Role and Applications of Crop Simulation Models

Crop simulation models use quantitative descriptions of ecophysiological processes to predict plant growth and development as influenced by environmental conditions and crop management that are specified for the model as input data (Table 13.2). Many models developed by a single researcher or laboratory are used for a single purpose and have a short life. Others evolve over time and are similar to modern software packages. Among the longer lived models that have seen widespread use in climate change research are apsim (Keating et al., 2003), the Cropping Systems Model (csm) series (Jones et al., 2003; Hoogenboom et al., 2004), CropSyst (Stockle et al., 2003) and epic (Meinardus et al., 1998).

The simplest models estimate daily growth through conversion factors for intercepted solar radiation to biomass, whereas complex models may simulate growth at a timescale of minutes and include routines to simulate key biochemical pathways of photosynthesis. Hay and Porter (2006) provide a general review of the physiological processes described in models, and Tsuji et al. (1998) describe multiple aspects of models, including soil and weather processes and example applications.

A typical model simulates assimilate production by estimating gross photosynthesis and then reducing the assimilate pool through respiration and senescence. The resulting net pool is then allocated to

Table 13.2. Examples of data inputs required for a typical crop model that runs with daily time steps.



Daily weather Maximum and minimum air temperatures Solar radiation


Dewpoint or vapour pressure deficit

Wind speed Soil properties Albedo

Runoff characteristics Infiltration characteristics

Initial water and nutrient levels

Crop management Sowing rate Row spacing Fertilization

Affect almost all plant and atmospheric processes and are also used to estimate soil temperatures Key for establishing potentials for photosynthesis and evapotranspiration. Data are often either unavailable or inaccurate Affects moisture levels in the soil profile and runoff Affects potential evapotranspiration. Average relative humidity is often reported but is a poor indicator of evaporative demand because of confounding with temperature Affects potential evapotranspiration

Reflectivity of soil to solar radiation. Affects soil temperature and evaporation

Used to estimate what fraction of precipitation is lost to runoff Used to estimate how moisture enters the profile, is distributed through soil layers, or drains out of the profile Establishes soil conditions for germination and subsequent growth. Preferably determined by soil horizons to the maximum depth of root development

Used to estimate initial stand of plants

Used to estimate light interception by crop canopy

Type, amount and date of application for any fertilizers different plant organs through partitioning rules. The rules assign priority to rapidly growing tissues such as leaves, with onset of reproductive growth representing a key developmental switch. Priorities also shift in order to satisfy the crop demand for water and nutrients. If supplies are limiting, more assimilate is allocated to root growth in order to increase extraction from the soil. Thus, under water or nutrient deficits, root growth may be favoured over leaf, stem or reproductive growth. Furthermore, nutrients may be mobilized from inactive tissues (e.g. older leaves) to organs with high demand.

The timing of key developmental stages such as seedling emergence, end of main stem leaf appearance, anthesis and physiological maturity are simulated using procedures that are analogous to the accumulation of growing degree days (heat units). As required for a given crop, however, the procedures may consider vernalization and photoperiod responses. Models for simulating root and tuber, forage and bioenergy crops are similar to those for seed and grain crops, but allocate assimilates to vegetative storage organs (Singh et al., 1998).

Water and nutrient budgets are usually modelled both for the plants and for the soil, requiring descriptions of root growth through the soil as well as the soil and atmospheric processes that affect water and nutrient dynamics.

Temperature responses

The main effects of temperature are modelled on assimilate production, phenology, soil processes and evapotranspiration. Relatively few models explicitly consider high temperature stresses causing abortion of reproductive structures or irreversible damage to vegetative organs. For models that estimate daily growth through a radiation use efficiency (RUE) approach, the potential RUE is adjusted by a simple temperature function. In the version of the ceres models implemented in the csm series, these temperature functions weight the daily maximum three times more than the minimum, on the assumption that daytime temperatures influence growth more than night-time temperatures. More complex models such as those using the Farquhar model may involve multiple temperature responses that are evaluated at scales of minutes, and the parameters are determined by measuring component physiological processes.

The occurrence of stages such as flowering and maturity is hastened by temperature, but interactions with vernalization (a requirement for cold temperatures prior to flowering) and day length can override the basic effect of temperature on development.

Physical and chemical processes affecting water and nutrient availability also respond to temperature. The net result is that the basic temperature responses described by models are more complex than one might expect.

Response to CO2

In RUE-based models, the main effect of CO2 is through a factor that scales RUE downwards or upwards, a key distinction being whether the crop has a C3 or C4 photosyn-thetic pathway. More complex models combine descriptions of diffusion of CO2 into the leaves and of the biochemical processes of photosynthesis.

