Lack of native species recovery following severe exotic disturbance in southern Californian shrublands

Authors


Cathlyn D. Stylinski (fax 619 594 7831; e-mail catd@sunstroke.sdsu.edu).

Summary

1. Urban and agricultural activities are not part of natural disturbance regimes and may bear little resemblance to them. Such disturbances are common in densely populated semi-arid shrub communities of the south-western US, yet successional studies in these regions have been limited primarily to natural successional change and the impact of human-induced changes on natural disturbance regimes. Although these communities are resilient to recurrent and large-scale disturbance by fire, they are not necessarily well-adapted to recover from exotic disturbances.

2. This study investigated the effects of severe exotic disturbance (construction, heavy-vehicle activity, landfill operations, soil excavation and tillage) on shrub communities in southern California. These disturbances led to the conversion of indigenous shrublands to exotic annual communities with low native species richness.

3. Nearly 60% of the cover on disturbed sites consisted of exotic annual species, while undisturbed sites were primarily covered by native shrub species (68%). Annual species dominant on disturbed sites included Erodium botrys, Hypochaeris glabra, Bromus spp., Vulpia myuros and Avena spp.

4. The cover of native species remained low on disturbed sites even 71 years after initial exotic disturbance ceased. Native shrub seedlings were also very infrequent on disturbed sites, despite the presence of nearby seed sources. Only two native shrubs, Eriogonum fasciculatum and Baccharis sarothroides, colonized some disturbed sites in large numbers.

5. Although some disturbed sites had lower total soil nitrogen and percentage organic matter and higher pH than undisturbed sites, soil variables measured in this study were not sufficient to explain variations in species abundances on these sites.

6. Non-native annual communities observed in this study did not recover to a predisturbed state within typical successional time (< 25 years), supporting the hypothesis that altered stable states can occur if a community is pushed beyond its threshold of resilience.

Introduction

More than 36% of the world's inhabitable land is exposed to extensive human disturbance due to urbanization and agricultural activities (Hannah et al. 1994). These disturbances can be categorized as (i) modifications of natural disturbance regimes (e.g. altered fire and grazing frequency) and (ii) the introduction of novel mechanical disturbances such as building and highway construction, soil excavation and landfill. The latter can be described as exotic, as they are not part of natural disturbance regimes and may bear little resemblance to them. While natural disturbances are often essential for the maintenance of native species diversity (Pickett & White 1985), exotic disturbance can alter successional processes due to the loss of soil nutrients, microflora, native seed banks and proximate seed sources and to the rapid invasion of exotic weedy species (Allen 1988; D’Antonio & Vitousek 1992). Such alterations can lead to the persistence of early seral species (Prose, Metzger & Wilshire 1987), reduction of native species cover and richness (Hironaka & Tisdale 1963; Lathrop 1983; Waaland & Allen 1987) and alteration of ecosystem processes (Vitousek & Walker 1989) and disturbance regimes (Mack & D’Antonio 1998). When human disturbance is severe and exotic species dominate the seed rain, changes in the trajectory of successional processes can result in the complete vegetation-type conversion of native plant communities (D’Antonio & Vitousek 1992). One dramatic example occurs in the western US, where tillage and changes in native disturbance regimes, along with prolonged drought, have led to the conversion of many Artemisia tridentata shrub communities to grasslands dominated by a single exotic grass, Bromus tectorum (Mack 1981; Billings 1990; Whisemant 1990).

In some cases native species can recover after severe anthropogenic disturbances, but typically only when such activities have occurred for many years or when they match natural disturbance regimes (Denslow 1985; Hobbs & Huenneke 1992). For example, Mediterranean Basin regions have a long history of urban and agricultural disturbance; thus annual herbaceous species endemic to these regions, such as Erodium spp., Avena spp. and Bromus spp., are well-adapted to human activities and readily invade mechanically disturbed sites in other areas of the world where such perturbations are relatively recent (Naveh 1967; Groves 1986).

