significance of summer fog and overcast for drought stress and

ORIGINALARTICLE

Significance of summer fog and overcastfor drought stress and ecologicalfunctioning of coastal Californiaendemic plant species

Douglas T. Fischer1*, Christopher J. Still2,3 and A. Park Williams2

INTRODUCTION

Fog has often been identified as a crucial water source for

plants in a variety of ecosystems world-wide, including a

number of those identified as global biodiversity hotspots

(Hamilton et al., 1995; Olson & Dinerstein, 1998). Biodiversity

hotspots where fog is thought to play a particularly important

role in species distributions include tropical montane cloud

1Geography Department, California State

University Northridge, Northridge,2Department of Geography and 3Institute of

Computational Earth System Science,

University of California, Santa Barbara, Santa

Barbara, CA, USA

*Correspondence: Douglas T. Fischer,

Geography Department, California State

University Northridge, Northridge, CA 91330-

8249, USA.

E-mail: [email protected]

ABSTRACT

Aim Fog drip is a crucial water source for plants in many ecosystems, including a

number of global biodiversity hotspots. In California, dozens of rare, drought-

sensitive plant species are endemic to coastal areas where the dominant summer

moisture source is fog. Low clouds that provide water to these semi-arid

ecosystems through fog drip can also sharply reduce evaporative water losses by

providing shade. We quantified the relative hydrological importance of cloud

shading vs. fog drip. We then examined how both factors influence the range

dynamics of an apparently fog-dependent plant species spanning a small-scale

cloud gradient.

Location The study area is on Santa Cruz Island off the coast of southern

California. It is near the southern range limit of bishop pine (Pinus muricata

D. Don), a tree endemic to the coasts of California and Baja, Mexico.

Methods We measured climate across a pine stand along a 7 km, coastal–inland

elevation transect. Short-term (1–5 years) monitoring and remote sensing data

revealed strong climatic gradients driven primarily by cloud cover. Long-term

(102 years) effects of these gradients were estimated using a water balance model.

Results We found that shade from persistent low clouds near the coast reduced

annual drought stress by 22–40% compared with clearer conditions further

inland. Fog drip at higher elevations provided sufficient extra water to reduce

annual drought stress by 20–36%. Sites located at both high elevation and nearer

the coast were subject to both effects. Together, these effects reduced average

annual drought stress by 56% and dramatically reduced the frequency of severe

drought over the last century. At lower elevation (without appreciable fog drip)

and also near the inland edge of the stand (with less cloud shading) severe

droughts episodically kill most pine recruits, thereby limiting the local range of

this species.

Main conclusions Persistent cloud shading can influence hydrology as much as

fog drip in cloud-affected ecosystems. Understanding the patterns of both cloud

shading and fog drip and their respective impacts on ecosystem water budgets is

necessary to fully understand past species range shifts and to anticipate future

climate change-induced range shifts in fog-dependent ecosystems.

Keywords

California, Channel Islands, cloud shading, drought stress, evapotranspiration,

fog, Pinus muricata, range limits, Santa Cruz Island, water balance.

Journal of Biogeography (J. Biogeogr.) (2008)

ª 2008 The Authors www.blackwellpublishing.com/jbi 1Journal compilation ª 2008 Blackwell Publishing Ltd doi:10.1111/j.1365-2699.2008.02025.x

forests (65–75 million ha in the Neotropics alone; Kappelle,

2004) and the more than 25% of the world’s coastal areas that

receive regular fog (T. Dawson, personal communication,

2007). Foggy coasts with remarkable levels of species diversity

and endemism include mediterranean-climate regions of Chile,

south-western Australia, southern Africa and California.

Global climate change and land-cover change are already

altering fog patterns in fog-dependent ecosystems, with as yet

unknown effects on the range dynamics of rare species

(Pounds et al., 1999, 2006; Still et al., 1999). In California,

one of the areas of highest endemic and relict species richness

is within the ‘fog belt’ along the immediate coast, which is

subject to persistent summer fog and overcast conditions

(Raven & Axelrod, 1978).

Coastal fog and overcast conditions both result from low

stratus cloud formation. The distinction is that ‘overcast’ is a

cloud layer above the surface that obscures all (or most of) the

sky, while ‘fog’ is in contact with the surface and reduces

visibility to below 1 km (National Weather Service, 2007).

Crucially, the same stratus cloud that is fog at one site on a

coastal mountain can manifest as an overcast layer for sites at

lower elevations (Fig. 1). Both fog and overcast can reduce the

severity of plant drought stress during the summer rainless

season, though via different mechanisms. Previous studies in

the California fog belt have shown that stratus clouds can

enhance the availability of water to the ecosystem during the

rainless summer through fog drip, and some have demon-

strated uptake of fog water by endemic plants (Azevedo &

Morgan, 1974; Jacobs et al., 1985; Ingraham & Matthews,

1995; Dawson, 1998; Burgess & Dawson, 2004; Cole, 2005;

Corbin et al., 2005; Kennedy & Sousa, 2006). In areas of

frequent fog and favourable canopy structure, such as coast

redwood forests, fog drip has been shown to contribute

hundreds of millimetres of water to the soil surface annually

(e.g., Jacobs et al., 1985; Dawson, 1998).

The other mechanism by which stratus clouds can poten-

tially ameliorate plant water stress is by reducing ecosystem

water loss. This effect has received less attention from

researchers (but see Burgess & Dawson, 2004). Stratus clouds

can retard water loss by absorbing and reflecting large amounts

of solar radiation, thus greatly reducing net radiation at the

surface. This cloud shading leads to lower air and leaf

temperatures, increased relative humidities and smaller vapour

pressure deficits. The overall effect of these changes in

microclimate is to substantially reduce atmospheric evapora-

tive demand and potentially decrease ecosystem water losses

during the dry season. These reductions in water stress are even

thought to have been consistent enough over long time periods

to have allowed the evolution of larger leaf sizes in insular

species within the fog belt (Hochberg, 1980). Many of the relict

and endemic species throughout California’s fog belt are quite

sensitive to drought (Raven & Axelrod, 1978), and so any

change in summer water stress, whether from rainfall, fog drip

or cloud shading, is likely to affect species range limits and

long-term population viability.

Fog climatology

Like most of California, the coastal fog belt is characterized by

a mediterranean climate, with cool, rainy winters and warm,

rain-free summers. The annual summer drought ranges from 3

to 7 months, with a longer duration further south. Summer

cloud cover is most common along the immediate coast, with

both frequency and duration decreasing further inland. The

tops of the summertime marine stratus clouds along the coast

of California are most commonly below 400–500 m, which

roughly corresponds to the base of the coastal inversion

(Leipper, 1994). As a result of summertime inversion base

height dynamics, coastal sites above 400 m or so experience a

decreasing frequency of fog, sites below c. 200 m experience an

increasing percentage of stratus cloud events as overcast rather

than as fog, and sites at elevations of roughly 200–400 m

experience the most frequent fog. Coastal mountains generally

restrict the inland movement of the marine stratus layer to

areas within a few to 20 km of the immediate coast (with the

exception of some larger drainages). During stratus cloud

events, these clouds tend to form and/or intensify overnight

and then ‘burn off’ during the next day, usually evaporating

progressively from inland areas towards the coast (Marotz &

Lahey, 1975; Filonczuk et al., 1995; Fischer & Still, 2007).

Fog belt biogeography

Dozens of plant species are endemic to the fog belt. An

accepted explanation for these narrow, coastal distributions is

that widespread drought-sensitive species suffered range col-

lapse as the interior of California became drier and more

climatically variable following uplift of the Coast Ranges

during the Pliocene and Quaternary (Raven & Axelrod, 1978).

Figure 1 Conceptual model of differences in spatial distribution

of coastal cloud shading vs. fog drip. In southern California

summer fog drip is generally concentrated on west-facing slopes at

elevations of 200–400 m a.s.l. (with significant variability related

to funnelling of clouds by local topography). Cloud shading

decreases inland from the coast and on ridges that extend above

the typical heights of persistent stratus clouds. Drought-sensitive

pines at the study site are most abundant in areas at high enough

elevation to receive extensive fog drip, and close enough to the

coast to receive extensive cloud shading. Other foggy ecosystems

are likely to have different spatial separation between the two

processes of cloud shading vs. fog drip.

