San Diego, California 92182-0057

2nd Interface Between Ecology and Land Development in Calfornia. J.E. Keeley, M.B. Keeley and C.J. Fotheringham, eds. USGS Open-File Report 00-62, Sacramento, California. In press.

Author for correspondence: E. B. Allen

email edith.allen@ucr.edu

tel. (909) 787-2123

Abstract. The coastal sage scrub (CSS) vegetation of southern California is becoming one of the most intractable vegetation types to restore. CSS is subject to weed invasion, fragmentation, frequent fire, nitrogen deposition, and other disturbances that have reduced the shrub density and increased the frequency of Mediterranean weeds. Natural recolonization of native shrubs and forbs back into disturbed CSS is often slow, especially where exotic weed abundance and disturbance frequency are high.

Where natural regeneration is slow, restoration will be necessary to return the native shrub and forb cover. Restoration trials with six species of shrubs showed that they would not establish from seed unless the Mediterranean annual grasses were removed. However, establishment of seedlings of Artemisia californica was successful if they were initiated as three month old greenhouse transplants. Transplant survival and growth rate were increased by weeding, and those transplants that survived the first growing season with grass competition survived into the second season. Thus weed management, including weed seedbank control, is the most critical factor for shrub establishment.

An additional problem related to weed invasion may be nitrogen deposition from automobile exhaust, which is up to 45 kg/ha/yr in the Los Angeles air basin. Anthropogenic N deposition causes increased weed productivity in other areas around the globe, and may in part explain the vegetation type conversion of CSS shrubland to annual grassland in southern California. Thus even our best efforts at restoration may fail in the most polluted areas, where we have measured extractable soil N up to 87 ug/g.

Reestablishing biodiversity is one of the most important goals of many restoration projects, but even the best stands of restored CSS have low biodiversity. CSS contains some 200 sensitive species, but restoration of many of these is limited by insufficient availability of seed, lack of knowledge of propagation, and of course the high cost associated with handling so many species. However, the greatest threats to biodiversity for CSS restoration are competition from weeds and possibly also elevated soil N in regions with N deposition. A restoration strategy for the region will include an understanding of what sites are most likely to respond to restoration, given problems inherent in areas that suffer from the highest air pollution and greatest weed invasion.

INTRODUCTION

The coastal sage scrub (CSS) vegetation of southern California is among the most intensively human-impacted vegetation types in the United States. For this reason it has become a focus for mitigation and restoration (Bowler 1990), driven by the legal requirements of the Endangered Species Act (Berger 1991). CSS contains about 200 sensitive plant species (Skinner and Pavlik 1994) and has lost both floral and faunal diversity in urban fragments (Soulé et al. 1988, Alberts et al. 1993). CSS has been subject to urbanization, agriculture, and invasion by exotic annuals to such a large extent that one estimate gives only 10% of the original vegetation in good condition (Westman 1981). Other estimates suggest approximately 40% of CSS has been lost to urban development and agriculture (Klopatek et al 1979), and a more recent estimate puts CSS loss at 66% (O'Leary et al 1992). The differences among the estimates may lie in the inability to estimate accurately how much of the natural potential vegetation is CSS. Alternatively, it may lie in the condition of the remaining vegetation, which is often very weedy. Estimates for various southern California counties indicate that up to 50% of the remaining CSS is degraded (O'Leary et al. 1992). Large expanses of CSS have been invaded by annual Mediterranean grasses and forbs (Freudenberger et al. 1987), and in many areas the shrub understory is dominated by exotic rather than native annuals (Stylinski and Allen 1999, Minnich and Dezzani 1998). The conversion of shrubland to Mediterranean annual grassland has occurred primarily in the last 60 years, with 90% loss in shrub cover close to urban areas in western Riverside County and replacement by annual grasses (Minnich and Dezzani 1998).

The large extent of degraded CSS, along with its diminishing land area and many sensitive species, has made it a target for restoration. While there are many restoration efforts in CSS, they are not so successful that they emulate the original vegetation (Bowler 1990). The objective of this report is to understand the ecological limitations to restoration of CSS. First we will examine the causes of vegetation type conversion in CSS, including anthropogenic disturbance, weed invasion, and their interactions. Then we will describe the processes of succession in CSS, especially as they have been modified by human influences. Succession is frequently used as a tool to promote restoration, so modifications in the successional process will necessarily require changes in management for restoration. Finally, we will focus on competition from weeds, inability to replace lost biodiversity, and nitrogen eutrophication as major limiting factors for restoration.

