Establishment of an emerging generalist pathogen in redwood forest communities


Patricia E. Maloney (tel. +1 530 754 894; fax +1 530 752 5674; e-mail


  • 1Phytophthora ramorum (causal agent of sudden oak death) is an emerging generalist pathogen in coastal forests of California and Oregon, USA, that causes lethal stem infections on oaks (Quercus spp.) and tanoak (Lithocarpus densiflorus) as well as non-lethal foliar infections on a broad range of trees and shrubs.
  • 2We studied P. ramorum over its known range in coastal redwood forests to determine forest compositional variables that are important to its epidemiology within the geographical area that it has already invaded. Redwood forests are dominated by coast redwood (Sequoia sempervirens), tanoak and California bay laurel (Umbellularia californica).
  • 3A total of 120 permanent plots (500 m2 each) were established in redwood forests at 12 sites within the main epidemic area in California. Over 5000 trees were mapped and examined for the presence of P. ramorum during spring 2002 and resampled in spring 2003.
  • 4Mean incidence of P. ramorum across all plots was 0.17 ± 0.01 in 2002 and 0.24 ± 0.02 in 2003. The highest infection levels by P. ramorum were found on California bay laurel (range 0.42–0.69) and tanoak (0.32–0.45). The highest levels of mortality were associated with tanoak and ranged from 0 to 66%, with 62.5% of that mortality associated with P. ramorum infection.
  • 5Disease incidence above 30% was most often associated with bay laurel importance value. In plots with few bay laurel stems, high disease levels were associated with the presence of understorey tanoaks. Bay laurel and small tanoaks are thought to represent the main source of inoculum for further spread of P. ramorum.
  • 6Differential host mortality due to this emerging generalist pathogen will exert considerable influence on redwood forest dynamics, with potentially dramatic shifts in forest composition and structure and subsequent cascading ecological and evolutionary effects.


Current events have highlighted the importance of understanding generalist parasites as causes of diseases that afflict humans, livestock and wildlife species (Daszak et al. 2000). There are clearly differences in the ecology, epidemiology, population biology and evolution of parasites with a broad host range as opposed to those with a very narrow host range (Woolhouse et al. 2001; Holt et al. 2003). This is especially true in the case of emerging or invasive organisms. Recent work on emerging zoonotic diseases has shown the need to understand all aspects of parasite life histories and the interactions among the various host organisms in order to adequately predict outbreaks, assess impacts and devise health policy strategies (Keeling & Gilligan 2000; Ostfeld & Keesing 2000).

Generalist plant pathogens, indigenous or exotic, have received less attention in the ecological literature than zoonotic diseases. Recent work on the role of native plant pathogens in influencing plant community structure has mostly focused on host-specific interactions such as tests of the Janzen-Connell hypothesis, soil-feedback loops or evolution of host resistance (see citations in Gilbert 2002). Ecological research on exotic plant pathogens has also primarily dealt with host-specific organisms such as Phytophthora lateralis (cause of Port-Orford cedar root disease) (Jules et al. 2002). As with generalist animal pathogens, generalist plant pathogens may have different ecological and evolutionary influences on hosts and communities (Gilbert 2002; Brown & Hastings 2003). For example, endemic root rot pathogens of forest trees (e.g. Heterobasidion annosum, Phellinus weirii) often affect hosts in multiple genera, but result in differential mortality between species and consequently exert considerable influence on plant successional patterns (Hansen & Goheen 2000). In one exceptional case, an exotic pathogen with a broad host range, Phytophthora cinnamomi, has altered the structure and species composition in the jarrah forests of western Australia, over large geographical areas (100 000s of hectares), converting them to grassland or shrubland since it was introduced in the 1920s (Weste & Marks 1987).

A recent example of an emerging generalist plant pathogen is Phytophthora ramorum, causal agent of sudden oak death (Rizzo & Garbelotto 2003). P. ramorum causes lethal stem cankers on oak (Quercus spp.) and tanoak (Lithocarpus densiflorus) in coastal forests of California and Oregon, USA. It also causes non-lethal foliar and twig infections on nearly all woody plant species and several herbaceous plants in these forests. The pathogen is currently found over c. 750 km from the Big Sur area of California to southern Oregon, but the main population of P. ramorum is mostly established over c. 450 km around the San Francisco Bay area (Rizzo & Garbelotto 2003). Within this coarse scale distribution, the presence of P. ramorum across the landscape is patchy, with many invaded forests juxtaposed with non-invaded forests (Rizzo & Garbelotto 2003). Molecular population genetic studies of P. ramorum indicate that the population structure is typical of an invasive species in North America, with very limited genetic variation among populations and a single genotype dominating over the entire geographical range in California and Oregon forests (Ivors et al. 2004). In addition, only one mating type (A2) has been found in these populations (Ivors et al. 2004; Prospero et al. 2004).

