Effects of Prescribed Burning on Two Perennial Bunchgrasses in the Bald Hills of Redwood National Park

by

Leonel A. Arguello

A Draft Thesis Presented to

The Faculty of Humboldt State University

In Partial Fulfillment of the Requirements for the Degree Master of Arts

December, 1994 Effects of Prescribed Burning on Two Perennial Bunchgrasses in the Bald Hills of Redwood National Park

by Leonel A. Arguello

We certify that we have read this study and that it conforms to acceptable standard of scholarly presentation and is fully acceptable, in scope and quality, as a thesis for the degree of Master of Arts

Major Professor

Approved by the Graduate Dean Abstract

Coastal grasslands in Redwood National Park, known locally as the Bald Hills, are dominated by introduced species. The most dominant and invasive is tall oatgrass, elatius. Increasing in cover and distribution rapidly since 1985, it has invaded many areas formerly dominated by native grass species. Park managers wish to develop prescribed burn strategies to inhibit tall oatgrass while not adversely affecting present levels of native species, in particular California oatgrass, Danthonia californica. In 1990 an investigation was undertaken to evaluate and compare the effects of spring and fall prescribed burning on tall oatgrass and California oatgrass. Results indicate that tall oatgrass is inhibited from further spread by both spring and fall burning. California oatgrass is severely reduced from spring burning and slightly reduced from fall burning. Spring burning is not an option for conducting prescribed burns because of the severe effect on the native California oatgrass. Fall burning may be more effective in controlling tall oatgrass if combined with other techniques. Further investigations are recommended utilizing these techniques.

iii Acknowledgments

I would like to thank Michael Mesler, for his support and encouragement throughout the years and for helping me complete this thesis. I am indebted to John Sawyer, Jerry Allen, and Terry Hofstra for all their constructive comments and suggestions in reviewing my drafts.

I thank my friends and co-workers at Redwood National Park for their support. In particular I wish to thank Vicki Ozaki, Jim Rogers, Jim Popenoe, the Research and Resource Management Support Crew, the RNP prescribed burn staff of 1991, the RNP Vegetation Management Staff, Beth Koltun, and Tony LaBanca for their assistance.

To my family, especially my mother, I am eternally grateful. You are my support and safety net when times are tough, and I would not have made it this far had it not been for all of you. A special thanks to Cara Smith, whose love, support, and encouragement in 1994 insured the completion of this thesis.

Lastly, I dedicate this work to my daughter Danielle. You kept my perspective in focus when I needed it the most by reminding me of what is really important in life.

iv

Table of Contents Page

Abstract iii Acknowledgments iv List of Tables viii List of Figures ix INTRODUCTION 1 Objective 3 Arrhenatherum elatius 4 Danthonia californica 7

STUDY AREA 1O Sampling Location 1O Vegetation 13 Climate 13 Historical Use 14

METHODS 16 General Design 16 Sampling 18 Burning 19 Data Collection 21 Estimates of Frequency 23 Cover Estimates 24 Data Analysis 24

v

Table of Contents (Continued) RESULTS 4O General Results 4O Repeated Measures Analysis of Mean Cover Data 4O Repeated Measures Analysis of Frequency Data . 51

DISCUSSION 67 CONCLUSION 74 LITERATURE CITED 75 APPENDIX A 79

vi List of Tables

Table Page 1. Generalized Phenological Sequence; 199O-1991 . . . 17 2. Fire Behavior for Spring Burned Plots 20 3. Fire Behavior for Fall Burned Plots 22 4. Modified Domin Index Cover/Abundance Scale . . . 25 5. Orthonormalized transformation matrices 34 6. Cover estimates by site, and treatment . . . . 42-43 7. Overall BURNSTAT effect for Cover data 47 8. BURNSTAT by YEAR Interaction 47 9. YEAR effect for Cover analysis 48 10. Oneway tests for Danthonia californica 5O 11. Frequency estimates by site, and treatment . . 53-54 12. Overall BURNSTAT effect for Frequency data . . . . 57 13. The AVERAGED BURNSTAT by YEAR interaction . . . . 59 14. The AVERAGED YEAR effect 59 15. BURNSTAT by YEAR interaction 62 16. YEAR effect for Frequency analysis 63 17. Paired t-test for Danthonia californica . . . 64-66

vii

List of Figures

Figure Page 1. Arrhenatherum elatius 5 2. Danthonia californica 8 3. Redwood National Park 11 4. Study sites 12 5. Location of cover plots 24 6. Schematic of line-point intercept method 27 7. Schematic of repeated measures design 30 8. Pooled means of Cover Estimates for D. californica 44 9. Pooled means of Cover Estimates for A. elatius 46 10. Pooled means of Frequency Estimates for D. californica 55 11. Pooled means of Frequency Estimates for A. elatius 56

viii INTRODUCTION

Redwood National Park (RNP) contains approximately 1,050 hectare (ha) of coastal grassland in an area called the Bald Hills. These grassland and adjacent oak woodlands support the greatest species diversity of any vegetation types in the park (United States Department of Interior (USDI) 1992). Since the arrival of white settlers to the Bald Hills in the 1850's, livestock grazing, cultivation, introduction of exotic , and fire suppression have reduced the presence and abundance of native species (USDI 1992). Presently, the dominant species are non-native forbs and grasses. The most common include Arrhenatherum elatius, Anthoxanthum odaratum, Holcus lanatus, Cynosurus echinatus, Bromus hordeaceous, Vulpia bromoides, Plantago lanceolata, and Rumex acetosella. The most common native species include Danthonia californica, Carex tumulicola, Elymus glaucus, Bromus carinatus, Lupinus bicolor, and Ranunculus occidentalis var. eisenii. D. californica is the only native grass found in significant numbers.

Management objectives for the park include "To restore and/or maintain the natural ecosystems of the park as they would have evolved without disturbance by human technology" (USDI 1987). The park's goal for the Bald

1 2 Hills is to maintain the diversity of plants and animals that prevailed when the area was first visited by Europeans. Although recreating the original mix of native species is not possible, the present assemblage of species is too heavily dominated by non-native species. The establishment of an assemblage of plant species closer to the original mix is central to the immediate management of the prairies. Several strategies are being investigated to accomplish this goal. One strategy involves the use of prescribed burning to inhibit and possibly reduce some of the aggressive non-native species.

One particularly aggressive non-native species is Arrhenatherum elatius (L.) Presl., tall oatgrass. A perennial bunchgrass, it has in most areas become the dominant grassland species, displacing native species and driving down species diversity where it heavily dominates. Its dominance is of concern to park managers. Inhibiting the spread of tall oatgrass may help return the grasslands closer to the original mix.

Experimental prescribed burns have been conducted in the park's grasslands since 1982. Data from these burns suggest that burning inhibits the spread of tall oatgrass into new areas (USDI 1992). Today, however, A. elatius is present in all of the grasslands and much of the 3 surrounding oak woodlands. The response of A. elatius to burning in areas where it is well established is unknown.

Timing of a burn can play a major role in determining how a species responds to fire (Daubenmire 1968). Fall is when RNP conducts prescribed burns in the Bald Hills. Spring burns have never been attempted because fall burning is generally accepted as the time of year to conduct burns. If A. elatius is adapted to fall burning, what effect would spring burning have? Can it be an effective tool against its spread? If so, how would spring burning affect D. californica? Any prescribed burn strategies developed must not negatively impact this important native species.

Objective

The objectives of this study were to evaluate and compare the effects of spring and fall burning on A. elatius and D. californica, in areas where they are dominant and coexist. The purpose was to determine if burning can inhibit A. elatius in areas where it is dominant. D. californica was included to ensure that any treatment adopted to inhibit tall oatgrass not adversely affect this important native. 4 Arrhenatherum elatius (L.) Presl. Tall Oatgrass

Arrhenatherum elatius is a tall (up to 2 m), tussock-forming, cool-season, perennial grass native to the Mediterranean region (Figure 1). It has been introduced in Australia, New Zealand, South Africa, Japan, and South and North America (Kernick 1978). In North America, it is common in meadows or prairies, open ground, and waste places from Newfoundland to British Columbia, south to Georgia, Tennessee, Iowa, Idaho, Utah, Arizona, and California (Chase and Hitchcock 1971). Its elevational range spans from sea-level to 3000 m in the Caucasus (Boissier 1884).

In the Bald Hills, A. elatius was first noted in Elk Camp Prairie in the early 1980's. Since then it has appeared throughout the Bald Hills grasslands, especially along ridges and near roads. Tall oatgrass predominates in abandoned, disturbed fields. It grows very well in loamy soils, often forming dense stands. These characteristics may explain why it has rapidly spread in the Bald Hills. Soils in the Bald Hills are loamy, and since 1982, after 13O years of ranching, have not experienced any widespread disturbance, i.e. fire, grazing, road building, or farming. 5

Figure 1. Arrhenatherum elatius (original drawings taken from herbarium specimen on file at Redwood National Park, Orick, California) 6 The rapid spread of tall oatgrass not only reduces species diversity in some areas, but it may decrease the quality of for resident herds of Roosevelt Elk (Cervus elaphus ssp. roosevelti). Fecal analysis conducted in the Bald Hills prairies, in 1986, showed that elk do not eat A. elatius (USDI 1992).

The effect of fire on tall oatgrass is not well documented in the literature. In one study, short tussock species were replaced by taller, more fire resistant species, such as tall oatgrass, following the cessation of grazing (Lloyd 1972). In 1976, the effects of prescribed burning on vegetation in a Ouercus coccifera garrigue in France were reported. Burning was conducted in the spring and fall every 2, 3, and 6 years. Results indicated that tall oatgrass increased or remained at pre- treatment levels for all treatments (Trabaud 1987).