Human disturbances are common in the densely populated semi-arid shrub communities of the south-western US. However, successional studies in these regions have been limited primarily to natural successional change and the impact of human-induced changes to natural disturbance regimes (Wells 1962; Zedler, Gautier & McMaster 1983; Whisemant 1990; Haidinger & Keeley 1993). Although shrublands in these regions are resilient to recurrent and large-scale disturbance by fire, they are not necessarily well-adapted to recover from exotic disturbances (Grubb & Hopkins 1986). The goal of this study was to investigate the impact of severe exotic disturbances due to urban and agricultural activities on native species composition and abundance in the semi-arid shrub communities of southern California. Our preliminary observations suggested that native shrub communities have a low resilience to these disturbances and that various types of urban and agricultural disturbances may lead to an exotic annual community with low native species richness for a long period of time. To explore this observation further, we gathered detailed historical information and vegetation composition and soil data on sites last disturbed 1–71 years ago. These sites were compared with nearby sites not exposed to severe exotic disturbance to determine the impact of this disturbance on native species recovery.

Materials and methods

Site description

Severe exotic disturbance was defined in this study as construction, landfill operations, heavy-vehicle activity, tillage (including levelling, clearing and planting) and soil excavation resulting in destruction of plant biomass and removal or dramatic alteration of surface soil. These forms of disturbance are distinct from large-scale disturbance endemic to the study community (i.e. fire) in that they preclude resprouting from root crowns, significantly deplete or remove the seed bank and dramatically alter soil structure. The complete loss of surface soil at filled and excavated sites probably led to successional processes that approximated primary succession because upper soil horizons were removed or buried deeply below the rooting zone.

Aerial photographs, historical records and interviews with local land managers helped to identify study sites and determine the type of disturbance and the time since the abatement of exotic disturbance. To minimize possible bias, sites were chosen from maps and aerial photographs prior to physical inspection. All disturbed sites were (i) subjected to exotic disturbance (> 1 ha) and then abandoned and (ii) not exposed to fire or chronic disturbance (i.e. grazing, mowing or off-road vehicle activity) since abatement of exotic disturbance. An undisturbed site was selected as close as possible to each disturbed site. Undisturbed sites included two Californian shrub community types: (i) dense chaparral dominated by two evergreen shrubs (Adenostoma fasciculatum and Salvia mellifera) and (ii) open coastal sage scrub dominated by evergreen and summer-deciduous shrubs (Artemisia californica, Eriogonum fasciculatum, Rhus integrifolia and Malosma laurina). If two or more disturbed sites were within 100 m, only one complementary undisturbed site was sampled. The term ‘undisturbed’ is relative here. Each of the 12 sites designated as undisturbed was probably exposed to slight perturbations (e.g. foot traffic) but showed no evidence of severe disturbance or fire since abandonment of the nearby disturbed site.

In total, 23 disturbed and 12 undisturbed sites were selected in San Diego, California (Table 1). Nine disturbed sites were designated as ‘compacted’ and were disturbed by construction and activity of military tanks and heavy vehicles. Eight disturbed sites were either tilled only or tilled and planted with crops (‘tilled’). The six remaining sites were subjected to severe soil alteration (‘filled-excavated’). Eight disturbed sites had neighbouring coastal sage scrub communities, while 15 disturbed sites were near chaparral communities. The distance to the nearest seed source for disturbed sites ranged from zero (adjacent source) to 1500 m (areas separated by clearings, roads or other disturbances). All sites were located on well-drained gravelly loam except site 1, which occurred on well-drained, coarse, sandy loam. Because flat undeveloped land is rare in southern California, most undisturbed sites were located on slopes (10–30°). Conversely, construction and agricultural practices typically require levelling, so that all disturbed sites were by definition on level ground.

Table 1.  Description of disturbed sites. Sites 1–8 were near undisturbed coastal sage scrub communities. Sites 9–23 were near undisturbed chaparral communities     
Site no.Age*Disturbance categoryHistorySoil typeDistance to nearest seed source (m)§
  • *

    Years since abatement of exotic disturbance.

  • Unless otherwise noted, cultivated sites were planted with grain, tomato and/or bean crops for more than a decade.

  • Redding = Redding Association; Rd.Ol = Redding–Olivenhain Association; Ci.Fa = Cieneba–Fallbrook Association (US Department of Agriculture Soil Survey).

  • §

    Zero distance indicates an adjacent seed source.