P. ramorum is primarily a pathogen of above-ground plant tissues. Foliage and twig infections of non-oak hosts appear to play a key role in the epidemiology of P. ramorum by serving as a source of inoculum (Davidson et al. 2002). The most likely dispersal propagules of P. ramorum, sporangia and chlamydospores, are readily produced on foliage, but have yet to be found on infected oak bark (Davidson et al. 2002). Hosts with non-lethal foliar infections may be important in transmission of P. ramorum, because such leaves may support abundant sporulation during the rainy winter months (Davidson et al. 2005; P. E. Maloney unpublished), and allow the pathogen to survive dry California summers as most host species in coastal forests are evergreen. California bay laurel (Umbellularia californica) in particular has been associated with P. ramorum infection on oak in mixed evergreen forests (Kelly & Meentemeyer 2002; Swiecki & Bernhardt 2002), although tanoak leaves and twigs may also support sporulation (J. M. Davidson unpublished observations; P. E. Maloney, unpublished observations).

Because of its wide host range, P. ramorum has become established in several different forest types in California and Oregon, including mixed-evergreen forests (dominated by coast live oak, Q. agrifolia) and coastal redwood forests (dominated by Sequoia sempervirens) (Rizzo & Garbelotto 2003). To date most monitoring of P. ramorum has been in coast live oak woodlands (e.g. Kelly & Meentemeyer 2002; Swiecki & Bernhardt 2002), with fewer studies in redwood forests. In addition, no studies have specifically examined disease incidence on non-oak hosts. Our objectives were to determine which compositional and structural variables are important to the epidemiology of P. ramorum in redwood forest communities. Given the diversity of hosts in redwood forest communities we consider (i) how overall disease incidence changes from year to year, (ii) how P. ramorum incidence varies between host species, (iii) whether specific forest community structures contribute to pathogen incidence, and (iv) how much tree mortality is associated with P. ramorum in redwood forests.

Materials and methods

study sites

A total of 120 permanent plots were established in redwood forests in the coast ranges of California from Sonoma county in the north (122° W, 38° N) to Monterey county in the south (121° W, 36° N). This represents the geographical range of the redwood forest type where coarse scale surveys have most often detected P. ramorum (Rizzo & Garbelotto 2003). Plots were located in Jack London State Park (Sonoma County), Marin Municipal Watershed and Samuel P. Taylor State Park (Marin County), Henry Cowell and Big Basin State Parks (Santa Cruz County), and Pfeiffer Big Sur and Julia Pfeiffer Burns State Parks (Monterey County). Plots in Jack London State Park were established across 66 ha. The Marin County plots were located in six different locations over a total of 104 ha. In Santa Cruz County, plots were established in the two areas over a total of 59 ha. In Monterey County, plots were located at three sites over a total of 184 ha. All sampling areas were within the boundaries of the known range of P. ramorum, but the locations of individual plots within the redwood forest type were chosen without regard for presence/absence of the pathogen.

Redwood forests are mixes of conifers and hardwoods and include tanoak, Douglas-fir (Pseudotsuga menziesii), California bay laurel and madrone (Arbutus menziesii). Minor hardwood species include black oak (Quercus kelloggii), big leaf maple (Acer macrophyllum) and coast live oak (Zinke 1988). All of the above species can be infected by P. ramorum in nature (Davidson et al. 2003).

Community structure of the study sites was similar, with some variation in species composition from north to south, mainly due to the presence or absence of less dominant hardwood associates (Table 1). The highest importance values (IV) on our plots were for redwood (range 48–67%), followed by tanoak (18–26%) and California bay laurel (5–15%) (Table 1). Relative density for redwood ranged from 23 to 46%, for tanoak from 35–42% and for bay from 8–26% (Table 1).