Two morphological types of this grass exist in the Bald Hills, non-bulbous (A. elatius var. elatius) and bulbous types (tuber oatgrass, A. elatius var. bulbosum (Willd.) Spenn.). In the bulbous type, the lowest stem internodes swell and form corms that contain regenerative buds. Personal observation indicate that the non-bulbous type is more common. 7 Danthonia californica Bolander. California Oatgrass

Danthonia californica is a long lived, perennial, cool- season, bunchgrass found throughout grasslands in the western United States (Figure 2). In North America it is found in open, grassy meadows, rocky ridges, coastal prairies, and middle elevations in the mountains, especially in ponderosa pine forests (Hitchcock and Cronquist 1973). Its distribution includes British Columbia and Alberta, south on both sides of the Cascades to southern California, east to Montana, and through the rocky mountains to New Mexico. It has also been introduced in Chile. D. californica spans an elevational range from sea level to 1700 m (Munz and Keck 1959). Its distribution in California is mainly in the coast ranges from Monterey north to the Oregon border, although it occurs east to the Nevada border, and south into the Sierra Nevada foothills (Beetle, 1947).

Danthonia californica is considered, by many authors, an important native species in California coastal grasslands including the Bald Hills (Davy 19O2; Cooper 196O; Heady et al. 1963; USDI 1992). It might have been the predominant grass species in northwestern Californian grasslands prior to European settlement (Davy 19O2). 8

Figure 2. Danthonia californica (original drawings taken from herbarium specimen on file at Redwood National Park, Orick, California) 9 While tall oatgrass has been increasing rapidly in the Bald Hills, D. californica has been on the decline (USDI 1992). Its distribution today is patchy and comprises no more than 50% cover in any area (USDI 1992). Although a cause and effect relationship between the relative abundance of the two species has not been established, many former D. californica dominated areas have been replaced by A. elatius or other non-native species (personal observation). There is concern with the decline of D. californica because of its former importance as a dominant native species.

The response of D. californica to burning is not well documented in the literature. Its distribution and prominence in California grasslands, however, prior to European settlement suggest that it is not adversely affected by fire. Data collected in RNP show that D. californica is not adversely affected by prescribed burning in the long term (USDI 1992). STUDY AREA

Sampling Location

Redwood National Park is located in extreme northwest California (Figure 3). The Bald Hills occur as discontinuous grassland and oak woodlands alternating with coniferous forest, along the ridge crest dividing the Redwood Creek and Klamath River drainages. The area contains approximately 1,O52 hectares of southwest facing prairies above Redwood Creek between 229 m and 729 m elevation. Within the boundaries of the park, the prairies begin 11 km from the Pacific Ocean and extend inland in a southeast direction for another 11 km.

Three sites, Slide Creek, Hospital Pasture, and Stagecoach, all within Childs Hill Prairie, were chosen for this study (Figure 4). Childs Hill Prairie begins about 17 km from U.S. 1O1. The prairie encompasses approximately 338 ha, and extends from 457 m elevation up to about 762 m. The three sampling sites are located along the upper portion of the prairie within 50 meters of the Bald Hills Road. Each site encompasses an area of approximately 0.5 ha.

10 11

Figure 3. Redwood National Park and the study area (on file at Redwood National Park, Orick, California, USDI 1992) 12

Figure 4. Study Sites within Childs Hill Prairie (on file at Redwood National Park, Orick, California, USDI 1992) 13 All three sites were similar in percent slope (0% to 20%), aspect (southwest), and distance to canopy (>100 meters). Soils in the three sites were similar and classified as fine-loamy (or loamy-skeletal), mixed mesic Ultic Haploxeralfs (personal observation).

Vegetation

The vegetation of upper Child's Hill prairie includes a mix of annual grasses, annual and perennial forbs, and perennial grasses. The dominant species include A. elatius, , Holcus lanatus, Anthoxanthum odaratum, D. californica, Elymus glaucus, Poa pratensis, Rumex acetosella, and Pteridium aquilinum. None of the study sites have been burned for at least 20 years. Livestock were removed in the area in 1982. A list of vascular plants is presented in Appendix A.

Climate

The Bald Hills area has a Mediterranean climate, characterized by mild temperatures, wet winters and dry summers. Mean daily high temperatures in July is 25°C and mean daily low temperature in January is 2°C (Sugihara and Reed 1987). Freezing temperatures can occur from late September to early May. In the summer, a strong oceanic 14 influence produces dense fog that may extend 24 km or more up the Redwood Creek Basin.

Rainfall is light during the summer months with 9Oó of the total precipitation falling from October-April. Annual precipitation during normal years range between 178-230 cm. A drought, however, during the course of the investigation reduced annual rainfall totals. From 1990 to 1992, annual precipitation ranged from 101 cm to 142 cm.

Historical Use

Evidence from archaeological studies indicate that Native Americans occupied the area for at least 4,500 years prior to 1850 (Benson 1983, Hayes 1985). The Chilula Indians inhabited the area and utilized various locations in the prairies and oak woodlands for village sites, seasonal camps, and/or ceremonial sites. They regularly set fire to the area to promote grass growth, to make gathering food and plant materials easier, and to attract wildlife (Grenier 1989).

White settlers immigrated to the Bald Hills during the gold rush of the 1850's. Between 1850 and 1870, 10 small ranches were established. Their livestock, 15 primarily cattle, horses, and mules grazed the area. Several ranches grew oats, potatoes, hay, corn, and wheat. All of these original ranches stopped operating by 1870. For a time, hostilities with the Chilula prevented settlers from operating a successful ranch. By 1880, however, all the remaining Chilula were forcibly removed to the Hupa reservation and the Bald Hills were open to settlement.

In the late 1880's, the Lyons family settled in the Bald Hills, southwest of Schoolhouse Peak, and started a profitable sheep ranch. A portion of their ranch included the study site. Sheep ranching flourished in the late 1800's and early 1900's. By 1940, however, the industry had declined, and ranchers eventually returned to raising cattle until park expansion in 1978. METHODS

General Design

This short-term study was conducted from 1990 to 1992. Vegetation data were collected in June of 1990 to determine the number of plots needed per site and treatment and to work out problems with data collection. Also in 1990, phenological stages of both species were followed to facilitate planning of the burns in 1991 (Table 1). To maximize differences between burns, the spring burn occurred during the onset of flowering for tall oatgrass. The fall burn occurred after both grasses had entered dormancy.

Vegetation sampling in 1991 was conducted in the first two weeks of June. The spring burn was conducted after data collection. Immediately after the spring burn, orchard fencing was erected around all plots to prevent ungulate grazing. The potential for deer and elk to graze regrowth in burned plots could have biased the results of the burning. The fall burn was conducted in November of 1991.

In June of 1992, all plots were re-sampled and the data prepared for analysis. Also in 1992, it was noted

16 17

Table 1. Generalized Phenological Sequence of A. elatius and D. californica in 1990.

Month D. californica A. elatius

January Leaves 1-3 inches Leaves 4 inches February Leaves 1-3 inches Leaves 4-5 inches March Leaves 3 inches Leaves 5-6 inches April Culms appearing Culms appearing May 1-15 Spikelets in boot Spikelets emerging stage from stem May 15-30 Spikelets emerging Flowering from stem June Flowering Flowering July 1-15 Plants beginning to Plants beginning to dry. Seeds maturing dry. Seeds maturing July 15 to August Seed scatter for Seed scatter for 15 spikelets. Plants most spikelets. dry Partial regreening with some stems flowering August 15 to Cleistogamous seeds Plants drying September released. Plant parts dry September Dry Dry October Partial regreening Partial regreening at base. Leaves at at base. Leaves at 2 inches 3 inches November through Leaf elongation to Leaf elongation to December 3 inches 4 inches 18 that the fencing used to prevent ungulate grazing had an unforseen effect at one of the study sites. Grassland rodents, particularly voles (Microtus californicus), were abundant at the Slide Creek site. After burning, they took refuge in the fenced control plots. Their grazing activities within the control plots and nearby treatment plots seriously undermined my ability to interpret results from this site. As a result, this site was not included in the analysis.

Sampling

At each site 30 plots were located where both species were growing together. The thirty plots were randomly divided into three groups of ten. Each group of ten was randomly assigned a treatment; fall burned, spring burned, or control plot.

Sampling plots measured 3 m by 3 m. The inner 2 m by 2 m area was identified as the sampling area. Sampling did not occur when the vegetation was wet, as the weight of the moisture tended to bend plants, especially tall oatgrass. Plots were not sampled on excessively gusty days because of the difficulty in obtaining accurate line- point data. To avoid trampling vegetation, sampling was done on an elevated wooden bench. Rejection criteria for 19 plot placement included a minimum of 20% cover and maximum of 80% cover (ocular estimation) per plot for both tall oatgrass and California oatgrass, locations within 5 m of roads, rehabilitated roads, gullies or creeks, rock outcrops, and forest edges.

Burning

The late-spring burn was conducted on June 17 and 18, 1991. Temperatures were 16°C, winds were out of the SW, and the last measurable precipitation, 1.6 cm was on May 30. Although the live fuel moisture was high (30%) and all of the vegetation green, the deep litter layer was able to sustain the flames. Averaged rates of spread and flame lengths were similar for all plots burned (Table 2). A light rain, 0.5 cm, fell on the 19th of June.