Plants also respond to elevated CO2 by reducing stomatal conductance, so most models also include an effect adjusting leaf or canopy conductance or transpiration per se (e.g. Tubiello and Ewert, 2002). In models that simulate a complete energy balance, reducing transpiration increases canopy temperature. Thus, an indirect effect of elevated CO2 is to warm the plants, which should further affect photosynthesis, respiration and development.

Differences among species and cultivars

Qualitatively, the most important physiological processes have proven to be similar across crop species. Furthermore, soil and atmosphere processes are largely species independent. Thus, differences among species are simulated mainly through changes in parameters rather than through fundamental differences in physiology. Exceptions include differences between C3 and C4 photosynthetic mechanisms, the nature of vernalization or photoperiod responses and how these affect phenology, and the ability of legumes to fix atmospheric N. Morphological constraints are also important, especially with regard to growth of seeds, storage roots or other economically important organs. Key parameters that distinguish among species include response curves for temperature and CO2, critical and maximal levels of nutrients, factors for sensitivity to water or nutrient deficits, and parameters for potential growth of leaves, stems, roots and seeds or fruits.

Parameters for differences among culti-vars can involve phenology, partitioning coefficients and reference organ sizes (e.g. maximal area of an individual leaf or mass of a seed). Phenology requires consideration of the relative duration of different phases, and responses to vernalization (if present) and photoperiod. Values of the parameters are usually determined through iterative parameter adjustment and comparison with observed data from field trials (e.g. Piper et al., 1996). This calibration process is problematic because it requires that detailed sets of accurate observations be available. The error inherent in data from field studies makes it difficult to discern whether differences between observed and simulated data are due to incorrect parameter values or to errors in the model per se. Various groups are exploring how to use information from genetics or genomics to parameterize cultivars more reliably (e.g. White and Hoogenboom, 1996; Yin et al., 2000; Messina et al., 2006).

Crop management

To simulate the growth of a crop, the model must know how the crop is to be grown, whether for a real world or hypothetical situation. Management information includes the date and manner of planting, the cultivar used, fertilization and irrigation practices, and for some crops, harvest practices (Table 13.2). Tillage and residue management may also be considered. The informa tion either establishes the initial conditions for the simulation or modifies aspects of the environment, such as through addition of N or water to the soil profile.

Basic application of crop models in climate change research

Assuming an appropriate model is at hand and a reference crop production scenario exists, simulating the effects of climate change mainly involves running the model for the weather and CO2 scenarios of interest. For a single site or region, the scenarios may be specified as fixed (e.g. an increase in daily mean temperature of 2°C) or relative (a 20% decrease in daily precipitation). These adjustments may be held constant over the crop cycle or varied. The choice depends on the objectives and the source of the climate change scenario. Because a season might be unrepresentative of long-term trends, simulations are usually run for 20 or more years. The requisite weather data may come from historical records or from weather generator software that reproduces the statistical properties of historic conditions (e.g. Mavromatis and Jones, 1998; Jones and Thornton, 2003).

A single set of runs can be compared to equivalent runs using unadjusted weather, thus providing one estimate of the potential impact of climate change on economic yield or a diverse range of other traits. None the less, such a comparison ignores the potential that producers will adapt their practices to the changing environment. We examine two hypothetical cases, one for soybean and planting dates and one for maize and N fertilizer response, to illustrate a few of the issues that may be relevant. Both studies assume an increase in CO2 from 380 ppm (the approximate level in 2005) to 580 ppm.

Soybean planting date

Crop response to planting date is readily modelled to examine how warming might affect the potential growing season. For temperate climates, logical expectations are that warming would allow earlier or later plantings, while elevated CO2 should increase growth and yield. However, warming accelerates development and causes earlier flowering and maturity, which would reduce growth, and at the higher temperatures in summer months, growth might decline further due to a decrease in photosynthesis and increase in respiration.

For Gainesville, Florida (latitude 29°38'N; elevation 10 m), the csm-cropgro-Soybean model predicts that very early plantings result in delayed flowering due to low temperatures, and, as expected, warming reduces the delay (Fig. 13.1a). By April, however, longer day lengths begin to slow development for both treatments. With an early May planting, the warming regime is predicted to slow flowering slightly due to supra-optimal temperatures. Note that the model assumes no effect of CO2 on phenology.

