Fire history of the San Francisco East Bay region and implications for landscape patterns

Abstract

The San Francisco East Bay landscape is a rich mosaic of grasslands, shrublands and woodlands that is experiencing losses of grassland due to colonization by shrubs and succession towards woodland associations. The instability of these grasslands is apparently due to their disturbance-dependent nature coupled with 20th century changes in fire and grazing activity.This study uses fire history records to determine the potential for fire in this region and for evidence of changes in the second half of the 20th century that would account for shrubland expansion. This region has a largely anthropogenic fire regime with no lightning-ignited fires in most years. Fire suppression policy has not excluded fire from this region; however, it has been effective at maintaining roughly similar burning levels in the face of increasing anthropogenic fires, and effective at decreasing the size of fires. Fire frequency parallels increasing population growth until the latter part of the 20th century, when it reached a plateau. Fire does not appear to have been a major factor in the shrub colonization of grasslands, and cessation of grazing is a more likely immediate cause. Because grasslands are not under strong edaphic control, rather their distribution appears to be disturbance-dependent, and natural lightning ignitions are rare in the region, I hypothesize that, before the entrance of people into the region, grasslands were of limited extent. Native Americans played a major role in creation of grasslands through repeated burning and these disturbance-dependent grasslands were maintained by early European settlers through overstocking of these range lands with cattle and sheep. Twentieth century reduction in grazing, coupled with a lack of natural fires and effective suppression of anthropogenic fires, have acted in concert to favor shrubland expansion.

Full text

CSIRO PUBLISHING
www.publish.csiro.au/journals/ijwf International Journal of Wildland Fire, 2005, 14, 285–296
Fire history of the San Francisco East Bay region and implications
for landscape patterns
∗
Jon E. Keeley
U.S. Geological Survey, Western Ecological Research Center, Sequoia-Kings Canyon Field Station,
Three Rivers, CA 93271-9651, USA; and Department of Ecology and Evolutionary Biology,
University of California, Los Angeles, CA 90095, USA. Telephone: +1 559 565 3170; fax: +1 559 565 3177;
email: jon_keeley@usgs.gov
Abstract. The San Francisco East Bay landscape is a rich mosaic of grasslands, shrublands and woodlands that is
experiencing losses of grassland due to colonization by shrubs and succession towards woodland associations. The
instability of these grasslands is apparently due to their disturbance-dependent nature coupled with 20th century
changes in fire and grazing activity.Thisstudyusesfirehistory records todetermine the potential for fireinthis region
and for evidence of changes in the second half of the 20th century that would account for shrubland expansion. This
region has a largely anthropogenic fire regime with no lightning-ignited fires in most years. Fire suppression policy
has not excluded fire from this region; however, it has been effective at maintaining roughly similar burning levels
in the face of increasing anthropogenic fires, and effective at decreasing the size of fires. Fire frequency parallels
increasing population growth until the latter part of the 20th century, when it reacheda plateau. Fire does not appear to
have been a major factor in the shrub colonization of grasslands, and cessation of grazing is a more likely immediate
cause. Because grasslands are not under strong edaphic control, rather their distribution appears to be disturbance-
dependent, and natural lightning ignitions are rare in the region, I hypothesize that, before the entrance of people into
the region, grasslands were of limited extent. NativeAmericans played a major role in creation of grasslands through
repeated burning and these disturbance-dependent grasslands were maintained by early European settlers through
overstocking of these range lands with cattle and sheep. Twentieth century reduction in grazing, coupled with a lack
of natural fires and effective suppression of anthropogenic fires, have acted in concert to favor shrubland expansion.
Additional keywords: anthropogenicfire regime; California Department of Forestry andFire Protection; fire climate;
fire suppression; grasslands; lightning-ignited fires; Native American burning; shrublands.
Introduction
The East Bay region of San Francisco, California (Fig. 1)
has had a long history of human impact beginning with
Native American occupation during the early Holocene
(Jones 1992). Throughout time humans have had an effect on
the local ecology, first through burning by the Native Ameri-
cans and later through a combination of burning and grazing
by the Euro-American settlers (McMinn 1916; Clarke 1959;
Roof 1971). Beginning in the early 20th century there has
been an effort to alter the course of impact on selected por-
tions of the landscape through the development of a series
of parks for purposes of conservation, fire protection and
recreation (Harris 1927; McBride 1970).
This landscape is a rich mosaic of grasslands, shrub-
lands and woodlands (Fig. 2), with limited coniferous forests
∗
This manuscript was written and prepared by a U.S. Government employee on official time and therefore is in the public domain and not subject to
copyright.
(McMinn 1916; McBride 1970; Edwards 2002). Currently
grasslands occupy the greatest proportion of the landscape
in the northern part of the East Bay (38% in Contra Costa
County) with decreasing proportions southward (29% in
Alameda and 17% in Santa Clara counties) (Huenneke 1989).
These are mostlydominated byalien annualgrasses and forbs.
Native perennial grasslands are restricted to a handful of sites
in the region and account for a small percentage of the total
grassland area (Edwards and Havlik 1984; Huenneke 1989).
Throughout the latter half of the 20th century, alien-
dominated grasslands in many of the protected conservation
areas of the East Bay have decreased markedly due to colo-
nization by shrubs and succession to woodland associations
(Fig. 3). The shrub Baccharis pilularis, known as baccharis
or coyote brush, is typically the first woody species to invade
10.1071/WF05003 1049-8001/04/030285

