Joint Management of Wheat Stem Sawfly, Fusarium Crown Rot, and Weeds: Assessing the Ecological Basis of a Total Systems Approach to Pest Management Strategies
We quantified interactions among wheat stem sawfly (Cephus cinctus), Fusarium crown rot and cheatgrass (Bromus tectorum) in their joint impact on spring wheat yield and assessed the effectiveness of modifying initial seeding density and crop variety as control practices for this pest complex. We have collected almost 4,000 samples at three fields across Montana in 2009 and 2010 and processed approximately 90%. We have developed a modeling framework to predict the impact of varying agronomic practices in the presence of pests by simultaneously estimating direct and indirect effects. Our partial results indicate significant yield reductions associated with cheatgrass and sawfly, and complex interactions between these two pests and in response to control practices.
- Quantify interactions among wheat stem sawfly, Fusarium and grassy weeds in their joint impact on crop yield and in response to simple control practices such as seeding density, herbicide management approach and crop variety. Develop a predictive computer model to assess the relative importance of seeding density, herbicide management approach and crop variety on crop yield and the population dynamics of wheat stem sawfly, Fusarium and grassy weeds. Develop and deliver an education and extension program to increase producers’ knowledge base on the ecological basis of multi-pest management strategies.
We conducted the second year of the study, evaluating the role of agronomic practices in the development of integrated management approaches for wheat stem sawfly, Fusarium and grassy weeds. In Montana, three fields located in production farms were selected and three spring wheat varieties (wheat stem sawfly resistant, drought tolerant and high yield) were seeded at three seeding rates (100, 200 and 400 plants m-2). Sulfosulfuron was applied at varying rates to generate a range of cheatgrass densities. All other broadleaf and grassy weeds were controlled with standard chemical treatments. Wheat stem sawfly abundance, Fusarium infection rates, as well as crop and weed emergence, abundance, biomass and seed production were evaluated using 0.485 m diameter rings. At the plot level, we obtained visual estimates of cheatgrass abundance and yield. To validate the responses observed in Montana, a reduced version of this experiment was conducted in Wyoming, Idaho and North Dakota.
Over 400 wheat and weed samples at each Montana site (1200+ total) and over 100 samples at out of state sites (350 total), in addition to plot level samples (a total of 351 at all six sites) have been collected for the 2009 and 2010 growing seasons. In Montana, plot level samples have been analyzed for yield and grain quality, and we finished weighing all cheatgrass dry biomass in the ring level samples. Fusarium infection levels of non-sawfly cut stems have been assessed for all samples, but sawfly abundance and yield were assessed for only two of the three sites so far.
Sawfly abundance in the samples was assessed by counting the numbers of sawfly cut stems in the sample, as well as splitting every non-cut stem and looking for sawfly larva or evidence parasitoids. We are now in the process of splitting the last 250 Montana samples and will then proceed to obtaining their yield estimates. We have also received complete data sets from two of the three out of state sites and anticipate receiving data from the last site in the next month.
We assessed Fusarium infection levels via visual scoring of the discoloration in every wheat stem in a ring sample. However, we cannot use this method for lodged stems because sawfly cutting in itself discolors the bottom node. Thus, our Fusarium results are confounded by sawfly as only non-sawfly cut samples have enough stems for an accurate Fusarium assessment. Therefore, we conducted an additional field experiment, where we collected 46 paired samples of 10 cut and 10 standing stems each. These samples where scored for Fusarium by our visual method and through quantitative polymerase chain reaction (qPCR). This allowed us to obtain an independent measure of infection to validate our visual method and assumptions on infection levels in cut stems. To our knowledge, assessing Fusarium levels under severe sawfly cutting has never been attempted before. All data from this experiment has been collected and analyzed.
To analyze our data and predict the outcomes of varying agronomic practices in the presence of multiple interacting pests, we developed a modeling framework where the residuals from a model of treatment effects on pests are added as a co-variate to a second model of their effects on crop. This procedure, which can be considered an extension of path analysis, allows simultaneous estimation of both the direct effects of the treatments on the crop and the indirect effects via increasing or decreasing pest levels. Interpreting the estimates in a Bayesian decision theory context produces the desired result. We have successfully derived appropriate statistical estimators and their variance properties for the normal distribution in a regression settings for two interacting pests (e.g. cheatgrass-Fusarium or cheatgrass-sawfly effects on yield), and in an ANOVA settings for a design matrix of agronomic practices in the presence of one pest. We are currently testing the behavior of these estimators in a more complicated network of multiple pests, non-linear relationships, and more complicated designs (multi-factor, split-plot, etc.).
Impacts and Contributions/Outcomes
Fusarium infection levels increased from our 2008 preliminary data to 2009 (Fig.1), and we expect them to be even higher for our 2010 data set. Inspection of the data suggests this increase may be related to moisture content as larger differences were observed in drier sites. We did not observe significant interactions between Fusarium and cheatgrass or sawfly in 2009, although we did observe a positive correlation between disease intensity and cheatgrass in the high yielding variety at the Amsterdam site in 2008. We believe the ambiguity of this relationship and the relative impact on yield may be resolved with better coverage, or range, of disease severity as observed in the field in 2010.
