Effects of Weed Communities in Conventional and Organic Agricultural Systems.

Final Report for GW06-026

Project Type: Graduate Student
Funds awarded in 2006: $7,536.00
Projected End Date: 12/31/2007
Grant Recipient: Montana State University
Region: Western
State: Montana
Graduate Student:
Principal Investigator:
Principal Investigator:
Fabian Menalled
Dept. of Land Resources and Environmental Sciences
Expand All

Project Information

Summary:

The negative effects of agriculture have become a growing concern. The goals of this study were 1) to increase our understanding of the importance of the distribution and diversity of weed communities in conventional and organic spring wheat systems, 2) to determine if these differences have an impact on yield, and 3) determine if it is possible to employ weed community structure as a sustainable management alternative. We found that organic and conventional system weed communities differ in both diversity and spatial arrangement, and that there is reason to believe that these inherent differences may have effects on crop yield.

Introduction

Agriculture is one of the most integral pieces in a functional society. It is also the largest industry on the planet (Clay, 2004). The ability of agricultural systems to produce adequate food has, in general, grown in parallel with an expanding world population (Tilman et al., 2002; Trewavas, 2002). In recent years, however, the negative effects of agriculture, such as soil and water contamination and erosion, have become a growing concern (Tilman, 1999; van der Werf and Petit, 2002; Biao et al., 2003). In order to meet the needs of the future, how can we continue to increase agricultural production without increasing the negative impacts of intensive agriculture?
In the United States, small farms tend to be more productive, per unit of land, than large farms (Clay, 2004). Farming small areas is, however, problematic for the farmer in that it is difficult to generate enough income with small farming operations (Clay, 2004). Thus, anything that decreases farm cost to the producer will increase the likelihood that small farming operations can adequately support the farmer. This, in turn, could lead to an efficiency increase if it became economically feasible for farming operations in the US to decrease in size.
Weed management using herbicides or cultivation often represents a large cost to farmers. Integrated weed management systems that combine biological, cultural, mechanical, and chemical control methods, have been identified as an alternative form of management that might lead to more sustainable systems (Clements et al., 1994) since they decrease costs to both the farmer and the environment. However, there are still knowledge gaps relating to ecological processes that could lead to the development of practical integrated weed management programs in agricultural systems. For example, relatively little is known about the potential importance of weed community characteristics under different agricultural management systems. Few studies, if any, have investigated the implications of differences in weed communities on crop plants and weed control. Diversity has been proposed as a critical component of stability in natural systems, allowing for nutrient accumulation and increases in productivity over time (Tilman et al., 1997). Would this apply in agricultural systems as well? This study examined two main components of the weed communities of conventional and organic spring wheat systems (spatial patterns and diversity) and attempted to relate these to effects on spring wheat productivity.
Spatial Patterns of Weed Communities
Knowledge of the spatial distribution of weeds within a field is useful for determining where and when to carry out weed control (Thornton et al., 1990; Wiles et al., 1992) and in making accurate estimates of yield loss due to weeds (Brain and Cousens, 1990; Thornton et al., 1990). Up to this point, most research has focused on the spatial distribution of weeds within conventional input systems where weed communities tend to have patchy distributions (Dieleman et al., 1999; Faechner et al., 2002; Hughes, 1990; Mortensen et al., 1995; Rew and Cousens, 2001; Wiles et al., 1992) that are most likely due to differential mortality rates.
Despite this focus on conventional input cropping systems, there are still knowledge gaps relating to weed spatial patterns in conventional fields, especially in relation to species richness and density within weed patches. Furthermore, the effects of weed diversity on yield loss have not been adequately investigated. Cardina et al (1997) suggested that within a patch, weed intraspecific competition reduces the competition between the weeds and the crop. However, little is known about the relative importance of weed density and interspecific competition among weeds in relation to yield loss. It follows that if weed aggregation is going to be considered as a determinant of weed impact on crop yield, the characteristics (composition, density and diversity) of the aggregations themselves should also be considered.
Organic cropping systems, on the other hand, remain relatively unexamined with respect to weed spatial patterns. Casual field observations and anecdotal evidence suggest that weeds are not as patchy in organic systems as they are in conventional ones. Thus, it is reasonable to think that there might be spatial aspects of weed community structure within an organic cropping system that would influence crop-weed interactions differently than in conventional input systems. In addition, knowledge of spatial patterns in the weed communities of organic systems might allow for the implementation of site specific management practices.