286 Int. J. Wildland Fire J. E. Keeley
San
Francisco
Marin
Solano
San
Mateo
Santa Cruz
Santa Clara
Alameda
Contra Costa
N
0 4 8 16 24 32 km
Fig. 1. The San Francisco East Bay counties included in this study.
Most local residents include only the two northern counties in the des-
ignation ‘East Bay’; however, Santa Clara is included here because it is
a natural southern extension of this ecological region.
Fig. 2. Vegetation mosaic of alien-dominated annual grassland, shrub-
lands and woodlands in the East Bay foothills (photograph by Oliver
Pearson, courtesy of Anita Pearson).
grasslands following a cessation of grazing and increased
fire protection (McBride 1964; McBride and Heady 1968).
These shrubs provide suitable sites for establishment of
woodland species that appear to progress through a series of
replacements (McBride 1974; Safford 1995). Baccharis col-
onization of grasslands is in fact a widespread phenomenon
throughout the central and northern coastal regions of Cali-
fornia (Elliott and Wehausen 1974; DaSilva and Bartolome
1984; Hobbs and Mooney 1986; Williams et al. 1987).
It appears that cessation of livestock grazing on areas
set aside for parklands has played a major role in the type
conversion of grasslands to baccharis shrublands (Harris
1927; McBride and Heady 1968; McBride 1974). In addi-
tion, a reduction in burning due to fire suppression policy is
thought by many to have also played a role in these landscape
changes (McBride 1970; Roof 1971; Edwards 2002; Russell
and McBride 2002). This model is consistent with fire-
suppression induced increases in forest density demonstrated
from historical photographs in many western US forested
landscapes (Gibbens and Heady 1964; Gruell 1983). These
forests historically burned for eons by frequent lightning
fires, but a century of effective fire suppression management
has all but eliminated fire, allowing increased density and
expansion of forests.
The vegetation changes in the San Francisco East Bay
(Fig. 3) have created landscapes with greater fuels and poten-
tial for higher intensity wildfires. This, coupled with urban
sprawl, has created an increased fire hazard, which is of pro-
found concern to many in the region because of the history
of devastating wildfires, e.g. the 1923 Berkeley Fire and the
1991 Tunnel Fire (Martin and Sapsis 1995). Management
requires a clear understanding of the role of fire on these land-
scapes, and how historical land use changes have affected fire
activity.
The purpose of this paper is to evaluate the extent to which
changes in fire activity can account for landscape changes.
In addition to documenting the known fire history for the last
half of the 20th century, I evaluate the sources of ignition
and changes in burning patterns that have accompanied rapid
population growth in the region and the extent to which his-
torical variations in climate, which are important drivers of
fire activity in other parts of the western USA (Westerling
et al. 2002), may have affected these fire history patterns.
Methods
Fire statistics were from the California Department of
Forestry and Fire Protection written records, variously named
Annual Fire Report, Fire Statistics or Wildfire Activity
Statistics (State of California Department of Forestry and
Fire Protection, 1931–2002). These data include all fires for
which action was taken and represent a much more com-
prehensive picture of fire activity than the recently available
electronic California Statewide Fire History database (e.g.
Keeley et al. 1999). Data before 1945, which included only
Santa Clara County, were reported as Clarke-McNary lands
and non-Clarke-McNary lands. The latter were largely local
rural areas and were not included in this study. After 1945 the

Fire history and landscape of San Francisco East Bay Int. J. Wildland Fire 287
(a)(b)
(d)
(c)
Fig. 3. Changes in the vegetation mosaic in the East Bay foothills during the latter half of the 20th century. (a) View south-east to part of San Pablo
Ridge just west of Inspiration Point in 1951; (b) approximately the same view as panel (a) in 1994; (c) view east to Nimitz Way north of Inspiration
Point in 1951; and (d) the same view as panel (c) in 1994 (photographs by Oliver Pearson, courtesy of Anita Pearson).
Clarke-McNary lands were separated into Zone I and Zone II
lands and after 1970 these were designated Direct Protec-
tion Areas (DPAs). Total area included in Clarke-McNary,
Zones I and II, and DPAs varied slightly through time and
thus all statistics were expressed as fire activity per unit area
protected, based on the areas reported in the annual fire sum-
maries. For years where area protected was not presented in
the database, the last known area reported was used. These
data are for wildfires and do not include prescription burning,
but they may include backfires used in suppression activities.
Climatic data for the Palmer Drought Severity Index
(PDSI) were downloaded from the National Oceano-
graphic and Atmospheric Administration website (http://lwf.
ncdc.noaa.gov/oa/climate/onlineprod/drought/xmgrg3.html,
accessed February 2004). This NOAA website presents
north central coast of California indices based on com-
bined values from all available stations in the region. Their
circumscribed areas roughly parallel the counties studied
here. Decadal census data by county were obtained from
http://www.census.gov/population/cencounts/ca190090.txt.
Preliminary analysis, before regression analysis, fitted a
smoothed curve to each scatter plot, generated with locally
weighted scatter plot smoothing (Wilkinson et al. 1996). In
instances where data did not closely follow a linear relation-
ship, the smoothed curve was presented.When log or log-log-
transforms produced a linear relationship, these regression
statistics were presented. Ordinary least-squares regres-
sion was used to investigate bivariate relationships between
fire activity and the chronological year January–December.
Statistics calculated from these regressions assume that the
residuals from the regression are normally distributed and
so residuals were regressed against the expected values for
a normal probability plot. Residuals from regressions gave
reasonably tight fits in these normal probability plots. Since
statistics calculated from ordinary least-square regression are
generally robust to departures from normality, this level of
approximation was considered acceptable for these analyses.
When investigating bivariate relationships between vari-
ables that occur in an ordered chronological sequence,
autocorrelation of variables in the time series complicates

