Agriculture Reference
In-Depth Information
key model components and the minimum data required for simulating crop yields
under different management practices. Research at the Kellogg Biological Station
Long-Term Ecological Research Site (KBS LTER) provides the opportunity to test
models of long-term changes in soil carbon, nitrogen leaching, crop yields, and
gaseous emissions from soil. Data from KBS LTER also provide an excellent con-
text for illustrating the utility and limitations of crop models, and we use these data
to show two examples of model applications: (1) an evaluation of nitrate leaching
as affected by nitrogen fertilizer management in a corn ( Zea mays L.) and alfalfa
( Medicago sativa L.) rotation and (2)  soil carbon dynamics under various tillage
systems. We also illustrate spatially connected processes by linking SALUS to digi-
tal terrain modeling.
Crop Models
Crop simulation models range from simple to complex. Simple models are often
adopted to estimate yield across large land areas based on statistical information
related to climate and historical yields and include little detail about the soil-plant
system. The more sophisticated physically based models are capable of providing
additional details on processes in the soil-plant-atmosphere system, but sophisti-
cated models demand detailed initial environmental and agronomic information
that may be unavailable in many situations.
Crop models may be either deterministic or stochastic. Deterministic models
provide a specific outcome for a certain set of conditions, with all plants and soil
within the simulation space assumed to be uniform. Stochastic models produce out-
comes that incorporate uncertainty due to spatial variability of soil properties, tem-
poral variability of weather conditions, abiotic and biotic factors not accounted for
in a deterministic model, and uncertainties of model logic and functions. However,
stochastic crop models are at an early stage of development and not used in DSSs
to our knowledge.
To overcome some of the problems of using deterministic crop models, soils
with known spatial variability can be grouped into small homogenous units and
the results aggregated to model yield at the whole-field scale. Similarly, running
simulations over multiple years with deterministic yields accounts for temporal
variability (Basso et al. 2007).
Deterministic crop models can be statistical, mechanistic, or functional
(Addiscott and Wagenet 1985, Ritchie and Alagarswamy 2002). Statistical
models—fitting a function to observed weather variables and crop regional yield
statistics to predict crop yield—were the first crop models used for large-scale yield
estimations. Average regional yields were regressed on time to reveal a general
trend in crop yields (Thompson 1969; Gage et al. 2015, Chapter 4 in this volume).
An example is the upward trend in crop yield over the past several decades due to
technological advancements in genetics and management, especially the increased
use of fertilizers. Thompson (1986) quantified the impact of climate change and
variability on corn yield in five U.S. states using a statistical model. In that study,
preseason precipitation (September-June), June temperature, and temperature and
Search WWH ::




Custom Search