Cheatgrass reduced yields significantly and non-linearly (Fig.2), where again our ability to fit the models is dependent on range of cheatgrass observed, and we anticipate greater amounts of cheatgrass in 2010 at both Havre and Ulm (where no cheatgrass emerged in 2009) sites. We also observed weak treatment effects for the competition between cheatgrass and wheat, with less effect on the drought tolerant cultivar and at high crop densities. However, these effects cannot be fully interpreted without considering sawfly effects on yields, which where considerable at all three sites, and its complex interaction with cheatgrass whose presence at the plot level reduced sawfly oviposition and modified cultivar preferences (Fig.3) but by weakening wheat through competition at the local (ring) level, also increases lodging and thus yield losses associated with sawfly (Fig.4).
As an example to the workings of our model, consider the following work flow for our Amsterdam site in 2009. First, we analyzed treatment effects on ring level yields as a two-factor ANOVA in a completely randomized block design (Table. 1.1).
These results indicate the drought tolerant cultivar was the highest yielding but was only marginally significant because of the large standard errors. Higher initial seeding rates slightly increase yields, but not significantly. This model completely ignores the presence of cheatgrass in the field, which significantly reduces yields (Fig.2). Thus, although results reflect the weedy conditions under which the data was collected, they cannot be fully interpreted. For example, the drought tolerant variety is best, but because it is a better competitor for weeds in a water limiting environment (Fig.5) and predictions based on this model may be erroneous when weeds are higher or lower in other situations. A common “solution” is to add a measure of cheatgrass in the field as a co-variate (Table 1.2).
As expected cheatgrass has a significant effect in reducing yield, but its inclusion in the model changes our inference for the treatments. The higher seeding rate has a negative coefficient. This is because estimates in a linear model are of treatments, accounting for all other variables in the model, so the negative coefficient reflects intra-specific competition, i.e. only the direct effect of increasing wheat density on yield, under 0 cheatgrass conditions. Our variety treatment now has no effect, because it does not account for differences in competitive abilities with cheatgrass. However, under our modeling frame work described in the previous section, we “annihilate” the co-variate before it is entered into the model, and thus we obtain estimates of the treatment’s direct effect on yield, while still accounting for their indirect effects through the co-variate (Table. 1.3).
In this model the p-values of the contrasts between high yield and sawfly tolerant and the drought tolerant cultivar are significantly lower than models 1 and 2, reflecting the differences in competitive abilities of these wheat varieties. Seeding rates estimates are again positive because they are now accounting for the inter-specific competitive advantage of higher initial seeding rates.
These results demonstrate the superiority of our model over some current approaches when predicting the outcomes of different agronomic practices in the presence of pests. Last, as the complexity of the model increases with multiple pests, and its utility and stability is data dependent, we expect to see improvement after incorporation of the 2010 data.
As part of our extension/outreach activities, we presented our preliminary results during the 2009 Montana State University Crop and Pest Management School, the 2009 Crop and Weed Annual Field Day and the 2009 Pest Management Tour. Results have been presented at the 2009 and 2010 Ecological Society of America, the 2009 Annual Meeting of the Montana Chapter of the American Statistical Association, the WERA-97 Diseases of Cereals meeting and the 2010 Joint Meeting of the Society for Range Management and the Weed Science Society of America.
- Increasing Fusarium disease severity at 3 Montana sites is related to moisture (sites ordered from left to right wettest to driest)
- Cheatgrass at the ring level in Amsterdam, MT 2009 increases wheat lodging (sawfly cutting of the stem). This may be related to structural strength of wheat stems under resource competition, which explains the lower effect observed in the drought tolerant cultivar (right panel), and higher effects at high seeding rates. Cheatgrass measured in grams is an index of biomass at harvest, corrected for sampling error with a visual estimate at peak biomass.
- Competitiveness of spring wheat for seeding rates (low, medium, or high), and cultivar (DT – drought tolerant, HY – high yield, SFT – sawfly tolerant) in 2009 as reflected in their ability to “suppress” cheatgrass biomass. The lower competitiveness of the high seeding rate may be a result of intra-specific competition or interactions with other pests. Cheatgrass measured in grams is an index of biomass at harvest, corrected for sampling error with a visual estimate at peak biomass.
- Coefficient values are for the differences between high yield variety (vrtyHY), sawfly tolerant (vrtySFT) and initial seeding rates (med, high) from drought tolerant variety at low seeding rates (Intercept)
- Cheatgrass non linearly reduces crop yield at two Montana sites in 2009. Cultural practice treatments such drought tolerance and increasing crop density have a limited effect in the presence of severe sawfly cutting and Fusarium infection.
- Sawfly attacks reduced, and preferences for different cultivars is modified by the presence of weeds at the plot level. Columns represent quartiles of cheatgrass volume (m3) observed early in the season in Amsterdam, MT 2009.
Montana State University
718 Leon Johnson Hall
Montana State University
Bozeman, MT 59717
Office Phone: 4069944783