Various methods have been used to map weed populations in farm fields. Discrete sampling is the most commonly used method, and involves collecting data from quadrats distributed on a grid (Rew et al., 2001). The data are then interpolated with techniques such as kriging to generate maps of weed distribution (Rew et al., 2001). Seedling counts from 0.25-m2 quadrats on a 10 m x 10 m grid have been shown to yield accurate maps of weed spatial distributions when kriging is applied (Heisel et al., 1996). However, determination of the proper grid size prior to sampling is essential for this type of technique, particularly when weeds are distributed in patches within a field, as larger grid sizes may lead to a decreased likelihood of recording any given patch of weeds (Rew and Cousens, 2001). In addition, grids are sometimes difficult to establish and the number of data points necessary to yield accurate maps is often very high (Rew and Cousens, 2001).
Continuous sampling is another option to assess weed spatial distribution (Rew and Cousens, 2001). With continuous sampling, the presence or absence of weeds is noted by an observer within the field while traveling in a continuous fashion. This data can then be analyzed using indices of aggregation to discern the general patterns of weed distribution. Quadrat variance techniques (Hill, 1973) can also be applied to continuous data, and can provide more insight into the spatial scale of patterns of vegetation. These types of analyses would be logical starting points in situations where the pattern of weed cover is unknown, and it is unnecessary to record discrete spatial patterns within the study area. In addition, these analyses would serve as some indication of whether the more intensive and detailed data collection needed to produce discrete spatial maps of weed cover would be useful.
Weed-Crop Competition
Interference is an important factor that affects individual plants, plant populations, and plant communities (Harper, 1977). The effects of weed-crop competition in agricultural settings have been thoroughly assessed in some respects. In a model based approach using the inverse hyperbolic equation (Cousens, 1985), Garrett and Dixon (1998) determined that increasing the competition resulted in a diminishing yield response. Many field studies have found that crop yield decreases with increased weed density (Moechnig et al., 2003; O'Donovan and Sharma, 1983; Tamado et al., 2002; Whish et al., 2002) and that weed biomass decreased with increased crop density (Weiner et al., 2001). Weiner et al (2001) also found that weed density decreased as a result of planting crops in a grid pattern as opposed to a row pattern. However, the relationship between weed diversity and crop yield is unclear. Although Clements et al (1994) questioned if a more diverse weed community, at a constant density, would have a decreased effect on crop yield as compared to a less diverse weed community, no empirical study has yet evaluated this issue. There have been recent studies that suggest that weed diversity has had either negative (Davis et al., 2005) or positive (Suarez et al., 2001) effects on crop yield. Both of these studies were observational in nature, however, and did not control for weed density. Therefore, the yield effects observed could have been due to other factors besides diversity. In one study, Murphy et al (2006) demonstrated that crop yield was not affected by management system, even in light of increased weed diversity in some systems. This suggests that weed species diversity may have no effect on yield (Murphy et al., 2006). The link between weed diversity and crop yield has yet to be experimentally tested.
To study the effects of diversity on crop yield, it would be necessary to manipulate weed diversity and examine the effects of this manipulation on target crop plants as well as on the weeds themselves. The effects of neighboring plants on the growth of an individual plant are commonly studied using a running radius neighborhood competition approach. In this approach, a target plant is centered within a circular area of fixed radius, and all other plants within that area are considered neighbor plants (Silander and Pacala, 1985). Although neighborhood size is sometimes set arbitrarily, Silander and Pacala (1985) found that the optimum neighborhood size for relating neighbor plants to target plant performance was quite small (5cm radius). This lends support to the idea that plants only compete strongly with their nearest neighbors (Crawley, 1997a).
The effects of neighbor plants on target plants can be quantified in several ways. Using linear models, Hickman (1979) was able to account for 48 to 73% of the variation in dry weight biomass of Polygonum using the mean distance to four neighbors. Weiner (1982) was able to account for over 80% of the variation in individual seed set using a model which incorporated both distance to neighbors and their competitive effects. The number of independent factors that could be used in these models is quite expansive, but it is usually limited by logistical considerations. One potential problem with these types of analyses is that they measure absolute differences in biomass, seed production, or yield, and generally do not account for initial size inequality due to differences in physiology between species. One measure that accounts for the efficiency of biomass accumulation of a plant given the size of the plant is the relative growth rate, or RGR (Radosevich et al., 1997). The RGR also serves as a good approximation of competitive ability (Hegazy et al., 2005; Holt and Orcutt, 1991; Wang et al., 2006).
Measurements of relative growth rate can be achieved by accounting for plant biomass at the time of planting and then measuring plant biomass at harvest. Relative growth rate can also be measured at many points throughout the growing season by measuring an attribute of the plant which can be correlated with biomass. Bussler et al (1995) found that plant volume was very closely correlated with plant biomass. Thus, by measuring plant volume at several points throughout the growing season, plant biomass could be estimated without actually harvesting the plant. In this way, relative growth rate of a target plant or species could be measured over a growing season and between points within the growing season.

Project Objectives:

Objective 1: Assess the spatial and temporal patterns of weed density and weed species richness in conventional input and organically managed spring wheat systems.

Objective 2: Quantify the effects of weed species richness and weed density on the growth and yield of spring wheat.

