2.2 Mechanistic Models
2.2.1 Non-Architectural Models
Mechanistic crop models are generally driven by photosynthesis. The rate of photosynthesis depends on the amount of light intercepted and the efficiency of this light to be absorbed and converted to photosynthate. Leaf area is therefore an important consideration as it directly af- fects the amount of light absorbed and as such photosynthesis and growth. The proportion of photosynthate distributed to certain organs such a leaves, stems, roots and grain is dependent on phenological stage. Therefore prediction of phenological stage is also of importance. Pro- gression of phenological stage of winter wheat is dependent on vernalisation, temperature and photoperiod, which also therefore have to be considered within the model. Potential growth is predicted usually over a time step of one day with limiting factors, such as temperature, nitrogen and water availability etc causing the expected response within the processes of crop growth.
Model Structure
In general mechanistic crop models have two components in which the important processes, mentioned, are described. These are plant and soil. Both of these components have further sub modules each of which deal with specific mechanisms. The sub modules of the plant module con- sider phenology (developmental stages), organ growth, and yield formation and the sub modules of the soil component consider root growth, water balance, nitrogen balance and soil transfers.
Most mechanistic crop models require the same inputs which include genetic information about the cultivar reaction to certain conditions. Management such as sowing depth and density and environmental factors such as temperature and solar radiation. The outputs are usually yield quality and quantity. The output is the quantification of above ground biomass usually in terms of quality and quantity of yield. There are various mechanistic crop models that simulate winter wheat growth, such as AFRCWheat (Weir et al. 1984, Porter 1993), CERES-Wheat (Ritchie and Otter 1984), SIRIUS (Jamieson et al. 1998b), SUCROS (Spitters et al. 1989) and STICS (Brisson et al. 2003a). They have each been built for specific purposes and therefore differ in their calculation of various plant growth and development processes, however the structure of these models is similar to the one described (Brisson et al. 2003b).
SIRIUS which is the simplest of AFRCWHEAT, CERES-wheat and SUCROS calculates grain yield and quality and nitrogen leaching and water and nitrogen uptake and assumes the canopy is a single entity, producing biomass as a product of light and RUE (radiation use efficiency). No calculation of yield components is included. By dealing with leaf layers it avoids the need to consider tillers and reduces the parameters required for calibration. Biomass accumulation is calculated from intercepted PAR (photosynthetic active radiation) and grain growth from simple partitioning rules. LAI is developed from a simple thermal time sub model. STICS was primarily designed to investigate agronomic and environmental impacts such as leaching at regional scale and is similar to SIRIUS in that it does not separate simulated ground biomass into organs and biomass accumulation is a product of intercepted light and RUE. SUCROS (Simple and Univer- sal Crop growth Simulator) simulates growth (rate of dry matter accumulation) based on CO2
assimilation (photosynthesis) of the canopy which is a function of incoming radiation and light. The rate of dry matter accumulation is a function of irradiation, temperature, crop characteristics and water supply. After subtraction of maintenance respiration, growth of leaf stem, root and storage organs are simulated. Biomass partitioning depends on crop development stage, which is computed as a function of temperature only. Different crops can be simulated by altering spe- cific input parameters. Influence on respiration can also be included by alerting environmental conditions such as temperature.
CERES-wheat and AFRCWHEAT consider the separation of biomass accumulation into separate organs within the canopy. They both simulate the process of crop growth and development by including the timing of phenological events during the life cycle of the crop and development of canopy and the interception of PAR and its use to fix carbon which is then converted to dry
matter. They both include an algorithm to reduce potential production via strategies of water and nitrogen (effects of other nutrients such as potassium and phosphorus and effects of weeds and pests are not considered although can be ‘added’ to the model ). Both models assume a linear relationship between rate of crop development and temperature. AFRCWHEAT includes the partitioning of photosynthesis, growth of leaf and stems, senescence biomass accumulation and root system dynamics and uses temperature to regulate growth. This model has been used to investigate effects of climate change at national scale and uses GIS technology. CERES- wheat has been applied at regional scale to estimate yield and forecasting and analysis of policy questions related to crop production and resource conservation and at the farm level for decision making and for multi year analysis for risk assessment. Its primary purpose was to predict alternative management strategies and tactics that affect yield at intermediate steps. It does this by simulation crop yield and focus on 3 import stages of growth, duration, rate and extent and the stress influence on such process( in terms of water and Nitrogen). These are beyond its initial goal which was to predict leaf number and sizes and quantify genetic and climate interactions.
Harnos and Kovacs (1999) compared CERES-Wheat, AFRCWHEAT2, CROPSIM and SU- CROS2 in order to select an appropriate model for climate change studies and found that al- though CERES-Wheat and AFRCWHEAT2 fitted the best with the historical data, used within the study, that all models showed different sensitivity to environmental parameters, creating dif- ferent simulated yields for the climate scenarios. From these results no decision was made on the most applicable model for this purpose and instead the inaccuracies associated with using these models, for such an application, were instead just highlighted. More worryingly when Jamieson
et al. (1998a) compared CERES-wheat, AFRCWHEAT2 and SIRIUS using observed UK grain
yields from well managed agricultural experiments, none of the models were found to accurately predict yield and substantial disagreement was found between the models’ predictions of both yield and yield loss due to water limitation. This disagreement between the models predictions was concluded to highlight the differences in the underlying hypothesis in the models. These comparisons highlight the fact that some models simulate different aspects of plant growth and development to differing degrees of accuracy and that the models although mechanistic should be used out of their ‘experimental scope’ with caution. In this study the ADEL-wheat model is used with the main aim to parameterise the model structure for predicting the EO signal, rather than predicting yield directly, and so these weaknesses common with such models are not so im- portant in the first instance. However, the improvement of the response of crop-growth models to environmental drivers is clearly an active area of research.
In general the models mentioned are constructed using mechanistic models at the level of organ growth, however as discussed when first introducing mechanistic models, empirical relationships are used, to describe certain relationships. For example the PAR extinction canopy coefficient in AFRCWHEAT2 is set as 0.44 and in CERES to be 0.85. This coefficient affects the rate of dry matter accumulation and as such the allocation of daily assimilates to leaves. The use of inaccurate coefficients within a relation affecting canopy development such as this, may possibly, lead to errors on the estimation of biomass production (Porter 1993).
It has been suggested that main parameters driving crop growth could be replaced or updated by estimations derived from remote sensing within the growing season (Moulin et al. 1998). The methods and models used for this coupling is discussed in section 2.3.
A recent advance in mechanistic crop models is to consider the architecture of the canopy result- ing in architectural which consider crop growth using empirical and mechanistic models. Such models are refereed to as Functional-Structural models.