288 Int. J. Wildland Fire J. E. Keeley
interpretation of the bivariate model. The Durbin-Watson
statistic indicates whether the residuals are correlated and
lack independence. Durbin-Watson statistics greater than 1
were considered sufficiently auto-correlated to require cor-
rection. This autocorrelation was corrected with the time
series technique of differencing, where the value of the vari-
able at time t −1 is subtracted from the value at time t. These
first-order difference values were then used in ordinary least-
squares regression and the Durbin-Watson statistic for these
regressions indicated that autocorrelation had been removed.
Results
Relative frequency of human-caused and lightning-caused
fires for the three East Bay counties are compared with
interior mountain ranges in the Sierra Nevada and south-
ern Rocky Mountains (Fig. 4). Human ignitions were several
times more common in the East Bay than in mountainous
areas to the east. In contrast, lightning-ignited fires were rare.
For every 100 000 ha of area under CDF protection the light-
ning fire frequency varied from 1.8 per decade in Alameda
County to 5.3 in Santa Clara County. East Bay lightning fire
frequency was about two orders of magnitude lower than
observed in interior mountain ranges.Acrossthe entire period
of record for all counties (1945–2002) there was an aver-
age of 3.4 lightning fires per 100 000 ha each decade for
the three counties combined. Most years were without any
0
200
400
600
800
1000
Fires/100 000 ha/decade
CC Al SC Pl Se Co
Bay area Interior mountainous
Human
Lightning
Fig. 4. Human and lightning-ignited fires in the three Bay Area coun-
ties compared with interior mountainous areas, including the Plumas
and Sequoia national forests in the Sierra Nevada Range of California
and the Coconino National Forest in central Arizona. Based on data
from the decade 1970–1979, Bay Area data from this paper, data from
the three national forests are from US Forest Service fire records.
lightning-ignited fires: in Contra Costa County 86% of the
years had no lighting-ignited fires, in Alameda the figure was
74% and in Santa Clara it was 60%. Thus, most fires in the
East Bay region were started by people.
0
30
60
90
120
150
0
30
60
90
120
150
1930 1945 1960 1975 1990 2005
Year
0
20
40
60
80
100
Fires/100 000 ha
(b)
(a)
(c)
Fig. 5. Annual fire frequency for the East Bay counties with the locally
weighted scatterplot smoothing. (a) Contra Costa County; (b) Alameda
County; and (c) Santa Clara County.

Fire history and landscape of San Francisco East Bay Int. J. Wildland Fire 289
Annual fire frequency showed a sharp rise after 1950 but
a leveling off of fires in the last decade of the 20th century
(Fig. 5). Considering the linear portion of this curve, ∼1950–
1985, there was a highly significant (P < 0.001) increase in
0
1200
2400
3600
4800
6000
(a)
0
1200
2400
3600
4800
6000
(b)
1930 1945 1960 1975 1990 2005
Year
0
3000
6000
9000
12 000
15 000
(c)
Area burned (ha/100 000 ha)
Fig. 6. Annual area burned for the East Bay counties with the locally
weighted scatterplot smoothing. (a) Contra Costa County; (b) Alameda
County; and (c) Santa Clara County.
fire frequency (r
2
=0.458, 0.823, 0.593, for Contra Costa,
Alameda and Santa Clara counties, respectively, n =58, 58,
72). For all counties combined over the period from 1945
to 2002 there was a highly significant positive relation-
ship between year and fire frequency (r
2
=0.752, P < 0.001,
n =58).
Area burned on the other hand exhibited relatively little
directional change over the period of record from 1945 to
2002 (Fig. 6). Santa Clara County, with records extending
back to 1931, did exhibit a significant (P < 0.01) negative
relationship, driven by the very high area burned during the
1930s. Throughout the record, annual area burned varied
by several orders of magnitude. Comparing annual burning
(Fig. 6) with decadal totals (Fig. 7), it is apparent that com-
monly 50% or more of the area burned during a decade may
come from just a single year. In all three counties the 1950s
were one of the lowest decades and the 1960s were highest.
In general, very little of the area protected by CDF burns
each decade. The peak was ∼30% for Santa Clara County
during the decade of the 1930s. Since then only ∼10% of
each county has burned each decade. Thus, the fire rotation
interval, or estimated time to burn the entire protected area,
is roughly 100 years.
Since nearly all ignitions were due to humans, the popula-
tion growth during this time period is illustrated (Fig. 8). Dur-
ing the latter half of the 20th century, population growth as a
function of area protected grew most sharply in Contra Costa
and Alameda counties. For the three counties combined there
was a highly significant (P < 0.001) relationship between
population density per unit of area protected and number of
fires for the years 1945–2002 (r
2
=0.706, differencing not
used because the Durban-Watson statistic =0.731). However,
unlike the patterns for annual fire frequency, which leveled
off during the 1990s (Fig. 5), population growth increased
linearly throughout the 20th century (Fig. 8). Thus, while
human population growth was strongly correlated with fire
frequency for much of the record, some changes occurred
in the behavior of people to put a cap on fire ignitions. Fac-
tors in the record that appear to have changed during this
period include changes in smoking and arson. Beginning in
the mid-1950s smoking steadily decreased as a source of igni-
tion (Fig. 9a).Arson as a proportion of all fire starts increased
during the first half of the record and then decreased and the
smooth curve closely follows a quadratic function (Fig. 9b).
Early in the record these intentional fires were known as
incendiary fires whereas today they are called arson fires.
This reflects changing attitudes towards fires, where early on
it was more sociologically acceptable to intentionally burn
wildland landscapes but today itis usually considered a crimi-
nal act. In recent years the primary causes of fires are vehicles
and equipment use (data not shown).
Although there are no obvious patterns with area burned
(Figs 6, 7), the size of fires changed markedly, with moderate
to large fires decreasing (Fig. 10a) and small fires increasing