Cooperators

Click linked name(s) to expand/collapse or show everyone's info
  • Bruce Maxwell
  • Fabian Menalled

Research

Materials and methods:

Objective 1 This part of the study was conducted on three production farms near Big Sandy, Montana, 48°10’44”N, 110°06’49”W, 824 m elevation. Soils in this area ranged from Telstad-Joplin Loam to Fort Benton Fine Sandy Loam (Bronec, 2003), and over the last five years annual precipitation averaged 28.83 cm. The study sites were spread across one conventional no-till and two organically managed farms, and all sampled fields were in spring wheat (Triticum aestivum) production at the time of sampling. Sampling took place in the early summer of 2005 and 2006. Each year, three conventional no-till and three organic fields were observed. The same farms were used for both years, but the location of fields changed between years due to crop rotation. The conventional no-till fields followed a winter wheat, fallow, spring wheat, fallow rotation and were subject to both fertilizer and herbicide use. The organically managed fields followed a spring wheat, Austrian winter pea (Pisum sativum) rotation with no fertilizer or pesticide inputs, and were plowed just prior to planting in both 2005 and 2006. All sampling took place before the conventional no-till fields had been sprayed with herbicides.
Since the primary goal of this study was to investigate the spatial distribution of weed communities in conventional no-till and organic spring wheat systems, it was necessary to ensure that the weed community would be sampled. In addition, it was also necessary to sample where patterns were likely to be found, in the direction of the majority of machine travel. Thus, the first transect was established perpendicular to crop rows such that it would intersect a portion of the field including a representative weed community. Two additional transects were then established at random distances greater than 15 m to the left or right of and parallel to the original transect. Each of the three transects measured 100 m in length and was approximately parallel to the orientation of the crop rows. Sampling along each transect was continuous using a 0.3 m wide by 0.99 m long frame. Within each frame, the percent cover of each weed species present was measured using ocular estimation by one observer. After sampling within a given field was completed, the location of each transect was geo-referenced using a Trimble Pro XRS Global Positioning System unit.
Data Analysis
Species richness and percent cover were compared between systems using a nested ANOVA where field was nested within system. Since year had no effect on either one of these variables, the species richness and percent cover data were pooled for 2005 and 2006. Two indices were used to test the null hypothesis that weeds are distributed randomly in both conventional and organic spring wheat systems; the index of dispersion, ID (Krebs, 1999) and the index of aggregation, Ia (Perry, 1998). Both of these indices were calculated based on percent cover data. The index of dispersion (ID) is calculated as follows:

where s2 and x-bar are the variance and mean of the data respectively. The index of aggregation (Ia) was calculated as follows:

where D = the observed distance to regularity of the data and Ea = the mean distance to regularity based on several randomizations of the data. Values of ID and Ia were calculated for each transect in each field for 2005 and 2006. This allowed for a detailed examination of each transect in which it could be determined if the spatial pattern of weed cover present varied from random. For a random population, the expected values for ID and Ia are 1. An ID or Ia value above or below 1 indicates aggregation or uniformity respectively. To test for more general differences in the ID or Ia values between fields, systems, and years, a nested ANOVA was used where field was a nested factor within system and the effects of year were investigated. Since year had no effect on ID or Ia values, data from 2005 and 2006 were pooled for the ANOVA and analyzed with field nested within cropping system. Data were transformed using the BoxCox routine in R 2.2.1 (CRAN, 2005) to minimize problems associated with non-constant variance and non-normality.
A combination of graphs of percent cover and three term local quadrat variance (3TLQV) plots (Hill, 1973) were then used to examine the spatial scale of pattern for each transect in each field. The 3TLQV method functions by calculating the average of the squared difference between block totals of adjacent trios at differing block sizes (b) using the following formula:

