In this chapter, two types of model uses related to SOC sequestration were demonstrated. In the first one, a model was used to perform computer experiments to evaluate SOC sequestration potential as affected by different management practices in a particular location in Mali. Although data were limited for this region, the example demonstrated the value of using available data to adapt a model before using it to evaluate alternative management practices in a region. When the computer experiment was conducted using the same DSSAT maize model, but without using local data to adapt the model, results were clearly inconsistent with known yield and runoff responses of RT vs. CT and to realistic changes in SOC. Obviously more data and more work are needed to improve predictions and confidence in simulated results. Nevertheless, the preliminary estimates of SOC sequestration obtained in this study are certainly reasonable. They fall within the range of values reported elsewhere (see West and Post, 2002; Pichot et al., 1981; Lal, 2000), and simulations of runoff and yield of maize in RT vs. CT are consistent with local data and knowledge.
The second example represents a new use of biophysical models. The motivation for this use was that reliable estimates of SOC sequestration will be needed if landowners enter into contracts in which they are paid to sequester an agreed upon amount of carbon. In this case, a model was used to integrate measurements over time to improve estimates of SOC sequestration using an ensemble Kalman filter. This procedure also improved model predictions over time (errors were reduced) as new measurements were used, and a model parameter was more accurately estimated for the particular field. Through the recursive combination of model predictions and new observations, the model was better adapted to predict SOC levels at the specific site.
The second example showed the need for knowing errors associated with model predictions, and it demonstrated how to include uncertainties (stochastic features) in models for this purpose. The demonstration of this data assimilation technique was limited to a single field and to the use of a simple model. However, this approach is amenable to the use of more complex models, including the DSSAT-CENTURY model used in the first example. Koo et al. (2003) and Bostick et al. (2003) showed that one can use the EnKF to assimilate both crop biomass and SOC measurements to improve estimates of SOC using more complex models. This approach lends itself to the use of remote sensing data to improve SOC estimates by assimilating biomass data over space and time. Although the case study presented was for a single field, the EnKF can be expanded to estimate SOC over large areas that would be required in a carbon contract. This capability is currently being developed, following similar developments in the field of hydrology (Graham, 2002). This approach appears to have potential for other practical applications as well.
Crop and soil models can be very useful tools in SOC sequestration studies. But, they can also be misused. The phrase "all models are wrong; some are useful" is important to keep in mind. The main theme of this paper was the importance of using local data, however scarce, with such models to help better understand and predict SOC sequestration responses to soil, climate, and management. The procedure used to integrate local data with models was referred to as model adaptation. One aspect of that adaptation process is the estimation of soil, crop, and management parameters that allow the model to predict important variables for application to the problem being addressed. This process is sometimes referred to as model calibration, which may invoke criticisms from those whose aim is to have models that do not require modifications in order to predict system performance. Certainly, more work is needed to improve model capabilities; this will always be true. However, the diversity and spatial variability of land management, soils, climate, and genetic composition of crops will always create challenges to those who need to tailor agricultural management systems to achieve goals of individual farmers and of society. In this chapter, it has been shown that existing models can be used effectively to better understand and predict SOC sequestration. The importance of both data and models needs to be recognized as efforts are made to improve knowledge and tools for use in science and policymaking.
Was this article helpful?