CAUSES OF VEGETATION TYPE CONVERSION IN CSS

For restoration to be successful, the causes of vegetation conversion must be known and remedied. The historic vegetation must also be known to set a restoration goal (White and Walker 1997). The causes of change in CSS are multiple; the most direct of these, urbanization and agriculture, have already taken 40% or more of CSS in southern California (Klopatek et al. 1979, O'Leary et al. 1992). Decline of the remaining CSS stands that are in poor condition is more complex to understand, and may be largely due to weed invasions. The invasion by Mediterranean annuals is given as a cause of loss of biodiversity in California grasslands (Pavlik et al. 1993), and may also be a cause for biodiversity losses in CSS. Invasive weeds in CSS include Mediterranean annual grasses in the genera Bromus, Avena, Vulpia, and Hordeum. Bromus rubens is especially widespread, either present or abundant in virtually all CSS stands across southern California (O'Leary pers. comm.). Other problem weeds of CSS include Cynara cardunculus, Carpobrotus edulis, and Chrysanthemum coronarium, which occur near the coast, species of Brassica, Erodium, and Centauria, Feniculum vulgare, Nicotiana glauca, and others. The restoration efforts reported here are in areas dominated by annual grasses because they are widespread and better represented in the published literature, although research and management on the other species is ongoing. Invasions of natural plant communities do not typically occur without some imposed disturbance (Fox and Fox 1986). Remnant stands of CSS have been subject to fragmentation, historic grazing by domestic livestock, increased fire frequency, and two major forms of air pollution, ozone and nitrogen deposition. Each of these causes of disturbance is briefly considered below.

Fragmentation by urbanization and agriculture may have promoted the spread of Mediterranean annuals in CSS. This is a naturally fragmented vegetation type, with patches surrounded by chaparral, native grassland, and oak savanna (Westman 1981). However, these adjacent native vegetation types would not promote the invasion of exotic species into CSS, as they themselves consist of native species. Many of them have herbaceous understory species in common with CSS. However, in some cases the surrounding native vegetation has become weedy, such as oak savannas with annual grass understorys that could act as seed sources for invasion to adjacent CSS. Urban weedlots and agricultural fields are primary sources of invasive weeds. CSS fragments surrounded by nonnative vegetation are subject to invasion of weeds, the extent of the invasion being related to the size of fragment, time since fragmentation, the biological characteristics of invading and invaded vegetation, and the disturbance regime (Alberts et al. 1993, Forman 1995). In one study, a pipeline corridor through the Santa Margarita Ecological Reserve was cited as the cause of weed invasion into the reserve (Zink et al 1995). Extensive fragmentation and disturbance corridors are relatively recent in their impact on southern California vegetation, compared to grazing which has occurred for several centuries.

Ecologists in southern California have discussed the impacts of grazing on CSS with little more than anecdotal information to back up competing hypotheses, since little published information is available. Burcham (1954) wrote that ranchers used fire in California sagebrush as a management tool, and may have induced past vegetation type conversion to annual grass by frequent burning. Fire was coupled with seeding by Bromus hordeaceous and other annuals in some areas. Some CSS may have been converted to grass long before local botanists and naturalists took note. However, the most intensive sheep and cattle grazing ended by the turn of the century (Robinson and Risher 1993), and any large scale effects of grazing were over by then. Shrub reinvasion has been observed where grazing was excluded (Wells 1962), so many grazed areas may have recovered after the animals were removed. However, not all areas that are dominated by exotic grasses returned to their original shrub vegetation (Freudenberger et al. 1987, Stylinski and Allen 1999 and the current vegetation of historically grazed shrublands may be the product of past grazing. For instance, the historically grazed Motte Rimrock Reserve and Sycamore Canyon Parks in western Riverside County have both large stands of exotic annual grasses and patchy shrublands with grass understorys. Other areas that were ranches in the past and are reserves today, such as the Western Riverside Multispecies Reserve or the Audubon Starr Ranch, have extensive stands of shrubland with primarily native species in the understory--but also extensive stands with weeds in the understory.