Plotting the variance at differing block sizes then allows one to examine the spatial pattern of the data in question. The 3TLQV plots for this study were generated using a maximum block size of 25 m. The maximum possible block size for this technique is 33% of the transect length, or 33 m in our case. We chose to use 25 m as the maximum block size to avoid smoothing of variance peaks at smaller block sizes, since we were interested in patterns occurring from the small to intermediate scale in our data. Semivariance analysis was not utilized due to the large number of vacant quadrats in the conventional no-till system. In such a case, when the data cannot be transformed to an approximately normal distribution, the validity of semivariance analysis is questionable (Rew et al., 2001). The ID, percent cover plots, and 3TLQV plots were all produced using PASSAGE 1.0 (Rosenberg, 2001). Ia values were generated using SADIE 1.22 (Perry, 2001) with 5967 randomizations. Nested ANOVAs were performed using R 2.2.1.
Objective 2 This study was conducted at the Montana State University Arthur Post Agronomy Farm in Bozeman, Montana, 45º 40’29” N, 111º09’14”W, 1423 m elevation, in 2005 and repeated in 2006. Mean annual precipitation for the last five years was 13.154 inches. Soil in the area is classified as an Amsterdam-Quagle Silt Loam (Brooker, 2002). This study was initiated in October 2004 at a 27.5 m by 35 m area in a field that had been fallow for three years prior to the start of this study. Using four weed species (green foxtail [Setaria viridis (L) Beauv.], wild oat [Avena fatua (L)], kochia [Kochia scoparia (L) Schrad.], and field pennycress [Thlaspi arvense (L)] ), and spring wheat (Triticum aestivum), a completely randomized full factorial addition series experiment was set up using combinations of 1, 2, 3, 4, and 5 species at four levels of density for a total of 124 one m2 plots. The approximate plot densities were 49, 100, 169, and 361 plants per square meter. Weed species were chosen to represent a combination of species which naturally co-occur in spring wheat systems in Montana (Chapter 2). Plots were placed with 1.5 m spacing, both within and between rows of plots. In the fall of 2005, the experimental area was shifted slightly to the East to minimize any problems caused by weed seed contamination from the previous year.
Plots were planted using 1 m2 cardboard templates in which a hole had been drilled for each location of seed placement, each hole being color coded to facilitate planting of different species. Holes were evenly spaced in a grid pattern on each template at either 5 cm, 7.5 cm, 10 cm, or 15 cm intervals, based on the density level of the given template. In the early fall, the area was tilled and then chained in preparation for planting. Flags were then placed at the corners of each plot. Each plot was planted following the same basic protocol. The soil surface of the plot was raked to provide a more uniform and level planting surface. The appropriate template was then placed on the ground and two 30 cm lengths of PVC pipe were driven through two large holes at the corner of the template and into the ground. This stabilized the template during planting and facilitated the proper placement of the same template the following spring when spring wheat would be planted. A length of dowel was pressed through each hole in the template to create a hole in the ground for seed placement which was approximately 2 cm deep. At least three seeds of the proper species were then dropped through each hole using a small metal funnel, followed by a small amount of soil. When planting was completed, the template was removed and the ground was closely inspected. At this point, any holes which had not been adequately filled with soil were filled.
The following spring, spring wheat was planted into each plot where it was a part of the species combination using the same template that had been used to plant weed species in the fall. The template was placed on the two PVC stakes and wheat was planted into the designated holes using the same process as in the fall. However, during the spring, plots were not raked in preparation of planting as this would disturb seed already in the soil.
During the early part of the growing season, each plot was visited frequently and thinned by hand in order to ensure that there was only one individual plant growing from each planting location. Other unwanted plants were also removed from the plots at this time. Throughout the growing season, the spaces between plots were mechanically kept weed free. No herbicides were used for weed control at any point during this experiment.
One target plant of each species in each plot was identified with a plastic pot tag after plots had been thinned. Spring wheat target plants were located as close to the center of the plot as possible. Target plants of other species were located wherever space allowed. Throughout the growing season, each target plant represented the center of a circular neighborhood of 16.5 cm radius. This neighborhood size was selected based on the assumption that plants only compete strongly with their nearest neighbors (Crawley, 1997; Stoll and Weiner, 2000), and due to the spacing between seed placements at the lowest density of planting(15 cm). Measurements for growth analysis were taken starting June 8, June 28, July 19 and August 17 in 2005, and June 13, July 5, July 25, and August 15 in 2006. Within each neighborhood, the largest width, width perpendicular to the largest width, and the height of each target plant were measured. In addition, the average height, maximum height and percent cover of each neighbor species within the neighborhood were also recorded. Average and maximum height were measured and percent cover estimated for each species within the neighborhood boundaries, regardless of whether or not it was rooted in the neighborhood. To ensure consistency, one person was responsible for all neighborhood measurements. At the end of the growing season in 2005 and throughout the growing season in 2006, several individual plants of each species located in separate plots were measured, harvested, dried, and then weighed to establish a relationship between plant volume and plant dry biomass so that RGR could be estimated non-destructively.