Coastal sage scrub
11TilledTilled onlyCi.Fa0
221FilledFilled with native soilRd.Ol30
328FilledSealed landfillRd.Ol600
411ExcavatedExcavated and levelledRedding1250
512ExcavatedConstructed but unused landfillRedding300
614CompactedBuilding constructionRedding500
743CompactedHeavy foot and vehicle trafficRedding1500
843CompactedBuilding constructionRedding1500
Chaparral
95TilledCultivatedRd.Ol100
105TilledCultivated (for 2 years)Rd.Ol50
1111TilledCultivatedRd.Ol0
1216TilledCultivatedRd.Ol1300
1325TilledTilled onlyRedding0
1435TilledCultivatedRedding0
1535TilledTilled onlyRedding0
1613FilledFilled with sandRedding1000
1730FilledSealed landfillRedding500
1813CompactedHeavy vehicle trafficRedding0
1913CompactedHeavy vehicle activityRedding1000
2040CompactedBuilding constructionRedding0
2140CompactedHeavy foot and vehicle trafficRedding0
2243CompactedBuilding constructionRedding500
2371CompactedBuilding constructionRedding0

Field sampling

All sites were sampled over a 6-week period from mid-March to April 1993. Percentage cover of shrubs was estimated by canopy intercept along four 50-m transects randomly placed within a 100 × 100-m sampling area. Percentage cover of herbaceous species and shrub seedlings (< 40 cm in height) was visually estimated to the nearest 1% in 20 0·5 × 1-m gridded quadrat frames randomly placed along the four transects (five per transect). Additional species not sampled in quadrats but within the sampling area were noted. Species were grouped by growth form (based on size, life span and woodiness of taxon). Plant nomenclature follows Hickman (1993).

Five soil samples were collected randomly at each site. Soil was cored to a depth of 15 cm. Samples were air-dried, sieved with a 2-mm mesh and analysed for total Kjeldahl nitrogen, percentage organic matter (by ignition), acid-extractable phosphorous, electrical conductivity and pH (two-to-one water extract) (Miller & Keeney 1982). Soil texture was determined by the Bouyoucos hydrometer method (Klute 1986). Values were expressed based on oven dry mass (65 °C to constant mass).

Multivariate statistical analysis

A two-group linear discriminant function analysis (LDFA) was used to determine the separation of group centroids of disturbed and undisturbed sites based on a linear combination (canonical variate) of growth form and soil variables (BIOSTAT II; Pimentel 1993). Equality of group centroids was tested with the F statistical approximation of Wilks’λ (Rao 1952). Multiple correlation between growth form abundance and soil characteristics on each site was tested with a canonical correlation analysis (CCA, BIOSTAT II; Pimentel 1993). Variables that did not meet the assumption of normality and homogeneity of variance (or variance–covariance matrix) were either log-transformed or arcsin-transformed. Multivariate normality was approximated with Wilks’ test for univariate normality (Legendre & Legendre 1983). Homogeneity of the variance–covariance matrix was tested with Box's modification of the likelihood function test (Krzanowski 1988). Only those variables that met both assumptions were used in LDFA and CCA.

Results

Total soil nitrogen and percentage organic matter were significantly lower on disturbed sites than on undisturbed sites (Table 2), most notably on tilled and filled-excavated sites (Fig. 1a,b). Filled-excavated sites had higher pH than all other sites (Fig. 1c). Other measured soil variables were not significantly different between disturbed and undisturbed sites (Table 2) or within the three disturbance types (data not shown).

