil layers (0-30 cm) and total soil profiles (0-100 cm; r = 0.75 and 0.71, respectively; SI 63
Fig. 1). 64
A closer examination of regional patterns illustrates specific gaps in our representation of 65
processes driving soil C cycling (Fig. 2). Some regions, especially in the tropics, have low 66
predicted soil C densities compared to soil C observations. These low biases suggest systematic 67
problems with modeling the physiochemical soil environment. Specifically, the CLM microbial 68
model does not simulate the physical protection of soil C or pH effects on soil microbial activity. 69