D. T. Fischer et al.

2 Journal of Biogeographyª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd

During this transition, summer rainfall was also eliminated

across most of California, with tremendous implications for

drought-sensitive species. One well-documented example of a

drought-sensitive species with such a relict distribution is the

coast redwood (Sequoia sempervirens D. Don). Many studies

have investigated the role of fog drip in maintaining stands of

coast redwood and other coastal species, particularly by

ameliorating the annual drought resulting from rainless

summers (Leyton & Armitage, 1968; Azevedo & Morgan,

1974; Ingraham & Matthews, 1995; Dawson, 1998; Burgess &

Dawson, 2004; Cole, 2005; Corbin et al., 2005). More recently

arrived or evolved species, such as Monterey pine (Pinus

radiata D. Don), may have spread in small pockets of equable

habitat along the coast during the Pliocene (Millar, 1986,

1999), expanding and contracting in response to repeated large

climate shifts throughout the Quaternary (Heusser, 1995).

After the Last Glacial Maximum, the southern range limits of

many California coastal species retreated tens to hundreds of

kilometres north as climate warmed (Chaney & Mason, 1930,

1933; Mason, 1934; Raven & Axelrod, 1978; Millar, 1999). All

of these fog-belt endemic species appear variously threatened

by the projected warming and drying of the climate in the

coming decades (IPCC, 2007). Also, regional climate model

simulations project changes to the upwelling regime along

California’s coast with climate warming (Snyder et al., 2003), a

phenomenon that has already been documented in other

eastern boundary coastal upwelling systems (Bakun, 1990;

Schwing & Mendelssohn, 1997; Mendelssohn & Schwing, 2002;

Snyder et al., 2003). Because fog formation is closely linked to

upwelling, these projections suggest alterations in coastal fog

dynamics. Changes in fog patterns, in addition to overall

drying and warming, may lead to large changes in the location

of suitable habitat for coastal endemic species.

Study goal

The goal of the current study was to investigate the separate

impacts of fog drip and cloud shading in maintaining the

ranges of coastal endemic species. Our first challenge in

meeting this goal was to quantify the degree to which shading

by clouds, independent of any fog drip, significantly reduced

summer ecosystem drought stress. We also quantified the

addition of water from fog drip and the resulting spatial

patterns in water availability. The second challenge was to

quantify the long-term effects of both cloud shading and fog

drip on the range dynamics of a fog-belt endemic species. We

utilized a combination of intensive field sampling, water

balance modelling and remote sensing to quantify these effects

in one part of the fog belt, on Santa Cruz Island off the coast of

southern California. Our study focused on a stand of the

drought-sensitive, fog-belt endemic species bishop pine (Pinus

muricata D. Don). This tree grows in a number of scattered

groves along the coast of central California, never more than

c. 30 km from the coast (Fig. 2a; Griffin & Critchfield, 1972;

Lanner, 1999). In the northern part of the range of bishop

pine, where rain and fog are greater and evapotranspiration is

generally lower, stands are larger and more numerous, and

trees are generally taller and grow more densely. Fossil records

of this and related species are likewise limited to the coast

(Axelrod, 1967; Raven & Axelrod, 1978; Millar, 1986). Our

study site on Santa Cruz Island focused on the southern-most

large stand of bishop pine.

Both the Materials and Methods and the Results sections are

organized by time-scale. We first discuss short-term monitor-

ing of water availability across our transect. These short-term

efforts to characterize spatial patterns combine both remote

sensing and extensive field measurements over 1–5 years. We

then describe the water balance modelling experiments we

conducted using 102 years of weather data to assess likely

long-term biogeographical impacts of these spatial patterns.

MATERIALS AND METHODS

Site description

Santa Cruz Island (SCI), the largest of the California Channel

Islands, is located c. 40 km south of Santa Barbara (Fig. 2a).

The island is 38 km long and is bisected by the rift valley of the

SCI fault. Steep east–west ridges rise both north and south of

the fault (Fig. 2b). Bishop pines grow in three stands on the

island, with approximate stand boundaries shown by white

polygons in Fig. 2c. These stands form the southern end of the

main distribution for this species and each stand is surrounded

by more xeric shrubland. (There is one small outlying stand in

northern Baja; Lanner, 1999). Within the SCI stand bound-

aries, pine densities are far from uniform. Our study was

confined to the large western stand. In this stand, the highest

pine densities occur near the centre (50–80% canopy cover).

Towards the eastern and western edges of the stand, pine

coverage decreases until there are only a few scattered trees in a

matrix of chaparral shrubs.

We installed a network of 12 weather stations spanning and

bracketing the western stand (Fig. 2b). The pines and the

weather stations are located primarily on north-facing slopes,

and site elevations range from 80–400 m. The stations are

numbered 1–12 from coast to inland and span a 7-km gradient

from the western-most to eastern-most pines within this stand

(Fig. 2b). Sites 1 and 10 were installed in the late 1990s, while

the remaining sites were installed between December 2003 and

April 2004.

Rainfall on SCI is highly variable (Fig. 3). Rainfall is also

highly seasonal, averaging more than 80 mm during each

month of the December–March rainy season (79% of the

average annual total of 509 mm falls during this period).

Rainfall averages <10 mm during each month of the May–

September dry season (4% of the average annual total falls

during these months). Rainfall is also highly variable for any

given month from year to year. On SCI, the standard deviation

of rainfall for a given month approximately equals the mean

rainfall for that month (except for February, which has a mean

rainfall of 116 mm and a standard deviation of 98 mm). The

interannual variability of annual rainfall is exceptionally high,

Fog and overcast conditions in coastal California

Journal of Biogeography 3ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd

similar to that of arctic and desert ecosystems that have much

lower total mean annual rainfall (Knapp & Smith, 2001).

Patterns in cloud cover

We evaluated the degree of cloud shading from fog and

overcast by analysing shortwave radiometer (pyranometer)

data from the weather stations along our transect (Licor

Li-200X, Li-190SB and Hobo S-LIA; LI-COR Environmental,

Lincoln, NE, USA and Onset Computer Corp., Bourne, MA,

USA). We then used remote sensing data to spatially extend

cloud cover information across SCI. Quarter-hourly and

average daily radiometer data from each station were used to

generate idealized ‘clear sky’ curves for each station. These

‘clear-sky’ radiation curves were manually generated as the

sum of two to three annual sine curves to fit the seasonal

pattern of observed clear days for each site (Fig. 4a). The

difference between modelled clear sky radiation and observed

radiation could then be attributed largely to cloud shading.

(While aerosols other than cloud droplets could affect surface

irradiance, their role is deemed insignificant in this area

compared with the frequent, thick stratus cloud layers.) We

calculated a daily ‘sunniness index’ for each station as 0–100%

of the expected clear-sky radiation for that station for that

date. On clear days, the sunniness index was 100% while on

cloudy days it frequently dipped below 30%.

Evapotranspiration modelling

Irrigation managers have standardized methods for calculating

evaporative demand as potential evapotranspiration (PET).

The PET is the amount of water that would evaporate from

(a)

(b)

(c)

Figure 2 Maps of Santa Cruz Island (SCI). (a) Location map has stars showing areas with stands of bishop pines (Pinus muricata D. Don; data

from Griffin & Critchfield, 1972). Note large gaps in range. (b) Elevation map of Santa Cruz Island (low elevations are dark, high elevations

light) shows our weather stations deployed on an east–west transect throughout the island’s main, western stand of bishop pines. Stations

are numbered 1–12, west to east. The island lies 30 km off the coast of southern California, is 38 km long, and reaches 753 m a.s.l. just

north of site 11. Site 10 is 437 m a.s.l. (c) Shading shows spatial variability in summer cloud cover, which is represented here as the average

percentage of summer mornings (3 July–30 September at 10:30 am Pacific Standard Time) that are cloudy for 5 years (2000–04) (from

MODIS satellite images; data from Williams, 2006). White polygons indicate approximate stand boundaries of bishop pine. In the western

stand, pine density is highest in the centre of the stand, with only a few scattered trees towards the eastern and western edges.



plants and soils under given conditions of temperature,

humidity, radiation and wind assuming unlimited water

availability. Many different PET formulae of varying complex-

ity have been developed over several decades. The most

rigorously tested are derivatives of Penman–Monteith (P-M),

which take into account temperature, relative humidity, solar

radiation, wind speed impacts on aerodynamic conductance

and vegetation properties that influence surface conductance

(McKenney & Rosenberg, 1993). These models are meant to be

run at hourly or daily time steps. The preferred model, when

high-quality weather data are available, is the latest P-M

derivative (Snyder & Eching, 2004; Allen et al., 2005). This

model (see also the Appendix) has been coded into a

downloadable, user-friendly spreadsheet (Snyder & Eching,

2004). We used this spreadsheet to estimate hourly reference

evapotranspiration from static site parameters and hourly

wind speed, solar radiation, temperature and dew point data.

The P-M equation we used calculates ‘reference ET’ (ETref)

for a ‘tall canopy’, similar to alfalfa (Snyder & Eching, 2004).