In late August, selected neighborhoods from each plot were harvested. The target plant was harvested and bagged separately. Other individuals rooted in each neighborhood were harvested and bagged by species, noting the number of individuals for each species. Only plants rooted within the neighborhood were harvested. All harvested plant material was then dried for 1 week until constant weight was achieved and then weighed. Seed production was estimated for all individual target plants of each species except kochia because seed production in kochia was incomplete at the time of harvest. Seed production for each species was estimated by establishing a relationship between seed weight and seed number using linear regression. Seeds from target plants were cleaned using a seed blower. Seed was then weighed and number of seeds produced per target plant was estimated using the regression equation. The number of seeds for wild oat target plants had to be estimated at harvest by counting the number of glumes on each plant, as some of the seed had already dropped.
Data Analysis
Predictors of Target Plant Biomass, Seed Yield, and Tiller Number
To begin first test the null hypothesis that species richness, via interspecific competition, would not account for any of the variation in the target plant dependent variables when neighborhood density was in the model, each dependent variable was regressed against neighborhood density, neighborhood species richness, and the interaction term between neighborhood density and neighborhood richness. This model was termed the combined neighbor species model (CNSM) because the independent neighborhood density variable did not take into account individual species contributions to neighborhood density and their separate effects on target plants.
Dependent variables were transformed as necessary to deal with normality/ non constant variance problems using the Box Cox routine in R 2.2.1 (CRAN, 2005). A p-value of 0.1 was used as the cutoff for statistical significance for the CNSM. If the CNSM model for spring wheat target plants included a positive regression coefficient for neighborhood species richness and a negative coefficient for neighborhood density, then one would reject the null hypothesis and conclude that diversity was having a positive influence on the competitive interaction between the crop and weeds. A similar conclusion could be drawn, but with more complicated stipulations, if a positive interaction term came into the model if the coefficient for neighborhood species richness and density were both negative.
To further tease apart neighborhood factors that influence interspecific competition on target plant biomass or yield, models containing the ratio of target plant biomass to neighbor biomass at the first growth analysis measurement (t1 ratio, a surrogate for the difference in emergence time between the target plant and its neighbors), the biomass of each neighbor species, and all of the individual species biomass interaction terms were examined for influence on the dependent variable for each species. This model was termed the by neighbor species model (BNSM), because the effects of each neighbor species on the target plant were taken into account. Schwarze’s Bayesian Information Criterion (BIC) was used to select variables which had the most effect on the dependent variables. The resulting model was called the best BNSM. The BIC was chosen due to its tendency to reject models that over-fit in favor of more simple models (Burnham and Anderson, 2004; Sullivan and Joyce, 2005).
The reasoning behind this two step approach was that the CNSM would separate the affects of species richness from density. The final BNSM would then give some indication of which species specific variables affected target plant performance. If any positive interaction terms between neighbor weed species variables came into the best BNSM for the target plant, this would indicate that some type of interspecific competition was occurring and decreasing the negative effects of competition on the target. Conversely, if negative interactions between neighbor species biomass variables were included in the best BNSM, this would indicate a negative species richness affect, perhaps due to increasing use of a particular resource when multiple species were present. The regression analysis and model selection were all performed in R 2.2.1. Analyses were conducted on 2005 and 2006 data separately. However, data from 2006 are emphasized in the results and discussion due to potential problems arising from inconsistent emergence and plant symptoms indicating potential residual herbicide effects in 2005. Inconsistent emergence may have skewed any competitive interactions occurring in the experiment in ways that would not have occurred naturally, i.e. weed species emerging much later than normal, crop species suffering from frost damage due to early planting. In addition, areas of residual herbicide were identified around the experimental area in 2005, and could have been affecting plants within the area. In 2006, the experimental area was shifted to an area where no residual herbicide had been detected.
Predictors of Relative Growth Rate
Relative growth rate (RGR) was calculated for each target plant using the classical method where RGR = ln(dry biomass t2) - ln(dry biomass t1) / (t2 - t1) and t was day of year. The period from the first to the third measurement was used to calculate the RGR of each target plant for spring wheat, kochia, green foxtail, and wild oat coinciding with the period of most active growth. The period from the first to the second measurement was used to calculate RGR for field pennycress, due to earlier senescence of target plants. Plant dry biomass was estimated based on a biomass/volume relationship using linear regression. RGR was then used as the dependent variable in a series of regression analyses to identify neighborhood variables influencing target plant RGR to test the same hypotheses as described above where target plant biomass, yield or number of tillers were the dependent variables. The neighborhood density variable was based on initial planting density because actual neighborhood density was not measured for all neighborhoods during the period of growth rate data collection. In addition to this regression analysis, an ANOVA with Tukey-Kramer procedure was also applied to the RGR data to examine the differences in mean RGR between species within each level of neighborhood species richness. Regression analyses, model selection, and ANOVA were all performed in R 2.2.1. This analysis was only carried out for the 2006 data.