290 Int. J. Wildland Fire J. E. Keeley
0
3000
6000
9000
12 000
15 000
(a)
0
3000
6000
9000
12 000
15 000
(b)
0
7500
15 000
22 500
30 000
37 500
(c)
Area burned (ha/100 000 ha)
Decade
1930 1950 1970 1990
Fig. 7. Decadal area burned for the East Bay counties. (a) Contra
Costa County; (b) Alameda County; and (c) Santa Clara County.
1930 1955 1980 2005
75 000
460 000
845 000
1 230 000
1 615 000
2 000 000
Population
(a)
r
2
⫽ 0.990
P ⬍ 0.001
r
2
⫽ 0.983
P ⬍ 0.001
r
2
⫽ 0.967
P ⬍ 0.001
1930 1955 1980 2005
(b)
1930 1955 1980
2005
(c)
Year
Fig. 8. Population growth for the East Bay counties expressed for the entire county (solid line) or per 100 000 ha of area protected
by the California Department of Forestry and Fire Protection (dashed line). (a) Contra Costa County; (b) Alameda County; and
(c) Santa Clara County.
(Fig. 10b).The rank order of all fire years showed that human-
ignited fires were distributed over a broad range with many
moderate size fire years (Fig.11a) whereas in most years there
was little or no area burned bylightning, save three significant
years (Fig. 11b). Changes in annual area burned as a function
of the proportion burned in brush or grassland for the counties
combined showed that, as the proportion of area burned by
grassland increased, the total area burned declined, and the
opposite pattern for brush (Fig. 12).
Using time series differencing for regressions of PDSI v.
fire activity showed very few significant relationships.
Specifically, the number of fires and area burned were not
significantly related to the seasonal PDSI (August–June), or
winter, spring or summer PDSI. The only significant relation-
ship was area burned by lightning fires, which was negatively
related to autumn PDSI (R
2
=0.120, P < 0.01), indicating
that lightning fire size was a function of the severity of the
autumn drought.
Discussion
The East Bay of San Francisco has a largely anthropogenic
fire regime with no lightning-ignited fires in most years, and
when they do occur they are generally on the higher peaks
in the region. This pattern is typical for coastal northern Cal-
ifornia (Keeley 1982). Thus, before the entrance of people
into the region, it seems unlikely that fire played a major role
in landscape patterns of vegetation distribution.
Although fire suppression policy has not excluded fire
from this region, it has been effective at maintaining simi-
lar levels of burning (Figs 6, 7) despite increasing ignitions
(Fig. 5). Fire frequency paralleled population growth until the
latter part of the century when it leveled off (Figs 5, 8). Sort-
ing out the factors responsible for this non-linear relationship
between fires and population density is complicated by the
fact that the causes of fires have changed over this period
(e.g. Fig. 9a). A significant contributor is more effective fire
prevention, evident by the reduction in incendiary/arson fires
(Fig. 9b). Other evidence of fire suppression effectiveness

Fire history and landscape of San Francisco East Bay Int. J. Wildland Fire 291
1945 1965 1985 2005
0
10
20
30
40
50
60
Fires due to smoking (%)
(a)
r
2
⫽ 0.739
P ⬍ 0.001
1945 1965 1985 2005
Year
0
10
20
30
40
50
Arson fires (%)
(b)
Fig. 9. Changes in fire ignition sources of (a) smoking and (b) arson
for the East Bay counties combined. Arson is a recent term (first used
in 1981); before then these were called incendiary fires.
is the trend towards increasing frequency of small fires
(Fig. 10a), although increased habitat fragmentation from
urban sprawl may also be a factor.
There is little evidence that changes in burning patterns
are a major factor in the widely reported colonization of
grasslands by shrubs and woodland elements (Fig. 3) during
the latter half of the 20th century (McBride 1964, 1970; Roof
1971; Edwards 2002; Russell and McBride 2002). Rather, the
elimination of grazing, as emphasized by McBride (1974;
McBride and Heady 1968), would seem to be a bigger factor.
This model, though, is at odds with other models for
grassland invasion by woody plants in the South-west, where
intense livestock grazing, not cessation of grazing, is tied
to what is often termed ‘brush encroachment’ of grasslands
(Archer 1994). These models, however, apply to two very
different ecosystems. In the absence of grazing, the very
high lightning fire frequency in the South-west (Fig. 4)
results in frequent grass fires that eliminate less fire-tolerant
1945 1965 1985 2005
Year
60
70
80
90
100
Fires ⬍ 4 ha (%)
(b)
r
2
⫽ 0.406
P ⬍ 0.001
0
5
10
15
Fires ⬎ 40 ha (%)
r
2
⫽ 0.573
P ⬍ 0.001
(a)
Fig. 10. Annual changes in percentage of (a) large fires and (b) small
fires for the East Bay counties combined.
shrubs. Grazing reduces grass fuels and thus fire frequency
declines, allowing shrub invasion. In California, natural fires
are infrequent enough to eliminate shrublands and thus graz-
ing introduces less of a change in fire frequency. These two
ecosystems also respond differently to grazing. In the South-
west, spiny shrubs such as Prosopis that invade grasslands
are noxious to livestock and generally avoided whereas, in
California, shrub colonizers are slightly palatable, and even
desirable in the summer dry season (Harris 1927; Sampson
and Jespersen 1963).
One of the important implications of this East Bay model
is that shrub colonization of grasslands (Fig. 3) occurs when
disturbance is removed from the system, implying that these
grasslands are disturbance-dependent. I hypothesize that,
just like the contemporary scene, the indigenous landscape
also comprised a mosaic of grasslands, shrublands, wood-
lands and coniferous forests, but because of limited natural
fires it had very limited grasslands and a decidedly greater
representation of woody associations than today. Thus, the

292 Int. J. Wildland Fire J. E. Keeley
0 102030405060
Rank of annual area burned
0
1000
2000
3000
4000
5000
6000
Lightning (ha)
(b)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
Humans (ha)
(a)
Fig. 11. Rank order fires by size for (a) human and (b) lightning
ignited fires for the East Bay counties combined.
widespread extent of contemporary grasslands is the result of
type conversion of woody vegetation to grasslands, a model
that has been applied to other parts of the state (Cooper 1922;
Wells 1962; Huenneke 1989; Keeley 1990; Hamilton 1997;
Holstein 2001), and applies to the creation of other grasslands
in the world (Bond et al. 2005; Keeley and Rundel 2005). In
the East Bay this model is supported by microfossil studies
of grass phytoliths. Over a dozen grassland sites examined
by Hopkinson (2003) showed that most sites had at one time
been dominated by more mesic woodland vegetation types.
While frequent fires are certainly capable of maintaining
grasslands free of woody species, the recent history for these
sites suggests that livestock grazing has been a more impor-
tant factor, and perhaps the dominant factor since the early
Euro-Americans introduced cattle and sheep ∼225 years ago.
However, grasslands in this region did not originate from
the European introduction of livestock, since the current
grassland distribution appears to be little changed from that
0 20 40 60 80 100
Brush burned (%)
5000
10 000
15 000
(a)
r
2
⫽ 0.799
P ⬍ 0.001
Area burned (ha/100 000 ha)
0 20 40 60 80 100
Grass burned (%)
0
5000
10 000
15 000
(b)
r
2
⫽ 0.244
P ⬍ 0.001
Fig. 12. Changes in annual area burned as a function of the propor-
tion burned in (a) brush or (b) grassland for the East Bay counties
combined. High Durbin-Watson statistics (>1.5) indicate substantial
temporal autocorrelation and thus all statistics are calculated after time
series differencing.
reported by the first Spanish explorers (DeNier 1928; Clarke
1959; Mayfield 1978), with the exception that a substantial
amount has been lost to development. In order for grasslands
to have been maintained in this environment by natural fires,
one would have to expect that every lightning ignited fire was
enormouslylarge,on the order of 10
5
–10
6
ha. Evenunder cur-
rent landscape patterns with extensive grasslands that cure to
form highly flammable fuels during the summer lightning fire
season, there is no evidence that they would commonly reach
this size. Indeed during the latter half of the 20th century, only
on three occasions did a lightning-ignited fireexceed 1000 ha.
It is even more unlikely that natural lightning-ignited fires
could have carved out the current grasslands from a shrub-
land or woodland landscape of more mesic fuels, since these
fuels have an even shorter fire season, generally later in the
fall after the summer lightning fire season.