In addition to direct effects of vegetation consumption and trampling, grazing animals also disperse seed (Burcham 1954, Malo and Suarez 1995). It is likely that grazers dispersed the weed seeds at a more rapid rate and to greater distances than other dispersal modes. Mediterranean annuals were already widespread in California grasslands by the mid 1800's and likely even earlier (Heady 1977). Because CSS was grazed by the earliest European settlers, weed seed also likely arrived in this vegetation type early on. However, the extensive type changes in CSS occurred primarily after the 1930's, after the heaviest grazing had been removed (Minnich and Dezzani 1998). These observations were based on the Forest Service Vegetation Type Map historic data from the 1930's and on aerial photographs from that time (Minnich and Dezzani 1998). The grazing animals may have "primed" the system by assuring that seeds were widely dispersed, but the extensive reproduction of weeds within a site may have continued after cattle and sheep were removed. Weed abundance may have increased in recent decades simply because it took the annuals several decades to become locally abundant after their seeds were dispersed. However, other factors in addition to time may have interacted to promote their reproduction and increase in local abundance in recent decades. Two additional important factors to consider for this region are fire and air pollution.

Increased fire frequency is another cause of decline of shrublands. A number of studies noted that short fire intervals cause conversion of shrubland to grassland (Zedler et al. 1983, Malanson 1985, Calloway and Davis 1993). The normal fire interval in CSS is about 30 years (Keeley and Keeley 1984, Westman and O'Leary 1986), but when the fire frequency becomes 2-3 years, the shrubs cannot regenerate. Urban parks in western Riverside Co., such as Box Springs Mountain and Mt. Rubidoux, that were dominated by CSS only 20 years ago are now largely annual grasslands because of fires that burn at 2-3 year intervals (Minnich 1988 and pers. comm.). Thus frequent fire can be blamed for the loss of shrubland in urban reserves where ignition sources are frequent. However, most remaining CSS does not burn this frequently, and even areas that have not burned in 30 years have experienced shrub loss (Minnich and Dezzani 1998). This leaves air pollution as the final discussion point to explain shrub loss.

Air pollution in southern California has two major components, ozone and nitrogen oxides. The ozone response of conifers has been well studied because they suffered mortality in the Transverse Ranges adjacent to the Los Angeles Air Basin (Miller et al. 1963). CSS shrubs are less sensitive to ozone than conifers (Westman 1990), but still responsive to ozone injury. Well-watered seedlings of CSS shrubs had reduced growth with 24 hr exposures of 150 ppb ozone, while lower exposures to ozone caused no measurable effects (Westman 1990). In another study, populations of Bromus rubens that were collected from areas with a history of ozone exposure showed no sensitivity to the same level of ozone, while populations from areas with clean air were sensitive (Westman 1991). These studies would suggest that ozone could contribute to CSS decline. But in fact, ozone levels are at their highest in summer when CSS shrubs are dormant, and are lower in spring. Most of the shrubs shed 50-90% of their leaves in the summer, and would not be affected by summer ozone highs of 150 ppb. While well-watered seedlings in the greenhouse with open stomates would be affected by ambient summer ozone levels, dormant shrubs in the field are unlikely to be affected. Therefore, ozone is not likely to be the most important component of air pollution affecting CSS.