Research results and discussion:

6. Results/Discussion
Objective 1. The un-pooled ID values (Table 1) indicate that weed percent cover was generally aggregated along transects in both the conventional no-tillage and organic systems. Based on the ID analysis for 2005, percent cover for all of the conventional no-tillage transects and seven of nine organic transects had patterns that were significantly different than random (P<0.001), with mean ID values of 2.9 and 4.8 for the conventional no-tillage and organic systems respectively. An outlying ID at Org-2, transect 3 (Table 1) was left in the analysis for 2005 because it represented a large patch of Canada thistle (Cirsium arvense l.). In 2006, percent cover in seven of nine conventional no-tillage transects and all of the organic transects had patterns that were significantly different from random (P<0.001) with mean ID values of 3.4 and 4.2 for the conventional no-tillage and organic systems respectively.
The un-pooled Ia values (Table 1) generated by the SADIE analysis suggest a general trend towards spatial aggregation along both the conventional no-tillage and organic transects. However, only two of nine conventional no-tillage transects and four of nine organic transects showed significant differences from a random distribution (0.025>Pa>0.975) in 2005 based on SADIE analysis methods (Perry, 1998). The significant departures from the random distribution all indicated an aggregated distribution (Pa < 0.025). In 2006, three of nine transects for both conventional no-tillage and organic systems indicated significant departures from the random distribution (0.025 > Pa >0.975). Of these significant departures, all indicated an aggregated distribution (Pa < 0.025).
When the 2005 and 2006 transect data were pooled after determining that there were no year effects, there was no difference in mean ID between the conventional no-tillage (mean = 3.14) and organic (mean = 4.55) systems based on the nested ANOVA (P = 0.31). In addition, the mean Ia values were not significantly different for the organic (mean = 1.70) and conventional no-tillage (mean = 2.38) systems (P = 0.11).
Three Term Local Quadrat Variance analyses were completed for all transects in all fields for both 2005 and 2006. However, only analyses from one field for each system in each year are presented as figures here. These fields are representative of the trends seen in all fields. The 3TLQV analysis indicated that for both conventional no-tillage and organic systems, percent cover showed multiple scales of pattern (Figs. 1 - 4). Examination of the graphs of percent cover for each transect (Figs. 1 - 4) reveals that the spatial patterns of weed percent cover exhibited by the conventional no-tillage system are different than the organic system. In both 2005 and 2006, the conventional no-tillage system transects had a baseline percent cover of zero, from which there were intermittent peaks in percent cover along each transect. In contrast, the organic system transects were characterized by an oscillating pattern of percent cover with very few zero cover quadrats. This suggests that the peaks in variance in the 3TLQV plots for conventional systems represent the scales of patchiness of weed cover as well as the scale of gaps in between patches.
Weed percent cover and species richness (Table 2) were significantly higher (P<0.001) in the organic system than in the conventional no-tillage system when considering either data from all quadrats, or data from quadrats in which weed species were present based on the results of the nested ANOVA.
Objective 2
Predictors of Target Plant Biomass, Seed Yield, and Tiller Number
Spring Wheat. In 2006, spring wheat dry biomass and seed yield were both negatively affected by increasing neighborhood density, and no significant effects of neighborhood species richness were observed (Table 4). The number of tillers per target plant was only affected by neighborhood density (Table 3). There were no significant interactions between neighborhood density and richness for any of the dependent variables. The next step was to look at the neighborhood species-specific factors responsible for the variation in spring wheat performance and thereby determine which, if any, were having an effect on spring wheat performance.
The best BNSM for spring wheat target plant dry biomass included the t1 ratio (relative initial size), spring wheat neighbor biomass, and field pennycress neighbor biomass as independent variables and accounted for 22% of the variation (Table 4). The best BNSM for spring wheat yield included only the t1 ratio, and accounted for 11% of the variation in yield (Table 4). The best BNSM for number of tillers per target plant included the t1 ratio, spring wheat neighbor biomass, and field pennycress neighbor biomass, and accounted for 33% of the variation (Table 5). None of the best BNSMs included any neighbor species biomass interactions, suggesting that no species-specific interactions were playing a direct role in final wheat target biomass, yield, or tiller production. The inclusion of spring wheat and field pennycress neighbor biomass terms in the best BNSMs for dry biomass and number of tillers suggests that early season competition has the most effect on spring wheat performance, as both spring wheat and field pennycress neighbors were present at the time of target emergence. The inclusion of the t1 ratio (target plant volume divided by total neighbor volume) also indicates that the early season competitive environment had important effects on spring wheat performance. Thus, the original null hypothesis could not be rejected based on our data, as no species interaction terms came into the best BNSM. For the 2005 data, neither neighborhood species richness nor density had any affect on spring wheat target plant dry biomass or yield indicating that other factors were determining the outcome of the interactions.
Weed Community. The same protocol for data collection and analysis was also performed for neighborhoods in which weed species served as target plants. Although that data is not presented in this paper, it is worth mentioning. The best BNSM for field pennycress dry biomass contained the t1 ratio, spring wheat, field pennycress, kochia, and green foxtail biomass, as well as the interaction terms for spring wheat x kochia biomass (negative) and kochia x green foxtail biomass (positive). This model accounted for 65% of the variation in field pennycress dry biomass. The best BNSM for kochia dry biomass included spring wheat, kochia, and wild oat biomass as well as the interaction term for kochia x wild oat biomass(negative). This model accounted for 25% of the variation in kochia dry biomass. The best BNSM for wild oat dry biomass included the t1 ratio, all individual neighbor species biomass variables, and the interaction terms for spring wheat x green foxtail biomass (negative), spring wheat x wild oat biomass (positive), field pennycress x kochia biomass (negative), and kochia x green foxtail biomass (negative). This model accounted for 80% of the variation in wild oat target plant dry biomass. The inclusion of negative interaction terms in these models suggests that when multiple weed species were present in a neighborhood, they increased the negative effects of competition between weeds were increased.
Predictors of Relative Growth Rate
Spring Wheat. Neighborhood species richness and neighborhood density did not come into the best model for RGR (Table 6). The best BNSM for spring wheat target plant RGR included only spring wheat neighbor biomass (Table 7). This model accounted for 10% of the variation in spring wheat RGR. These results suggest that intra-specific competition had the greatest effect on spring wheat RGR, and that neighboring weed species were having no significant influence on spring wheat during the period of most active growth.
Weed Community. The ANOVA with Tukey-Kramer procedure indicated that there were several differences in RGR between species within richness levels (Table 8). Kochia and wild oat generally had the highest RGR across levels of species richness, suggesting that Kochia and wild oat were superior competitors in the crop-weed community. However, as species richness increased, the competitive abilities of all species in the neighborhoods became more similar (Table 8). As with the regression models, the comparison of RGR within species across richness levels indicated that neighborhood species richness had very little effect on individual species RGRs.