Table 2.  Comparison of site variables of undisturbed and disturbed sites (mean ± SEM). P-values are from t-test or Mann–Whitney U-test. NS is not significant at P < 0·10
Site variablesUndisturbed sites (n = 12)Disturbed sites (n = 23)P-value
Total soil N (g kg–1)1·3 ± 0·10·9 ± 0·10·006
% organic matter3·6 ± 0·22·9 ± 0·20·05
pH5·8 ± 0·26·4 ± 0·2NS
Electric conductivity (mS)0·1 ± 0·00·2 ± 0·0NS
Extractable P (mg kg–1)12·5 ± 6·925·0 ± 6·9NS
% sand60·9 ± 2·058·9 ± 3·2NS
% clay15·2 ± 1·317·7 ± 1·9NS
% native shrubs68·2 ± 6·815·6 ± 3·20·0001
% native shrub seedlings2·4 ± 0·60·9 ± 0·20·006
% native forbs14·7 ± 3·68·7 ± 1·90·06
% native grasses0·6 ± 0·41·3 ± 0·8NS
% exotic forbs12·2 ± 3·430·5 ± 3·10·0009
% exotic grasses22·8 ± 5·627·3 ± 3·7NS
% exotic shrubs0·1 ± 0·10·1 ± 0·1NS
Exotic species richness (ha–1)6·1 ± 1·010·0 ± 0·50·002
Native species richness (ha–1)18·9 ± 1·69·0 ± 1·00·0001
Figure 1.

Changes in (a) total soil nitrogen (g kg–1), (b) percentage organic matter and (c) pH on undisturbed (U), compacted (C), tilled (T) and filled-excavated (F-E) sites (mean + 1 SEM). See Table 1 for description of disturbance types. Different letters indicate significant difference (Scheffé's test; P < 0·05).

Nearly 60% of the cover on disturbed sites consisted of exotic herbaceous species (grasses and forbs), while undisturbed sites only had 35% exotic cover and instead were dominated by native shrub species (68% cover; Table 2). Mean cover of native shrubs and of native shrub seedlings was significantly lower on disturbed sites. Although the most prevalent forb, Hemizonia fasciculatum, was found on both sites (Table 3), cover of native forbs was also reduced on disturbed sites. Exotic forb cover was significantly greater on disturbed sites, with Erodium botrys and Hypochaeris glabra being the dominant forbs. Several exotic grass species, namely Bromus hordeaceus, B. diandrus, Avena barbata, A. fatua and Lolium multiflorum, also had greater coverage on disturbed than undisturbed sites; however, cover of exotic grasses was not significantly different on these sites overall. The most common grasses on both sites were B. madritensis ssp. rubens and Vulpia myuros. Cover of native grasses and of exotic shrubs was low on all sites. Cover of native and of exotic herbs was not significantly different among compacted, tilled and filled-excavated sites (data not shown).

Table 3.  Percentage cover of dominant herbaceous species (> 1%) on undisturbed and disturbed sites (mean ± SEM). Mean coverages are also presented for different disturbance categories
 SpeciesUndisturbed sites (n = 12)Disturbed sites (n = 23)Compacted sites (n = 9)Tilled sites (n = 8)Filled and excavated sites (n = 6)
Exotic forbs
 Erodium botrys4·2 ± 2·412·3 ± 3·018·0 ± 5·810·7 ± 3·95·9 ± 4·6
 Hypochaeris glabra4·4 ± 1·77·5 ± 2·17·6 ± 2·38·8 ± 5·05·8 ± 3·6
 Melilotus indicaNil2·2 ± 1·60·7 ± 0·7Nil7·2 ± 5·7
 Silene gallica0·3 ± 0·21·2 ± 0·70·2 ± 0·21·0 ± 0·72·8 ± 2·8
 Lactuca serriolaNil1·2 ± 0·8Nil3·5 ± 2·2Nil
Exotic grasses
 Bromus madritensis ssp. rubens7·7 ± 2·47·6 ± 1·37·3 ± 2·39·3 ± 2·45·8 ± 1·6
 B. hordeaceus2·5 ± 0·94·0 ± 1·45·6 ± 3·14·4 ± 1·70·9 ± 0·6
 B. diandrus0·1 ± 0·12·4 ± 1·50·0 ± 0·03·0 ± 2·15·3 ± 5·3
 Vulpia myuros6·6 ± 3·16·8 ± 1·67·8 ± 2·67·1 ± 2·64·6 ± 3·4
 Avena barbata0·9 ± 0·53·5 ± 1·02·6 ± 0·96·7 ± 2·30·6 ± 0·4
 A. fatuaNil1·6 ± 0·81·3 ± 1·01·2 ± 0·62·7 ± 2·7
 Lolium multiflorumNil1·0 ± 0·61·8 ± 1·3Nil0·9 ± 0·9
 Agrostis viridis4·3 ± 4·3NilNilNilNil
Native forbs
 Hemizonia fasciculatum1·9 ± 0·72·1 ± 0·83·0 ± 1·40·8 ± 0·42·6 ± 2·4
 Filago californica1·9 ± 0·80·4 ± 0·20·8 ± 0·4Nil0·5 ± 0·5
 Navarretia hamata1·2 ± 0·70·0 ± 0·00·1 ± 0·1NilNil
 Juncus bufonius var. bufonius1·0 ± 0·70·7 ± 0·31·3 ± 0·80·5 ± 0·40·3 ± 0·3
 Lotus sp.Nil1·1 ± 1·1NilNil4·4 ± 4·4
 Daucus pusillus1·0 ± 0·50·4 ± 0·20·7 ± 0·50·4 ± 0·3Nil