Table 1.  Relative density, basal area (BA) and importance value (IV) of tree species ≥ 1.37 m tall and ≥ 1 cm d.b.h.
SpeciesSonomaMarinSanta CruzMonterey
DensityBAIV (%)DensityBAIV (%)DensityBAIV (%)DensityBAIV (%)
  1. Note: Importance values were calculated for each species as (relative density + relative basal area)/2.

Acer macrophyllum Big-leaf maple 3.1  2.4 2.7 0.4   0.3 0.3 1.6  0.2 0.9
Arbutus menziesii Pacific madrone 1.0  1.6 1.3 2.5   4.3 3.4 1.9  1.4 1.7
Lithocarpus densiflorus Tanoak40.0  10.826.835.5  2.318.935.4  7.221.3
Pseudotsuga menziesii Douglas-fir 4.3  4.9 4.6 3.7   5.1 4.4 1.4  6.4 3.9
Sequoia sempevirens Redwood23.7 73.248.441.7  77.759.740.8 86.463.646.3 87.867.1
Umbellularia californica California bay26.0  4.215.1 8.6   1.8 5.218.5  2.510.517.4  4.110.7
Other hardwood associates 1.7  5.4 3.5 0.12 < 0.01 0.06 0.2  0.7 0.4 0.09  0.6 0.3
Total density (inds. ha−1) 878  1065  738  671 
Basal area (m2 ha−1) 107    96  175  142 

When and how P. ramorum first invaded our study areas is not known. Tanoak mortality was first noted in Marin, Santa Cruz and Monterey counties in the mid-1990s (Rizzo & Garbelotto 2003). Therefore, P. ramorum has probably been in the vicinity of our plots, if not actually on trees within individual plots, for at least 10–15 years.

plot and pathogen sampling

At each study site, 30 plots (500 m2, i.e. 25.2 m diameter) were established with their location based on species composition that most closely resembled stands in the central and southern redwood forest subregions (Sawyer et al. 2000). Once stands were located at a site, a random starting point c. 25 m from the edge of a redwood stand or trail was chosen for the first plot. Additional plots were 100 m from the previous plot and were placed by randomly selecting a compass bearing of 0, 90, 180 or 270 from the initial location, so long as the bearing kept us within the redwood forest type. Plots at a number of the sites were not contiguous, as the vegetation type often changed within a few hundred metres. Locations of all plots were recorded using global positioning system (GPS) equipment (GARMIN, Olathe, KS, USA).

Within each plot, woody plant species were identified and diameter at breast height (d.b.h.) was taken for all individual stems ≥ 1.0 cm d.b.h. and 1.4 m tall. All tree positions (azimuths and distance) from the plot centre were recorded and mapped. Plant status was noted (live or dead) and each plant was designated a crown position (understorey, intermediate, codominant, dominant or emergent). Signs and symptoms of P. ramorum were recorded and symptomatic bark or leaf tissue was collected for laboratory isolation and identification by standard methods (Davidson et al. 2003). The presence of other pathogens and insects was also noted and collections were taken where necessary. The slope and aspect of each plot and visible signs of past fire (i.e. basal fire scars) were also recorded. Percentage maximum solar radiation input was calculated using slope and aspect for latitude 40° N (Buffo et al. 1972). In 2003, a subsample of randomly selected uninfected host species from all plots (c. 35% of the total trees plot−1) was censused again and the fate of individuals that were alive in 2001–02, despite confirmation that they were infected with P. ramorum, was determined.

Disease incidence was based largely on positive isolation of P. ramorum from symptomatic host plants. However, some tanoaks (< 9%) were given P. ramorum positive designations based on the presence of symptoms and disease status of a nearest conspecific neighbour. Most tissue from these individuals was dead and it is difficult to recover P. ramorum from older necrotic tissue using culture techniques.

statistical analyses

All statistical analyses were conducted with the software program JMP, version 4.0 (SAS 2001). A one-way analysis of variance (anova) was used to determine differences in disease incidence between years (time). A Wilcoxon rank sum test was used to determine differences in percentage infection between tanoak and bay laurel. Chi-square analyses were used to determine if all size classes of bay laurel and tanoak were equally likely to be infected by P. ramorum. We tested correlations of P. ramorum incidence with species variables (bay laurel and tanoak) and host density. A stepwise multiple linear regression analysis was used to develop a model for redwood forest communities that used stand structure, composition and environmental variables to predict disease incidence (number of infected trees/total number of all host trees in a plot). The stepwise criterion was run with forward direction with P ≤ 0.05 for F for variable entry. A check of collinearity in our regression model was done employing leverage plots and bivariate scatterplots. For parametric analyses, assumptions of normality and homogeneity of variances were checked and met.