Since most vegetation transpires at different rates from fields of

alfalfa, a ‘crop coefficient’ is used to scale ETref up or down to

estimate PET for a specific vegetation type (Dunne & Leopold,

1978; Costello et al., 2000). Pines transpire significantly less

water than alfalfa, even if they are well watered (due to lower

hydraulic and stomatal conductivity and lower leaf-level

photosynthetic capacity; Larcher, 2003). Careful measurements

Figure 3 Monthly (bottom, grey line) and water year rainfall

measured at the main ranch on Santa Cruz Island, 1904–2005.

The water year runs from 1 October to 30 September. Note the

extreme variability in both monthly and annual rainfall. Of

1224 months, only 717 had any recorded rainfall. Fifty-eight

individual months had more than 50% of mean annual

precipitation of 509 mm (dashed line). More than 10% of the

century’s total rain came in the rainiest 13 months.

(a)

(b) (c)

Figure 4 Effects of cloud on solar radiation and evapotranspiration. (a) Temporal variability in short-wave radiation (SWR, 0.4–1.1 lm)

at site 7, which has the most complete meteorological record (December 2003–December 2005). Mean daily SWR (from quarter-hourly

measurements, light grey line) shows many days with dramatic reductions from modelled clear-sky conditions (top line). These reductions are

large, particularly around the summer solstice, even when examined through a relatively long 14-day moving average. (b) Temperature,

humidity, SWR and reference evapotranspiration (ETref; Snyder & Eching, 2004) are strongly affected by coastal stratus and fog events.

Environmental data are shown for a sequence of three contrasting days at site 7 on Santa Cruz Island: a clear day, a partly foggy day and

a mostly overcast/foggy day. Note the sharp reduction in solar radiation and ETref with the onset of fog late on 8 June followed by uniformly

high relative humidity and decreases in temperature variation (raised minimum and lowered maximum). Radiation, humidity and

temperature plus wind speed are the main drivers of reference evapotranspiration (ETref). (c) Differences in cloudiness account for

most of the difference in ETref between sites 1 and 12 (over the period 17 March–12 July 2005, during which driving data are available

at both sites). Daily SWR at both sites was expressed as a percentage of the expected clear-sky value (the ‘sunniness index’). This percent-

age at site 12 minus that at site 1 is plotted on the x-axis. Only a few days had substantially more cloudiness at site 12 than 1 (the days

with negative cloud gradient values). The y-axis shows the percentage reduction in daily ETref at site 1 compared with ETref at site 12.



of another maritime pine species (Pinus pinaster Ait.), when

well watered, showed transpiration rates to be 60% of P-M

ETref (Granier & Loustau, 1994). We have adopted this value

to estimate PET for bishop pines (and the other plants in the

surrounding chaparral community, all of which have similar or

lower water requirements; Costello et al., 2000). In the

remainder of this paper, we will use ETref to refer to the

values predicted by the P-M equations and PET to mean ETref

values that have been scaled down to 60% to estimate

evapotranspiration for bishop pine stands. (The precise value

used to scale PET affects patterns of water availability across

the transect in absolute but not relative terms).

The dense network of weather stations across our transect

allowed us to quantify the gradient in evaporative demand. We

calculated hourly PET for each of the sites that had a

functional datalogger for the warm season (March–October) of

2005. We chose this time period because it brackets the period

of maximum drought stress. It also includes the period of

maximum tree growth rates as observed by dendrometers

installed across the transect; in 2005 and 2006, the maximum

growth rates occurred between March and July, with variations

related to the timing of final rainfall events of the preceding

rainy season (Williams, 2006).

Water balance modelling

In order to extend our analysis across more years than those

for which we had weather data with high spatial and temporal

resolution, we used 102 years of monthly temperature and

rainfall records to drive a simple water balance model. This

model accounts for interactions among rainfall, PET and soil

moisture storage, and it estimates both actual evapotranspi-

ration (AET) and Deficit, which is a measure of drought stress.

Running this water balance model for the last century provided

insights into how cloud patterns may have driven spatio-

temporal differences in water availability and drought stress.

The water balance model we used was modified from

Thornthwaite & Mather (1955, 1957), and was driven by

monthly precipitation and estimated monthly PET. All other

terms were calculated from these two quantities as follows (in

units of mm of water per unit area; see the Appendix).

Incoming precipitation left the system via run-off or evapo-

transpiration with some intermediate soil moisture storage. On

a monthly basis, a small percentage (3%) of precipitation was

diverted immediately to run-off as storm flow (Dunne &

Leopold, 1978). This low percentage diverted to monthly

storm flow was, if anything, a conservative underestimate that

increased the relative importance of rain water compared with

fog drip in the model. It was a similar value to that used in an

earlier calibrated run-off study in coastal California (Fischer

et al., 1996), and model results were not sensitive to this

parameter up to storm flow values of at least 20%. The rest of

the precipitation became available to increase AET up to the

evaporative demand limit of PET.

In wet months, excess precipitation (after AET equalled

PET) recharged stored soil moisture until field capacity was

reached. After that, any remaining precipitation was diverted

to run-off (above and beyond storm flow). We estimated a

maximum soil moisture storage of 500 mm of water in the

rooting zone (equivalent to a functional rooting depth of 1 m

with 50% soil porosity), based on soil textures, rooting depths

estimated from soil pits and road cuts, and standard tables

(Dunne & Leopold, 1978).

In dry months, when precipitation was less than PET, AET

equalled precipitation plus withdrawn soil moisture. Soil

moisture was made available for AET as a negative exponential,

so that it became increasingly difficult to draw moisture from

the soil as it dried (Thornthwaite & Mather, 1955, 1957;

Dunne & Leopold, 1978; Fig. 5b). This negative exponential

roughly captured the increasingly negative soil water potential

that occurs as soil water content is decreased. If there was not

sufficient plant-available soil moisture, the shortfall between

PET and AET was termed Deficit. Note that in the water

balance model we initially made no allowance for additions of

water to the soil by fog drip, so that the only water input was

from rainfall.

Many water balance studies include additional terms that we

omitted. In particular, canopy interception is generally less

important in this ecosystem than in many others as the rain

comes in a relatively few, intense storms. Groundwater flow

and delayed run-off were deemed unimportant for determin-

ing plant-available water on the relatively steep slopes that

typify SCI and the Californian coast ranges.

Constructing long-term time-series data

The water balance model required monthly inputs of precip-

itation and PET. Rainfall at the main ranch on SCI has been

measured continuously since 1904 (Fig. 2, Laughrin, 2005).

However, rainfall can vary significantly over short distances in

the coast ranges of California due to the rugged topography.

To account for this spatial variation in rainfall, we regressed

monthly totals at sites 1 and 10 against the long-term record

from the main ranch (n = 76 for the period from 1998 to 2005,

with gaps). Site 10 received the same amount of rainfall as the

ranch (100.4%, R2 = 0.92). Site 1 received 70% as much

rainfall as the ranch (R2 = 0.93). For Site 12, we regressed daily

rainfall totals at Site 12 to daily totals at Site 10 (as a proxy for

the ranch). Site 12 received 92% as much rainfall as Site 10

(n = 99 rainy days, 6 April 2004 to 11 November 2005,

R2 = 0.96). For the model runs below we used rainfall

estimated for each site (i.e. ranch rainfall multiplied by 0.70,

1.0 or 0.92 for Sites 1, 10 and 12, respectively).

There are no long-term records of ETref for our site, and

weather records sufficient to calculate P-M ETref extend back

only to 1998 (Wall, 2005). Fortunately, there are many

alternative formulae for calculating ETref, including the

Blaney–Criddle (B-C) formulation, which requires only mean

monthly temperature and latitude (Brouwer & Heibloem,

1986). We calculated monthly B-C ETref using mean monthly

temperatures from Santa Barbara, 40 km north across the

Santa Barbara Channel, from 1898 to 2005 (National Weather



Service, 2005). We also calculated hourly P-M ETref from 1998

through 2005 (with some gaps) for sites 1 and 10 and for 2005

at site 12. When we regressed monthly B-C ETref against the

monthly sum of hourly P-M ETref, the agreement was very

good (Fig. 5c). We then used these regressions to hindcast

monthly ETref (and PET) time series for sites 1, 10 and 12 back

to 1904.

Representing fog drip at each site was more difficult. We

collected fog drip at each site along the transect 2004–05 and

found consistent spatial patterns related to local topography,

generally increasing with elevation (Fischer & Still, 2007).

Temporal patterns were more difficult, as the summer of 2005

had much more fog drip than 2004. To put these 2 years of

spatially explicit data into a longer temporal context, we

employed nearby airport weather records. Williams (2006)

developed an index of summer fogginess based on the number

of hours per day (between midnight and 10 am) when weather

observers at Oxnard and Santa Barbara airports reported cloud

bases of different heights. These airports are 59 km east-north-

east and 46 km north of site 10, respectively, across the Santa

Barbara Channel and within 11 m of sea level. Summer stratus

cloud layers tend to be relatively level across the Santa Barbara

Channel so that months with many hours of low cloud on the

mainland generally also have many hours of low cloud on SCI.