Although exotic species richness was significantly higher on disturbed sites, native species richness was twice as great on undisturbed sites (Table 2). Site 10 (farmed for only 2 years) had the highest richness and percentage cover of native shrubs of all disturbed sites (nine species and 24·5%, respectively). A total of 140 species was recorded on all study sites; a complete species list is provided in Davis (1994).

The most abundant shrub species on undisturbed sites were Adenostoma fasciculatum, Salvia mellifera, Artemisia californica and Eriogonum fasciculatum (Fig. 2). With the exception of E. fasciculatum (common on both disturbed and undisturbed sites), these species together made up less than 5% cover on disturbed sites. Even for the nine disturbed sites adjacent to shrub-dominated undisturbed sites (Table 1) a distinct transition in shrub abundance was readily apparent (e.g. Figure 3). Note that these sites bordered intact stands of both coastal sage scrub and chaparral and included tilled and compacted disturbances.

Figure 2.

Changes in percentages of dominant shrub species (mean + 1 SEM) on undisturbed (U), compacted (C), tilled (T) and filled-excavated (F-E) sites. Eriogonum fasciculatum and Baccharis sarothroides were the only native shrub species common on disturbed sites. Bars with an asterisk above them indicate significant difference within a species (Scheffé's test; P < 0·05).

Figure 3.

Site 3 was disturbed in 1917 by the construction of a 3230-ha military training facility and was abandoned in 1922. This 71-year-old site typifies species composition observed on disturbed sites in this study. It is dominated by European forbs and grasses including Erodium botrys, Hypocharis glabra, Bromus spp., Vulpia spp. and Avena spp. Only one native shrub, Eriogonum fasciculatum, was common. Note the distinct edge between the military facility site (left) and the relatively undisturbed chaparral community (right).

In addition to E. fasciculatum, the only other native shrub common on disturbed sites was Baccharis sarothroides (Fig. 2). While E. fasciculatum was prevalent on many compacted sites, B. sarothroides was most common on filled-excavated sites. Tilled sites contained both species. These results suggest that (i) E. fasciculatum may colonize more readily following exotic disturbances that do not remove upper soil horizons, and (ii) B. sarothroides is possibly more tolerant of disturbances that lead to poor soil conditions (i.e. low soil nitrogen and high pH of filled-excavated sites). In fact, site 16, which had the lowest concentration of soil nitrogen (0·03 g kg–1), was colonized only by B. sarothroides and Melilotus indica, a nitrogen-fixing exotic herb. Despite this possible relationship between soil characteristics and the abundance of these two shrub species, there was no significant multiple correlation between soil variables and plant growth form when tested with canonical correlation analysis (P = 0·39). Thus soil variables (when analysed simultaneously) did not adequately explain differences in growth form abundance on disturbed sites compared with undisturbed sites.

The group centroid of undisturbed sites was significantly different (P = 0·0002) from that of disturbed sites when nine site variables were analysed with LDFA. The single canonical variate created by LDFA was defined by increasing native shrub cover and, to a lesser extent, decreasing exotic forb cover (Table 4); undisturbed and disturbed sites were well separated along this gradient (Fig. 4). Note that soil variables were weakly correlated with the canonical variate. Compacted, tilled and filled-excavated sites did not separate out along the canonical variate (data not shown).