Fire history and landscape of San Francisco East Bay Int. J. Wildland Fire 293
One possibility is that the native grazing/browsing fauna,
including deer (which still maintain sizeable populations
in the East Bay), antelope and elk, maintained extensive
grasslands by excluding woody plant invasion before Euro-
American contact. Several factors make this unlikely. This
region was heavily dominated by Native American settle-
ments, with over 100 village sites and more than 2000
inhabitants in the East Bay area at the time of contact (Cook
1957). Hunting would have had a significant impact on the
size of these herds. In addition, herbivory per se is not respon-
sible for conversion of woody associations to grasslands but
rather the over-stocking of rangelands with livestock.
A more likely factor in the origin and maintenance of these
grasslands was the frequent use of fire by the high density of
NativeAmericans in the East Bay.There are many reasons for
believing that these Native Americans managed their envi-
ronment with fire in order to expand grasslands and other
herbaceous associations over woody vegetation (Bean and
Lawton 1973; Lewis 1973; Keeley 2002). Indirect evidence of
Native American burning in this region are the frequent fires
documented for the 18th and 19th centuries from fire-scarred
redwoods just across the bay in Muir Woods (Jacobs et al.
1985), and in other redwood forests farther north (Finney and
Martin 1989; Brownand Swetnam 1994) and south (Stephens
and Fry 2005). These fire scar records of 10–20 year inter-
vals are not consistent with the very limited lightning fire
potential for the East Bay (Fig. 4) and the rest of the north
coast region(Keeley 1982) and likelyreflect NativeAmerican
burning.
Frequent burning by Native Americans was an effective
means of altering landscapes to provide more readily utiliz-
able resources, since open herbaceous vegetation enhanced
both wildlife habitat and seed resources relative to closed
canopy woodlands (Timbrook et al. 1982; Keeley 2002).
Countless studies have shown that even resprouter-dominated
shrubland communities in this region can be readily type-
converted to herbaceous associations with repeat fires at
2–10 year intervals (Sampson 1944, 1952; Burcham 1957;
Nichols et al. 1984; Heady and Child 1994). Native Ameri-
cans were skilled in the use of fire to type convert shrublands
to grasslands; however, those grasslands were substantially
different in composition from those currently dominating
the region. Today most grasslands in California are com-
posed of alien assemblages of annual grasses and forbs from
the Mediterranean Basin. Type conversion of shrublands
by Native Americans would have favored native perennial
bunchgrasses as well as a rich assemblage of native annual
and perennial forbs.
European colonization of the East Bay resulted in the
rapid decimation of Native American populations and loss
of their fire management activities. However, fire was used
by the Spanish, and later the Mexicans, to expand grazing
lands in the region (Roof 1971). Since hides and tallow,
not meat, were the primary products being produced, it was
often productive to overstock these rangelands. Thus, fire
and grazing combined to maintain a quasi-equilibrium of
disturbance-maintained grasslands.
Conclusions
The limited role of natural fires in the East Bay, coupled
with the disturbance-dependent character of grasslands in the
region, suggest that before human entry into the region the
landscape mosaic heavily favored shrublands and woodlands
with coniferous forests and smaller pockets of grasslands. A
projected time line of these changes is suggested in Table 1.
Prior to human entry into California the only potential
source of disturbance would havebeen thegrazing and brows-
ing mammals. Potentially the diverse Pleistocene fauna of
the region would have been an important disturbance factor
that may have played a role in the maintenance of grasslands
(Edwards 1990, 1992, 1995). The cooler more mesic Pleis-
tocene climates, however, would have favored faster recovery
by mesic woodland elements (Keeley 2002), working against
widespread grassland formation, but to be sure, we lack a
clear quantitative picture of how natural grazing and browsing
during this epoch would have affected vegetation patterns.
Humans entered this region in the late Pleistocene and
rapidly decimated much of the native fauna (Martin 1984),
thus removing this disturbance factor. Early Holocene popu-
lations likely had limited impact on vegetation patterns due to
low population density and cool, mesic conditions. As pop-
ulations grew, coupled with the mid-Holocene drying and
transition to a seed-based economy(Jones 1992), the uniquely
newdisturbance factor offire became an important vegetation
management tool. NativeAmericanburning greatlyexpanded
grassland distribution. This impact would have been light at
first but, based on the population density of the region at the
time of contact with Europeans in the late 18th century, it
likely became very intense. It appears that the contemporary
distribution of grasslands had already been established by
the time of contact with Spanish settlers. Landscape burn-
ing did not die with the demise of Native American cultures
since fire was an important feature of European agropastoral
land management practices (Pyne 1995). However, 19th and
20th century development increased urban sprawl into water-
sheds of dangerous fuels (Hornbeck 1983), and thus there was
an increasing pressure on fire prevention and suppression
(Clar 1959).
Throughout the 20th century there was an increasing need
to manage much of this landscape for recreation and con-
servation. With this came a reduction in livestock grazing
resulting in successional changes towards shrublands and
woodlands (Fig. 2). These changes are commonly referred