A total of 5342 trees were mapped and examined for P. ramorum. Mean incidence of P. ramorum across all plots (incidence per plot = number of P. ramorum infected stems/total stems) was 0.17 ± 0.01 (± 1 SE) in 2002 and 0.24 ± 0.02 in 2003. Mean P. ramorum incidence differed significantly between years (F1,239 = 10.11, P = 0.0017) with a 52% change in disease incidence from 2002 to 2003. Only 22% and 20% of the 120 plots were negative for P. ramorum in 2002 and 2003, respectively. The majority of plots with no incidence of P. ramorum were located in Big Basin S.P. and Julia Pfeiffer Burns S.P. (Fig. 1). Overall, 36 of 120 plots had disease levels ≤ 0.10, but 52% of plots with low or no disease incidence were adjacent to plots with moderate to high disease (Fig. 1). The remaining 47% of plots with no P. ramorum were clustered on the landscape, but were within a few kilometres of infested forests (Fig. 1).

Figure 1.

Disease incidence of Phytophthora ramorum in 2002 and 2003 from plots sampled at sites located in Sonoma, Marin, Santa Cruz and Monterey counties, California. Arrows indicate plots grouped in a contiguous area. Grouped plots with no P. ramorum in Santa Cruz and Monterey counties were at Big Basin S.P. and Julia Pfeiffer Burns S.P., respectively.

For the dominant overstorey tree species on plots where P. ramorum was identified, disease incidence ranged between 0.42 and 0.69 on bay laurel, 0.32 and 0.45 on tanoak and 0.03 and 0.10 on redwood. Mean percentage of infected bay laurel (0.55 ± 0.3) was significantly higher than that of tanoak (0.35 ± 0.3) (Wilcoxon rank sum test, n = 214, P = 0.002). Phytophthora ramorum cankers were found on all sizes of tanoak, but were significantly more common on smaller trees than on larger sized trees, especially individuals between 1 and 5 cm d.b.h. (χ2 = 30.20, P < 0.0001, d.f. = 4) (Table 2). Leaf lesions caused by P. ramorum on bay laurel were found on all sizes of trees but were significantly more common on trees 20 cm d.b.h. and smaller (χ2 = 20.29, P = 0.0004, d.f. = 4) (Table 2). Infection levels on Douglas fir ranged between 0.01 and 0.08 and were solely found on understorey trees. Infection levels for tree species that were not found at all study sites included Douglas fir (0.02, 0.01 and 0.08 at Jack London S.P., Marin Municipal Water District, and Henry Cowell S.P., respectively), madrone (0.07 at Jack London S.P.), black oak (0.05 at Jack London S.P.) and coast live oak (0.0, 0.02, 0.05 and 0.14 at Jack London S.P., Samuel P. Taylor S.P., Henry Cowell S.P., and Pfeiffer Big Sur S.P., respectively).

Table 2.  Disease status of bay laurel (non-lethal foliar host) and tanoak (lethal stem, branch and foliar host) from 2002 and 2003 censuses from all four study locations by diameter size classes. Pr positive = Phytophthora ramorum positive confirmation
Diameter (cm)Bay LaurelTanoak
Number Pr positive (%)TotalNumber Pr positive (%)Total
1–5182 (59.67)305489 (38.38)1274
5.1–10105 (68.18)154 99 (31.83) 311
10.1–20130 (60.47)215 47 (22.93) 205
20.1–40 74 (48.68)152 61 (26.29) 232
≥ 40.1 19 (38.78) 49 43 (29.45) 146
Total510 (58.3)875739 (34.1)2168

There was no relationship between disease incidence and overall stem density in a plot (r = −0.0017). Phytophthora ramorum incidence was significantly correlated with bay laurel importance value (n = 120, r = 0.57, P = 0.01) and tanoak basal area (n = 120, r = −0.37, P = 0.01). Correlations between disease incidence and density of bay laurel (r = 0.47) and tanoak (r = −0.22) were not as strongly correlated as the above variables for these two species.

