Management and Restoration of Grasslands on Yellow Island, San Juan Islands, Washington, USA

Peter W. Dunwiddie[1]

Abstract

A native grassland dominated by Roemer's fescue (Festuca idahoensis var. roemeri), great camas (Camassia leichtlinii), and a diversity of other forbs has been the focus of a variety of experiments on restoration techniques, as well as studies tracking ecological changes since 1981. Investigations in existing grasslands have primarily focused on responses of native and non-native species to prescribed burns. Following each of 3 burns, responses of different species are complex, varying in direction, magnitude, and duration. A second series of studies has focused on developing effective means for controlling and removing invading trees and shrubs, and on limiting non-native grasses and forbs. Methods have included a variety of manual, mechanical, and chemical techniques. We have also tested several approaches for restoring native grassland species in areas where they had been excluded by competing woody plant growth. Even when abundant native seed sources exist in close proximity, non-native species usually establish more quickly following removal of trees and shrubs, and continue to dominate for many years. Out-planting of propagated plants has proven most effective in rapidly re-establishing native species. Greatest success has been achieved in establishing a dense fescue matrix that excludes invasive species.

Key words: Camas, fescue, fire, herbicides, long-term studies

Introduction

Native grasslands, prairies, and oak savannahs are rare in the lowlands of Puget Sound and the Gulf Islands (Ecoregional Plan). Most were converted to agriculture, destroyed by development, or overgrown by trees and shrubs over the last 150 years (Hall and Crawford 1996). Most that remain are small fragments that have been degraded by invasive non-native species. Yellow Island is a 4.5 ha island purchased by The Nature Conservancy in 1980, and managed as a nature preserve. Native grasslands dominate the east and west ends of the island (Fig. 1). They are of particular ecological interest due to the rich diversity of forbs set in a matrix of predominantly native grasses.

Since its acquisition by the Conservancy, annual monitoring and research have primarily focused on the maintenance and restoration of grassland communities. Restoration experiments were begun in 1984 to limit the expansion of shrubs and trees into the grasslands. Subsequently, studies were initiated to develop methods for restoring grassland vegetation in areas that were predominantly occupied by woody species. Restoration began on a more extensive scale in 1998, and is continuing up to the present.

The studies conducted on the island's grasslands over the last 20 years provide insights into many aspects of the ecology, management, and restoration of this rare vegetation type. The long duration of several of the studies affords a unique perspective on community dynamics. This paper will present data from several on-going vegetation studies that examine impacts of prescribed burning, manual control of woody plants, as well as the use of several herbicides. Results of repeated treatments, and data on responses of individual species are synthesized to draw conclusions that may be applicable to restoration and management efforts in similar vegetation assemblages elsewhere in the Puget Sound - Gulf Islands region.

Study Site

The grasslands of Yellow Island are dominated by Festuca roemeri (Roemer's fescue), Camassia leichtlinii (Great camas), Ranunculus occidentalis (Western buttercup), and a diversity of other native forbs and grasses (Table 1). Several non-native species are also widespread but generally provide little cover, including Holcus lanatus (Velvetgrass), Hypochaeris radicata (Cat's ear), and Rumex acetosella (Sheep sorrel). Lichens and mosses offer significant cover as well in many areas, especially prior to burning.

The central portion of the island is dominated by a forest of Pseudotsuga menziesii (Douglas-fir) and Abies grandis (Grand fir), with a dense understory of Symphoricarpos albus (Snowberry) and Holodiscus discolor (Ocean spray). Other common woody plants on the island include Rosa nutkana (Nootka rose) and Prunus cerasus (Sour cherry). There are two Quercus garryana (Garry oak) trees on the island, neither of which have successfully reproduced. Invasive non-native shrubs are absent.

There are no resident mammalian herbivores on the Yellow Island. Deer occasionally swim over from adjacent islands, but do not remain due to the absence of fresh water. Mink and river otters are the only other mammals that are found in the terrestrial environment.

Figure 1. Aerial orthophoto of Yellow Island, showing locations of plots and transects reported in the text.

Table 1. Frequency of most common species in native grasslands on Yellow Island. (Non-native species indicated with asterisk.)