Nitrogen deposition has not previously been examined as a contributor to CSS decline, but up to 45 kg/ha/yr N are deposited in the mountains adjacent to the Los Angeles Basin (Bytnerowicz and Fenn 1996). Nitrogen deposition is known to cause vegetation type conversions to low diversity grasslands in other regions, notably the Netherlands (Bobbink and Willems 1987), which have up to 85 kg/ha/yr of N deposited. In southern California some 90% of the N arrives as dry fall (particulate and ion deposition to surfaces) during the dry mediterranean-climate summer. Of the total N deposited, 80-90% of this is nitrate that originates from internal combustion engines, the remainder is ammonium from agricultural origins (Bytnerowicz et al 1987). N deposition is so high that soils with more than 80 ppm extractable N occur in the polluted regions near Riverside during the dry season, an extremely high value that is unknown from natural wildland soils (Allen et al. 1996). In greenhouse experiments both the annuals and the shrubs had rapid rates of growth and N uptake at higher fertilization rates, but two shrub species, Artemisia californica and Encelia californica, began to die after 6-12 months in soil with 30-50 ppm N (Allen et al. 1996 and P. Padgett, unpublished). Although we have yet to explain this effect of high N on the shrubs, they apparently suffer from the physiological effects of high soil N. First individual leaves turn brown and die, then branchlets, then entire branches in the greenhouse. In the field we have anecdotal observations that shrub mortality in native stands is higher in polluted than unpolluted regions. This corroborates the findings that shrub loss since the 1930's was also greater in urban than rural areas (Minnich and Dezzani 1998). Thus among the several causes of loss of CSS in natural reserves, N deposition may be important in air-polluted regions. Nitrogen deposition would be most noticeable in those reserves that do not burn frequently or are not subject to other forms of frequent disturbance.

For restoration to be successful, N deposition will need to be reduced and N levels that have been built-up in soils will need to be immobilized. Along with ozone, concentrations of nitrogen oxides in the atmosphere are becoming lower (EPA 1997), although N eutrophication continues to be a problem. High levels of N deposition are less of a problem in the coastal than inland areas of southern California (EPA 1997), because onshore breezes move the pollution inland. Thus the inland sage scrub of Riverside County is likely more threatened by n deposition than the CSS vegetation near the coast. Nitrogen must be removed or immobilized for restoration of eutrophied soils, as discussed below.

SUCCESSION IN CSS

To understand the processes and limits of restoration of CSS, we must study successional processes in CSS. Many restorationists use succession as a tool to enhance their restoration activities, which can be termed passive restoration if natural succession is the only method used (Fig. 1). Conversely, in situations where natural processes will not return the system to the desired state, or where natural processes, such as recolonization, are too slow, active restoration is required. Succession in CSS has been studied primarily in the context of fire (Zedler et al 1983, Westman and O'Leary 1986, O'Leary 1990, Keeley and Keeley 1984). Recovery from fire that occurs at natural, decades-long intervals is rapid. A flush of native annual forbs (sometimes mixed with exotic grasses and forbs) during the first year or two is followed by recovery of native shrubs from the seed bank, seed dispersal, resprouting, and seeds produced by the resprouters. Ecologists would normally not propose to restore CSS after such a burn, because the natural regeneration is rapid enough. Active restoration is needed if natural processes will not return the desired vegetation, or if natural processes are too slow for the management goal.

Sometimes land managers seed exotic perennial grasses in a mistaken attempt to hasten reestablishment of vegetative cover (Conard et al. 1995). This is done on steep slopes at the edge of urban/suburban areas where soil erosion is feared, but the natural ability of native mediterranean shrublands to regenerate after fire is very high (Westman and O'Leary 1986). Experiments in chaparral have shown no conclusive evidence that seeding with exotic grasses either helps or hinders native vegetation establishment, or decreases soil erosion (Conard et al. 1995). Similar results are not available for CSS, so it is not known whether seeding exotic grasses will reduce the ability of native CSS shrubs to regenerate after fire.

Although CSS recovery from fire is rapid, recovery from other disturbances that cause type conversion to annual grassland is slow or does not occur. Some of the instances of this were cited above, e.g., the grasslands of Los Angeles and Riverside County have only been slowly recolonized (Freudenberger et al. 1987). The pipeline through the Santa Margarita Reserve was dominated by exotic annuals for 20 years (Zink et al. 1995). The first author of this paper revisited the reserve in 1997, and noted that Eriogonum fasciculatum is higher in density and cover than when Zink et al (1995) did their observations in 1992. However, Eriogonum is still virtually the only native shrub, and the understory is still dominated by exotic annuals. Thus it is premature to say the pipeline is dominated by native CSS vegetation. Succession has not returned this pipeline to its former undisturbed vegetation composition.