The frequency with which cloud bases are observed below the

altitude of a given site is related to the amount of fog drip,

because fog drip can only occur at sites higher than the local

cloud base. The fogginess index based on these airports is not a

perfect proxy for fog drip at our sites on SCI for several

reasons, but it does allow at least some insight into how our

recent fog drip records fit into the long-term picture for the

local region. Further, fog collection and solar radiation data

collected at site 7, as well as tree-ring widths on Santa Rosa

Island, are significantly and positively correlated with this

fogginess index (Williams et al., 2008). The fogginess index is

available for 49 years during which sufficient observations

were recorded (1944–2005 with some gaps). For cloud bases of

400 m and lower, the fogginess index varied from 3.3 to 8.3 h

per day. The summers of 2005 and 2004 were 32% and 10%

foggier than average, respectively. To estimate ‘typical’ fog drip

inputs, we adjusted downward the cumulative 2005 and 2004

summer fog collection values at each site by 32% and 10%,

respectively. The results from both years were nearly identical

(± 10 mm) at each site (data from Fischer & Still, 2007; typical

values are listed in Table 1).

There was no strong correlation between the fog index and

either of our long-term data sets (rainfall and PET), so

building a continuous fog time series was problematic. Instead,

we added the same ‘typical’ fog drip values to the water balance

in every year. We added that annual fog drip in the model as

extra monthly precipitation from June to September (distrib-

uted as 10%, 40%, 40%, 10%, respectively) to match observed

seasonal patterns of fog drip (Goodman, 1985; Estberg, 2001;

Ruiz, 2005; Fischer & Still, 2007). Given the limited data

(a)

(b)

(c)

Figure 5 (a) Water balance calculations (following Thornthwaite & Mather, 1955) from averages of 1904–2005 rainfall (R) and potential

evapotranspiration (PET). Note that rainfall is restricted to the late autumn to early spring period. Stored soil moisture (SM) peaks in

early spring and declines over the dry season. Note that SM represents an amount of water remaining in the soil, while other variables

are fluxes of water per month. SM is distinct from plant-available soil moisture, as described in the text. Deficit is the difference between PET

(black squares) and actual evapotranspiration (AET, grey circles), and is not plotted explicitly. (b) Water balance SM as a function of

cumulative deficit. In wet soils near the field capacity (500 mm in this case), SM drawdown to meet evaporative demand is nearly

linear 1 : 1. As the soil dries, SM becomes proportionally less available. When the soil dries to 50% of field capacity it only releases 50%

of the monthly deficit applied to it. At 20% (100 mm) it releases only 20% of the deficit applied. (c) Monthly reference evapotrans-

piration (ETref) from the Santa Barbara temperature record (Blaney–Criddle, grey lines) vs. summed hourly ETref calculated from site

temperature, humidity, radiation and wind speed (Penman–Monteith, black lines) when available driving data exist at sites 1, 10 and 12.

The Blaney–Criddle ETref captures the seasonal trend and many of the excursions (R2 = 0.92, 0.93 and 0.96, respectively). Note the

overlapping scales on the y-axis, and the increase in ETref from site 1 (coastal) to site 12 (7 km further inland).



available and large uncertainties involved, more complex

modelling of temporal variability in fog drip did not seem

justified and would be likely to introduce additional errors in

our analysis.

Long-term modelling experiments

We ran the water balance model for the three sites (sites 1,10

and 12 from Fig. 2a) for the period 1904–2005 to look for

temporal and spatial patterns in cumulative annual Deficit. All

runs used the observed spatial patterns in rainfall across the

stand. There were four runs in total. Run A was the reference

case, with which the other runs were compared. For this run,

no fog drip was added at any site, and the effect of cloud

shading on PET was held constant at all three sites (using

modelled PET from the least cloudy, inland site – site 12).

Thus, differences in Deficit across sites for this run were due

solely to rainfall variations as described in the previous section.

For model runs B and C, fog drip and cloud-driven PET

variations were modelled separately to quantify their individ-

ual impacts, while run D incorporated both fog drip and PET

variations to simulate their combined impacts on water

balance. Specifically, run B incorporated reduced PET at sites

1 and 10 as observed in the weather record (due to increased

relative humidity and cloud shading at these more coastal

sites) but did not incorporate fog drip at any site. Run C used

PET from site 12 (inland) at each site but also included

‘typical’ summer fog drip amounts at each site. Run D

included both fog drip and PET variations across the three sites

(results are summarized in Table 1).

RESULTS

Spatial and temporal patterns of cloud cover

We wanted to know whether differences in frequency and

intensity (opacity) of cloud shading across the transect were

significant, and if so whether they appreciably reduced water

loss through evapotranspiration. Cloud cover varies seasonally.

The radiation record from site 7 (near the centre of the stand

and having the most complete record) shows that the

frequency and thickness of fog and cloud cover varied

substantially over time (Fig. 4a). There were periods of

pronounced reductions in mean daily solar radiation

(W m)2), particularly around the June solstice (locally referred

to as the ‘June gloom’). These radiation reductions were large

and persistent, even when viewed over a 14-day running

average.

Summer cloud shading decreases from the coast inland. This

gradient of decreasing cloud shading from the coast is

illustrated in Fig. 2c by a composite of morning MODIS

satellite images across five summers (averaged across 2000–04).

The resulting image shows the percentage of days that are

cloudy at 10:30 am Pacific Standard Time (PST) (Williams,

2006). At the western coast, c. 50% of days were cloudy at

10:30 am, while in the centre of the stand that figure decreased

to 35% and to <30% near the last few pines at the eastern

(inland) edge of the stand (Fig. 2c). This metric serves as a

proxy for overall cloud shading because clouds and fog tend to

form or intensify overnight, and burn off towards the coast the

next morning, so that the spatial cloud patterns captured by

morning MODIS images are representative of mean daily

cloud patterns (Fischer & Still, 2007). The radiometer records

for summers 2004–05 from our weather stations across the

western stand show the same pattern of a steady decrease in

Table 1 Site characteristics and model results. Model run

(a) takes into account differences in rainfall between the sites and

serves as a reference point for subsequent model runs. Deficit is

the difference between potential evapotranspiration (PET) and

actual evapotranspiration (AET). Model run (b) shows that

reduced coastal PET decreases both average and maximum

drought stress by about a third compared to just 7 km further

inland. Run (c) shows that the average amount of fog drip received

at higher-elevation sites is sufficient to reduce drought severity by

19–36% (compared to run a). Run (d) includes both fog drip and

PET reduction. The combined effect is a 36–56% reduction in

average deficit. At site 10 (which receives both persistent cloud and

significant fog drip) only 1989–90 had deficits > 400 mm, while

such severe droughts were common at site 1 (little fog drip) and

site 12 (little cloud cover).

Site characteristics Site 1 Site 10 Site 12

Distance inland (km) 2.1 7.1 9.2

Elevation (m) 61 437 387

Avg. annual PET (mm) 681 766 916

Avg. annual rain (mm) 356 509 468

(as % of Ranch rain) 70% 100% 92%

Est. typical fog drip (mm) 10 139 178

(a) Reference case (holding PET constant, without fog drip)

Avg. AET (mm) 344 468 439

Avg. deficit (mm) 572 448 477

Max. deficit (mm) 809 753 768

(b) Effect of DPET only (without fog drip)

Avg. AET (mm) 339 449 439



Reduction of Avg. deficit by DPET 40% 29% –%

Reduction of Max. deficit by DPET 31% 21% –%

(c) Effect of fog drip only (holding PET constant)

Avg. AET (mm) 354 597 609



Reduction of Avg. deficit by fog drip 2% 29% 36%

Reduction of Max. deficit by fog drip 1% 19% 24%

(d) Effects of DPET and typical fog drip

Avg. AET (with fog drip) (mm) 349 570 609

Avg. deficit (with fog drip) (mm) 332 197 307

max deficit (with fog drip) (mm) 551 441 582

Reduction of Avg. deficit by DPET

plus fog drip

42% 56% 36%

Reduction of Max. deficit by DPET

plus fog drip

32% 41% 24%

No. annual deficits > 400 mm

(1904–2005)

21

years

2

years

20

years



cloud frequency and opacity from west to east for all hours of

the day (not shown).

Spatial and temporal patterns of evapotranspiration

Calculated values of ETref across the transect showed that,

with small exceptions related to local aspect, ETref increased

consistently from the coast to inland. Throughout the 2005

warm season (March–October), hourly calculations of ETref at

each site showed that seasonal mean ETref was 25–30% lower

at the coastal sites than at the inland sites.