Species / % Freq.
Camassia leichtlinii (Great camas) / 100
Festuca idahoensis (Idaho fescue) / 97
Ranunculus occidentalis (Western buttercup) / 97
Vicia americana/sativa (Vetch) / 97
Cerastium arvense (Field chickweed) / 93
*Holcus lanatus (Velvetgrass) / 87
Lomatium nudicaule (Naked desert parsley) / 87
Achillea millefolium (Yarrow) / 80
*Hypochaeris radicata (Hairy cat’s ear) / 80
Castilleja hispida (Hairy paintbrush) / 73
Luzula campestris (Field woodrush) / 73
Erythronium oregonum (White fawn-lily) / 73
*Rumex acetosella (Sheep sorrel) / 67

It is likely that prior to settlement of the San Juan Islands by Europeans in the mid-1800s, Yellow Island was frequently visited by Native Americans. Several neighboring islands are known to have been major collecting sites for camas and other important native food plants (Gunther 1930). The abundance of Camassia leichtlinii, Fritillaria lanceolata (Chocolate lily), and other edible native plants on Yellow Island, as well as the presence of several shell middens and possible camas roasting ovens, suggest this island also received similar use. It therefore seems probable as well that Yellow Island was frequently burned, a practice that was often associated with maintenance of camas beds (Boyd 1999). Several large, open-grown Douglas-fir trees in excess of 250 years old exhibit multiple fire scars, confirming the past occurrence of at least occasional fire (Agee and Dunwiddie 1984).

The frequency of fires would have declined in the latter 19th century as Native American use of the area declined. As in other Puget Lowland prairies, woody plants began to invade the grasslands in the absence of frequent fire (Boyd 1999). An early 20th-century photograph of Yellow Island provides independent corroboration of significant changes in the island's vegetation over the last century (Agee and Dunwiddie 1984). It reveals a mostly open, herbaceous landscape, with scattered large trees. The central, forested core of dense trees and shrubs is not evident, and appears to have developed primarily in the last 60-80 years. Only one historical fire has been documented on the island, sometime in the 1930s.

Yellow Island was settled in the late 1940s by a couple who lived there until the 1970s. They built several small cabins and maintained a garden, but appear not to have significantly altered the natural successional processes that were occurring. There is no evidence that they introduced livestock, set fires, or cut trees to a significant degree. The cabins were built entirely from wood that washed up on the beaches, and there are no stumps to suggest a history of cutting.

Today the island attracts numerous visitors, especially during the spring flowering season, which extends from late March to early June. Impacts from foot traffic is slight, as visitors are limited to a few narrow trails. A resident caretaker interacts with the public, and performs some ecological restoration and maintenance tasks.

Monitoring and Research Methods

In 1981, three plots were installed in the grasslands to detect potential invasion by woody species, particularly Symphoricarpos albus. Data from two of these (Plots 1 and 2) are presented in this paper. Percent cover of all vascular plants, lichens, mosses, bare ground, and litter was recorded in a series of ten 1-m2 permanent quadrats installed in each plot. A wire was lowered 25 times into each quadrat, and all species that touched the wire at each pin placement were tallied. These plots later were incorporated into other studies, as described below.

With only a few interruptions as noted in the figures, all plots have been monitored annually. To minimize potential errors introduced by different observers, I have monitored all plots myself. Variations due to phenological differences have been reduced by conducting all monitoring during the 2nd or 3rd weeks of May.

Numerous additional plots have been installed since 1981 to investigate various management activities. All have been monitored using the same 1-m2 quadrats as in Plots 1 and 2. In 1984, a 2 x 5m plot was installed to track natural re-establishment of grassland vegetation following removal of three dense, fast-growing Pseudotsuga menziesii trees. This plot was monitored through 1997, and because of its location within the perimeter of the 1987 and 1996 burns, it also was exposed to these treatments as well.

In 1987, ten 1-m2 quadrats were established along a 90-m transect (Transect 1) that bisects the grassland on the eastern end of the island. This transect was set up to detect changes in response to a prescribed burn of ca. 1 ha that was carried out in late July of 1987. This was the first fire known to have occurred in this grassland in at least 50 years. Plots 1 and 2 were in a portion of the eastern grassland that was left unburned, and served as reference plots for this burn. A slightly larger prescribed burn of ca. 1.5 ha was conducted in 1996 that included all of the area burned in 1987, as well as an additional area that included Plot 1. Plot 2 continued to serve as an unburned reference. In 1998, just the quadrats arrayed along Transect 1 were burned for a third time. All three burns were conducted in late July or early August. Fire intensities were not recorded, but all resulted in nearly complete combustion of all fine fuels. No other treatments were carried out in these study plots during this period.

A transect of 35 contiguous 1-m2 quadrats (Transect 2) was established in May, 1998 to track the success of efforts to restore a forested area to grassland. Presence/absence was recorded for all species in each quadrat. Additionally, the number of individuals was recorded for several species. Prior to installation of the transect, most trees and all shrubs were cut and removed from a ca. ¼ ha unit, leaving an open canopy of large Pseudotsuga trees. Some of the thick needle and twig litter layer was removed in a burn in September, 1998, and much of the remainder was removed by raking. Most shrubs resprouted following cutting; these were treated with foliar applications or stump-painting of the herbicides Roundup or Krenite. Revegetation was begun in March, 1999, and consisted of out-planting nursery-grown Festuca roemeri plugs at a density of ca. 10/m2. Small numbers of forb plugs, primarily Ranunculus occidentalis, were also out-planted.

A 28-m transect of 0.2 x 0.5m quadrats spaced at 1m intervals (Transect 3) was established in 1998 to investigate techniques for removing invasive non-native grasses, primarily Dactylis glomerata (Orchardgrass) and Holcus lanatus. As with Transect 2, presence/absence of all species was recorded in each quadrat. In April, 1999, an herbicide (Poast) that selectively impacts a number of tall grasses was applied. Festuca roemeri plugs were out-planted in April, 2000 to help re-establish a graminoid matrix on the site.

Results

Fire Management

Responses of individual taxa to the three burns varied among one another in direction, magnitude, and duration. Following the 1987 burn on Transect 1, cover of Festuca roemeri declined significantly compared to the unburned plots, and did not recover to pre-burn levels for about 7 years (Fig. 2a). A similar but less pronounced decline occurred in 1996 following the second burn on this transect. Fescue cover had only partially returned towards pre-burn levels when the transect was burned for a third time in 1998. Cover was once again reduced following this burn, and has gradually increased since then. The first burn occurred in Plot 1 in 1996, where fescue exhibited a dramatic decline that was similar to the response observed following the first burn on Transect 1. Here, too, cover has subsequently increased at a rate of about 10 percent per year over the last 5 years.

A second dominant grassland species, Camassia leichtlinii, has exhibited a more varied response following burning (Fig. 2b). In the year following the first burn on Transect 1, camas cover increased, contrasting with the decreases that occurred that year in the two unburned plots. A similar pattern occurred again following the burn in 1998. However, in 1996, camas cover declined in both Plot 1 and on Transect 1 following the burn, as well as in the control plot. In the

Figures 2a – 2d. Percent cover of Festuca roemeri, Camassia leichtlinii, Ranunculus occidentalis, and Holcus lanatus in burn and control plots. Transect 1 (solid diamonds) – burns in 1987, 1996, 1998 indicated with black arrows; Plot 1 (open squares) – burn in 1996 indicated with clear arrows; Plot 2 (open triangles) – no treatment.

two cases where camas cover increased following burning, the response persisted for only 1 year. Photographs of the burned and control sites suggested that flowering also was stimulated in the burned site.

A third common native species, Ranunculus occidentalis, exhibited yet a different response to the three burns (Fig. 2c). In 1987 and 1998, there was no difference in buttercup cover between the burned and unburned sites. However, both cover and flowering (apparent in photographs) increased following the 1996 burn, especially in Plot 1.

The response of non-native species also varied between burns and among species. Holcus lanatus, a weedy perennial grass, occurs frequently in some of the grasslands (Fig. 2d). Cover of this species increased for a year or two following the 1987 and 1996 burns. However, in 1998, cover declined the year following the burn. Overall, this species exhibited considerable inter-annual variation. Rumex acetosella (Sheep sorrel) showed a similar pattern, increasing in cover following the first two burns, but not in 1998 (Fig. 3a). A third non-native species, Hypochaeris radicata, exhibited no consistent response following burning (Fig. 3b).