We graphically compared the influence of tanoak and bay laurel in the context of P. ramorum incidence levels (sensuHolt et al. 2003) (Fig. 2). There was a weak negative correlation between bay laurel importance value and tanoak basal area (n = 120, r = −0.28) (Fig. 2). At the scale of our plots, P. ramorum could become established at nearly any combination of bay laurel importance value and tanoak basal area (Fig. 2). Disease incidence above 30% was most often associated with bay laurel importance value, although a number of plots had high incidence of P. ramorum with little or no bay laurel present (Fig. 2). In plots with few bay laurel stems, higher disease levels were associated with smaller tanoaks (Fig. 2).

Figure 2.

Bay laurel importance value vs. tanoak basal area in relation to ranges of Phytophthora ramorum incidence (0 to ≥ 0.61).

A disease incidence model was constructed using bay laurel importance value and tanoak basal area as the two key predictor variables. Solar radiation was used as an indirect measure of temperature and moisture conditions. In a previous study (Swiecki & Bernhardt 2002), canopy openness was correlated to disease incidence in coast live oak forests. We assessed openness or closure (density) of a canopy by the number of codominant, dominant and emergent tree species within a plot. We employed stepwise multiple regression to predict disease incidence for redwood forest communities, Loading X1, bay laurel importance value, X2, tanoak basal area, X3, % maximum solar radiation and X4, number codominant, dominant and emergent trees into the model. Inclusion of the first three yielded Y = 0.50 + 0.007X1 − 0.10X2 − 0.03X3; r2 = 0.412 (F3,116 = 27.19, P < 0.0001), which explained 41% of the variation in P. ramorum incidence.

Cumulative mortality across the 120 plots ranged between 0 and 66% for tanoak, 0 and 76% for redwood, 0 and 42% for madrone, and 0 and 28% for bay laurel (Fig. 3). In tanoak, 62.5% of the cumulative mortality was associated with P. ramorum infection and 37.5% associated with other biotic organisms and environmental factors (e.g. light, water and nutrient limitation) (Fig. 3). Average cumulative mortality of tanoak was much higher in P. ramorum-associated plots in Santa Cruz (9.6%) and Monterey (15%) counties than in plots with no P. ramorum in those locations (1.4% and 0.57% for Santa Cruz and Monterey, respectively) (Fig. 3). Other organisms associated with tanoak included various root and canker rot fungi and bark beetle and ambrosia beetle species. Following re-census in 2003, the mortality rate of P. ramorum-infected tanoaks (those confirmed positive for disease and alive in the 2001–02 census) was determined to be 6.0% year−1. The cause of death of most bay laurels was not readily apparent, but approximately half of the dead bay laurels were associated with opportunistic wood boring insects; in addition Armillaria spp. (root pathogen) and Ganoderma spp. (heart rot fungi) were also associated with bay laurel. Madrone leaves and branches are very susceptible to P. ramorum, but field confirmations of the pathogen on overstorey trees are difficult to attain due to our inability to obtain samples from high in the crown. Therefore, madrone mortality associated with P. ramorum may be significantly underestimated. Seedlings and saplings of madrone exposed to P. ramorum have high mortality rates (50–66%) in field and laboratory experiments (Maloney et al. 2004). No mortality of redwood was attributed to P. ramorum; c. 90% of dead redwood trees were understorey, light-suppressed individuals (≤ 15 cm d.b.h.).

Figure 3.

Average percentage cumulative mortality of tanoak, bay laurel, madrone and redwood from plots in Sonoma, Marin, Santa Cruz (P. ramorum and no P. ramorum plots) and Monterey (P. ramorum and no P. ramorum plots). Asterisks (*) indicate plots where species were not present. Numbers above tanoak bars indicate percentage of cumulative mortality of tanoak associated with P. ramorum, all other species have no P. ramorum-associated mortality.


We have presented a 2-year look at disease incidence of P. ramorum in redwood forests throughout the main range of this emerging pathogen in California. Establishment and spread of P. ramorum across the landscape is a function of pathogen life history and dispersal, host community composition and susceptibility, stand structure and climatic conditions. By infecting multiple host species, P. ramorum was found to be established on plots within a range of plant community structures. Although bay laurel appears to be the main host for this pathogen, P. ramorum could become established in tanoak forests on plots with little to no bay laurel (Fig. 2). Although not common in California forests, establishment of the pathogen in tanoak forests without bay laurel appears to be the situation in forests in southern Oregon (Rizzo et al. 2005).

