The influence of heterogeneity on soil microbial processes in agroecosystems - Microbial Ecology in Sustainable Agroecosystems - page 53

temperature are the primary controls on diel soil temperature variation (Hillel, 1998). This
indicates that aggregation bias should decrease as soil albedo and moisture content rise.
In contrast, as agronomic management practices are implemented to promote soil drying
(e.g., to increase seedbed temperature ahead of row crop planting in humid regions), the
diel soil temperature variance becomes increasingly important.
Several biogeochemically important soil process-environmental driver functions are
more complex than monotonic functions and require individual consideration for inclu-
sion in process-based modeling. For instance, the relationship between soil moisture and
soil microbial activity is controlled by two separate factors. At low soil moisture content
(approximately less than field capacity), the diffusion of soluble substrates such as simple
organic compounds and enzymes limits microbial activity, whereas at soil moisture con-
tents greater than field capacity microbial activity tends to be limited by the diffusion of
soil O 2 . These two opposing controls on microbial activity result in a discontinuous micro-
bial response function to soil moisture content (Linn and Doran, 1984) with an optimum
near field capacity (~60% water-filled pore space) and minimal activity at the extreme soil
moisture contents ( Figure 3.3a ) .
For complex functional responses, the question then becomes, how does aggregating
soil moisture content across space or time influence estimation bias (i.e., aggregation bias)?
To explore this question, I use soil moisture and ammonium (NH 4 + ) concentration data
and a simple model to describe soil nitrification rates (Parton et al., 1996):

Nitrification Rate = [NH4+] * WFPSf f * Temperature f * pH f (3.3)

where [NH 4 + ] is the micrograms of 1 N KCl extractable NH 4 + as nitrogen per gram of dry
soil; WFPSf f is the water-filled pore space activity coefficient ( Figure 3.3a ) ; temperature f is the
temperature activity coefficient; and pH f is an activity coefficient with respect to soil pH. To
contrast the influence of a complex functional soil process response with that of a simple
monotonic response (e.g., soil temperature), I limit the nitrification model to include soil mois-
ture and NH 4 + concentration only and assume a constant temperature and pH for simplicity.
For the soil data, I use data collected from the Kellogg Biological Station's Long-Term
Ecological Research (KBS-LTER) site located in southwest Michigan in the United States.
Specifically, I use data from an experiment comparing four annual row crop manage-
ment systems that vary in management intensity (e.g., tillage, fertilizer, and pesticide
inputs). Soil cores (five 2.5-cm inner diameter [i.d.] × 30 cm deep) have been collected
monthly (1989-2001) from each replicate ( r = 6) of each treatment ( n = 4). Within-replicate
soil samples are composited and subsampled for moisture and inorganic N (NO 2 − , NO 3 − ,
and NH 4 + ).
For modeling simplicity, one might desire to average soil data across replicates to
estimate nitrification rates for each treatment on each sampling date (i.e., the aggregated
estimate) rather than using the individual replicate data to predict nitrification and then
average plot-level estimates to obtain treatment-level nitrification rates (i.e., the unbiased
estimate). Doing so creates a treatment-level aggregation bias that is not related to treat-
ment-level soil moisture variance ( Figure 3.3b ) . The aggregation bias is also not simply
related to soil moisture content (data not shown). This random aggregation bias precludes
any simple correction for variance like that available with simple monotonic functions
(e.g., soil temperature vs. microbial activity). The importance of the aggregation bias
observed in the nitrification example presented is relative to the desired precision of esti-
mate. The long-term (12 years of data) aggregation bias has a distinct central tendency of
no net bias ( Figure 3.3b ) ; however, the aggregation bias over shorter periods of time (e.g.,