In a larger scale study of 23 anthropogenically disturbed sites throughout San Diego County, none of the sites were statistically similar to adjacent stands of native shrubs even 71 years after disturbance (Stylinski and Allen 1999). The study compared each disturbed site to an adjacent undisturbed site (Fig. 2), and concluded that the similarity index between disturbed and undisturbed sites did not increase over time. The sites included chaparral and coastal sage scrub, which both behaved similarly under anthropogenic disturbance. Two native shrub species were abundant on a few of the sites, Eriogonum fasciculatum which is a component of both chaparral and CSS, and Baccharis sarothroides, which colonizes natural edges of riparian disturbances. All 23 sites had a dense ground layer dominated by Mediterranean annual grasses and forbs. The disturbances included earthen fills, compaction, scraping, and tillage. They were all chosen as intensive or long term disturbances that largely removed the seed bank and also altered the soils. Thus, there was no evidence of traditional "Clementsian" succession to a prior vegetation type. The combination of loss of seed bank, invasive annuals, and possibly also altered soils appears to have caused a vegetation type conversion in CSS as well as chaparral in these 23 sites.

Certainly there are examples, both published and anecdotal, of CSS shrubs recolonizing disturbed areas after tillage or grazing (Freudenberger et al. 1987, Wells 1962). Superficially, these areas may even look like CSS if the shrub density is high enough, especially from aerial photographs. Some areas of San Diego County that were farmed in the early part of this century had shrub cover equal to surrounding unplowed hillsides after several decades (T. Scott unpublished aerial photographs). However, ground truth surveys have shown that the understory is not nearly as diverse as the surrounding undisturbed vegetation, and has a high cover of Mediterranean annuals (Scott, unpublished observations). Eriogonum fasciculatum and Baccharis sarothroides appear to be exceptional shrubs that can invade the dense annual grasses, as noted in other areas (Kirkpatrick and Hutchinson 1977, Williams and Hobbs 1989), but the resultant plant community cannot be described as native CSS. Even less disturbed CSS may have a high cover of exotics, such as nature reserves with high foot traffic. In fact, some of the "undisturbed" CSS stands that Stylinski and Allen (1999) studied were quite weedy, having a relatively high cover of exotic annuals. They differed from the disturbed in that native herbaceous species were also abundant in the understory and there was a higher diversity of native shrubs. CSS may often be a weedy vegetation type, even where it has not been subject to intensive disturbance. Large scale surveys have not been done to show where the best and poorest condition CSS occur, so it is not possible to speculate the extent and location of poor condition CSS. However, our preliminary observations during travels throughout the region suggest that the weedier stands of CSS tend to occur inland under higher N deposition as well as drier (lower humidity) summer conditions. This stands as a hypothesis to be tested using large scale surveys and remote sensing techniques. For the purposes of CSS restoration, we are interested in selecting reference areas as goals for restoration. If the native areas have a large proportion of exotic species, they do not serve well as a reference area. In addition, if natural stands are weedy, then it will be difficult for the restorationist to improve upon natural conditions. Restored lands may initially have a low proportion of weeds because of the efforts of restorationists, but how do we keep the weeds out without sustained and expensive effort?

LIMITATIONS TO RESTORATION IN CSS

During the last five years we have undertaken numerous restoration trials in CSS that will be used to illustrate the limits to restoration. Among these are studies on weed competition, restoring biodiversity, and problems of high soil nitrogen. These are the ecological limitations. Even before facing these there are political, economic, and social limitations. However, we will not consider these here, and will discuss only the ecological limitations.

Restorationists have long considered nurse plants for restoration, especially where the native species are too slow growing for rapid establishment to stabilize soil. One of the dilemmas of restorationists who work with steep slopes is that they must get both a rapid cover to prevent erosion, and establish native plants. Many native plant species are slow growing compared to the horticultural plants used to stabilize soil. A few years ago we were approached by the California Department of Transportation to determine whether Trifolium hirtum , a fast-growing, non-native legume, acts as a nurse plant to establish CSS vegetation on roadsides. Trifolium hirtum is used throughout the state to stabilize roadsides, but in southern California this annual typically persists only for a few years. Thus it seems like it should be a good nurse plant, in addition to the fact that as a legume that could contribute N to the poor soils often created by the road building process. We tested the interactions of Artemisia californica with T. hirtum, and learned that T. hirtum acted as a competitor even under the lowest density of one plant of each species per 25 X 25 cm (Marquez and Allen 1996). Competition from T. hirtum was so intense even at the lowest density that A. californica biomass was reduced by an order of magnitude. Noncompetitive, mechanical means must be taken where slope stabilization is an issue, such as mulching or furrowing, to assure the establishment of slower growing CSS shrubs and at the same time reduce soil erosion.

