Downsizing an Integrated Crop Management Field Study Affects Economic and Biological Results

and correctly conducted, long-term field studies identify profitable cropping systems; assess conservation tillage In recent years, there has been increased interest in long-term, practices; obtain information on direct, residual, and field-scale cropping systems research to improve pest management, cumulative treatment effects; identify strategies for eroto protect air and soil quality, and to increase or maintain growers’ profits. However, these studies require large tracts of land, sizeable sion control and pest management; and develop regulalabor forces, and substantial inventories of equipment, which make tions for reduced pesticide use (Cady, 1991; Young et them very expensive to conduct. Because of recent concerns about al., 1994a, 1994c; Alldredge and Young, 1995). These reducing field research costs, this study compares economic and biostudies generally involve several agencies and must be logical results from an original complete 6-yr integrated cropping interdisciplinary or multidisciplinary to identify impormanagement (ICM) systems field study to results from several downtant interacting agronomic, economic, and biological resized experiments, which were components of the complete study. lationships (Martin et al., 1991; Schweizer et al., 1988; Compared with the original ICM study, the downsized experiments Young et al., 1994c). Integrated crop management studreduced the number of treatment replications from four to three, ies are rarely attempted because they are expensive to reduced the number of crop rotation cycles from two to one (from 6 conduct; require large areas of land; and utilize considerto 3 yr), or only grew one crop per rotation each year. The effect of able labor to plant, harvest, and maintain plots, and to downsizing on the profitability analysis and the statistical (biological) analysis were similar. Reducing replications altered both profitability collect data (Cady, 1991; Young et al., 1994b). Reducing and biological conclusions less than reducing the number of rotation the size of long-term field studies is desirable if downsizcycles. Reducing crop rotation cycles markedly altered treatment profing does not sacrifice biological and economic informaitability rankings compared with the complete study. Growing only tion. There are numerous ways in which researchers one crop in a rotation per year was the most detrimental to biological can downsize long-term, large-scale ICM studies. These results and entirely precluded computing mean annual cropping sysinclude reducing treatments, data collected, plot size, tem profitability. This empirical study supports the importance of disciplines involved, as well as decreasing replications, replicating treatments fully over time, over space, and over crop rotaduration of the study, and land area required based on tional positions. number of crops grown each year within a rotation. For example, the original proposal for the Pacific Northwest (PNW) ICM study called for 432 subplots rather than L large-scale field studies that address the final 144. The proposal was reduced by eliminating integrated pest management (IPM) and/or intea tillage regime and fertility rates (Young et al., 1994b). grated crop management (ICM) agro-ecosystems are Statistical theory guides researchers in determining important for several reasons. When properly designed sample sizes for testing hypotheses or estimating parameters (Neter et al., 1996). Most sample size information W. Wei, presently American Express Travel Related Services, 7740 is available for simple experimental designs, but may N. 16th St., Phoenix, AZ 85020; and J.R. Alldredge, D.L. Young, and F.L. Young, USDA-ARS, Washington State Univ., Pullman, WA Abbreviations: WML, weed management level; ww, winter wheat; sw, 99164. Senior authorship is shared equally; the first author conducted spring wheat; sb, spring barley; sp, spring pea; MP, main plot; rep., the initial statistical analysis. Received 8 Nov. 1999. *Corresponding replication; PNW, Pacific Northwest; EMS, error mean square; LSD, author ([email protected]). least significant difference; ICM, integrated crop management; IPM, integrated pest management. Published in Agron. J. 93:412–417 (2001). WEI ET AL.: DOWNSIZING AN INTEGRATED CROP MANAGEMENT FIELD STUDY 413 of that rotation. Farmers typically allocate land equally to not be applicable to complex interdisciplinary long-term rotated crops each year to smooth seasonal labor and machinfield studies. The purpose of this research is to demonery demands and to capture the risk reducing effect of a diversistrate, using actual data, what biological and economic fied crop mixture. All main plots and all subplots within main information would be lost if the experiment is reduced plots were