Burn consumption was greater than 75% for all plots. Although the litter and thatch layer were completely burned in most plots, the culms and inflorescences of A. elatius were not, due to their height, high moisture content, and low flame lengths. Exposed basal internodes of most A. elatius were severely burned while D. californica was completely consumed.

20

Table 2. Average Fire Behavior for Spring Burned Plots by Plot Number and Site

Stagecoach Flame Length (m) Rate of Spread

26 O.30 28.2 m/hr 20 O.61 32.2 m/hr 16 O.45 22.1 m/hr 11 O.91 30.2 m/hr 23 O.84 40.2 m/hr 27 O.61 34.2 m/hr 9 O.61 24.1 m/hr 8 O.53 42.2 m/hr 18 O.91 46.3 m/hr 1 O.99 36.2 m/hr Hospital Pasture 26 O.61 26.2 m/hr 28 O.99 38.2 m/hr 2 1.06 28.2 m/hr 17 1.14 54.3 m/hr 5 O.91 30.2 m/hr 8 1.22 28.2 m/hr 6 1.07 24.1 m/hr 25 1.07 24.1 m/hr 23 O.99 34.2 m/hr 3 1.07 26.2 m/hr 21 The fall burn was conducted on November 7, 1991. Temperatures were 20°C, winds were out of the NW, and the last measurable precipitation, 0.7 cm, was on November 3. Dead fuel moisture were recorded at 15% for that day. Averaged rates of spread and flame lengths were similar for all plots (Table 3). Burn consumption ranged from 60 to 80 percent for all plots.

The fall burns did not burn the vegetation as completely as did the spring burns. Moisture from precipitation the week before prevented burning of the vegetation near the soil surface. The day after burning, a light rain (0.84 cm) fell over the area.

Data Collection

Two different sampling techniques were utilized. One technique estimated frequency via a modified version of the Line-Point Intercept method developed for Redwood National Park (Veirs and Goforth 1988). The other method employed estimates of cover using a modified Domin index (Mueller-Dombois and Ellenberg 1974). Both were used concurrently within each plot in 1991 and 1992. In 1990, only the Line-Point Intercept method was used.

22

Table 3. Fire Behavior for Fall Burned Plots by Plot Number and Site

Stagecoach Flame Length (ft) Rate of Spread

3 O.67 201.2 m/hr 2 O.67 126.7 m/hr 4 O.46 126.7 m/hr 6 O.58 98.6 m/hr 29 O.58 98.6 m/hr 14 1.43 not recorded 25 1.43 not recorded 24 0.76 161.O m/hr 28 O.76 94.5 m/hr 15 0.67 140.8 m/hr Hospital Pasture 15 0.76 110.6 m/hr 1 0.76 100.6 m/hr 7 0.85 144.8 m/hr 16 0.58 82.5 m/hr 24 0.76 273.6 m/hr 20 O.67 60.4 m/hr 29 O.58 38.2 m/hr 14 0.58 42.2 m/hr 30 0.58 72.4 m/hr 13 O.67 108.6 m/hr 23 Cover Estimates

Estimates of cover were made using a circular 0.125 m2 frame as recommended by Saenz (1983) in this vegetation type. Circular plots were permanently located with rebar in the center of the plot (Figure 5). For each circular plot, percent cover of each species was recorded by ocular estimation. Cover is defined as the "percentage of the ground included in a vertical projection of imaginary polygons drawn about the total natural spread of foliage of the individuals of a species" (Daubenmire 1968). Cover classes utilized a modified Domin Index cover/abundance scale (Table 4). Cover values of one and two were omitted because estimates could not be made accurately. Litter, bare ground, or rock were recorded in those portions of the plot that were unvegetated.

Estimates of Frequency

The modified Line-Point Intercept method utilized a 2 m by 2 m plot, through which three, parallel transects were located and sampled (actual plot sizes measured 3 m by 3 m to provide a buffer around each plot). The transects were fastened to a 2 m by 2 m pvc frame, and were spaced evenly at the 0.5 m, 1.0 m, and 1.5 m 24

Figure 5. Location of Cover Plots 25

Table 4. Modified Domin Index Cover/Abundance Scale

Cover class Percentage Range Midpoint

3 Less than 1% .5

4 1% - 4% 2.O

5 5% - 10% 7.5

6 11% - 20% 15.5

7 21% - 33% 27.O

8 34% - 50% 42.O

9 51% - 75% 63.O

10 76% - 90% 83.O

11 91% - 100% 95.5 26 positions (Figure 6). Rebar placed at the outer corners of the buffer permanently located the plots and anchored the frame when sampling.

Sampling began at the upper right transect and proceeded down each transect in the same direction and on the same side (Figure 6). A .635 cm diameter(1/4 inch) range pole was dropped from a vertical position every 15 cm along the transect, excluding the first 15 cm and last 20 cm. In this way, 12 points were sampled per transect for a total of 36 points per plot. At every point, each taxon touching the range pole was recorded once only. Litter, rock, or bare ground were recorded when encountered on unvegetated points. Species seen within the 3 m by 3 m frame, but not sampled, were recorded as 'also seen'.

Data Analysis

Two data management programs were used to prepare field data for analysis. "Prairie Analysis", a series of programs that prepares data for statistical analysis, was used for estimates of cover. This program converts cover classes to the midpoint value of its corresponding 27

Figure 6. Schematic of Line-Point Intercept Method 28 percentage range. From these midpoint values, relative, absolute, and mean cover estimates for all species sampled were generated. Absolute cover values were used for further statistical analysis.

"Transect" (Veirs and Goforth 1988), a program designed to analyze Line-Point intercept data, was used to generate values of absolute and relative frequency (called dominance) for all points sampled. Estimates for absolute frequency were used for further statistical analysis.

Statistical analysis of the data was conducted using repeated-measures analysis of variance. A repeated measures design was selected because of its sensitivity both to changes in response over time and differential changes in response within treatment groups over time (Stevens 1986). These designs are much more powerful than completely randomized designs because they control for individual differences among subjects (individual plots in this study). Repeated measures blocks on each subject thus removing from the error term variability among subjects due to individual differences. Repeated measures analysis is an appropriate statistical procedure for this study because parameter estimates, i.e. cover and frequency values, of Arrhenatherum and Danthonia were repeatedly sampled in the same plots over time. 29 The specific analysis used was a doubly multivariate repeated measures design with two between-subjects factors and one within-subjects factor. A schematic of the design is given in Figure 7. The dependent variables are the cover and frequency values recorded every year for both A. elatius and D. californica. The number of dependent variables depends on the data set. For the cover data, there are only four dependent variables, AREL91, AREL92, DACA91, and DACA92. For the frequency data, there are six dependent variables, AREL90, AREL91, AREL92, DACA90, DACA91, and DACA92. The independent variables are burning treatments (spring burn, fall burn, and control) and year sampled (1990, 1991, 1992).

All repeated measures analyses involve within- subjects factors. A within variable is one on which all the subjects are repeatedly measured (Stevens 1986). In the present design, the within factor is YEAR. The within variable can have several levels by which each subject is measured. In this study the levels of YEAR depended on the sampling technique being analyzed. For example, cover data were collected for only two years, thus there are only two levels for the within-subject factor YEAR. For the frequency data, there are three years of data, thus there are three levels of the YEAR variable. 30

Figure 7. Schematic of study design 31 There are two between-subjects factors, burn status (BURNSTAT) and SPECIES. A between-subjects variable is simply a grouping or classification variable such as sex, age, social class. BURNSTAT refers to the treatments used, spring burning, fall burning, and controls. Thus there are three levels to this factor. SPECIES refer to the species analyzed. Since there are two species analyzed, there are two levels to this factor. The first between-subjects factor used to sub-divide the data set is SPECIES. In the analysis, the names AREL and DACA are given to differentiate sets of dependent variables. BURNSTAT is used to further breakdown the two data sets into three sub-groups each, according to the treatment received. Because the data set was designed with both species already separated, SPECIES was not used as a between-subjects factor in the analysis.

Since multiple measures are made on the same plot over time, repeated measures analysis uses special procedures that incorporate dependencies within an experimental unit (Norusis 1988). For example, the plots in the frequency data set were measured yearly for three years. Performing three paired t-tests, 1990 vs 1991, 1991 vs 1992, 1990 vs 1992 may seem to be the simplest analysis but in fact is not the best strategy. First, the three pairwise tests are not statistically independent; 32 and second, an overall test of the hypothesis that there is no difference between years is not available. To circumvent this problem, repeated measures transforms the original k dependent variables into k-1 variates. The variates represent linear combinations, sometimes called contrasts, of two original dependent variables that maximally separate them. Which two variables used depends on the contrast specified. These new variates are used only in the analysis involving the within-subjects variable, YEAR. In addition, a contrast corresponding to the overall mean of the dependent variables is always formed. This contrast is used to analyze between-subject factors.

There are many transformations available. The one utilized in this study is the difference contrast, which compares each level of a factor to the average of the levels that precede it. Thus the transformations in the frequency data, for example, creates two variates. One based on the difference of 1991 to 1990, and one based on the difference of 1992 to the average of 1991 and 1990. The names of these variables are called ADIFOV1 and ADIF01V2 for A. elatius and DDIF01v2 and DDIF01v2 for D. californica. Table 5 shows the orthonormalized transformation matrices from MANOVA used to create the new variates from the original dependent variables. The first 33 contrast for each species is the sum of the original variables, and represents the average response of over all years. It is used in tests involving the constant and the between-subjects factor, BURNSTAT. The next columns show the transformed variates used in tests involving the within-subjects variable YEAR and its interaction with

BURNSTAT.

