use for describing C and N cycling typically have at most 10 or 20 chemical flows. It
is not mathematically possible to write solvable equations that link 10,000 independent driver variables to 10 or even 100 response variables. We must either collapse
down the ‘omics data set or increase the number of processes expressed in the model,
or some combination of both, to match the numbers of independent and dependent
variables. Here, too, approaches are developing to solve this problem (Treseder et al.
2011). Identifying functional guilds (e.g., Moorhead and Sinsabaugh 2006) possibly
by using big data techniques such as network analysis to identify clusters of
organisms that respond in similar ways to environmental drivers (moisture, substrates, etc.) offers great potential (Barberán et al. 2012; Lennon et al. 2012). Such
tools provide a more empirical definition of “functional group” than using phylogeny as the only guide, or by classifying organisms purely functionally—i.e., “nitrifiers,” a group that now includes bacteria, archaea, and even fungi, but different
groups of ammonia oxidizers respond to environmental drivers quite differently. In
the carbon cycle, of course, “heterotroph” encompasses everything from Escherichia
coli to Homo sapiens, not a useful functional group definition!
The next challenge is model parameterization. To develop and run a model, you
must know the values of the parameters that go into the model. If specific microbial
populations are to drive processes in a model, we need to know the size and
dynamics of those populations. But, we don’t have a time machine—we can’t
know what those populations will be at some future time when climate has changed.
Equally, we can’t sample DNA from every hectare on Earth to map populations
spatially. Requiring measured data over both future time and space to parameterize a
large-scale model would therefore be impossible-squared. Including actual population sizes as drivers in a process model limits its scope to a fine-resolution, shorttime-scale, mechanistic exploration of how processes function. That isn’t a criticism,
but it is a constraint.
We can get past that constraint to some degree by understanding what drives
those microbial communities well enough to reasonably predict what they will be in
the future, or in other locales, and so include them as explicit terms in an equation
describing a process. But, if we can do that, then we also understand their behavior
well enough to collapse them out of the equation!
Instead, approaches to capturing microbial processes in models—to count the
votes of the “microbe party”—have “modelled past” the microbes by assuming that
microbial communities are in equilibrium with their environment. This allows
microbial dynamics to be collapsed into model equations implicitly as rate constants
and response functions that relate the environmental drivers to process rates
(Schimel 2001). But what do we do when conditions are changing and we cannot
assume that communities are in equilibrium with their environment and resources?
If the goal is to better describe biogeochemical processes at large scales, and
under changing environmental conditions, we need the insights into process dynamics and linkages that only fine-scale analyses and models can provide, but then we
must figure out how to distill out the essence of those phenomena that make a
difference at the macro-scale.
4 The Democracy of Dirt: Relating Micro-Scale Dynamics to Macro-Scale Ecosystem. . .
93
is not mathematically possible to write solvable equations that link 10,000 independent driver variables to 10 or even 100 response variables. We must either collapse
down the ‘omics data set or increase the number of processes expressed in the model,
or some combination of both, to match the numbers of independent and dependent
variables. Here, too, approaches are developing to solve this problem (Treseder et al.
2011). Identifying functional guilds (e.g., Moorhead and Sinsabaugh 2006) possibly
by using big data techniques such as network analysis to identify clusters of
organisms that respond in similar ways to environmental drivers (moisture, substrates, etc.) offers great potential (Barberán et al. 2012; Lennon et al. 2012). Such
tools provide a more empirical definition of “functional group” than using phylogeny as the only guide, or by classifying organisms purely functionally—i.e., “nitrifiers,” a group that now includes bacteria, archaea, and even fungi, but different
groups of ammonia oxidizers respond to environmental drivers quite differently. In
the carbon cycle, of course, “heterotroph” encompasses everything from Escherichia
coli to Homo sapiens, not a useful functional group definition!
The next challenge is model parameterization. To develop and run a model, you
must know the values of the parameters that go into the model. If specific microbial
populations are to drive processes in a model, we need to know the size and
dynamics of those populations. But, we don’t have a time machine—we can’t
know what those populations will be at some future time when climate has changed.
Equally, we can’t sample DNA from every hectare on Earth to map populations
spatially. Requiring measured data over both future time and space to parameterize a
large-scale model would therefore be impossible-squared. Including actual population sizes as drivers in a process model limits its scope to a fine-resolution, shorttime-scale, mechanistic exploration of how processes function. That isn’t a criticism,
but it is a constraint.
We can get past that constraint to some degree by understanding what drives
those microbial communities well enough to reasonably predict what they will be in
the future, or in other locales, and so include them as explicit terms in an equation
describing a process. But, if we can do that, then we also understand their behavior
well enough to collapse them out of the equation!
Instead, approaches to capturing microbial processes in models—to count the
votes of the “microbe party”—have “modelled past” the microbes by assuming that
microbial communities are in equilibrium with their environment. This allows
microbial dynamics to be collapsed into model equations implicitly as rate constants
and response functions that relate the environmental drivers to process rates
(Schimel 2001). But what do we do when conditions are changing and we cannot
assume that communities are in equilibrium with their environment and resources?
If the goal is to better describe biogeochemical processes at large scales, and
under changing environmental conditions, we need the insights into process dynamics and linkages that only fine-scale analyses and models can provide, but then we
must figure out how to distill out the essence of those phenomena that make a
difference at the macro-scale.
4 The Democracy of Dirt: Relating Micro-Scale Dynamics to Macro-Scale Ecosystem. . .
93