We tested the role of clouds in reducing ETref at individual

weather stations. For example, clouds significantly reduced

ETref at site 7 for three adjacent days in June 2005 (Fig. 4b).

Hourly shortwave radiation observations show that 7 June was

clear all day, while 8 June was cloudy in the afternoon, and 9

June was cloudy for most of the day. Temperature varied

diurnally by 5�C on 7 June and early on 8 June, but then

temperature dropped sharply as the cloud layer arrived in the

late afternoon. Temperature variation was much more muted

for the rest of the period. Similarly, relative humidity showed a

moderate daily cycle until the arrival of the cloud layer, at

which point relative humidity rose to 100% and remained

unchanged for the rest of the period. These three variables,

plus wind speed, are used in the P-M ETref formulation

(Snyder & Eching, 2004; Allen et al., 2005).

The reduction in daily total ETref from 7 June (4.5 mm) to

9 June (1.5 mm) was 66%. This was due to the combined

effects of reduced solar radiation, reduced midday tempera-

tures and increased relative humidity, all of which were driven

by increased clouds (wind speeds averaged c. 3 m s)1 on both

days). Reduced short-wave radiation not only limits the energy

required for vaporization of water inside the leaf and at the soil

surface, but it reduces surrounding air temperature and vapour

pressure deficit. To test the relative impact of each weather

component on ETref, we substituted hourly 9 June variables

into the 7 June calculations one at a time. Low solar radiation

(from 9 June) reduced ETref by 40% (holding all other

variables at 7 June values). Increased relative humidity reduced

ETref by 22%. Substituting temperature from 9 June (lower at

midday, but higher at night) reduced ETref by 2%. Therefore,

the primary driver of changes in ETref at a given site, under the

relatively narrow ranges of temperature and relative humidity

that typify our sites and the fog belt in general, is variation in

solar radiation. Aerosols other than cloud droplets also reduce

incoming solar radiation, but in this area of frequent thick

stratus cloud cover their effects are comparatively small.

The spatial gradient in ETref across the transect is also

driven primarily by variations in radiation. Across a landscape,

radiation variations are strongly driven by aspect, but among

our sites, primarily on north-facing slopes, we hypothesized

that the strongest driver of radiation differences (and therefore

differences in ETref) was the observed gradient in cloud

shading. To test this hypothesis, we compared the daily

coastal/inland gradient in ETref with the daily cloud-driven

gradient in solar radiation (comparing sites 1 and 12, at

opposite ends of the transect). We used the longest period of

continuous hourly data, 17 March–12 July 2005 (which also

generally coincided with the period of maximum tree growth,

mid-March to mid-July; Williams, 2006). We calculated a daily

cloud shading gradient as the sunniness index (0–100%) at site

12 (inland) minus site 1 (coastal). We also calculated a daily

ETref gradient as the percentage difference between ETref at

site 12 minus site 1 (Fig. 4c). The cloud shading gradient

explained most of the difference in daily ETref. Mean daily

ETref over this period was 4.9 mm at site 12, and 4.0 mm at

site 1. Departures from the linear regression (of cloud shading

gradient vs. ETref difference) are primarily on days with

unusual altitude effects (e.g. a strong temperature inversion

between Site 1 at 61 m and Site 12 at 387 m, or rain only at the

higher site). The small positive intercept of the regression

indicates a slightly higher ETref at site 12 even when cloud

shading is equal, resulting from the slightly greater continen-

tality of this site. In summary, ETref is significantly lower near

the coast than just a short distance inland. Despite differences

in elevation and continentality, most of that difference can be

attributed to the observed differences in solar radiation

resulting from increases in frequency and intensity of cloud

shading near the coast.

Spatial and temporal patterns of fog drip

Fog water was collected and volumes were logged at each site

with harp-style passive fog collectors (Fischer & Still, 2007).

There were large spatial differences in the amounts of fog water

collected during the 2005 dry season (10 May to 16 October;

Fig. 6a). Site 1, at the lowest elevation, received a negligible

amount, while sites 10 and 12 received substantial fog water

inputs (Fig. 1). At site 10 we calibrated the fog collector against

a throughfall collector placed immediately under a mature pine

canopy. The regression estimated 30 mm of throughfall per

litre of collected fog water, R2 = 0.74 (Fischer & Still, 2007).

This calibration allowed us to estimate how much fog drip a

mature pine might collect at each site, irrespective of current

pine abundance at each site. Under areas of similarly dense

pine canopy, the fog water collected at site 10 could generate of

the order of 175 mm of throughfall, or 34% of the mean

annual rainfall at this site, using the above calibration.

Assuming a similar relationship between throughfall and the

fog collector at site 12 (which collected more fog water in

fewer, larger events in 2004 and 2005; Fischer & Still, 2007),

throughfall from fog drip might approach 225 mm. It should

be noted that the long narrow leaves of conifers are more

effective at collecting fog drip than the leaves of many other

plants (Goodman, 1985).

Previous work has shown that fog drip can penetrate into

the soil and increase soil water potential (Cole, 2005). We

installed paired sets of soil water potential sensors from depths

of 2–15 cm both inside and outside the tree canopy at site 7

(Fig. 6b). The two soil profiles are c. 10 m apart. The soil

sensors outside the canopy had gone off scale as the soil dried

(i.e. below )1.5 MPa) within 2 weeks after the last rain of the



season on 9 May 2005. By contrast, under the tree canopy,

repeated fog drip events throughout the summer kept the

upper soil moist into mid-August. We dug a soil pit under a

canopy on an adjacent slope on 2 September 2005, and found

the soil quite wet to the touch at the bottom of the soil profile

(50 cm depth). This was 4 months after the last rain.

Water balance for an ‘average’ year

Modelling plant water availability during the growing season is

more complicated than just assessing annual rainfall and PET.

Other important factors include the timing of rainfall and

seasonal changes in soil moisture storage. For areas with highly

variable rainfall, year-to-year changes in soil moisture must

also be considered. Given the extreme variability of rainfall

throughout southern California both within and among years

(Fig. 3), it is important to consider a broad suite of both wet

and dry years in order to estimate how observed differences in

quantities like PET actually affect availability of water to

plants. Years with very high rainfall in a few storms may not

provide as much summer moisture as years with moderate

rainfall that occurs more evenly, or that falls later in the season.

Water availability in low-rainfall years is strongly dependent

on how wet the previous year was. Also, in very wet years,

changes in PET have less of an impact on plant drought stress.

Tracking this temporal variability required a running water

balance in which monthly soil moisture carried over from

1 month to the next across years.

To calculate long-term effects of the observed spatial pattern

of cloud shading and fog drip, we used a running water balance

driven with monthly rainfall and PET time-series data from

1904 to 2005. Figure 5a shows the water balance for an

‘average’ year at site 10. Rainfall and PET are long-term site

averages from these time series. Rainfall averages

509 mm year)1 and is concentrated during the winter months,

with almost no rain from May to October. Rainfall surplus

(beyond evaporative demand) allows for recharge of soil

moisture from December to March. Run-off is a minor part of

the water balance in the ‘average’ year. PET averages

766 mm year)1; it is low during the winter and peaks in late

summer. AET peaks in spring as temperature and insolation

are rising, but before soil moisture becomes limiting. AET is

higher than rainfall from April to October, by an amount that

equals the drawdown of soil moisture for that month. Still, the

available soil moisture is not sufficient to match the evapo-

rative demand (Fig. 5b). Deficit is greatest in August, but

remains large from June to October.

Both cumulative annual AET and cumulative annual Deficit

have been shown to be ecologically meaningful predictors of

plant distributions for species, including other Californian

conifers (Stephenson, 1998). We tested AET from this running

water balance model as a predictor of tree growth based on a

tree-ring chronology developed on torrey pines (Pinus torre-

yana ssp. insularis Haller) on adjacent Santa Rosa Island, and

found AET to be a better predictor of tree growth than rainfall

(Williams, 2006; Williams et al., 2008). Annual Deficit appears

to correlate with mortality risk for bishop pines as well (see

Discussion).

We compared site 10, near the heart of the stand where pine

density is highest, with sites 1 and 12, which bracket the coastal

and inland ends of the stand as described previously. We

compared ‘average’ water balances for the three sites, holding

(a)

(b)

Figure 6 Fog drip inputs and the impact of fog drip on soil moisture. (a) Cumulative fog water collected at seven sites (using the

harp fog collector pictured) over the 2005 summer dry season shows that higher elevations (black lines, 300–440 m) receive more fog

than lower sites (60–200 m). Site 7 is in a topographic funnel south of the main transect and gets significantly more fog than the other sites.