Analysis proceeds in a stair-step fashion with multivariate testing preceding the univariate tests. Because both the multivariate and univariate tests are used, half the experimentwise level of significance is used for each test, as recommended by Stevens (1986). Instead of testing at the .05 level, the alpha level is set at the .025 level.

The first factor analyzed is the between-subjects variable BURNSTAT. Both species are analyzed simultaneously in the multivariate test to determine if each species, A. elatius and D. californica, responds differently to treatments. Univariate tests that follow determine which species, if any, had significant differences between treatment groups. The null hypothesis is that for each species, average response does not differ between treatments. 34

Table 5. Orthonormalized transformation matrices showing dependent variables on left margin and transformed variables along upper row. Two matrices, one for the frequency analysis and one for the cover analysis, are shown.

Transformed Variables for Analysis of Data

Frequency ACONS ADIF0V1 ADIFO1V2 DCONS ADIF0V1 DDIF01V2 AREL90 .577 -.707 -.408 .000 .000 .000 AREL91 .577 .707 -.408 .000 .000 .000 AREL92 .577 .000 .816 .000 .000 .000 DACA90 .000 .000 .000 .577 -.707 -.408 DACA91 .000 .000 .000 .577 .707 -.408 DACA92 .000 .000 .000 .577 .000 .816

Cover ACONS ADIFO1V2 DCONS DDIF1V2 AREL91 .707 -.707 .000 .000 AREL92 .707 .707 .000 .000 DACA91 .000 .000 .707 -.707 DACA92 .000 .000 .707 .707 35 The next step in the analysis is testing the within- subjects variable YEAR, and its interaction with BURNSTAT. These tests are conducted on the k-1 transformed variables, not the original dependent variables. The first effect tested is the BURNSTAT by YEAR interaction. The multivariate test checks the responses of both A. elatius and D. californica simultaneously to changes occurring within treatment groups, from one year to the next. In effect, this test compares the effect of no burning to spring burning to fall burning across years. The univariate tests that follow analyze the transformed variables separately to determine for which species and across which years significant changes occurred within treatments.

An overall test for the YEAR effect is also conducted. This effect tests the hypothesis that no significant changes occurred, for either species, from one year to the next, irrespective of treatment group membership. The transformed variables are used again to test this hypothesis. The multivariate test looks at both species simultaneously while the univariate tests look at the transformed variables of both species individually.

The last part of the repeated measures analysis reported are AVERAGED tests of significance. These tests 36 are done on the same within variable effects, YEAR and the BURNSTAT by YEAR interaction. The difference is that instead of using the transformed variables to test for significance, as before, the tests are based upon the pooled error sums of squares and associated degrees of freedom of the transformed variables for each species. For example, for A. elatius, the hypothesis and error sums of squares for ADIFOV1 and ADIF01V2 are pooled to yield a single univariate test of significance. As a result, these tests are more general than the tests on the transformed variables. For this reason, the reporting of these tests in the results section will precede the tests of the transformed variables.

The results of the AVERAGED tests determines if the analysis should proceed to individual testing of the transformed variables. Decisions were made for A. elatius and D. californica separately. Thus if D. californica showed no significance in the AVERAGED univariate tests, then only A. elatius will be further tested.

If any of the univariate tests are significant, then post-hoc tests can be used to determine for which treatment and year significant differences exist for either or both species. The procedure used depends on how 37 severely the assumption of sphericity is violated in the repeated measures analysis (discussion on this and other assumptions of repeated measures analysis will follow). Stevens (1986) recommends using pairwise procedures because they are: 1) easily interpreted, 2) are quite meaningful, and 3) powerful. He believes the Tukey procedure is appropriate for repeated measures designs that either satisfy or do not severely violate the assumption of sphericity, that is e >.70 (Stevens 1986).

If e < .70, then a multiple dependent t test utilizing the Bonferroni adjustment to alpha (.05/[k(k-1)/2] where k is the number of treatments), is a more appropriate test. Cover data, in the analysis, did not violate the assumption of sphericity, consequently the Tukey procedure was used for post-hoc testing. Sphericity in the frequency data was not tenable thus multiple dependent t tests with the Bonferroni adjustment were utilized.

The assumptions for univariate repeated measures analysis are more rigorous than those for multivariate repeated measures analysis. As a result I will discuss the assumptions needed for the univariate approach. The assumptions are: 1) Independence of the observations 2) Normality

3) Sphericity (sometimes called circularity) 38 The first two assumptions are also required for the multivariate approach, but the sphericity assumption is not. In this study, random assignment of treatments to individual plots ensures independence of observations. Repeated measures is fairly robust against violations of normality, as are ANOVA and MANOVA (Steven 1986). The assumption of sphericity, however, merits special attention.

In a repeated measures design, a sufficient condition for the univariate model to be valid is that, for each effect, the variance-covariance matrix of the transformed variables used to test the within-subjects effect have covariances of 0 and equal variances. This assumptions is often called the symmetry condition or .sphericity. If it is not tenable, the F ratios from the univariate results may be positively biased. A significance test for the sphericity condition is reported in the SPSS output. Its observed level of significance is based on a chi-square approximation. If sphericity is tenable, then Mauchlys test for sphericity will not be significant. The extent to which sphericity is violated is reflected in a

parameter called epsilon 'E'. Violations of sphericity can be tempered by multiplying the degrees of freedom with

E, yielding an "honest" type I error rate. Results from several studies show that this approach keeps the actual 39 alpha close to nominal alpha. Sphericity was not tenable in tests involving frequency data, thus the value of epsilon were used to adjust the degrees of freedom. RESULTS

General Results

The late spring burn had a deleterious effect on D. californica. In all plots sampled, D. californica was severely reduced in both cover and frequency or eliminated. In contrast, A. elatius seemed unaffected by the spring burn. The results of the fall burn was less clear for either species. Resprouts were seen in all fall burned plots for both species within a week of the burns.

Repeated Measures Analysis of Mean Cover Data

Danthonia californica Hospital Pasture Site - The mean cover for control plots at this site showed a decrease from 43.2% cover to 39.7% cover from 1991 to 1992. Fall plots decreased from 29% to 23.6% mean cover. Spring plots decreased from 41.2% to 2.9% cover.

Stagecoach - Control plots at this site showed and increase from 32% to 35.4% mean cover. Fall plots decreased from 32.8% mean cover to 29.2%. Spring plots also decreased in mean cover from 29.4% to 0.4%.

40 41

Arrhenatherum elatius Hospital Pasture - Control plots at this site increased in mean cover from 17.8% in 1991 to 25.8% in 1992. Fall plots decreased from 28.9% mean cover to 27.2%. Spring plots decreased from 19.7% to 15.4% mean cover.

Stagecoach - Control plots at this site increased in mean cover from 26.8% in 1991 to 31.9% in 1992. Fall plots increased slightly from 26.2% mean cover to 26.6%. Spring plots decreased slightly from 29.2% to 29% mean cover. For each species, basic statistics were generated and shown by site, treatment, and year (Table 6).

Because the magnitude and direction of change for each species at the two sites were similar, individual site data for each species were pooled and analyzed (Figure 8). The pooled means for D. californica in spring plots were 35.3% in 1991, and 1.7% in 1992. For fall plots the pooled means were 30.9% in 1991, and 26.4% in 1992. For control plots the pooled means were 37.6% in

1991, and 37.5% in 1992. The overall mean cover for the entire population of D. californica sampled declined from

34.58% in 1991 to 21.86% in 1992.

42

Table 6. Cover values for both D. californica and A. elatius for each site & treatment.

HOSPITAL PASTURE

Danthonia californica

S.E. Mean Mean Std Dey Variance N Spring Plots 1991 41.15 5.47 17.30 299.23 10 1992 2.85 1.04 3.27 10.73 10

Fall Plots 1991 29.00 5.57 17.60 309.83 10 1992 23.60 5.22 16.51 272.60 10

Control Plots 1991 43.20 3.83 12.10 146.40 10 1992 39.60 3.43 10.84 117.60 10

Arrhenatherum elatius

S.E. Mean Mean Std Dey Variance N Spring Plots 1991 19.70 5.57 17.63 310.73 10 1992 15.35 5.82 18.41 339.06 10

Fall Plots 1991 28.85 2.47 7.80 60.89 10 1992 27.25 3.83 12.11 146.57 10 Control Plots 1991 17.75 4.51 14.28 203.79 10 1992 25.75 3.46 10.95 119.90 10

43

Table 6. Continued

STAGECOACH

Danthonia californica

S.E. Mean Mean Std Dey Variance N Spring Plots 1991 29.35 5.09 16.11 259.50 10 1992 0.45 0.19 0.60 0.36 10

Fall Plots 1991 32.80 4.59 14.53 211.01 10 1992 29.20 3.15 9.95 99.01 10

Control Plots 1991 32.00 4.08 12.91 166.67 10 1992 35.45 4.24 13.39 179.36 10

Arrhenatherum elatius

S.E. Mean Mean Std Dey Variance N

Spring Plots 1991 29.20 3.15 9.95 99.01 10 1992 29.00 4.98 15.76 248.33 10 Fall Plots 1991 26.20 2.32 7.33 53.68 10 1992 26.55 3.06 9.69 93.86 10 Control Plots 1991 26.80 4.62 14.62 213.84 10 1992 31.85 2.98 9.41 88.56 10 Figure 8. Mean percent cover estimates for D. californica from 1991 to 1992 for each of the three treatments. Data pooled from both sites. 45 The pooled means for A. elatius in spring plots were 24.5% cover in 1991 and 22.2% in 1992 (Figure 9). In fall plots, the pooled means were 27.5% cover in 1991, and 26.9% in 1992. For control plots, the pooled means were 22.3% in 1991, and 28.8% cover in 1992. The overall mean cover for the entire population of A. elatius increased slightly from 24.8% in 1991 to 26.O% in 1992.

Analyzed first was the overall effect of burning regime on cover. The multivariate result was significant (Table 7), indicating an overall effect. The univariate tests showed a strong treatment effect for D. californica (P < .001) but not for A. elatius (P > .05). Thus for California oatgrass, the treatments had an overall effect on its cover. Tests involving the BURNSTAT by YEAR interaction were analyzed next. The significant multivariate test indicates an overall interaction between treatments and year sampled for both species (Table 8). The univariate tests that followed determined which species changed significantly. For D. californica, the changes in cover within treatment groups from 1991 to 1992 was highly significant (P < .001). In contrast, A. elatius showed no significant changes (P > .05). Figure 9. Mean percent cover estimates for Arrhenatherum elatius from 1991 to 1992 for each of the three treatments. Data pooled from both sites. m

47

Table 7. Overall BURNSTAT effect for Cover Data. ACONS refers to A. elatius, DCONS refers to D. californica.

EFFECT .. BURNSTAT

Tests Involving Between-Subjects Effects. Adjusted Hypothesis Sum-of-Squares and Cross-Products

ACONS DCONS ACONS .028 DCONS .085 .776

Multivariate Tests of Significance (S = 2, M = -1/2, N = 27 )

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F Pillais .34159 5.87035 4.00 114.00 .000 Hotellings .50607 6.95847 4.00 110.00 .000 Wilks .66176 6.41987 4.00 112.00 .000 Roys .33149

Univariate F-tests with (2,57) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F ACONS .02757 1.65476 .01379 .02903 .47486 .624 DCONS .77646 1.87252 .38823 .03285 11.81781 .000

Table 8. BURNSTAT by YEAR Interaction. ADIF1V2 refers to A. elatius from 1991 to 1992, and DDIF1V2 refer: to D. californica from 1991 to 1992.

EFFECT .. BURNSTAT BY YEAR

Tests Involving Within-Subjects Effects Adjusted Hypothesis Sum-of-Squares and Cross-Products

ADIF1V2 DDIF1V2 ADIF1V2 .044 DDIF1V2 .130 .720

Multivariate Tests of Significance (S = 2, M = -1/2, N = 27 )

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F Pillais .60660 12.40727 4.00 114.00 .000 Hotellings 1.36026 18.70357 4.00 110.00 .000 Wilks .41467 15.48178 4.00 112.00 .000 Roys .56923

Univariate F-tests with (2,57) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F ADiF1V2 .04407 .56476 .02204 .00991 2.22397 .117 DDIF1V2 .71969 .54842 .35985 .00962 37.40029 .000

48

Table 9. Overall YEAR effect for Cover estimates.

EFFECT .. YEAR

Test Involving Within-Subjects Effects Adjusted Hypothesis Sum-of-Squares and Cross-Products

ADiF1V2 DDiF1V2 ADiF1V2 .005 DDIF1V2 -.053 .536

Multivariate Tests of Significance (S = 1, M = O, N = 27 )

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F Pillais .52565 31.02836 2.00 56.00 .000 Hotellings 1.10816 31.02836 2.00 56.00 .000 Wilks .47435 31.02836 2.00 56.00 .000 Roys .52565

Univariate F-tests with (1,57) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F ADIF1V2 .00519 .56476 .00519 .00991 .52421 .472 DDIF1V2 .53569 .54842 .53569 .00962 55.67624 .000 49

The overall YEAR effect was tested next. Results of the multivariate tests were significant (Table 9) indicating an overall change from 1991 to 1992. As with the previous two tests, significant changes were seen for D. californica (P < .001) but not for A. elatius (P > .05) .

On the basis of these results, further analysis for A. elatius was not needed since no significant changes occurred from 1991 to 1992. For D. californica, however, post-hoc analysis was needed to determine which treatments significantly differed from 1991 to 1992.

Post-hoc tests revealed no significant differences for either control or fall plots (Table 10, P > .05). Spring plots, however, did show a significant decrease in cover (P < .001). It is clear that for D. californica, the drastic decline in cover within spring plots was responsible for all significant repeated measures tests. 50

Table 10. Oneway ANOVA tests for Danthonia californica for each of the three treatments.

Effect..Danthonia californica by Year

Spring Plots

Sum of Mean F F Source D.F. Squares Squares Ratio Prob.

Between Groups 1 11289.6000 11289.6000 73.2928 .0000 Within Groups 38 5853.3000 154.0342 Total

17142.90Fall Plots Fallali

Sum of Mean F F Source D.F. Squares Squares Ratio Prob.

Between Groups 1 202.5000 202.5000 .9315 .3406 Within Groups 38 8261.1000 217.3974 Total 39 8463.6000

Control Plots

Sum of Mean F F Source D.F. Squares Squares Ratio Prob.

Between Groups 1 .0250 .0250 .0002 .9902 Within Groups 38 6193.2500 162.9803 Total 39 6193.2750 51

Repeated Measures Analysis of Frequency Data

Danthonia californica

Hospital Pasture - The means for the control plots at this site showed a decrease in frequency from 90.3% in

1990 to 85.8% in 1992. Fall plots decreased from 77.2% in

1990 to 63.1% in 1992. Spring plots decreased in mean frequency from 92.0% in 1990 to 24.2% in 1992.

Stagecoach - Control plots at this site showed a decrease in mean percent frequency from 78.7% in 1990 to

71.9% in 1992. Fall plots decreased in mean frequency from 81.9% in 1990 to 70.6% in 1992. Spring plots also decreased from 75.6% mean frequency in 1990 to 17.8% in

1992.

Arrhenatherum elatius

Hospital Pasture - Control plots at this site showed an increase in mean frequency from 61.2% in 1990 to 67.5% in 1992. Fall plots decreased in mean frequency from

79.7% in 1990 to 79.4% in 1992. Spring plots also decreased from 69.2% in 1990 to 66.7% in 1992.

Stagecoach - Control plots at this site increased in mean frequency from 79.2% in 1990 to 88.9% in 1992. Fall plots also increased in mean frequency from 77.2% in 1990 52 to 81.1% in 1992. Spring plots increased in mean frequency from 76.4% in 1990 to 77.2% in 1992. For each species, basic statistics by site, treatment, and year are shown in Table 11.

Because the magnitude and direction of change for both species at the two sites were similar, site data were pooled by treatment and analyzed (Figure 10). The pooled means for D. californica in spring plots were 83.8% in 1990, 79.6% in 1991, and 21.O% in 1992. In fall plots the pooled means were 79.5% in 1990, 76.1% in 1991, and 66.8% in 1992. The pooled means for control plots were 84.5% in 1990, 80.6% in 1991, and 78.9% in 1992.

For A. elatius, the pooled means for spring plots were 72.8% in 1990, 74.9% in 1991, and 71.9% in 1992 (Figure 11). In fall plots the pooled means were 78.5% in 1990, 80.7% in 1991, and 80.3% in 1992. In control plots, the pooled means were 70.2% in 1990, 73.5% in 1991, and 78.2% in 1992.

The multivariate test involving BURNSTAT was significant. Burning regime had an overall effect on both species (Table 12). The univariate test that followed was insignificant for A. elatius (P > .05). There was no 53 Table 11. Frequency values for D. californica and A. elatius for each site and treatment; 1990 to 1992.

HOSPITAL PASTURE

Danthonia californica

S .E. Year Mean Mean Std Dey Variance N

Spring DACA90 91.94 2.85 9.02 81.32 10 DACA91 87.22 2.53 7.99 63.80 10 DACA92 24.17 3.61 11.42 130.44 10

Fall DACA90 77.22 6.32 19.98 399.06 10 DACA91 75.28 6.06 19.17 367.66 10 DACA92 63.06 4.76 15.05 226.46 10

Control DACA90 90.28 2.59 8.20 67.26 10 DACA91 86.95 2.23 7.06 49.82 10 DACA92 85.83 2.91 9.21 84.76 10

Arrhenatherum elatius S.E. Year Mean Mean Std Dey Variance N

Spring AREL90 69.16 4.94 15.63 244.30 10 AREL91 72.78 3.20 10.13 102.56 10 AREL92 66.67 2.81 8.88 78.89 10

Fall AREL90 79.72 2.93 9.26 85.84 10 AREL91 81.11 3.23 10.21 104.25 10 AREL92 79.44 4.23 13.37 178.65 10

Control AREL90 61.11 5.12 16.20 262.35 10 AREL91 65.28 5.03 15.89 252.58 10 AREL92 67.50 4.78 15.10 228.13 10

54 Table 11. Continued

STAGECOACH

Danthonia californica

S .E. Mean Mean Std Dey Variance N

Spring DACA90 75.56 5.33 16.86 284.29 10 DACA91 71.95 5.55 17.54 307.65 10 DACA92 17.78 5.62 17.77 315.80 10

Fall DACA90 81.94 4.33 13.68 187.26 10 DACA91 76.94 4.42 13.98 195.56 10 DACA92 70.56 4.60 14.53 211.23 10

Control DACA90 78.66 2.72 8.61 74.06 10 DACA91 74.17 2.72 8.59 73.83 10 DACA92 71.94 3.03 9.57 91.66 10

Arrhenatherum elatius

S .E. Year Mean Mean Std Dey Variance N

Spring AREL90 76.39 3.19 10.08 101.59 10 AREL91 76.94 3.99 12.63 159.58 10 AREL92 77.22 4.68 14.80 219.13 10

Fall AREL90 77.22 6.12 19.37 375.01 10 AREL91 80.28 5.06 16.01 256.20 10 AREL92 81.11 4.42 13.97 195.11 10

Control AREL90 79.23 3.31 10.47 109.61 10 AREL91 81.67 3.64 11.50 132.32 10 AREL92 88.89 3.07 9.71 94.26 10 Figure 10. Frequency values for D. californica from 1990 to 1992 for each of the three treatments. Data pooled from both sites. Figure 11. Frequency values for A. elatius from 1990 to 1992 for each of the three treatments. Data pooled from both sites. m

57

Table 12. Overall BURNSTAT effect for Frequency Data. ACONS and DCONS refer to Arrhenatherum and Danthonia respectively.

EFFECT .. BURNSTAT

Tests Involving Between-Subjects Effects. Adjusted Hypothesis Sum-of-Squares and Cross-Products

ACONS DCONS

ACONS .550 DCONS .260 1.690

Multivariate Tests of Significance (S = 2, M = -1/2, N = 27 )

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F

Pillais .34039 5.84536 4.00 114.00 .000 Hotellings .46648 6.41411 4.00 110.00 .000 Wilks .67287 6.13450 4.00 112.00 .000 Roys .29554

Univariate F-tests with (2,57) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F

ACONS .54981 8.17658 .27490 .14345 1.91639 .156 DCONS 1.69032 7.77735 .84516 .13644 6.19414 .004 58 overall difference between the spring burned, fall burned, or control groups. For D. californica, however, the treatment groups were significantly different at the .01 level.

The multivariate test was significant for the AVERAGED BURNSTAT by YEAR interaction, indicating a significant difference between the way both species responded to the treatments over time (Table 13). A highly significant change for D. californica occurred (P < .001) but was insignificant for A. elatius (P > .05 after adjustment).

Results from the AVERAGED test of significance for YEAR effect shows multivariate significance (Table 14). A highly significant change occurred between years for D. californica (P < .001) but not for A. elatius (P > .05).

Because no significant differences were seen for any univariate tests involving A. elatius, further testing was unnecessary. Further testing of D. californica, however, was needed.

The next level of testing for D. californica, involved the same two effects, YEAR and the BURNSTAT by YEAR interaction. Unlike the AVERAGED tests of 59

Table 13. The AVERAGED BURNSTAT by YEAR interaction from 1990 to 1990.

EFFECT .. BURNSTAT BY YEAR

Tests involving 'YEAR' Within-Subject Effect. Adjusted Hypothesis Sum-of-Squares and Cross-Products

AREL DACA

AREL .145 DACA .597 4.114

AVERAGED Multivariate Tests of Significance (S = 2, M = 1/2, N = 55 1/2)

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F

Pillais .73942 16.71721 8.00 228.00 .000 Hotellings 2.44076 34.17068 8.00 224.00 .000 Wilks .28387 24.77268 8.00 226.00 .000 Roys .70646

Univariate F-tests with (4,114) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F

AREL .14494 1.67929 .03624 .01473 2.45992 .049' DACA 4.11391 1.73955 1.02848 .01526 67.40040 .000

1 Adjusted D.F is (3,80) and Significant F = 3.28.

Table 14. The AVERAGED YEAR effect from 1990 to 1992.

EFFECT .. YEAR

Adjusted Hypothesis Sum-of-Squares and Cross-Products

AREL DACA

AREL .077 DACA -.541 5.080

AVERAGED Multivariate Tests of Significance (S = 2, M = -1/2, N = 55 1/2)

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F

Pillais .75885 34.85052 4.00 228.00 .000 Hotellings 2.97664 83.34595 4.00 224.00 .000 Wilks .24939 56.63708 4.00 226.00 .000 Roys .74782

Univariate F-tests with (2,114) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F

AREL .07673 1.67929 .03836 .01473 2.60437 .078 DACA 5.07975 1.73955 2.53987 .01526 166.44834 .000 60 significance, however, these tests involved transformed variables. The variates used were DDIFOV1 and DDIF01V2. DDIF0V1 compares 1991 data against 1990 data. DDIF01V2 compares 1992 data against the average of 1990 and 1991.

Results for the multivariate test, BURNSTAT by YEAR interaction was significant (Table 15). The results of the univariate tests indicate that there was no significant change between 1990 and 1991 (P > .05). In 1992, however, a significant change occurred (P < .001).

For the YEAR effect, the multivariate test was significant (Table 16). Univariate F tests show that significant changes occurred for D. californica across all three years (P < .001).

Post-hoc analysis, utilizing multiple dependent t tests was used to determine for which treatment group and between which years significant changes occurred for D. californica. All paired t-tests were evaluated at the .05/3 level of significance.

Results from spring plots indicate that D. californica, significantly decreased in frequency every year (Table 17). Declines in fall plots were also significant across all years. Control plots significantly 61 decreased from 1990 to 1991 and the decrease from 1990 to 1992 was also significant. From 1991 to 1992, however, the decrease in frequency was not significant.

62

Table 15. BURNSTAT by YEAR Interaction of Transformed Variables. DDIF0V1 contrasts 1990 to 1991. DDIF01V2 contrasts the average of 1990-91 with 1992.

EFFECT .. BURNSTAT BY YEAR

Test Involving the Within-Subjects Effects Adjusted Hypothesis Sum-of-Squares and Cross-Products

DDIF0V1 DDIF01V2

DDIF0V1 .004 DDIF01V2 .092 4.110

Multivariate Tests of Significance (S = 2, M = -1/2, N = 27 )

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F

Pillais .75358 17.23101 4.00 114.00 .000 Hotellings 2.99150 41.13313 4.00 110.00 .000 Wilks .24971 28.03270 4.00 112.00 .000 Roys .74919

Univariate F-tests with (2,57) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F

DDIF0V1 .00351 .33331 .00175 .00585 .30005 .742 DDIF01V2 4.11040 1.40624 2.05520 .02467 83.30459 .000

63

Table 16. YEAR Effect for Transformed Variables. DDIF0V1 contrasts 1990 with 1991. DDIF0IV2 contrasts the average of 1990-91 with 1992.

EFFECT .. YEAR

Tests Involving the Within-Subjects Effect Adjusted Hypothesis Sum-of-Squares and Cross-Products

DDiF0V1 DDiF01V2

DDiF0V1 .235 DDiF01V2 1.067 4.845

Multivariate Tests of Significance (S = 1, N = 0, N = 27 )

Test Name Value Approx. F Hypoth. DF Error DF Sig. of F

Pillais .78674 103.29513 2.00 56.00 .000 Hotellings 3.68911 103.29513 2.00 56.00 .000 Wilks .21326 103.29513 2.00 56.00 .000 Roys .78674

Univariate F-tests with (1,57) D. F.

Variable Hypoth. SS Error SS Hypoth. MS Error MS F Sig. of F

DDiF0V1 .23512 .33331 .23512 .00585 40.20907 .000 DDiF01V2 4.84462 1.40624 4.84462 .02467 196.36970 .000

64

Table 17. Paired samples t-test for Danthonia californica. Tested at the .05/3 level of significance.

Spring plots

Paired samples t-test: DACA90 DACA91

Variable Number Standard Standard of Cases Mean Deviation Error

DACA90 20 83.7495 15.616 3.492 DACA91 20 79.5835 15.407 3.445

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

4.1660 4.459 .997 .959 .000 4.18 19 .001

Paired samples t-test: DACA90 DACA92

Variable Number Standard Standard of Cases Mean Deviation Error

DACA90 20 83.7495 15.616 3.492 DACA92 20 20.9745 14.904 3.333

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

62.7750 17.391 3.889 .351 .129 16.14 19 .000

Paired samples t-test: DACA91 DACA92

Variable Number Standard Standard of Cases Mean Deviation Error

DACA91 20 79.5835 15.407 3.445 DACA92 20 20.9745 14.904 3.333

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

58.6090 19.008 4.250 .214 .365 13.79 19 .000

65

Table 17. Continued.

Fall Plots

Paired samples t-test: DACA90 DACA91

Variable Number Standard Standard of Cases Mean Deviation Error

DACA90 20 79.5830 16.840 3.766 DACA91 20 76.1105 16.356 3.657

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

3.4725 4.117 .921 .970 .000 3.77 19 .001

Paired samples t-test: DACA90 DACA92

Variable Number Standard Standard of Cases Mean Deviation Error

DACA90 20 79.5830 16.840 3.766 DACA92 20 66.8060 14.904 3.333

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

12.7770 10.289 2.301 .797 .000 5.55 19 .000

Paired samples t-test: DACA91 DACA92

Variable Number Standard Standard of Cases Mean Deviation Error

DACA91 20 76.1105 16.356 3.657 DACA92 20 66.8060 14.904 3.333

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

9.3045 9.883 2.210 .804 .000 4.21 19 .000 66

Table 17. Continued.

Control Plots

Paired samples t-test: DACA90 DACA91

Variable Number Standard Standard of Cases Mean Deviation Error

DACA90 20 84.4700 10.121 2.263 DACA91 20 80.5565 10.076 2.253

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

3.9135 3.643 .815 .935 .000 4.80 19 .000

Paired samples t-test: DACA90 DACA92

Variable Number Standard Standard of Cases Mean Deviation Error

DACA90 20 84.4700 10.121 2.263 DACA92 20 78.8880 11.590 2.592

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

5.5820 4.786 1.070 .912 .000 5.22 19 .000

Paired samples t-test: DACA91 DACA92

Variable Number Standard Standard of Cases Mean Deviation Error

DACA91 20 80.5565 10.076 2.253 DACA92 20 78.8880 11.590 2.592

(Difference) Standard Standard 2-Tail t Degrees of 2-Tail Mean Deviation Error Corr. Prob. Value Freedom Prob.

1.6685 3.968 .887 .942 .000 1.88 19 .075 DISCUSSION

From the results it would appear that burning does not control, via reduction of cover or frequency, tall oatgrass. No significant changes were detected in any of the univariate F-tests. These tests, however, do not reveal possibly important trends in the data. Closer examination of cover and frequency values for A. elatius from 1991 to 1992, show a consistent increase in tall oatgrass in control plots for all sites (Figures 9 and 11). In contrast, slight decreases in tall oatgrass were detected for both spring and fall burned plots for both sites from 1991 to 1992 (Figures 9 and 11). These numbers suggest that burning restricts tall oatgrass from increasing cover and frequency. While these numbers are not significant, they are consistent with trends suggested from adjacent RNP monitoring plots. In these nearby monitoring plots, burning every three years has prevented the spread of tall oatgrass, while unburned monitoring plots have seen tall oatgrass increase from 0% to 35% cover in 9 years (USDI 1992).

For California oatgrass, the results of the investigation strongly suggest that burning in the spring results in its near elimination. It is unlikely that the precipitous decline in cover and frequency from 1991 to

67 68

1992 can be attributable to unknown environmental factors, because neither control or fall plots displayed similar declines. No effects from grazing or other disturbance were noted. The small declines in cover and frequency in fall plots were entirely consistent with trends detected in other long term monitoring plots sampled by RNP staff. D. californica declines slightly the first year after burning but quickly rebounds in the second year. Fall burning has no lasting deleterious effect on California oatgrass.

Why was D. californica so intolerant to spring burning and not fall burning? Why was A. elatius tolerant of both regimes? There are several clues which may point to the answers. The morphology of A. elatius as compared to D. californica may protect it from severe fire damage or death. A. elatius does not have the basal rosette of leaves characteristic of many bunchgrasses. The leaves of A. elatius are spread over the lower half of the culm with a density much less than D. californica. When the fire burned through A. elatius, the low leaf density allowed the heat to dissipate quickly. In addition, the absence of a basal rosette lowered the duration of the fire near the base of the plants. Heat and flames did not linger at the base, rather it moved steadily over the plant. The 69 result was that heat from the fire did not approach lethal temperatures.

The root and crown morphology of A. elatius is similarly well adapted to burning. The basal stem internodes, from which new growth occurs, are mostly buried under the soil. Heat from a passing fire may burn and damage the top layer of internodes. Indeed many of the upper stem internodes were killed by the passing fire. Those underneath and buried, however, were not killed and resprouted within a week of burning (personal observation). The fire was unable to damage the lower internodes because of their position under the soil surface.

The basal rosette of D. californica is much more dense than A. elatius and includes much of the previous years growth. As a result, the leaves concentrate the heat and prolong the fire near the base of the plant. In addition, root and crown morphology of D. californica may not adequately protect it from fire. The perennating zone is exposed above the soil surface. Fire passing over the plant destroys all the above ground biomass including the perennating zone. With its energy being primarily used for sexual reproduction, burning at this time causes maximum damage. 70

Fall burning, on the other hand, is conducted when plants are dormant and most of their energy concentrated in the roots. As a result, burning may not kill as often as spring burning. The small declines seen in D. californica are probably the result of burning after rain has caused some regrowth to occur. When a burn passes over, there is a small loss or decline that lasts for one growing season. Because fall burning is not planned in the same area every year, there is little chance that adverse effects from fall prescribed burning can become significant on a large scale for D. californica.

Although fall burning did not reduce the cover or frequency of A. elatius, it may be possible to use other methods in combination with burning to reduce its cover or frequency. Some of these other methods might include the use of herbicides, selective grazing, and/or the establishment of native perennial bunchgrasses via reseeding or transplanting after prescribed burning.

The use of herbicides to control A. elatius has been firmly established in the literature (Ayres 1981, 1985;

Birnie 1983; O'Keefe 1981; Samuel 1985). The most recent contribution to the literature was work done in Oregon by

Tanphiphat (1989). In this study, translocation of glyphosate to dormant corms was greatest when the 71 herbicide was applied after 4 to 5 leaves had emerged, generally after February. Concentrations used were 1.2 kg/ha and 2.5 kg/ha. If herbicides are used, the method of application will be an important consideration. Spraying can kill native species as well as the target species. Application of the herbicide, to selectively target A. elatius by utilizing its height, would be easiest. Whether sufficient translocation of the herbicide is possible without spraying needs investigation. After application, the area could be prescribed burned to remove any residual chemicals.

The use of selective or prescription grazing can also be used to control A. elatius and simultaneously promote D. californica. In studies reviewing the response of A. elatius to grazing, Pfitzenmeyer (1962) states that tall oatgrass is not tolerant to grazing. Gibson (1988) found that sheep grazing in dune grasslands caused a decline in the vigor of dominant, perennial grasses, in particular A. elatius. Hope-Simpson (1940), Wells et ál. (1976) and King (1977) all showed A. elatius to be an aggressive, tenacious, late successional species, after the cessation of grazing. Although grazing appears to reduce and control A. elatius, it does not eliminate it. A. elatius is polymorphic and can survive in areas subjected to 72 frequent grazing (Mahmoud et ál. 1975), albeit at lower density and cover values.

D. californica, on the other hand, appears to be tolerant of moderate grazing. Observations by Heady et ál. (1963) indicate that under moderate grazing, California oatgrass stools readily and forms resistant sod. Less desirable perennial and annuals decrease as the sod forms. Their observations also indicate that the cattle graze D. californica lightly during its flowering and seed shatter stage. They cite this observation as the reason why this species can survive under heavy grazing. They, nevertheless recommend that a management system to favor this species through grazing, defer grazing until after its seed has set. Amme (personal communication) also suggests using rotational grazing to encourage D. californica while simultaneously inhibiting A. elatius. Prairie areas can be grazed 'hard' for one full year and then 'rested' for several more years. Quantification of 'hard' and 'rested' merits further investigation. In this way changes will be more gradual over the long term.

Rotational grazing can be used in conjunction with fall prescribed burning. Once the desired levels of A. elatius and D. californica are achieved, fall prescribed 73 burning, could be used to prevent reinvasion of A. elatius, while maintaining D. californica.

Whatever the management scheme used, reseeding or transplanting with D. californica or any other native grass is desirable. Reseeding or transplanting can give native species a chance to out-compete non-native species such as A. elatius for space and light. Collections of native seed can be made in the Bald Hills. Although the delayed germination of D. californica makes it difficult for nurseries to generate sufficient amounts of seed, other native grasses such as Elymus glaucus or Bromus carinatus are more prolific seed producers and should be used. Tests evaluating the response of A. elatius to fall burning, herbicide use, grazing, and reseeding in all combinations would help determine the best course of action for park managers. CONCLUSION

A. elatius was not adversely affected by either spring burning or fall burning. Neither its frequency nor cover values changed significantly as a result of the treatments. Burning tall oatgrass, however, did prevent increases recorded in adjacent control plots. D. californica was adversely affected by spring burning. Dramatic drops in cover and frequency were recorded. In fall plots, estimates of cover did not significantly change, however, estimates of frequency decreased significantly. This decrease was small and is not expected to last for more than one growing season. Cover and frequency in control plots did not significantly change for D. californica. I concluded that fall burning can be used to maintain A. elatius at its present level. In conjunction with other techniques, fall burning may effectively maintain it at lower percentages of cover and frequency. Spring burning offers no advantage over fall burning in controlling A. elatius. The effect of spring burning on D. californica, however, is so dramatic that it should not be considered an option for burning in the Bald Hills prairies.

74 LITERATURE CITED

Ayres, P. 1981. Investigations on the Growth of Arrhenatherum elatius var. bulbosum (Willd.) Spenn with Reference to the Effect of Tillage, Autumn Regrowth and Reproduction by Seed. Association of Applied Biologists Conference. Grass Weeds in Cereals in the United Kingdom. 6:77-81. . 1985. The Response of Onion Couch (Arrhenatherum elatius var. bulbosum (Willd.) Schub & Mart.) to glyphosate and other foliage applied herbicides. Crop Protection 4(2):263-271. Beetle, A.A. 1947. Distribution of the native grasses of California. Hilgardia 17(9):309-357. Benson, J.R. 1983. Archaeological test excavations at four sites in Redwood National Park, Humboldt County. Redwood National Park, Arcata, California. Birnie, J.E. 1983. A preliminary study on the timing of glyphosate application for control of onion couch. Tests of Agrochemicals and Cultivars. Annals of Applied Biology 102. Supplement 4:108-109. Boissier, E. 1884. Flora Orientalis, Lyons. Chase, A. and Hitchcock, A.S. 1971. Manual of the Grasses of the United States: Second Edition. Dover Publications, Inc. New York. Cooper, D.W. 1960. Fort Baker ranges returned to champagne grasses. Journal of Range Management. 13:203-205. Daubenmire, R. 1968. Ecology of fire in grasslands. Advances in Ecological Research. 5:209-266. Davy, J.B. 1902. Stock ranges of northwestern California: notes on the grasses and forage plants and range conditions. U.S. Department of Agriculture, Bureau of Plant Industry Bulletin 12. Gibson, D.J. 1988. The Relationship of Sheep Grazing and Soil Heterogeneity to Plant Spatial Patterns in Dune Grassland. Journal of Ecology. 76:233-252. Grenier, K.H. 1989. Vegetation patterns in grasslands of Redwood National Park, California. M.A. Thesis, Humboldt State University, Arcata, California. 75 76

Hayes, J.F. 1985. An analysis of Redwood National Park artifacts. U.S. Department of the Interior, National Park Service, Redwood National Park, Crescent City, Ca. Heady, H.F., D.W. Cooper, J.W. Rible and J.E. Hooper. 1963. Comparative forage values of California oatgrass and soft chess. Journal of Range Management. 16(2):51-54.

Hickman J.C. Ed. 1993. The Jepson Manual. Higher Plants of California. University of California Press. Berkeley and Los Angeles, California.

Hitchcock, L. and A. Cronquist. 1973. Flora of the Pacific Northwest. University of Washington Press. Seattle, Washington. Hope-Simpson, J.F. 1940. Studies on the Vegetation of the English Chalk. VI. Late Stages in Succession Leading to Chalk Grasslands. Journal of Ecology. 28: 386-402. Kernick, M.D. 1978. Technical Data Sheet No.2 Graminae (=) Arrhenatherum sp. In: Ecological Management of Arid and Semi-arid Rangelands in Africa and the Near Middle East (EMASAR-Phase II). 4:63-75.

King, T.J. 1977. The Plant Ecology of the Ant-Hills in Calcareous Grasslands. H. Success on the Mounds. Journal of Ecology. 65:257-278. Lloyd, P.S. 1972. Effects of Fire on a Derbyshire Grassland. Community Ecology. 53:915-920. Mahmoud, A.S., Grime, J.P. and Furness, S.B. 1975. Polymorphism in Arrhenatherum elatius (L.) Beauv.ex J.& C. Presl. New Phytologist. 75:269-279.

Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and methods of vegetation ecology. John Wiley & Sons, Inc., New York. Munz, P. and D. Keck. 1959. A California Flora with Supplement. University of California Press, Berkeley, California. Norusis, M.J. 1988. SPSS/PC+ Advanced Statistical Version 2.O. SPSS, Inc. Chicago, Illinois. 77 O'Keefe, M.G. 1981. The control of perennial grasses by pre-harvest applications of glyphosate. Proceedings in the Association of Applied Biology Conference, Grass Weeds in Cereals in the United Kingdom. 6:137- 144. Pfitzenmeyer, C.D.C. 1962. Biological Flora of the British Isles: Arrhenatherum elatius (L.) J.& C. Presl. Journal of Ecology. 50:235-245. Saenz, L. 1983. Ouercus garryana woodland/grassland mosaic dynamics in northern Ca. M.S. Thesis. Humboldt State University, Arcata, California.

Samuel, A. M. 1985. Chemical and cultural control of Arrhenatherum elatius var. bulbosum. Aspects of Applied Biology. 9:273-279. Stevens, J. 1986. Applied Multivariate Statistics for the Social Sciences. Lawrence Erlbaum Associates. Hillsdale, New Jersey. Sugihara, N.G, and Reed, L.J. 1987. Vegetation Ecology of the Bald Hills oak woodlands of Redwood National Park. Technical Report 21, Redwood National Park. Tanphiphat, K. 1989. Biology and Control of Tuber Oatgrass. Ph.D Thesis, Oregon State University, Corvallis, Oregon. Trabaud, L. 1987. Experimental Study on the Effects of Prescribed Burning on a Quercus coccifera L. Garrigue: Early Results. The Role of Fire in Ecosystems, SPB Academic Publishing, 4:97-121. United States Department of the Interior. 1987. Statement for Management. Redwood National Park, Crescent City, California.

. 1992. Bald Hill Vegetation Management Plan, Redwood National Park. National Park Service, Redwood National Park, Crescent City, California. Veirs, S.D., Jr. and Goforth, D. 1988. A line point transect method for long term monitoring of shrub and grassland or forest understory vegetation and a personal computer program for data analysis. Unpublished draft. CPSU Davis, University of California, Davis, Ca. 78 Wells, T.C.A., Sheail, J., Ball, D.F. and Ward, L.K. 1976. Ecological Studies on the Porton Ranges: Relationships Between Vegetation, Soil, and Land Use History. Journal of Ecology. 64:589-626. APPENDIX A - Species List

79 80

Prairie plant species list

FORBS i Achillea millefolium L. Common Yarrow n Agoseris grandiflora (Nutt.) Greene Mountain Dandelion n Aphenes arvensis (Nutt.) Rydb. Lady's Mantle i Anagallis arvensis L. Scarlet Pimpernel n Apocynum androsaemifolium L. Dogbane n Aster chilensis Nees Aster i Bellis perennis L. English Daisy n Brodiaea elegans Hoover Harvest Brodiaea n Calandrinia ciliata (R. & P.) DC. Red Maids n Carex tumulicola Mkze. Foothill Sedge n Cerastium arvense L. Field Chickweed i Cirsium vulgare (Savi.) Ten. Bull Thistle n Claytonia perfoliata Donn. Miner's Lettuce i Daucus carota L. Queen Anne's Lace n minutum Lehm. Minute Willow i Erodium cicutarium (L.) L'Her. Stork's Bill n Eschscholzia californica Cham. California Poppy i Galium aparine L. Bedstraw n Galium triflorum Michx. Bedstraw i Geranium dissectum L. Cut-leaved Geranium i Geranium molle L. Cranesbill n Gnaphalium purpureum L. Purple Cudweed n Hemizonia congesta ssp. tracyi (Babc. & Hall) Keck Tarweed i Hypericum perforatum L. Klamath Weed i Hypochaeris radicata L. Hairy Cat's Ear n Juncus bufonius L. Toad Rush n Juncus tenuis Willd. Slender Rush i Linum bienne P. Mill. Narrow-leaved Flax n Lotus micranthus Benth. Bird's Foot n Lotus purshianus (Benth.) Trefoil Clem. & Clem. n Lupinus bicolor Lindl. Annual Lupine n Lupinus rivularis Dougl. ex Lindl. Riverbank Lupine n Luzula comosa E. Meyer Wood Rush n Madia gracilis (Sm.) Keck Gumweed n Marah oreganus (Torr. and Gray) Wild Cucumber T.J. Howell. n Perideridia kelloggii (Gray) Math. Kellog's Yampah n Plagiobothrys nothofulvus (Gray) Gray Pop-corn Flower i Plantago lanceolata L. English Plaintain 81

n Potentilla gracilis Dougl. ex Hook. Cinquefoil

n Prunella vulgaris L. Selfheal n Ranunculus occidentalis Buttercup

i Rumex acetosella L. Sheep Sorrel

i Sherardia arvensis L. False Bedstraw

i Silene gallica L. Common Catchfly

n Sisyrinchium bellum Wats. Blue Eyed Grass

i Sonchus asper (L.) Hill Prickly

Sow-Thistle

i Stellaria media (L.) Vill. Chickweed

i Taraxacum officinale Weber Common Dandelion

i Trifolium dubium Sibth. Shamrock

n Trifolium microdon H. & A. Valparaiso

Clover

i Trifolium subterraneum L. Subclover

n Trifolium variegatum Nutt.

n Triteleia hyacinthina (Lindl.) Greene White Brodiaea

n Vicia americana Muhl. American Vetch

i Vicia benghalensis L. Vetch

i Vicia sativa L. Common Vetch

n Viola adunca Sm. Western Dog

Violet

n Viola glabella Nutt. Smooth Yellow

Violet

n Wyethia angustifolia (DC.) Nutt. Mule's Ears

GRASSES

i Agrostis capillaris L. Colonial

Bentgrass i Agrostis stolonifera

var. major (Gaudin) Farw. Redtop

i Aira caryophyllea L. Silver Hairgrass

i Anthoxanthum odoratum L. Sweet Vernal

Grass

i Arrhenatherum elatius (L.) Beauv. Tall Oatgrass

ex J. & C. Presl

i Avena barbata Pott ex Link Slender Wild Oat

n Bromus carinatus H. & A. California Brome

i Bromus diandrus Roth. Ripgut Grass

i Bromus hordeaceus L. ssp. hordeaceus Soft Chess

i Cynosurus echinatus L. Dogtail

i Dactylis glomerata L. Orchardgrass

n Danthonia californica Bol. California

Oatgrass

n Elymus glaucus Buckl. Western Ryegrass

n Festuca californica Vasey California

Fescue

n Festuca idahoensis Elmer Idaho Fescue

n Festuca rubra L. Red Fescue

^ i Holcus lanatus L. Velvetgrass 82 i Lolium multiflorum Lam. Italian Ryegrass i Lolium perenne L. Perennial Ryegrass i Phalaris aquatica Canary Grass i Phleum pratense L. Timothy Grass i Poa annua L. Annual Bluegrass i Poa pratensis L. Kentucky Bluegrass i Vulpia bromoides (L.) S.F. Gray Six Weeks Fescue i Vulpia myuros var. Foxtail Fescue hirsuta (Hack.) Asch. and Grae. i Vulpia myuros (L.) K. Gmel. Rat's-tail Fescue

Nomeclature follows Jepson (1993). n = native species, i = introduced