randomly assigned in the first year. For the subsein size, duration, and design. A long-term ICM field quent 5 yr, crop rotation, tillage system, and WML remained study in southeastern Washington State was used for the same as in the first year. One complete cycle of the design this purpose. Specifically, we will examine the consetook 3 yr. The full 6-yr study allowed each rotation to be quences of reducing the number of cycles in which every evaluated through two complete rotation cycles (Table 1). crop is present in the rotation every year from two Downsized Experiments cycles (6 yr) to one cycle (3 yr), reducing the number of replications (blocks) from four to three, and reducing For the purpose of this manuscript, the ICM study that was the number of crops within a rotation that are grown conducted for 6 yr with 12 main plots (every crop grown every every year. Results obtained from actual data are comyear) and 36 subplots in each of four replications (blocks) is pared with those predicted by statistical theory using considered the complete, long-term study. Experiments were then downsized by reducing the years, or reducing the number several measures of information loss. Economic preferof crops grown each year, or reducing the number of replicaence rankings based on profitability also are compared tions. The first and second 3-yr cycles (Years 1,2,3 and Years for the complete and downsized revisions. 4,5,6) represent one downsized data set formed by reducing the duration of the experiment. A second data set was formed MATERIALS AND METHODS by reducing the number of crops and was done by deleting some main plots (MP) yet maintaining the three-crop rotation Integrated Crop Management Description sequence over the full 6 yr. This resulted in three downsized The original 32-ha study site was located 5 km northwest sets of data comprised of MP series 1-4-7-10, MP series 2-5of Pullman, WA. The long-term annual precipitation for this 8-11, and MP series 3-6-9-12 (Table 1). The last data set was region averages 540 mm yr2. The average over the 6-yr study formed from four downsized sets of data from all possible period (1985–1986 to 1990–1991) was 495 mm. In general, the combinations of three of the four replications (i.e., Rep. 1,2,3; last 3 yr of the trial were wetter than the first 3 yr and averaged Rep. 1,3,4; Rep. 2,3,4; and Rep. 1,2,4). Each of the downsized 530 and 460 mm, respectively. Detailed field procedures, bioexperiments contains complete information on a cropping syslogical data analysis (Young et al., 1994c; Boerboom et al., tem, but has fewer observations than the complete experiment. 1993), and economic assessment (Kwon et al., 1995; Young Additional downsized experiments are possible, but those preet al., 1994a) have been well documented. sented will demonstrate the effect of downsizing. The ICM study was designed as a randomized complete block using a split-plot arrangement with four replications; Statistical Model each replication consisted of 12 main plots or cropping sysThe linear model within each year for this split plot randomtems. The main plots were subdivided into three economically ized complete block design (Neter et al., 1996) is: distinct weed management levels (WML). Main plots represented particular combinations of crop rotations and tillage Yijk 5 m 1 ri 1 tj 1 dij 1 bk 1 (tb)jk 1 εijk [1] systems. The two crop rotations were a 3-yr rotation and a continuous wheat (Triticum aestivum L.) rotation (Table 1). for i 5 1,2,3,4, j 5 1,2, . . .,12, k 5 1,2,3, where m is the overall The two tillage levels were a conventional tillage system and mean for yield or net return, di is the effect due to the ith a conservation tillage system (a combination of reduced-till replication (block), tj is the effect of the jth level of MP (rotaand no-till systems depending on the previous crop). A croption with crops and tillage systems), dij is the error component ping system was a combination of crop rotation [i.e., 1: winter for the MP, bk is the effect of the kth level of subplot WML, wheat (ww)–winter wheat (ww)–spring wheat (sw); 2: winter (tb)jk is the interaction effect of the MP and the subplots, and wheat (ww)–spring barley (sb)–spring pea (sp)], tillage system εijk is the error component for the subplot. Note that other (1: conservation; 2: conventional), and weed management models may be considered to represent the data. For instance, level (1: minimum; 2: moderate; 3: maximum). For computing one might decide that various block 3 treatment interactions per hectare profit, a cropping system in each year of the study occur, so these terms could be removed from the subplot error and explicitly included in the model. We have no reason to was assumed to have one-third of the land area in each crop Table 1. Crop rotation and tillage system for the 6-yr ICM project at Pullman, WA.