(b) Soil water potential was logged hourly at site 7 in two profiles 10 m apart, just outside (O, grey lines) and inside (In, black lines)

the tree canopy (dotted lines are closer to the surface). Surface soils (0–15 cm) dried substantially prior to the last rains of the wet

season (28 April and 9 May). The soil outside was drier than the permanent wilting point ()1.5 MPa) within 2 weeks of the last rains.

Repeated fog drip events throughout the summer (grey bars) kept the soil under the tree canopy moist until late August. Daily fog drip

at site 7 is estimated from fog collection volumes (see text).



rainfall constant and varying only PET (Fig. 5c; variations in

PET reflect decreasing cloud shading from the coast inland).

Lower PET values towards the coast resulted in decreased AET

in winter. This allowed increased soil moisture recharge, which

increased soil moisture availability all year round, and

therefore increased dry season AET. Monthly Deficits were

smaller throughout the dry season, due to both decreased PET

and increased AET. This pattern illustrates how PET can affect

AET and Deficit, but it is important to remember that the use

of an ‘average’ year is problematic given the extreme interan-

nual variability of rainfall in California (McGuirk, 1982;

Schonher & Nicholson, 1989; Fig. 2).

Long-term running water balance

Figure 7 shows 10 years of running water balance estimates for

site 10, spanning two of the most extreme years of the century.

Water year 1983 (WY 1983, 1 October 1982 to 30 September

1983), associated with a very large El Nino–Southern Oscil-

lation (ENSO) event, was the fifth wettest year of the century.

Soil moisture reached the estimated rooting zone capacity of

500 mm, and AET remained high throughout the summer. In

contrast, WY 1989 and WY 1990 had the fifth lowest and

second lowest rainfall totals of the century, respectively.

Further, each water year during the period 1987–1991

experienced <80% of the long-term average of 510 mm

rainfall, resulting in a prolonged drought that was unprece-

dented in the historic record. (It should be noted that the

historic record up to the mid-1980s shows substantially fewer

and less intense droughts than inferred for the previous 300–

500 years based on tree-ring analyses; Michaelsen et al., 1987;

Haston & Michaelsen, 1994, 1997). The largest annual Deficits

of the century were in 1989 and 1990 (Fig. 7). (Water year

rainfall was analysed using water years ending 30 September so

as to keep entire rainy seasons together. Annual Deficit was

analysed using calendar years, as monthly Deficits often

continue from summer through the autumn.) Stored soil

moisture was extremely low in these years, especially WY 1990,

so that AET was essentially the same as monthly rainfall.

Modelled PET did not vary substantially over the 10-year

period (Fig. 7).

We also examined the long-term (1904–2005) running water

balance for three sites (1, 10 and 12) to quantify interactions

among rainfall, fog drip and PET in determining annual

Deficit, which we took as a proxy for drought stress and

mortality risk. As described previously, run A was the reference

case (no fog drip and cloud shading/PET held constant across

sites); run B included variations in cloud shading/PET but did

not include fog drip; run C included ‘typical’ amounts of

summer fog drip at each site but held cloud shading/PET

constant across sites; and run D included both fog drip and

variations in cloud shading/PET.

In run B (effects of cloud shading on PET only), average

Deficits decreased by 29% at site 10 and 40% at site 1 relative

to run A (control case; Table 1). At site 1, the reduction in PET

compared with site 10 was sufficient to make up for the 30%

lower rainfall received at the low elevation of site 1. Average

Deficit at site 1 across the entire record was 572 mm under run

A; reduction in Deficit from cloud shading under run B

regularly exceeded 200 mm at this site. This made Deficits at

site 1 comparable to site 10, even though the low elevation of

site 1 reduces rainfall by 30%. The reduction in Deficit at both

sites was greater in dry years than wet years, reducing overall

climatic variability. Another measure of variability we exam-

ined is the recurrence interval of severe droughts. At site 1, the

1987–91 drought resulted in the highest modelled Deficit of

561 mm. At site 12, this extreme Deficit was exceeded in

30 years out of 102, and by as much as 207 mm. This

comparison shows that observed changes in PET over a short

distance can have large impacts on climatic variability and

consistency of habitat suitability over time.

In run C (effects of typical fog drip only), the low elevation

of site 1 resulted in negligible fog drip, while sites 10 and 12

both received large amounts (139 and 178 mm, respectively)

resulting in reductions in Deficit of 29% and 36% at sites 10

and 12, respectively (Table 1, Fig. 8). The reductions in

average Deficit from fog drip are similar in magnitude

to those from reduced PET, but occur primarily at the

Figure 7 Running water balance for site 10, Santa Cruz Island 1983–92. This decade started with one of the rainier years on record during

the 1982–83 El Nino event. The most severe multi-year drought of the century began in 1987, culminating in the driest year of the

century in 1990. Potential evapotranspiration (PET) is relatively consistent from year to year. Actual evapotranspiration (AET) varies

much more, depending on both monthly rainfall (R) and soil moisture (SM). Deficit (Def) is the difference between PET (black

squares) and AET (grey circles). Annual Def is also shown in grey bars plotted against the right axis; it varies greatly from year to year.



higher-elevation stations (sites 10 and 12) rather than at the

more coastal ones. Including fog drip in the water balance

shifts the prediction of the most drought-prone habitat to low-

elevation coastal habitat, with both high-elevation sites being

more mesic.

In run D (combined effects of fog drip and cloud shading),

the impacts of fog drip and cloud shading were essentially

additive (despite minor differences related to timing and

interactions among variables). Annual Deficits were reduced

by about a third at higher elevation (sites 10 and 12) as a result

of higher fog drip. Annual Deficits were reduced by about a

third near the coast (sites 1 and 10) by reduced PET from

increased cloud cover. Sites 1 and 12 showed similar levels of

Deficit for most years. Site 10 (with both abundant fog drip

and low PET) had the lowest mean Deficit (a 56% reduction in

mean Deficit from run A), and the lowest frequency of

droughts exceeding any given threshold (Fig. 8; Table 1).

DISCUSSION

Drought sensitivity of bishop pines

Bishop pine, like many rare and endemic species of the fog

belt, is sensitive to drought. The most severe 2-year drought of

the last century on SCI (and in most of CA) occurred in WY

1989–90 (Fig. 3; WY 1989 is 1 October 1988 to 30 September

1989). This drought was followed by widespread pine mortality

across the island over the next few years – estimated as 70–90%

of individuals (Walter & Taha, 1999). The previous large 2-

year drought of WY 1976–77 was also followed by widespread

(a)

(b)

(d)

(c)

Figure 8 Annual deficits for three sites on Santa Cruz Island under four model runs. Deficit is the difference between potential evapotrans-

piration (PET) and actual evapotranspiration (AET), with drier years extending towards the bottom of the graph. Run (a) is the reference

case, ignoring fog drip, and using PET from site 12 (inland) for each site. Differences in deficit result from lower and higher rainfall at sites

1 (coastal) and 10, respectively. Run (b) incorporates reduced PET from greater cloud cover at sites 1 and 10 along with observed rainfall

patterns, but still has no fog drip. Note the marked decrease in severity of both average and extreme droughts at both sites compared to (a).

Run (c) ignores PET, using site 12 PET at each site, but adds ‘typical’ summer fog drip at each site. The low elevation of site 1 results in

negligible fog drip, while sites 10 and 12 both receive large amounts resulting in large reductions in deficit. Run (d) includes both fog drip

and reduced PET. Annual deficits are reduced by about a third at higher elevation (sites 10 and 12) by higher fog drip. Summer deficits are

about a third lower near the coast (sites 1 and 10) due to reduced PET from cloud cover. The result is that sites 1 and 12 have similar rates

of drought frequency and severity, and site 10 (with both abundant fog drip and low PET) has the least drought.



pine mortality (M. Carroll, J.R. Haller, L. Laughrin, personal

communication, 2006). In both instances, the proximal cause

of the die-off is reported to be bark beetle infestation. Bark

beetles, fungi and other pathogens often cause mortality in

pines in the years following drought because the weakened

trees are unable to produce sufficient pitch and defensive

chemicals (Richardson, 1998). During severe droughts, even

relatively healthy pines may succumb to disease as large

concentrations of weakened trees provide abundant breeding

ground for pathogens (Richardson, 1998; Breshears et al.,

2005; White et al., 2006).

Range limits are necessarily determined by a balance

between recruitment and mortality. On the recruitment side,

seedlings are abundant throughout the stand in most years,

even following drought when mature trees are dying en masse

(Wehtje, 1991; Walter & Taha, 1999). Prior to the removal of

feral sheep in the 1990s, pine recruitment was somewhat

limited as most seedlings were eaten (Hobbs, 1980). However,

despite intensive sheep grazing, starting in the 1850s, a 1929

aerial photograph suggests that overall range limits of the

western stand that we studied (Fig. 2) were similar to the

present day (J. Howarth, personal communication). This

suggests that range limits are less influenced by recruitment

than mortality.