to as shrub invasion or brush encroachment of grasslands.
Alternatively, this is perhaps best viewed as a natural recolo-
nization of grasslands that have been maintained by millennia
of human disturbance. There is little evidence that changes

294 Int. J. Wildland Fire J. E. Keeley
Table 1. Proposed time line of changes in vegetation and disturbance history for the East Bay expected from the patterns of natural and
human fires on these landscapes
Period Human impact Expected or observed vegetation
Late Pleistocene to Limited Native American populations. Fire regime Woodland-dominated landscape with mosaic of shrublands,
mid-Holocene dominated by lightning with fire rotation intervals forests and patches of grassland.
on the scale of centuries.
Mid-Holocene to late Increasing density of Native Americans and increasing Grassland-dominated landscape in a mosaic of shrublands
18th century dependence on plant products with frequent use of and woodlands maintained by high fire frequency.
fire for landscape management. Expected fire rotation Dominated by a combination of native perennial grasses
interval on the order of a decade or less. and native annual forbs.
19th century Euro-American settlement with heavy livestock grazing Grassland-dominated landscape maintained by replacing
Fire frequency probably lower than earlier periods. frequent fire with heavy grazing. Natives replaced
Expected fire rotation interval perhaps several by alien annual grasses and alien annual forbs.
decades, maybe longer.
20th century Increasing protection, reduction of lifestock grazing. Gradual recolonization of grasslands by native shrubs
Increasing fire frequency due to population growth (baccharis) and trees, also invasion by alien shrubs
and urban sprawl coupled with increasing effectiveness (brooms).
of fire suppression. Fire rotation intervals very long on
the order of a century or more.
in fire activity during the latter half of the 20th century can
explain these successional changes.
These successional changes in vegetation have created
landscapes with increased fire hazard from the greater fuel
loads (Russell and Tompkins 2005). Adding to this prob-
lem is urban sprawl with its ever expanding wildland–urban
interface, and this increasing periphery means an increas-
ing number of people at risk (Russell and McBride 2002).
Shrublands with their enhanced fuel loads produce more
intense fires that are more difficult to suppress and result
in somewhat larger fires (Fig. 12a). On the other hand, the
alien dominated grasses cure by early summer and likely
expand the fire season far longer than the shrubland fuels.
Regardless of fuel type, however, the primary risk is asso-
ciated with extreme weather events. Under these severe fire
weather conditions, fire spread is extremely rapid and the
area has a history of devastating fires (Martin and Sapsis
1995). These, however, have all been relatively small fires
that involved fuels at the wildland–urban interface. Fuels far
removed from this interface zone (Fig. 2) played very little
role in these conflagrations. Thus, it would seem the most
cost-effective approach to fire hazard reduction should be
focused at the interface zone and here the problem is often as
much due to exotic fuels (Gallagher 2004) as it is to natural
successional processes (Fig. 3).
Acknowledgements
I thank Joe McBride for detailed comments on an earlier ver-
sion of the manuscript, Peter Hopkinson for a copy of a chap-
ter from his dissertation, J. Kent for detailed information on
East Bay Regional Parks management history, Anita Pearson
for permission to use historical photographs by Oliver
Pearson, Anne Hopkins-Pfaff for the map, Hugh Safford,
Laura Baker, Steve Edwards, Greg Schneider, and William
McClung for stimulating discussion and Joe Engbeck whose
kind invitation to speak to the East Bay Regional ParksAsso-
ciation rekindled my long-standing interest in the fire history
of the region.
References
Archer S (1994) Woodland plant encroachment into southwestern
grasslands and savannas: rates, patterns and proximate causes. In
‘Ecological implications of livestock herbivory in the west’. (Eds
M Vavra, WA Laycock, RD Pieper) pp. 13–68. (Society for Range
Management: Denver, CO)
Bean LJ, Lawton HW (1973) Some explanations for the rise of cultural
complexity in native California with comments on proto-agriculture
and agriculture. In ‘Patterns of Indian burning in California: ecology
and ethnohistory’.(Ed. HT Lewis)pp.v–xlvii. (Ballena Press:Menlo
Park, CA)
Bond WJ, Woodward FI, Midgley GF (2005) The global distribution of
ecosystems in a world without fire. New Phytologist 165, 525–538.
doi:10.1111/J.1469-8137.2004.01252.X
Brown PM, Swetnam TW (1994) A cross-dated fire history from coast
redwood near Redwood NationalPark,California. Canadian Journal
of Forest Research 24, 21–31.
Burcham LT (1957) ‘California range land: An historic-ecological
study of the range resources of California.’ (State of California,
Department of Natural Resources,Divisionof Forestry: Sacramento)
Clar CR (1959) ‘California government and forestry from Spanish days
until creation of the Departmentof Natural Resources in 1927.’(State
of California, Department of Natural Resources: Sacramento)
Clarke WC (1959)The vegetation cover of the San Francisco Bay region
in the early Spanish period. MA Thesis, University of California,
Berkeley.
Cook SF (1957) The aboriginal population of Alameda and Con-
tra Costa counties, California. Anthropological Records 16,
131–156.
Cooper WS (1922) ‘The broad-sclerophyll vegetation of California.
An ecological study of the chaparral and its related communities.’
Carnegie Institution of Washington, Publication No. 319.
Da Silva PG, Bartolome JW (1984) Interaction between a shrub, Bac-
charis pilularis subsp. consanguinea (Asteraceae), and an annual