Transplanted seedlings suffered from competition in stands of annual grasses, and germinants from planted seed fared even more poorly. In a seeding experiment with six CSS shrub species (Eriogonum fasciculatum, Salvia mellifera, S. apiana, Lotus scoparius, Encelia farinosa, A. californica) there were no survivors unless the annual grasses were cleared (Schultz 1996), and the same pattern was seen for germinants of A. californica (Eliason and Allen 1997). Conversely, all six seeded species survived at rates of 1-5% of seedlings per 10 X 10 cm plot if the grasses and all other vegetation was cleared (Schultz 1996). The study by Schultz (1996) was done in Riversidean sage scrub, while that of Eliason and Allen (1997) was in Diegan sage scrub. However, both studies produced the same conclusions regarding the need to remove grasses if seeding, rather than transplanting, is the preferred technique for restoration. Weeding is seldom done adequately because of the high cost. In one example where weeding was part of the planning process, the California Department of Transportation removed all exotics for 2-3 years from planted stands at a mitigation site in Santee (John Rieger, pers. comm.). This resulted in a healthy stand of CSS shrubs with little herbaceous understory that has persisted for more than 7 years in this condition.

Results of these grass/shrub competition experiments help to explain why reinvasion of CSS species is so slow in many converted grasslands. The annual grasses are dense and highly competitive, and native seeds that disperse into these grasslands have a diminished chance of survival. Some researchers have hypothesized that they may invade natural disturbances, such as animal diggings. Comparisons between gopher exclosures and control plots showed no increase in shrub germinant survival with gopher digging (Schultz 1996). However, these were small scale plots measured over two growing seasons. In large annual grasslands over many years some establishment of shrubs would be expected, as was seen by Stylinski and Allen (1999) with the recolonization of Eriogonum fasciculatum. Schultz (1994) also measured the dispersal of shrub seeds into grassland, and of the six species only E. fasciculatum dispersed any seed at all beyond the canopy of the parent shrub. Of these six, E. fasciculatum has the only winged seed that might be wind dispersed. Seed dispersal is poorly known for CSS shrubs (Keeley 1991), but appears, like many other vegetation types, to be largely limited to the immediate surroundings of the parent plant. Thus competition from annual grasses is one of the greatest limitations to restoration of CSS shrubs, but these studies and other anecdotal restoration studies (Bowler 1990, Hillyard and Black 1987) indicate that shrubs can be reestablished if sufficient efforts are made.

A larger problem than shrub establishment may be restoration of the diverse understory of CSS. Most of the species declared sensitive by the California Native Plant Society are herbaceous annuals and perennials, including some fire following annuals (Skinner and Pavlik 1994). These have been included in restoration seed mixes, but with varied success. For instance, some large stands of Lasthenia spp. were seeded successfully at Crystal Cove State Park (Hillyard and Black 1987 and Hillyard, pers. comm.). At several sites throughout San Diego County, less common species such as Hemizonia fasciculatum, Eriophyllum confertifolia, and others were successfully seeded, or species including Chamaesyce albomarginata, Solanum douglasii, Gnaphthalium califonicum, and Heliotropum curassavicum, appeared in the soil seedbank from salvaged topsoil (Marquez, pers. obs.), but most restored CSS sites are largely dominated by shrubs with an understory of annual grasses.

For the most part, seeded sites contain only a fraction of the biodiversity of natural CSS stands, which may have some 60-70 species per hectare. The low diversity is in part related to the low numbers of species prescribed in the seed mix, but also because of competition from exotic annuals, possibly coupled with N deposition in some areas. Thus our main goal of restoring biodiversity is thwarted by the high cost and low availability of seed, the lack of mandate to use more than a few species in the seed mixture, and weed competition. Inability to collect, process, and germinate seed is another limitation, but California botanists have learned much about native seed germination (e.g., Keeley 1991), and even more unpublished information is available from local botanical gardens and firms that process and sell seed. To improve the chances of increasing diversity, the restorationist needs to take advantage of soil seed banks and of the ability of seeds to disperse naturally.