In the model of Holt et al. (2003), bay laurel and tanoak (as well as redwood and other co-occurring plant species) could be considered substitutable hosts, i.e. the pathogen can invade all combinations of host densities outside a certain threshold level. Threshold host densities for P. ramorum establishment in these forests are very low (Fig. 2). As a foliar pathogen of large evergreen trees, P. ramorum lesions are very small in proportion to the size of the host, therefore changes in host stem density are less likely to have a significant effect on pathogen incidence than would a pathogen that causes whole plant mortality (Burdon & Chilvers 1982). In the case of bay laurel, P. ramorum needs only a single tree in which to become established on a site because most transmission may be within the tree canopy. An indirect measure of plant tissue (i.e. leaf area), such as importance value, rather than stem density, takes into account the potentially similar influence of one large and several-to-many smaller bay laurel trees and was a better indicator of disease incidence on our plots. The negative association with tanoak basal area suggests the importance of smaller tanoak trees, which can support P. ramorum sporulation on their branches, in the establishment of the pathogen. The stems of larger tanoaks, in contrast, do not support sporulation of the pathogen (J. M. Davidson, unpublished observations) and such trunk infections may serve as ecological dead ends for P. ramorum (cf. zoonotic diseases, such as bubonic plague and Lyme disease, in which humans serve as dead-end hosts, Keeling & Gilligan 2000; Woolhouse et al. 2001).

However, as it appears that most infected tanoaks eventually die, sublethal infections of non-oak hosts may be required for P. ramorum to persist indefinitely in infested forests. Theory based on single-host pathogens suggests that evolution of virulence on a host is usually inversely related to the degree to which the pathogen can reproduce from the host (Woolhouse et al. 2001). Long-term evolution of generalist pathogens can be difficult to predict. Given the high capacity for reproduction from bay laurel leaves, P. ramorum should evolve non-lethal infections on this host. Although P. ramorum can reproduce from tanoak trees, there may be little selection pressure for decreased virulence against tanoak on many sites because of the presence of bay laurel.

The range in susceptibility of coexisting plant species may also lead to P. ramorum-mediated competition that influences future successional patterns in these forests (Hudson & Greenman 1998). In the short term, the species that appear to be most affected by mortality are tanoak and madrone (Maloney et al. 2004). Even though other biotic and abiotic agents that cause mortality are present, P. ramorum is accelerating tanoak mortality in many locations. Hunter (1997) reported the annual mortality rate of tanoak to be c. 2.1% in a Douglas fir-tanoak forest in northern California, much lower than our 6.0% year−1 in P. ramorum-infested forests. Phytophthora ramorum-associated mortality of overstorey tanoaks has opened gaps in many of the plots. Such disturbances provide establishment opportunities for all associated tree species in this forest community. However, the continued presence of P. ramorum may alter recruitment dynamics by favouring non-hosts and foliar hosts in gaps. Tanoak sprouts prolifically and successfully recruits into canopy gaps (Hunter 1997), but its susceptibility to P. ramorum may diminish its recruitment success. Bay laurel, as a persistent foliar host with no associated P. ramorum-mediated mortality, may be favoured in gaps. Redwood recruitment may also benefit from tanoak mortality; to date, the pathogen has only been observed causing needle lesions, cankers on small branches and tip dieback of sprouts of this species.

Our results suggest that the rate of community change in redwood forests will be accelerated by P. ramorum intensification and subsequent mortality of tanoak. Phytophthora ramorum incidence exceeds 0.30 in many locations, yet the distribution of this pathogen is still patchy across the landscape, with high incidence plots often very close to plots with little or no disease. The complete absence of P. ramorum from several locations appears to be historical, i.e. the pathogen has simply not reached these areas yet, rather than some intrinsic factor preventing establishment. Plots with low disease levels were not compositionally or environmentally different than those with high disease levels. We therefore expect disease incidence to continue to increase on many of these plots over the next few years.


We thank K. Huryn, R. Albright, M. Voigt, A. Wickland, G. Slaughter, J. Davidson, T. Burt, K. Falk, and L. Douhan for laboratory and field assistance. The comments of J. M. Davidson, E. M. Hansen, and three anonymous reviewers on the manuscript are appreciated. We thank the California State Parks and the Marin Municipal Water District for allowing research on their lands. This research was funded by the USDA Forest Service Pacific Southwest Research Station and the Gordon and Betty Moore Foundation.