Literature Cited
Biao, X., W. Xiaorong, D. Zhuhong, and Y. Yaping. 2003. Critical impact assessment of organic agriculture. Journal of Agricultural & Environmental Ethics 16:297-311.
Brain, P., and R.D. Cousens. 1990. The effect of weed distribution on predictions of yield loss. journal of Applied ecology 27:735-742.
Bronec, R.L. 2003. Soil survey of Chouteau County area, Montana Natural Resources Conservation Service.
Brooker, J.W. 2002. Soil survey of Gallatin County area, Montana Natural Resources Conservation Service.
Burnham, K.P., and D.R. Anderson. 2004. Multimodel inference: Understanding AIC and BIC in model selection. Statistical Methods and Research 33:261-304.
Clay, J. 2004. Agricultural trends and realities, p. 11-44 World agriculture and the environment : a commodity-by-commodity guide to impacts and practices. Island Press.
Clements, D.R., S.F. Weise, and C.J. Swanton. 1994. Integrated weed management and weed species diversity. Phytoprotection 75:1-18.
Cousens, R.D. 1985. A simple model relating yield loss to weed density. Annals of applied biology 107:239-252.
CRAN. 2005. R. Release 2.1.1. CRAN.
Crawley, M.J. 1997. Life History and Environment, p. 72-131, In M. J. Crawley, ed. Plant Ecology. Blackwell Science, Malden.
Davis, A.S., K.A. Renner, and K.L. Gross. 2005. Weed seedbank and community shifts in a long-term cropping systems experiment. Weed Science 53:296-306.
Dieleman, J.A., D.A. Mortensen, and A.R. Martin. 1999. Influence of velvetleaf (Abutilon theophrasti) and common sunflower (Helianthus annuus) density variation on weed management outcomes. Weed Science 47:81-89.
Faechner, T., K. Norrena, A.G. Thomas, and C.V. Deutsch. 2002. A risk-qualified approach to calculate locally veraying herbicide application rates. Weed Research 42:476-485.
Harper, J.L. 1977. Population biology of plants Academic Press.
Hegazy, A.K., G.M. Fahmy, M.I. Ali, and N.H. Gomaa. 2005. Growth and phenology of eight common weed species. Journal of Arid Environments 61:171-183.
Heisel, T., C. Andreasen, and A.K. Ersboll. 1996. Annual weed distributions can be mapped with kriging. Weed Research 36:325-337.
Hill, M.O. 1973. The intensity of spatial pattern in plant communities. Journal of ecology 61:225-235.
Holt, J.S., and D.R. Orcutt. 1991. Functional-Relationships of Growth and Competitiveness in Perennial Weeds and Cotton (Gossypium-Hirsutum). Weed Science 39:575-584.
Hughes, G. 1990. The problem of weed patchiness. Weed Research 30:223-224.
Krebs, C.J. 1999. Ecological Methodology. 2 ed. Benjamin/Cummings, Melno Park.
Moechnig, M.J., D.E. Stoltenberg, C.M. Boerboom, and L.K. Binning. 2003. Empirical corn yield loss estimation from common lambsquarters (Chenopodium album) and giant foxtail (Setaria faberi) in mixed communities. Weed Science 51:386-393.
Mortensen, D.A., G.A. Johnson, D.Y. Wyse, and A.R. Martin. 1995. Managing spatially variable weed populations. ASA-CSSA-SSSA soil specific management for agricultural systems:397-415.
Murphy, S.D., D.R. Clements, S. Belaoussoff, P.G. Kevan, and C.J. Swanton. 2006. Promotion of weed species diversity and reduction of weed seedbanks with conservation tillage and crop rotation. Weed Science 54:69-77.
O'Donovan, J.T., and M.P. Sharma. 1983. Wild oats, competition, and crop losses. Wild oat action committee proceedings:27-37.
Perry, J.N. 1998. Measures of spatial pattern for counts. Ecology 79:1008-1017.
Perry, J.N. 2001. SADIE. Release 1.22. Perry, J.N., Harpenden.
Radosevich, S., J. Holt, and C. Ghersa. 1997. Weed Ecology: Implications for Management. second ed. John Wiley and Sons, New York.
Rew, L.J., and R.D. Cousens. 2001. Spatial distribution of weeds in arable crops: are current sampling and analytical mehods appropriate. Weed Research 41:1-18.
Rew, L.J., B. Whelan, and A.B. McBratney. 2001. Does kriging predict weed distributions accurately enough for site-specific weed control? Weed Research 41:245-263.
Rosenberg, M.S. 2001. PASSAGE. Release 1.0. Rosenberg, M.S., Tempe, AZ.
Silander, J.A., and S.W. Pacala. 1985. Neighborhood predictors of plant performance. Oecologia 66:256-263.
Stoll, P., and J. Weiner. 2000. A neighborhood view of interactions among individual plants, p. 11-27, In U. Diekmann, et al., eds. The Geometry of Ecological Interactions. Cambridge University Press.
Suarez, S.A., E.B. de la Fuente, C.M. Ghersa, and R.J.C. Leon. 2001. Weed community as an indicator of summer crop yield and site quality. Agronomy Journal 93:524-530.
Sullivan, J., and P. Joyce. 2005. Model selection in phylogenetics. Annual Review of Ecology Evolution and Systematics 36:445-466.
Tamado, T., L. Ohlander, and P. Milberg. 2002. Interference by the weed Parthenium hysterophorus L. with grain sorghum: influence of weed density and duration of competition. International Journal of Pest Management 48:183-188.
Thornton, P.K., R.H. Fawcett, J.B. Dent, and T.J. Perkins. 1990. Spatial weed distribution and economic thresholds for weed-control. Crop Protection 9:337-342.
Tilman, D. 1999. Global environmental impacts of agricultural expansion: The need for sustainable and efficient practices. Proceedings of the National Academy of Sciences of the United States of America 96:5995-6000.
Tilman, D., C.L. Lehman, and K.T. Thomson. 1997. Plant diversity and ecosystem productivity: Theoretical considerations. Proceedings of the National Academy of Sciences of the United States of America 94:1857-1861.
Tilman, D., K.G. Cassman, P.A. Matson, R. Naylor, and S. Polasky. 2002. Agricultural sustainability and intensive production practices. Nature 418:671-677.
Trewavas, A. 2002. Malthus foiled again and again. Nature 418:668-670.
van der Werf, H.M.G., and J. Petit. 2002. Evaluation of the environmental impact of agriculture at the farm level: a comparison and analysis of 12 indicator-based methods. Agriculture Ecosystems & Environment 93:131-145.
Wang, G.Y., M.E. McGiffen, and J.D. Ehlers. 2006. Competition and growth of six cowpea (Vigna unguiculata) genotypes, sunflower (Helianthus annuus), and common purslane (Portulaca oleracea). Weed Science 54:954-960.
Weiner, J., H.W. Griepentrog, and L. Kristensen. 2001. Suppression of weeds by spring wheat Triticum aestivum increases with crop density and spatial uniformity. Journal of Applied Ecology 38:784-790.
Whish, J.P.M., B.M. Sindel, R.S. Jessop, and W.L. Felton. 2002. The effect of row spacing and weed density on yield loss of chickpea. Australian Journal of Agricultural Research 53:1335-1340.
Wiles, L.J., G.W. Oliver, A.C. York, H.J. Gold, and G.G. Wilkerson. 1992. Spatial-distribution of broadleaf weeds in North-Carolina soybean (Glycine-Max) fields. Weed Science 40:554-557.