Even though some CSS sites have been nearly replaced by exotic annual grasses for 20 years or more (Minnich and Dezzani 1998), they may still have a native seedbank. Richard Minnich observed vegetation recovery after a May burn in the Box Springs Mountains, Riverside, compared to a fall burn. These mountains lost most of their native vegetation cover during the last 20 years, but primarily native species came up after the spring burn (Minnich, unpublished observations). This occurred because the current year standing crop of weed seed burned almost completely, since the seeds had not yet shattered from their seedheads. For the fall burn, the seeds had already shattered, and enough were in soil cracks and protected from the fire that they germinated in large numbers at the onset of the winter rains. The longevity of native seed banks is well known for CSS and chaparral, since these vegetation types have decades-long fire cycles. Thus the restorationist can take advantage of this knowledge to promote diversity by using prescribed fire in the spring. This poses two problems. One is obtaining permission for prescribed fire in a pyrophobic society, and the other is keeping the grasses from recolonizing over the longer term. This would require longterm maintenance, which is seldom available. Thus a high diversity stand may not be possible except in special circumstances where grass competition is controlled.

The second chance that the restorationist has to improve diversity is to take advantage of natural colonization. From the many examples given above, it should be clear that this can occur only if weed competition can be reduced, because only a few unusual species colonize dense stands of annual grasses. In a study of roadside restoration on Interstate 15 in San Diego County, we examined sites up to 18 years old to determine how many native species would colonize and increase stand richness. The advantage of using a roadside for an ecological experiment, is that topsoil is not saved and there is no seedbank of any species initially. Of 36 locations, no site had more than 12 species planted originally. After 10 years, a few sites had 15-16 additional colonizing species that were not among the planted species (Fig. 4, Allen et al 1993). These included such native annuals as Cryptantha intermedia, Filago arizonica, Plantago spp., and others. Not all sites had native colonizers, only those that were adjacent to native vegetation. Sections of the roadside that ran through the city or suburbs had only exotic colonizing species, which occurred throughout. Thus we can expect native species to colonize only in those restoration sties that are surrounded by native vegetation. Isolated restoration or mitigation sites will have low richness even with the greatest efforts, as do small isolated natural patches of vegetation in CSS and chaparral (Soulé et al. 1988).

Soils that have been eutrophied by nitrogen deposition may also prove a formidable task for restorationists. Obviously the first task would be to reduce the production of nitrogen oxides, and this is occurring to some degree (EPA 1996). However, the amount of nitrogen that is coming down is still at agricultural levels, and is not likely to decline rapidly. We still know relatively little about the effects of high soil N on shrub longevity, and in addition nitrogen oxides have direct detrimental effects on leaf cuticles (Wellburn 1990, Allen et al. 1996). Thus we would not choose those sites that have the highest N deposition for restoration sites, but the upper limits of N deposition for negative effects on CSS communities are not known.

In the Netherlands where the N deposition rate is as much as 90 kg/ha/yr, annual haying in the early summer has been used successfully to increase pasture biodiversity (Bobbink and Willems 1991). However, cutting later in the growing season was not as effective. This technique was successful because pastures have become dominated by one species of perennial grass native to Europe, Brachypodium pinnatum, which when removed allows other species to grow. Haying could potentially be used as a restoration technique in type-converted CSS, but only if done early enough in the season to remove the current year contribution to the seedbank. Once the shrubs are reestablished, haying is of course no longer possible.

Where soils have high available N, the overriding restoration goal is to immobilize it and make it unavailable to plants. In our semiarid system external additions of organic matter may help to solve the excess fertility problem by immobilizing the high levels of inorganic N. This was accomplished by using mulch with a high C/N ratio in a restoration plot on a pipeline disturbance with relatively high extractable N. Bark mulch, with high C/N, was more effective than straw mulch in promoting the growth of planted seedlings of Artemisia californica (Zink and Allen 1998). Bark also promoted growth of soil saprobic fungi more than straw or unmulched soil, and the fungi are responsible for immobilization of nitrogen. At this time the mechanisms by which N immobilization promoted shrub growth over grass growth are not clear. Changes in forms of N (ammonium vs. nitrate) with mulching, differential abilities of shrubs and grasses to take up different N forms, or ability to take up N in soils with lower concentrations of N, may all come into play.