Table 4.  Linear discriminant function analysis of nine environmental variables for undisturbed and disturbed sites. Only one canonical variate (CV) is shown because there were only two groups. Its significance was determined with Barlett's χ2 estimate of Wilks’λ. Coefficients were standardized with means = 0 and standard deviations = 1. Large correlations indicate the importance of a variable in defining the canonical variate
Site variablesStandardized coefficients of CVCorrelation
% native shrubs0·7710·862
Total soil N (g kg–1)0·3880·384
% native shrub seedlings0·0730·339
% organic matter0·0200·245
% native forbs and grasses0·1070·232
% sand0·0040·053
% exotic grass–0·053–0·083
Electric conductivity (mS)–0·316–0·140
% exotic forbs–0·191–0·437
Eigenvalue2·110 
χ232·34 
Degrees of freedom9 
Probability0·0003 
Figure 4.

Linear discriminant function analysis plot of disturbed and undisturbed sites using nine site variables. Sites are plotted on the first canonical variate axis which is defined by increasing cover of native shrubs and decreasing cover of exotic forbs. Diamond and cross symbols are disturbed and undisturbed sites, respectively.

Using posterior probability of group membership (Geisser classification corrected for chance; Titus, Mosher & Williams 1984), two undisturbed sites were misclassified into the disturbed group, while two disturbed sites were misclassified into the undisturbed group. The two misclassified undisturbed sites had open overstorey canopies and a high cover of exotic herbs relative to native shrubs. One site was a chaparral stand with several vernal pools invaded by exotic herbs, and the other was a sparsely populated coastal sage scrub community on a steep slope. The misclassified disturbed sites (numbers 19 and 21) were both compacted sites and had very high cover of E. fasciculatum. Site 21 was also apparent in the LDFA plot (Fig. 4) as the disturbed site among a cluster of undisturbed sites.

Percentage similarity (Czekanowski 1909) between each disturbed site and its reference site was not significantly related to age (years since disturbance) when analysed with a linear regression (Fig. 5a). While some sites were more than 40% similar to adjacent undisturbed sites, this moderate similarity was due to the occurrence of E. fasciculatum and B. sarothroides on disturbed sites and to the high cover of exotic herbs on undisturbed sites (Table 2 and Fig. 2). Moreover, as age of disturbed sites increased, native shrub cover did not increase and exotic herb cover did not decrease even 71 years after abatement of exotic disturbance (Fig. 5b,c). Native shrub cover on disturbed sites was not related to seed source proximity (P = 0·39).

Figure 5.

A chronosequence of vegetation change on sites exposed to exotic disturbance analysed with linear regression analysis. (a) Age (years since abatement of disturbance) of disturbed sites vs. Czekanowski's percentage similarity (Czekanowski 1909); percentage similarity was calculated for each disturbed site and the nearest undisturbed site. (b) Age vs. cover of native shrubs on disturbed sites. (c) Age vs. cover of exotic herbs on disturbed sites. All relationships were not significant when tested with a linear regression (P < 0·05). Cross symbols are compacted sites, filled symbols are tilled sites, and unfilled symbols are filled and excavated sites

Discussion

Severe exotic disturbances dramatically affected succession and led to vegetation-type conversion of two California shrublands to exotic annual communities with low native species richness. In this study, the cover of native species on disturbed sites did not approach that of native communities even seven decades after the abatement of disturbance. A distinct edge was clearly apparent when disturbed exotic-herb communities were adjacent to undisturbed chaparral and coastal sage scrub communities. Of disturbed sites, the compacted areas had the highest cover of native shrubs but contained only one shrub species. Site 18 had the largest native shrub richness and is the most likely to return to its predisturbed state, possibly because it had been tilled for only 2 years prior to abandonment. However, this site was exceptional. There appeared generally to be no return to native species composition with time for sites disturbed by compaction, tillage, soil excavation and filling. These results agree with White (1966), who noted poor native perennial grass and chaparral shrub recruitment and a preponderance of exotic herbaceous species on an abandoned agricultural site in California that had not been grazed or burned for 27 years. In the same region, Stromberg & Griffin (1996) also reported stable communities of exotic annuals on abandoned cultivated fields, despite the presence of native perennial grasses on adjacent uncultivated sites. Likewise, Zink et al. (1995) found low native coverage in a coastal sage scrub community 11 years after construction of an underground water pipeline.