The yield responses suggest that the beneficial effects of elevated CO2 roughly balance the detrimental effects of temperature up to early May, but subsequently,


Planting (day of year)

— Historical temperature conditions — 1.5/3.0°C

Planting (day of year)

— Historical temperature conditions — 1.5/3.0°C

1 Oct

ld iel yi

1 1000

Planting (day of year)

380 ppm/Historical temperatures 380 ppm/1.5/3.0°C — 580 ppm/Historical temperatures — 580 ppm/1.5/3.0°C

Fig. 13.1. Simulated response of soybean to planting date (February-September) at Gainesville, Florida. (a) Flowering response under +1.5°C daily maximum, +3.0°C daily minimum air temperature versus historical temperature conditions from 1988 to 2001. (b) Seed yield response under +380 ppm CO2 or +580 ppm CO2 and +1.5°C daily maximum, +3.0°C daily minimum air temperature versus +380 ppm CO2 or +580 ppm Co2 with historical temperatures from 1988 to 2001. Simulations are from csm-cropgro-Soybean for cultivar 'Bragg' with irrigations applied as needed to avoid water deficit.

elevated CO2 provides a small but consistent benefit equivalent to 5-10% of the yield expected for historical conditions (Fig. 13.1b). For plantings around 1 April, additional yield benefit might be obtained by substituting a later-flowering cultivar.

Maize response to warming, elevated CO2 and N

Maize crop growth was simulated for 25-year periods at Palmira, Colombia, an equatorial location (latitude 3°29'N; elevation 965 m) with a mean annual temperature of 25°C. A September planting date was used, corresponding to the onset of the rainy season. The crop was assumed to be rainfed, fertilized at 50, 100 or 200 kg N/ha, and otherwise well managed.

Seed yield declines with increasing temperature for the 200 kg/ha N at ambient (380 ppm) CO2 (Fig. 13.2a) and elevated CO2

(Fig. 13.2b), but not at the other two N levels. Warmer temperatures promote early flowering (Fig. 13.2c), so a portion of the temperature effect on yield relates to the shorter growth duration (Fig. 13.2d). One interpretation of the response to N is that at lower N levels, yield is limited by N and not assimilate production. Alternatively, assumptions about how to model interacting temperature and N stresses in the csm-ceres-Maize model may merit review.

Coupling GIS to crop models

GIS and simulation models complement each other for data management, analysis and presentation. Simulation models have traditionally been used on a site-specific basis, but the coupling to GIS is appealing because it permits the possibility for simultaneous investigation of spatial and temporal





12 3 4 Temperature increase (°G) N (kg/ha): -5o —- 1oo»« 2oo

12 3 4 Temperature increase (°G) N (kg/ha): -5o —- 1oo»« 2oo aOOO



12 3 4 Temperature increase (°G) N (kg/ha): -5o-— 1oo.-.-. 2oo

12 3 4 Temperature increase (°G) N (kg/ha): -5o-— 1oo.-.-. 2oo ee e2

Temperature increase (°G)

5a e2 ee

Flowering (days after planting) N (kg/ha):-5o—• 1oo«-. 2oo

Fig. 13.2. Simulated response of maize to 50, 100 or 200 kg/ha N fertilization at Palmira, Colombia under temperature increases over historical conditions of +0.5°C to +5.5°C from 1978 to 1997. (a) Seed yield response under +380 ppm CO2. (b) Seed yield response under 580 ppm CO2. (c) Days to flowering (response same for all N and CO2 levels). (d) Seed yield response versus days to flowering under +380 ppm CO2. Simulations are from csm-ceres-Maize for cultivar 'Cargill 111S' assuming rainfed conditions.

Temperature increase (°G)

5a e2 ee

Flowering (days after planting) N (kg/ha):-5o—• 1oo«-. 2oo

Fig. 13.2. Simulated response of maize to 50, 100 or 200 kg/ha N fertilization at Palmira, Colombia under temperature increases over historical conditions of +0.5°C to +5.5°C from 1978 to 1997. (a) Seed yield response under +380 ppm CO2. (b) Seed yield response under 580 ppm CO2. (c) Days to flowering (response same for all N and CO2 levels). (d) Seed yield response versus days to flowering under +380 ppm CO2. Simulations are from csm-ceres-Maize for cultivar 'Cargill 111S' assuming rainfed conditions.








ee e2

phenomena. Visualization of model summary outputs, for example yield response, via a GIS also adds an extra dimension. As a result, there has been a rapid growth in the number of applications interfacing GIS and simulation models since the late 1980s (Hartkamp et al., 1999). Multiple examples now exist of crop models, typified by the Decision Support System for Agrotechnology Transfer (dssat) family, and linked to GIS at a range of spatial scales from field to region (see summary table in Hartkamp et al., 1999). Simulations run over large geographical regions extend the model outputs to areas that have not been validated, so serve more as a sensitivity analysis for the model rather than a precise calculation. However, such assessments do permit the possibility for the evaluation of multiple scenarios in relative terms within a spatial framework. The HarvestChoice project (HarvestChoice, 2009a) is taking such an approach, attempting to simulate yield potential of major crops on a continent-wide basis under a range of differing technological scenarios (see HarvestChoice, 2009b). Availability of highly disaggregated data sets, both spatial and temporal, is fundamental to this approach, and although progress is being made, several challenges still remain.

Was this article helpful?

0 0

Post a comment