Obviously it will not be possible to apply bark mulch to vast expanses of CSS that have excesses of nitrogen. It is possible that excess nitrate may eventually be lost, if air pollution can be decreased. For soil N to increase permanently, there must be an increase in organic matter plus soil microorganisms to bind the nitrogen. Our measurements of southern California soils do not show significant increases in soil organic matter in high N sites (E. Allen et al., unpubl.). Wedin and Tilman (1996), working in a mesic grassland, also have shown no accumulation of organic matter even after years of fertilization. The most likely reason for this is increased rates of mineralization in high N soils, which keeps carbon from building up in the soil. Thus many soils that have been impacted by N deposition will have higher levels of inorganic N, but organic N buildup will be slow and small. An exception to this is sites that are invaded by nitrogen fixing plants, such as Myrica faya in the poor volcanic soils of Hawaii (Vitousek and Walker 1989). These soils have permanently increased organic N, but in this case the source of N input is organic litter rather than inorganic N deposition.

There is a great deal of reason to hope that vegetation that has been altered by N deposition can be at least partially restored, because inorganic N may be lost from the system after inorganic input declines. Studies on N deposition and N fertilization can give us some clues about how restoration of these may begin. Available N and N mineralization rates are typically higher under grasses than shrubs (Ingham et al. 1989). Thus practices that slow mineralization and bind inorganic nitrogen, such as mulching, and planting with species that promote slower mineralization (Wedin and Tilman 1996) would push the system back to a lower available N . Thus revegetation efforts need to consider more than reestablishing the vegetation, they need to manage and restore N dynamics.

The overriding limitation to restoration of CSS is replacing the biodiversity. This will become increasingly difficult as competition from exotics becomes greater, and as nitrogen deposition possibly makes the exotics even more competitive. Limitations of seed dispersal and remnant native soil seed bank must also be taken into consideration if revegetation of a community high in richness is to succeed. Sometimes this will take a better understanding of community and ecosystem dynamics than we now have. For instance, studies of R. Minnich (unpublished) showed that native seed banks can be rejuvenated by timed burning that removes the current year exotic seed crop. Or it may require a better understanding of the nitrogen dynamics of native and exotic species than we have now. Restoration of CSS is a relatively recent activity, and as argued by Bowler (1990), we will not know whether we have truly restored CSS until it has matured to the point where we can reapply the natural fire regime. The resultant community must then respond to fire as does a natural community. We can use these limitations in a positive sense to choose those sites that have the best potential for restoration. For instance, sites with a potential native seed bank, sites that are adjacent to existing vegetation and can receive dispersed seed, sites with relatively low N deposition, and sites where weed invasion can be kept under control, are the best potential candidates for restoration. Sites that have more severe environmental problems may be chosen for political or social reasons as restoration sites, but these cannot be truly restored given the limitations discussed here.

ACKNOWLEDGMENTS

The research synthesized here was supported by the National Science Foundation Division of Environmental Biology, USDA Competitive Grants Program, and the California Department of Transportation.

Figure Captions

Fig. 1. Comparison of active vs. passive restoration.

Fig. 2. Similarity index (Czekanowski's) for adjacent disturbed and undisturbed CSS and chaparral sites ranging in 1-71 years since disturbance. Data from Stylinski and Allen (1999).

Fig. 3. Growth responses of Artemisia californica to different manipulated densities of Mediterranean annual grasses after one growing season in a field experiment. Data from Eliason and Allen (1997).

Fig. 4. Richness of native plant colonization over time on a revegetated California roadside. Data from Allen et al. (1993)

Fig. 1. Comparison of active vs. passive restoration.

Fig. 2. Similarity index (Czekanowski's) for adjacent disturbed and undisturbed CSS and chaparral sites ranging in 1-71 years since disturbance. Data from Stylinski and Allen (1999).

Fig. 3. Growth responses of Artemisia californica to different manipulated densities of Mediterranean annual grasses after one growing season in a field experiment. Data from Eliason and Allen (1997).

Fig. 4. Richness of native plant colonization over time on a revegetated California roadside. Data from Allen et al. (1993)