Fire history and landscape of San Francisco East Bay Int. J. Wildland Fire 295
grass, Bromus mollis (Poaceae) in coastal California. Madroño 31,
93–101.
DeNier FL (1928) Robert Livermore and the development of Livermore
Valley to 1860. MA Thesis, University of California, Berkeley.
Edwards SW (1990) The East Bay’s richest grassland: A Pleistocene
relict? Four Seasons 8(4), 23–32.
Edwards SW (1992) Observations on the prehistory and ecology of
grazing in California. Fremonita 20, 3–11.
Edwards SW (1995) Notes on grazing and native plants in central
California. Four Seasons 10, 61–66.
Edwards S (2002) A botanical exploration of the Tilden Park Area of
San Pablo Ridge, Contra Costa County, CA. Four Seasons 11(4),
3–52.
Edwards SW, Havlik N (1984) Some natural bunchgrass preserves in
the San Francisco Bay Area. Four Seasons 7, 5–14.
Elliott HW, III, Wehausen JD (1974) Vegetational succession on coastal
rangeland of Point Reyes Peninsula. Madroño 22, 231–238.
Finney MA, Martin RE (1989) Fire history in a Sequoia sempervirens
forest at Salt point State Park, California. Canadian Journal of Forest
Research 10, 1451–1457.
Gallagher N (2004) ‘Thousands of invasive trees cleared in UC Berke-
ley fire project.’ University of California, Berkeley, Press Release,
21 September 2004.
Gibbens RP, Heady HF (1964) ‘The influence of modern man on the
vegetation of Yosemite Valley.’ (Division of Agricultural Sciences,
University of California: Berkeley)
Gruell GE (1983) ‘Fire and vegetative trends in the northern Rockies.
Interpretations from 1871–1982 photographs.’ USDA Forest Ser-
vice, Intermountain Forest and Range Experiment Station General
Technical Report INT-158. Odgen, UT.
Hamilton JG (1997) Changing perceptions of pre-European grasslands
in California. Madroño 44, 311–333.
Harris MR (1927) A grazing management plan for the Berkeley Hills.
MS Thesis, University of California, Berkeley.
Heady HF, Child RD (1994) ‘Rangeland ecology and management.’
(Westview Press: San Francisco)
Hobbs RJ, Mooney HA (1986) Community changes following
shrub invasion of grassland. Oecologia 70, 508–513. doi:10.1007/
BF00379896
Holstein G (2001) Pre-agricultural grassland in central California.
Madroño 48, 253–264.
Hopkinson PJM (2003) Native bunchgrass diversity patterns and
phytolith deposits as indicators of fragmentation and change in
California Coast Range grassland. PhD Dissertation, University of
California, Berkeley.
Hornbeck D (1983) ‘California patterns: a geographical and historical
atlas.’ (California State University: Northridge)
Huenneke LF (1989) Distribution and regional patterns of Californian
grasslands. In ‘Grassland structure and function: California annual
grasslands’. (Eds LF Huenneke, HA Mooney) pp. 1–12. (Kluwer
Academic Publishers: Dordrecht)
Jacobs DF, Cole DW, McBride JR (1985) Fire history and perpetua-
tion of natural coast redwood ecosystems. Journal of Forestry 83,
494–497.
JonesTL (1992) Settlement trends along the California coast. In ‘Essays
on the prehistory of maritime California’. (Ed. TL Jones) pp. 1–37.
(Center for Archaeological Research, University of California:
Davis)
Keeley JE (1982) Distribution of lightning and man-caused wildfires
in California. In ‘Proceedings of the symposium on dynamics and
management of Mediterranean-type ecosystems’. (Eds CE Conrad,
WC Oechel) pp. 431–437. USDA Forest Service, Pacific South-
west Forest and Range Experiment Station General Technical Report
PSW-58. Berkeley, CA.
Keeley JE (1990) The California valley grassland. In ‘Endangered
plant communities of southern California’. (Ed. AA Schoenberg)
pp. 2–23. (Southern California Botanists: Fullerton, CA)
Keeley JE (2002) Native American impacts on fire regimes of the
California coastal ranges. Journal of Biogeography 29, 303–320.
doi:10.1046/J.1365-2699.2002.00676.X
Keeley JE, Rundel PH (2005) Fire and the Miocene expansion of
C
4
grasslands. Ecology Letters 8, 683–690. doi:10.1111/J.1461-
0248.2005.00767.X
Keeley JE, Fotheringham CJ, Morais M (1999) Reexamining fire
suppression impacts on brushland fire regimes. Science 284, 1829–
1832. doi:10.1126/SCIENCE.284.5421.1829
Lewis HT (1973) ‘Patterns of Indian burning in California: ecology and
ethnohistory.’ (Ballena Press: Ramona, CA)
Martin PS (1984) Prehistoric overkill: the global modal. In ‘Quaternary
extinctions: a prehistoric revolution’. (Eds PS Martin, RG Klein)
pp. 354–403. (University of Arizona Press: Tucson)
Martin RE, Sapsis DB (1995) A synopsis of large or disastrous wildland
fires. In ‘The Biswell symposium: fire issues and solutions in urban
interface and wildland ecosystems’. (Eds DR Weise, RE Martin)
pp. 35–38. USDA Forest Service, General Technical Report PSW-
GTR-158. Washington, DC.
Mayfield DW (1978) Ecology of the pre-Spanish San Francisco Bay
area. MA Thesis, San Francisco State University, San Francisco.
McBride JR (1964) Invasion of East Bay Regional Parks grassland
by Baccharis pilularis DC. MS Thesis, University of California,
Berkeley.
McBride JR (1970) Plant succession in the Berkeley Hills. PhD
Dissertation, University of California, Berkeley.
McBride JR (1974) Plant succession in the Berkeley Hills, California.
Madroño 22, 317–380.
McBride JR, Heady HF (1968) Invasion of grassland by Baccharis
pilularis DC. Journal of Range Management 21, 106–108.
McMinn HE (1916) The vegetation of Hamilton Gulch, California.
MA Thesis, University of California, Berkeley.
Nichols R, Adams T, Menke J (1984) Shrubland management for
livestock forage. In ‘Shrublands in California literature review
and research needed for management’. pp. 104–121. Contribu-
tion No. 191. (Water Resources Center, University of California:
Davis)
Pyne SJ (1995) ‘World fire.’ (Henry Holt: New York)
Roof J (1971) Growing California’s field wildflowers. Four Seasons
3(4), 2–24.
Russell WH, McBride JR (2002) Vegetation change and fire haz-
ard in the San Francisco Bay area open space. In ‘Proceedings
of the California’s 2001 wildfire conference: ten years after the
East Bay Hills Fire; October 10–12, 2001, Oakland, CA’. (Eds
KS Blonski, ME Morales, TJ Morales) pp. 27–32. Technical Report
35.01.462. (University of California Forest Products Laboratory:
Richmond)
Russell W, Tompkins R (2005) Estimating biomass in coastal Bac-
charis pilularis dominated plant communities. Fire Ecology 1,
20–27.
Safford HD (1995) Woody vegetation and succession in the Garin
Woods, Hayward Hills, Alameda County, California. Madroño 42,
470–489.
Sampson AW (1944) ‘Plant succession and burned chaparral lands in
northern California.’Agricultural Experiment Station Bulletin 685.
(University of California: Berkeley)
Sampson AW (1952) ‘Range management. Principles and practices.’
(John Wiley & Sons: New York)
Sampson AW, Jespersen BS (1963) ‘California range brushlands
and browse plants.’ Agricultural Experiment Station Manual 33.
(University of California: Berkeley)

296 Int. J. Wildland Fire J. E. Keeley
Stephens SL, Fry DL (2005) Fire history in coast redwood stands in
the northeastern Santa Cruz Mountains, California. Fire Ecology
1, 2–19.
Timbrook J, Johnson JR, Earle DD (1982) Vegetation burning by the
Chumash. Journal of California and Great Basin Anthropology 4,
163–186.
Wells PV (1962) Vegetation in relation to geological substratum and
fire in the San Luis Obispo quadrangle, California. Ecological
Monographs 32, 79–103.
http://www.publish.csiro.au/journals/ijwf
Westerling AL, Gershunov A, Cayan DR, Barnett TP (2002) Long
lead statistical forecasts of area burned in western U.S. wildfires
by ecosystem province. International Journal of Wildland Fire 11,
257–266. doi:10.1071/WF02009
Wilkinson L, Blank G, Gruber C (1996) ‘Desktop data analysis with
SYSTAT.’ (Prentice Hall: Upper Saddle River, NJ)
Williams K, Hobbs RJ, Hamburg SP (1987) Invasion of an annual
grassland in northern California by Baccharis pilularis ssp. con-
sanguinea. Oecologia 72, 461–465. doi:10.1007/BF00377580