The role of soil variables in the establishment and persistence of exotic annual species was not clear in this study. Although some disturbed sites had lower fertility, measured soil variables were not sufficient to explain vegetation patterns on these sites. Furthermore, differences in soil characteristics on disturbed and undisturbed sites may have been the result of species composition (e.g. rapid turnover and low-fertility litter of annuals; Davidson, Stark & Firestone 1990) rather than the cause of it. Additionally, although not measured, soil compaction, reduced surface heterogeneity and lack of good microsites for seed germination on these mechanically disturbed sites may have limited native shrub seedling recruitment after disturbance.

Invasion by annual Mediterranean Basin herbs such as Erodium spp., Avena spp. and Bromus spp. may also reduce the resilience (as defined by Westman 1986) of Californian shrub communities to exotic disturbance. Mediterranean Basin annuals are successful invaders in semi-arid regions of the US because they have evolved under similar environmental conditions and with extensive human disturbances (Naveh 1967; Groves 1986). These species are excellent dispersers and can germinate before native shrub seedlings and perennial herbs (Schultz, Launchbaugh & Biswell 1955; Bartolome 1979; Keeley 1991), so they are probably the first to colonize disturbed sites. In contrast, excepting Eriogonum fasciculatum and Baccharis sarothroides, seeds of coastal sage scrub species may not disperse far from the parent plant (Schultz 1996). Limited seed dispersal together with loss of root crowns and soil seed banks on disturbed sites in this study probably contributed to low native shrub recruitment. Once established, Mediterranean Basin annuals grow rapidly, quickly depleting water, soil nutrient and light availability (Schultz, Launchbaugh & Biswell 1955; DaSilva & Bartolome 1984; Davis & Mooney 1985; Gordon et al. 1989; Eliason & Allen 1997). Both E. botrys and B. diandrus, common on disturbed sites in this study, were reported to reduce soil water potential and seedling growth of a Californian oak species (Gordon et al. 1989; Gordon & Rice 1993). Exotic grasses and forbs were also common on undisturbed sites in our study, constituting more than a third of average cover. Historical grazing, increased fire frequency prior to perturbation of nearby disturbed site (Zedler, Gautier & McMaster 1983; Minnich & Dezzani 1998) and atmospheric nitrogen deposition (Allen et al. 1997; Padgett et al. 1999) may have promoted invasion of undisturbed communities by exotic annual species.

Some Californian shrub species can establish successfully in dense annual communities. For example, Baccharis pilularis can invade abandoned pasture and annual grasslands under high moisture conditions (DaSilva & Bartolome 1984; Williams, Hobbs & Hamburg 1987). Likewise in this study, two native shrubs, B. sarothroides and E. fasciculatum, successfully established in some annual-dominated sites. Eriogonum fasciculatum is a common ‘weedy invader’ of disturbed xeric sites (Kirkpatrick & Hutchinson 1977; Steward & Webber 1981; Zedler & Zammit 1989; DeSimone & Burk 1992). Its tolerance of a wide range of moisture conditions (Steward & Webber 1981) suggests that it can compete successfully with annual species for limited soil water. Adenostoma fasciculatum also occurs along a wide moisture gradient (Steward & Webber 1981; Keeley 1986); however, its limited recruitment on severely disturbed sites in this study may be due to its partial dependence on fire for successful germination (Stone & Juhren 1953; Keeley 1991). While Salvia mellifera can also germinate without a fire-related stimulus, its seeds must be exposed to light (Keeley 1986); high cover of herbs on disturbed sites probably limited light availability. Like E. fasciculatum, Artemisia californica has been reported to colonize after less intense human disturbances, such as frequent burning, grazing and construction of fire breaks and road cuts (Wells 1962; Kirkpatrick & Hutchinson 1980; Zedler, Gautier & McMaster 1983; Haidinger & Keeley 1993). However, unlike B. pilularis, it has limited recruitment on abandoned annual-dominated pastures (Grubb & Hopkins 1986).

Patterns observed in this study are similar to those found in other woody ecosystems where exotic disturbance interacts with exotic biological invasions, resulting in conversion of an entire community type (Hobbs 1991; D’Antonio & Vitousek 1992; Hobbs & Huenneke 1992; Vitousek et al. 1997). Examples of conversion include Central American and South American dry and mesic forests to African perennial grasslands (Parsons 1972), Asian tropical forests to open grasslands dominated by Imperata cylindrica (Whitmore 1984), Sonoran desert woodland to monocultures of Cenchrus ciliaris (African buffelgrass; Yetman & Burquez 1994) and Great Basin shrub and perennial grass communities to annual grasslands dominated by Bromus tectorum (Mack 1981; Billings 1990; Whisemant 1990). Denslow (1985) proposes that communities such as these often do not recover from human disturbance when such activity is recent and unlike natural disturbance regimes. This provides an opportunity for better-adapted invasive species. For example, many tropical rain forests are adapted to small-scale disturbances (tree falls) and are converted to grasslands when disturbed by intensive and extensive agriculture (Gómez-Pompa, Vázquez-Yanes & Guevara 1972). As shown in this study, disturbed native-perennial communities in California can be replaced by Mediterranean annual grasslands, possibly because these annual species are better adapted to recover from agricultural and urban activities that remove root crowns and soil seed banks. In contrast, native communities are resilient to exotic disturbances when these events resemble historical disturbance regimes (Denslow 1985). For example, Canadian boreal forest is adapted to fire, large-scale budworm infestation (Choristoneura fumiferara) and high mortality and thus is also resilient to extensive logging (Denslow 1985). Tree species of temperate forests have large populations, extensive distributions and great seed longevity and consequently are better adapted to large-scale and intense disturbances than many tropical forests (Gómez-Pompa, Vázquez-Yanes & Guevara 1972).

Invasion by annual exotic species often accelerates natural fire frequencies, further promoting fire-adapted annuals (D’Antonio & Vitousek 1992). Increased fire frequency may promote replacement of woody species with annuals in Californian chaparral and coastal sage scrub communities (Zedler, Gautier & McMaster 1983; Westman 1986; Haidinger & Keeley 1993; Minnich & Dezzani 1998). However, in this study the absence of fire or grazing up to 71 years suggests that chronic disturbance is not necessary for the establishment or persistence of these exotic annual communities. Persistence of exotic species has been observed elsewhere on abandoned lands, including farmland in southern Florida (Doren, Whiteaker & LaRosa 1991) and coastal California (Stromberg & Griffin 1996), mines in Wyoming (Waaland & Allen 1987), military camps in arid regions of California (Lathrop 1983) and logged rain forests in the Far East (Whitmore 1984).

In conclusion, the present study supports the hypothesis that succession does not necessarily follow a trajectory towards a predisturbed state (Holling 1973; Connell & Sousa 1983; Allen 1988) if a community is pushed beyond its threshold of resilience or amplitude (Westman 1986). Once a threshold has been crossed to a more degraded state, a return to native composition and structure within a practical time scale is not likely (Holling 1973; Grubb & Hopkins 1986; Friedel 1991). Exotic disturbance is often the impetus of a shift to a new state (Denslow 1985). The persistence of annual species and the lack of shrub recruitment within typical successional time (< 25 years; Hanes 1977; Keeley & Keeley 1984) on disturbed sites in this study suggest that severe disturbances due to urban and agricultural activities surpass the amplitude of both chaparral and coastal sage scrub shrublands, converting them to exotic annual communities with occasional stands of Eriogonum fasciculatum and Baccharis sarothroides.

Acknowledgements

We thank the Staff Civil Engineer Department at NAS Miramar, the San Diego City and County Planning Departments and the landowners who provided historical data and access to sites. Valuable comments on the manuscript were provided by Michael Allen, Dave Bainbridge, Sandra DeSimone, Neil Stylinski and a very thorough anonymous referee. Support for this study was provided by grants from the California Transportation Department, California Native Plant Society, Rancho Santa Fe Garden Club and National Science Foundation Division of Environmental Biology.

Received 30 July 1998; revision received 26 April 1999

Ancillary