On the mortality side, the die-off of the early 1990s appears

to have been spatially selective. A 1983 photograph shows large

numbers of mature pines between sites 11 and 12 prior to the

1987–91 drought (M. Carroll, personal communication). In

2006, none of those trees remained alive, or even standing

dead, and all of the pines east (inland) of site 11 appear to have

established since 1990. Also, the entire area west (coastward) of

site 4 contains few, if any, pines that pre-date the last drought.

Around sites 7 and 10 older pines are fairly common, although

the oldest pines we could find to core at both sites date back to

just 1952. This suggests another die-off and recruitment pulse

following the severe drought of the late 1940s (Fig. 3).

Drought-mediated mortality appears to be the main driver

of range limits at least in the western stand, with some range

expansion in wet years, and range reduction (or density

reduction, especially near the edges) following droughts.

Direct effects of fog drip and cloud shading

Fog drip is taken up by coastal pines and contributes to tree

growth. Uptake of fog water has been demonstrated using

isotopic techniques (Fischer, 2007). Increased growth is

apparent in a tree-ring chronology developed from torrey

pines 15 km west of site 1 on adjacent Santa Rosa Island

(Williams et al., 2008). This chronology (1920–2004) was

compared with a rainfall-based water balance model that

explained most of the variability in tree growth (R2 = 0.60).

However, more than half of the remaining variability could be

explained using proxies for fog drip and cloud shading

(derived from nearby airport weather observations, 1944–

2005). Importantly, cloud shading was shown to significantly

increase growth even when looking only at clouds too high to

provide fog drip. In addition to statistical correlation, the

increased growth from fog drip can also be observed directly in

the form of numerous ‘false bands’ in tree cores from both

Santa Rosa Island and SCI. These bands occur when a tree

whose summer growth has slowed reverts to rapid growth in

the same season (Schulman, 1947). Since growth is so tightly

linked to water availability in these systems, and no rainfall was

recorded during these summers, we interpret these false bands

as indicating additions of fog water to the ecosystem The

presence of numerous false bands within the tree ring record

from SCI, combined with the strong correlation of annual

growth and a fog proxy, suggest that fog drip relieves drought

stress significantly across individual years and multiple decades

(Williams, 2006; Fischer, 2007; Williams et al., 2008).

Long-term effects

Higher PET due to reduced cloud shading at the inland edge of

the stand (site 12) is potentially significant as a range-limiting

factor for this drought-sensitive species. Given similar amounts

of soil water at the beginning of a drought, trees growing under

a higher PET regime will run out of water before trees growing

where PET is lower (see Results: ‘Water balance for an

‘‘average’’ year’).

Lower rainfall at the coastal end of the stand where

elevation decreases (Site 1) is also potentially range-limiting.

However, the results of run B show that reduced coastal PET

(attributable largely to increased cloud shading) essentially

compensated for the 30% lower rainfall (compared with site

10) leaving no large differences in deficits between those two

sites.

Both fog drip and cloud shading can sharply reduce drought

frequency and drought severity. Run B shows that lowered PET

from cloud shading not only reduced mean Deficit, but also

reduced climatic variability. Run C suggests the same is true of

fog drip, although more rigorously assessing the strength of

that effect will require more long-term data on fog drip. Run D

shows that cloud shading and fog drip were somewhat spatially

independent, and, importantly, that their effects may be

additive in only limited locations.

We have used Deficit as a proxy for drought stress, and, by

extension, as a proxy for mortality risk. Deficit is a useful proxy

for mortality risk, but the relationship is nonlinear. Mortality

in most years is very low, with only a few trees succumbing to

insects or disease. As drought becomes more severe, more and

more pines weaken and become susceptible to disease. As a

result, host/pathogen population dynamics and stand connec-

tivity patterns become important factors in the spread of

epidemic disease (Richardson, 1998; White et al., 2006). To

better assess mortality risk, then, it makes less sense to examine

average Deficit than to look at the return interval of extreme

events. The two most extreme annual Deficits at site 10 were

just over 400 mm (1989–90). This level of Deficit, associated

with the severe pine die-off of the early 1990s, seems an

appropriate threshold for analysis. At site 10, near the middle

of the stand, this level of Deficit was modelled in just those 2 of



the last 102 years. By contrast, at sites 1 and 12, modelled

Deficits of this severity (> 400 mm) occurred in 20 or more

years out of 102 (Table 1, Fig. 8d). This same pattern of

reduced drought severity at site 10 also holds when comparing

the frequency of severe 2-year droughts (measured as summed

monthly Deficit for two calendar years). There were at least 13

2-year droughts at both sites 1 and 12 that exceeded the driest

2-year drought at site 10. There were at least 23 3-year

droughts at both sites 1 and 12 that exceeded the driest 3-year

drought at site 10. The greatly reduced frequency of severe

drought at site 10 suggests that the middle of the stand

provides suitable habitat much more consistently over time

than do areas near the edges. This habitat consistency, or

reduced frequency of severe droughts, due to the combined

effects of fog drip and cloud shading, is a critical habitat

feature for long-term population viability.

Comparing these model results with reality, we observed

that pine density is much higher around site 10 (50–80%

cover) than near sites 1 and 12 (a few widely scattered young

trees in chaparral). Growth rates and longevity (from tree

cores) are also much higher at site 10 than at sites 1 and 12

(not shown). For a drought-sensitive species like bishop pine

this spatial pattern only makes sense when compared with

drought frequency under run D. Each of the other model runs

fails to produce a credible spatial pattern of moisture

availability; it is only by incorporating both cloud shading

effects (reduced PET near the coast) and fog drip (estimated

typical water additions each summer) that the model matches

well with the observed pattern of pine density.

CONCLUSIONS

Fog is recognized as an important component of the hydro-

logical cycle in many ecosystems that are of great conservation

concern (Olson & Dinerstein, 1998). The same low stratus

clouds that produce fog drip where they intercept vegetation

also influence ecosystem hydrology by cloud shading. Low

clouds shade plants, both where they intercept vegetation, and

at lower elevations beneath stratus clouds. This shading

reduces PET and water loss by reducing net radiation,

increasing relative humidity and decreasing air temperature.

We evaluated the potential ecological effects of cloud

shading and the resulting spatial differences in PET using a

water balance model. The model tracked monthly ecosystem

water availability over the last century at three sites across a

stand of endemic coastal pines. We compared tree mortality

risk among these sites by analysing the frequency and severity

of drought stress, and examined how cloud patterns mediate

drought stress and thereby influence range dynamics. We show

that for low-elevation sites within the California fog belt, the

effects of cloud shading can be much greater than those of fog

drip. Even high-elevation sites that receive substantial fog drip

may benefit as much from shading as from fog drip – this

would be particularly true for species whose leaves are less

efficient than conifers at collecting fog droplets. Either effect

alone can reduce annual drought stress by a third compared

with adjacent less cloudy areas, and together these effects can

reduce drought stress by more than half. The shading aspect of

coastal fog and its potential impact on vegetation growth and

range dynamics has not previously been quantified, but should

be considered in future fog-vegetation research.

Inferred range dynamics of the western stand of bishop pine

on SCI over the last century suggest that pines can disperse

outward from the centre of the stand in favourable years, but

that drought frequently kills off pines near the periphery of the

stand. Further, there are micro-refugia from severe drought

near the centre of the stand. This pattern was not predicted by

water balance modelling until the separate spatial patterns of

both fog drip and cloud shading were taken into account.

While this study focused on bishop pine, the same cloud-

influenced patterns of episodic drought at coastal and inland

range edges presumably affect many other species endemic to

the Californian fog belt. Similar large differences in moisture

availability over relatively small changes in distance and

altitude also exist in other semi-arid, fog-inundated ecosystems

(e.g. other mediterranean-climate coasts and the lomas of the

Atacama Desert; Rundel et al., 1991). Even tropical montane

cloud forests, which receive much more precipitation and are

much wetter, are drying in places (Pounds et al., 1999, 2006).

As global climate change and deforestation alter cloud patterns

in fog-dependent ecosystems around the world (e.g. Still et al.,

1999; Lawton et al., 2001), increasing attention and effort are

focusing on conserving their disproportionately large share of

global biodiversity and species endemism. Conservation efforts

in these ecosystems will be better able to anticipate range

dynamics of drought-sensitive species by considering the

spatially distinct hydrological impacts of both fog drip and

cloud shading.

ACKNOWLEDGEMENTS

This work was supported by the Andrew W. Mellon Foun-

dation and the University of California, Santa Barbara. The

fieldwork was performed at the University of California

Natural Reserve System’s Santa Cruz Island Reserve, on

property owned and managed by The Nature Conservancy.

We are grateful to J. B. Wall at CSU Northridge for long-term

weather data, and to L. Laughrin and staff at the reserve for

weather records, long-term perspective and field support.

K. McEachern, S. Chaney and staff at Channel Islands

National Park and Island Packers helped with extensive

transportation to the study sites. M. Carroll provided invalu-

able information and photographs from many years of careful

observation of the ecosystem. The manuscript was improved

by helpful comments from O. A. Chadwick, B. Tiffney, M.

Scholl, G. MacDonald and an anonymous referee. We thank

all the following for their assistance with fieldwork: R.

Richardson, E. Thorsos, T. Eckmann, M. Kelly, S. Battersby,

J. Bernstein, A. Hartshorn, K. Cover, W. Amselem, E. Mason,

R. Roff, N. Wilson, E. Edwards, J. Rumbly, J. Hecht, N. Nagle,

P. Dennison, J. Baker, K. Hertel, T. Laverne, all of the

Geography 194 students and especially C. Ebert.



REFERENCES

Allen, R.G., Walter, I.A., Elliott, R. & Howell, T.A. (2005) The

ASCE standardized reference evapotranspiration equation.

American Society of Civil Engineers, Reston, VA.

Axelrod, D.I. (1967) Evolution of the Californian closed-cone

pine forest. Proceedings of the Symposium on the Biology of

the California Islands (ed. by R.W. Philbrick), pp. 93–150.

Santa Barbara Botanic Garden, Santa Barbara, CA.

Azevedo, J. & Morgan, D.L. (1974) Fog precipitation in coastal

California forests. Ecology, 55, 1135–1141.

Bakun, A. (1990) Global climate change and intensification of

coastal ocean upwelling. Science, 247, 198–201.

Breshears, D.D., Cobb, N.S., Rich, P.M., Price, K.P., Allen,

C.D., Balice, R.G., Romme, W.H., Kastens, J.H., Floyd, M.L.

& Belnap, J. (2005) Regional vegetation die-off in response

to global-change-type drought. Proceedings of the National

Academy of Sciences USA, 102, 15144–15148.

Brouwer, C. & Heibloem, M. (1986) Irrigation water man-

agement: irrigation water needs. Agriculture Department,

Food and Agriculture Organization of the United Nations,

Rome. Available at: http://www.fao.org/documents; last

accessed 22 October 2008.

Burgess, S.S.O. & Dawson, T.E. (2004) The contribution of fog

to the water relations of Sequoia sempervirens (D. Don):

foliar uptake and prevention of dehydration. Plant Cell and

Environment, 27, 1023–1034.

Chaney, R.W. & Mason, H.L. (1930) A Pleistocene flora from

Santa Cruz Island, California. Studies of the Pleistocene pal-

aeobotany of California, pp. 1–24. Carnegie Institution,

Washington, DC.

Chaney, R.W. & Mason, H.L. (1933) A Pleistocene flora from

the asphalt deposits at Carpinteria, California. Studies of the

Pleistocene palaeobotany of California, pp. 45–79. Carnegie

Institution, Washington, DC.

Cole, E.S. (2005) Root and whole plant growth responses to soil

resource heterogeneity in coastal dune shrubs of California.

PhD Dissertation, Department of Ecology, Evolution and

Marine Biology, University of California, Santa Barbara.

Corbin, J.D., Thomsen, M.A., Dawson, T.E. & d’Antonio, C.M.

(2005) Summer water use by California coastal prairie

grasses: fog, drought, and community composition. Oeco-

logia, 145, 511–521.

Costello, L.R., Matheny, N.P., Clark, J.R. & Jones, K.S. (2000)

A guide to estimating irrigation water needs of landscape

plantings in California. University of California Cooperative

Extension, California Department of Water Resources,

Sacramento, CA.

Dawson, T.E. (1998) Fog in the California Redwood forest:

ecosystem inputs and use by plants. Oecologia, 117, 476–485.

Dunne, T. & Leopold, L.B. (1978) Water in environmental

planning. W.H. Freeman and Co., New York.

Estberg, G. (2001) Intercomparison of fog water deposition

between two sites in proximity to Pinus torreyana. The

Second International Conference on Fog and Fog Collection

(ed. by R.S. Schemenauer and H. Puxbaum), pp. 189–191.

FogQuest, St John’s, Canada.

Filonczuk, M.K., Cayan, D.R. & Riddle, L.G. (1995) Variability

of marine fog along the California coast. Climate Research

Division, Scripps Institution of Oceanography, University of

California, San Diego, La Jolla, CA.

Fischer, D.T. (2007) Ecological and biogeographic impacts of fog

and stratus clouds on coastal vegetation, Santa Cruz Island,

California. PhD Dissertation, Department of Geography,

University of California, Santa Barbara.

Fischer, D.T. & Still, C.J. (2007) Evaluating patterns of fog

water deposition and isotopic composition on the California

Channel Islands. Water Resources Research, 43, W04420.

Doi: 10.1029/2006WR005124.

Fischer, D.T., Smith, S.V. & Churchill, R.R. (1996) Simulation

of a century of runoff across the Tomales watershed, Marin

County, California. Journal of Hydrology, 186, 253–273.

Goodman, J. (1985) The collection of fog-drip. Water

Resources Research, 21, 392–394.

Granier, A. & Loustau, D. (1994) Measuring and modelling the

transpiration of a maritime pine canopy from sap-flow data.

Agricultural and Forest Meteorology, 71, 61–81.

Griffin, J.R. & Critchfield, W.B. (1972) The distribution of forest

trees in California. Pacific SW Forest and Range Experiment

Station, USDA Forest Service, Berkeley, CA.

Hamilton, L.S., Juvik, J.O. & Scatena, F.N., eds (1995) Tropical

montane cloud forests. Tropical montane cloud forests,

pp. 1–23. Springer, New York.

Haston, L. & Michaelsen, J. (1994) Long-term central coastal

California precipitation variability and relationships to El

Nino–Southern Oscillation. Journal of Climate, 7, 1373–

1387.

Haston, L. & Michaelsen, J. (1997) Spatial and temporal var-

iability of southern California precipitation over the last

400 yr and relationships to atmospheric circulation patterns.

Journal of Climate, 10, 1836–1852.

Heusser, L. (1995) Pollen stratigraphy and paleoecologic

interpretation of the 160-k.y. record from Santa Barbara

Basin, hole 893a. Proceedings of the Ocean Drilling Program,

scientific results, Santa Barbara Basin; covering Leg 146 of the

cruises of the vessel JOIDES Resolution, Santa Barbara

Channel, California, Site 893, 20 September–22 November

1992 (ed. by J.P. Kennett and L. Haskins). Texas A & M

University, ODP Ocean Drilling Program, College Station,

TX.

Hobbs, E. (1980) Effects of grazing on the northern population

of Pinus muricata on Santa Cruz Island, California. The

California Islands: proceedings of a multidisciplinary sympo-

sium (ed. by D.M. Power). Santa Barbara Museum of Nat-

ural History, Santa Barbara, CA.

Hochberg, M.C. (1980) Factors affecting leaf size of the chap-

arral shrubs Ceanothus megacarpus, Dendromecon rigida,

and Prunus ilicifolia on the California islands. Master’s

Thesis, Department of Botany, University of California,

Santa Barbara.



Ingraham, N.L. & Matthews, R.A. (1995) The importance of

fog-drip water to vegetation – Point-Reyes Peninsula, Cali-

fornia. Journal of Hydrology, 164, 269–285.

IPCC (2007) Intergovernmental Panel on Climate Change fourth

assessment report. World Meteorological Organization,

United Nations Environmental Program. Available at:

http://www.ipcc.ch/ipccreports/assessments-reports.htm

(last accessed 10 December 2007).

Jacobs, D.F., Cole, D.W. & McBride, J.R. (1985) Fire history

and perpetuation of natural Coast Redwood ecosystems.

Journal of Forestry, 83, 494–497.

Kappelle, M. (2004) Tropical montane forests. Tropical forests

(ed. by J. Burley, J. Evans and J.A. Youngquist), pp. 1782–

1793. Elsevier, Oxford.

Kennedy, P.G. & Sousa, W.P. (2006) Forest encroachment into

a Californian grassland: examining the simultaneous effects

of facilitation and competition on tree seedling recruitment.

Oecologia, 148, 464–474.

Knapp, A.K. & Smith, M.D. (2001) Variation among biomes in

temporal dynamics of aboveground primary production.

Science, 291, 481–484.

Lanner, R.L. (1999) Conifers of California. Cachuma Press, Los

Olivos, CA.

Larcher, W. (2003) