Risk, Rate, and Impact of Medusahead Invasion of California Savannas

Risk, Rate, and Impact of Medusahead Invasion of California Savannas

Summary

This graduate student project is integrated into a larger pre-existing USDA Western SARE project (SW06-038) intended to develop grazing-based methods for medusahead (Taeniatherum caput-medusae) control. A preliminary survey of medusahead (Mh) across the Dunnigan Hills region of Yolo County was initiated to refine an Mh mapping technique using Landsat imagery, which will allow future modeling of Mh distribution based on environmental characteristics. Spatial mapping of “precision grazing” and Mh distribution over treated pastures in Yolo County was conducted and available results are presented. Final results will be integrated with the parent project for maximum impact.

Objectives/Performance Targets

Objective 1. Create a method to determine the risk that a pasture or a land management unit will be invaded by Mh. Use method to prioritize areas for prevention of new Mh infestation in participating ranches.
Objective 2. Improve effectiveness of precision grazing to control Mh by determining how patch size and density affect utilization by livestock and subsequent Mh spread.
Objective 3. Establish thresholds for Mh control by assessing the effects of Mh spatial distribution and patch size on loss of grazing capacity and biodiversity.

Accomplishments/Milestones

To achieve Objective 1, I will use ground-based surveys in to calibrate Mh level in Landsat satellite images of California, and use maps of predicted Mh from these images to model Mh level and changes in Mh level over time against environmental factors (soil, topography, weather, and management).
To refine my procedure, I conducted a preliminary survey of about 100 observations of % medusahead coverage in the Dunnigan Hills geographic area of Yolo County, CA, in September 2007. Observation resolution was roughly 3600 to 10,000 m2. These observations were matched to Landsat 7 images taken in May 2007 (when Mh was green and most other annual grasses and forbs had senesced). Observations and the data from the visible spectra from the Landsat images were used to calibrate a discriminant analysis function that accurately predicted Mh level (none, present and less than 15%, and present but over 15%) in 82% of “test” observations (observations not used for the calibration). A canonical correlation plot of the observations and the Landsat RGB spectra is shown in Fig. 1. Incorrect predictions were “false positives” (Mh predicted to be higher than I actually observed). These may have been due to the late date of the survey (September) which occurred well after the date that the image was captured (May); it is possible that Mh actually existed in the false positive sites but was grazed beyond recognition by the time of my September survey. This procedure is promising because its preliminary accuracy was quite high and is likely to increase with more data (only 110 observations were made in the preliminary survey), and overestimating the risk from Mh (having false positives) is preferable to underestimating the risk (which would have occurred with false negatives – predictions of less Mh than actually existed).
With this data I developed a map of predicted Mh level in the Dunnigan Hills in May 2007 (Fig. 2). The map appears to overpredict Mh level in perennial/cultivated crop areas and in a sub-region infested with thistle (Centaurea spp.). This occurred because both types of vegetation are green at the same time as the Mh and were not included in my preliminary ground survey. This will be remedied by gathering more observations in such areas and utilizing data from Landsat images captured later in the year (when crops and thistles are still green but Mh is senesced).

As part of the larger parent project (Western SARE SW06-038), a precision grazing trial was initiated in Yolo county in spring of 2007. The trial consists of 12 plots of 0.5 acres that are infested with Mh. Each plot is grazed by sheep to varying levels of utilization (ranging from 50-85%) quickly (10 days) or slowly (20 days). Grazing occurs at a time that we hypothesized would disrupt Mh seed production. Half of each plot is reseeded with a forage mixture in the fall after grazing.
To achieve objective 2, I surveyed (or will map) the spatial distribution of forage biomass (before and after grazing in 2007), botanical composition (at peak green mass in 2007 and 2008 (Fig. 3), and % Mh covaerge (at peak green mass in 2007 and 2008, before grazing in 2007, and after grazing in 2007; Fig. 4). Survey observations were used to create kriged maps of each variable and estimate forage utilization and pre-graze %Mh across all the plots. These maps were used to explore the correlation between grazing and Mh distribution.
Mean absolute coverage of Mh and Mh across all plots at peak green mass was over 50% (Fig. 3). Desirable forage species provided less than 30% absolute coverage. Botanical composition of these plots will be surveyed again in spring 2008 to assess the plot-level change in % Mh and % non-Mh coverage, and relate these changes to the patch size and density of Mh in the previous year (2007).
Preliminary analysis showed a three-way interaction between % forage utilization by sheep, duration of grazing, and the variability (patch intensity) of Mh that supports my hypothesis that livestock will avoid grazing Mh as it becomes easier to select non-Mh forage: this occurs when Mh coverage is more variable within a pasture, when forage utilization is lower, and when duration of grazing is longer. Preliminary analysis showed that as the Mh variability (standard deviation of % Mh coverage) across each pasture increased, sheep were indeed more likely to avoid utilizing high Mh areas as forage utilization went down and the duration of grazing increased in the Yolo County study site in 2007 (three-way interaction significant at the 0.01 level).
Across all utilization levels, precision grazing reduced end of season dry matter in Mh infested areas by about 30-40% compared to infested areas protected from grazing and infested areas where sheep had access but did not graze heavily (Fig. 5). Extreme differences in precision grazing utilization level had weakly significant effects on post-graze and end-of-season % Mh coverage (Table 1); compared to 80% utilization level, Mh tended to “bounce back” to much higher end-of-season at precision grazing when utilization was applied at the 50% level (Table 2). Precision grazing reduced Mh seed density by 75%. Utilization level (from 50 to 85%) had no clear effect on Mh seed density, but a seed viability analysis has not yet been conducted. It is possible that more heavily grazed Mh produced less viable seed.

Achieving this objective will require results from Objective 2 to be completed and nutritional analysis of collected forage samples. I expect to begin processing samples for analysis in late 2008 and the results from all precision grazing trials will be completed by spring 2009.

Impacts and Contributions/Outcomes

Achieving Objective 1 will provide ranchers in mapped areas with information on potential Mh infestation on their property, which will help inform their decisions regarding prioritization of Mh control efforts. Preliminary results from Objective 2 were already shared with participants of the larger parent project (SW06-038) at our annual meeting, and this greatly aided in the visualization and increased precision of results for the larger parent project. All results will be eventually integrated into the outreach and education program for the parent project, ensuring they are presented and disseminated to a wide audience. We intend to also integrate this information with the online Univ. of California Cooperative Extension Weed Research and Information Center.

Collaborators: