Agriculture Reference
In-Depth Information
rainfall in July and August were closely correlated with corn yield variations from
the trend. Recently, Gage et al. (2015, Chapter 4 in this volume) incorporated cli-
mate effects into regional yield trends with the use of a Crop Stress Index (CSI).
This approach significantly improved predictions of historical yields of corn and
soybean.
In general, the results of statistical models cannot be extrapolated to other
places and time periods because of variation in soils, landscapes, and weather
not included in the population of information from which the statistical relation-
ship was derived. Furthermore, the impact of agricultural technology cannot be
extrapolated over space and time. Despite these limitations, statistical models
can provide many insights about past yields and historical influences (Gage et
al. 2015, Chapter 4 in this volume) and can be used to inform the other kinds of
models.
Mechanistic models are based on known physical, chemical, and biological pro-
cesses occurring in the soil-plant-atmosphere continuum. Soon after computers
became available, mechanistic models were developed to simulate photosynthetic
processes such as light interception, uptake of carbon dioxide (CO
2
), carbon allo-
cation to different plant organs, and loss of CO
2
during respiration, as well as the
dynamics of soil water including infiltration, evaporation, drainage, and root uptake.
Mechanistic models describe processes at fine time scales (e.g., photosynthesis
and transpiration processes) but a large amount of input information is required to
execute them. Uncertainties in some assumptions make mechanistic model out-
comes less certain and often make them less useful to those outside of the model
development group (Basso et al. 2012a). Mechanistic models are rarely adopted to
solve problems; rather, they are often used for academic purposes to gain a better
understanding of specific processes and interactions.
Functional models are based on empirical functions that approximate complex
processes, such as a crop's interception of energy using plant leaf area (as an indica-
tor of biomass) and radiation use efficiency (as a measure of biomass produced per
unit of radiation intercepted). This type of function is relatively simple and usually
produces reasonable results when compared to field measurements, although it has
uncertainties related to the fraction of biomass partitioned to roots and nonlinear
photosynthetic responses to light. Another example is the simulation of potential
evapotranspiration using the well-known functional Penman or Priestley-Taylor
equations, which have been used successfully for decades although they are highly
simplified compared to mechanistic evapotranspiration models.
Functional crop models use simplified equations and logic to partition simulated
biomass into various plant organs, which are integrated to estimate total biomass
and yield. Functional models primarily use “capacity” concepts to describe the
amount of water available to plants as compared to using “instantaneous rate” con-
cepts from soil physics. The difference between the upper and lower limits of soil
water-holding capacity determines the amount of water available to plants.
Functional models typically use daily time step inputs for weather and man-
agement variables such as precipitation, solar radiation, temperature, irrigation,
and fertilizer use. Low data input requirements make these models attractive when
detailed data on biophysical processes are lacking. These models, when properly
