Two distinct approaches have generally been used to situate interdisciplinary studies of agricultural systems (Shennan et al., 1991; Drinkwater, 2002). One approach uses simulated agricultural systems, which are simulated, replicated experiments designed to answer specific questions. Simulated agricultural systems are often set up at agricultural research stations but can also be located on farms. The second approach uses in-place existing agricultural systems, systems that are already in operation for production purposes. These are more commonly used to make comparisons across specific landscapes or to understand actual systems in-situ. Existing agricultural systems can range from working farms (Drinkwater et al., 1995, Needelman et al., 1999) to larger agricultural landscapes (Auclair, 1976) and watersheds (David et al., 2009). As with other experimental decisions, choosing which type of design to use will depend on the project goals and hypotheses. Each approach has strengths and limitations.
Simulated agricultural systems consist of blocked and replicated treatments, generally at an agricultural research station, and are design to answer specific systems-level questions. Existing agricultural systems consist of operations that are already in place and allow for the study of real-world systems, but with varying degrees of experimental control.
In some cases, the appropriate research site(s) may be obvious because the questions or goals specify the design. For example, if researchers were studying the fate of legume-derived nitrogen or carbon after incorporation of green manures, they would set up a simulated, replicated system because the research would involve stable isotope tracers, which would be difficult to use on a working farm. Research on the impacts of farm-scale vegetation and crop rotation on insect pests and natural enemies would need to be conducted on existing farms due to the scale of the processes being studied.
Other questions, such as those exploring the relationship between plant species diversity and cover crop performance, could be addressed using a range of experimental designs, depending on the hypotheses to be tested. For example, research examining the effects of cover crop species composition on biomass production, weed suppression or N2O emissions could be conducted in simulated plots at a research station. However, research evaluating cover crop performance using farmer-developed criteria such as compatibility with rotations, ease of management, and reliability would be conducted on existing farms. Frequently, a hybrid approach that includes field stations as well as working farms or landscapes may be the most effective design, because the advantages and limitations of different types of research venues are often complementary.
Historically, the simulated agricultural system model has been more widely used. Table 3.1 summarizes some of the largest and most recent replicated, on-station systems experiments in the United States comparing various production systems.
Simulated agricultural systems offer a number of advantages over existing agricultural systems. For one, researchers can compare management systems while reducing variability in soil type, management history, farmer skill, surrounding habitat or microclimate. Second, promising innovative cropping systems that are not currently in use by farmers can be studied. Finally, simulated systems allow for investigations over time of changes that occur after implementation of new management schemes, since all systems begin with well-defined time-zero conditions.
However, these strengths are tied to certain limitations. Since data are obtained from a single location, the effect of varying environmental conditions (e.g., soil texture or landscape- level biodiversity) on agricultural systems cannot be studied. Also, research is sometimes limited to the study of a few sets of practices to represent a given type of production system (Liebhardt et al., 1989; Temple et al., 1994), whereas in reality, farmers can choose from many variations on a central theme. It can also be challenging to achieve optimal management of all systems being compared, particularly when innovative, farmer-developed management regimes are compared to more common management systems. Even with detailed advice from farmers, farm crews may not have the skills or equipment to effectively simulate farmer practice. Furthermore, agricultural systems developed at research stations cannot adequately address landscape-scale ecosystem processes or certain socioeconomic questions. Finally, while simulated agricultural system experiments are well suited for examining early effects of management on soils, it is financially and logistically difficult to maintain these experiments over decades; thus, obtaining long-term data can be difficult.
The use of existing systems such as farms, watersheds or agricultural landscapes as study sites has been less common than the use of simulated systems, perhaps because the limitations are often thought to exceed the advantages. There is a continuing misperception among agronomists that hypotheses cannot be tested using existing sites since many factors cannot be controlled across farms. On-farm studies can be more logistically difficult due to lack of control over the research sites (for example, farmers sometimes change their plans and may forget to contact researchers). Furthermore, existing farms are usually less accessible and more dispersed, and therefore more costly to study, compared to sites at an experiment station. However, uncontrolled variation and some of the other challenges can be dealt with through appropriate research design and site selection.
The most important advantage of using existing systems as research sites is that the systems are realistic in terms of scale, management practices and farmer constraints and so provide a unique opportunity to study agricultural processes under “real world” conditions. Another advantage is that site selection can be used to either minimize or increase environmental variability to test hypotheses and achieve results that apply across a wider range of environments and conditions. For example, Needelman et al. (1999) selected farms with soil textural differences to investigate interactions between management practices and soil texture. Furthermore, transition effects can be avoided by selecting well-established farms with differing management regimes, which are probably closer to steady-state conditions.
Transition Effect: An interim time period in which an agricultural system undergoes a shift from one management system to another and experiences production losses due to this shift. For example, when farmers switch from less diverse, chemical-based farming to a more biologically diverse approach with reduced inputs, they usually experience reduced yields and increased pest pressure during the first few years.
Alternatively, research that addresses changes through time can test hypotheses by using sites that have been managed for varying durations. Finally, some agricultural system properties that are influenced by landscape-scale characteristics require the use of farm-scale study sites (Letourneau et al., 1996; Elias et al., 1998), as do studies of socioeconomic processes.