Participation Summary

Research Outcomes

No research outcomes

Education and Outreach

Participation Summary:

Education and outreach methods and analyses:

Presentations

“Spatial Distribution of Weed Communities in Conventional and Organic
Spring Wheat Systems” Western Society of Weed Scientists Annual Meeting, Reno, NV, March 14, 2006

“Spatial distribution of weeds in conventional and organic spring wheat agroecosystems” Ecological Society of America Annual Meeting, Memphis, TN, August 8, 2006

“Effects of weed community species richness on the performance of spring wheat: A neighborhood competition approach” Ecological Society of America Annual Meeting, San Jose, CA, August 9, 2007

Publications

“Diversity, Spatial Patterns and Competition in Conventional No-tillage and Organically Managed Spring Wheat Systems in Montana,” MS Thesis, Fredric W. Pollnac, Department of Land Resources and Environmental Sciences, Montana State University, 2007

“Spatial Patterns, Species Richness and Percent Cover in Weed Communities of Organic and Conventional No-Tillage Spring Wheat Systems,” Pollnac, F., Rew, L., Maxwell, B.D., and Menalled, F.D., Weed Research (Submitted)

“Weed Community Species Richness and Crop Performance: A Neighborhood Approach,” Pollnac, F., Maxwell, B.D., and Menalled, F.D., Weed Research (In Preparation)

Outreach

Tour of research plots at Montana State University Agronomy Field Day 2006

Montana State University Crop and Pest Management School 2006

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.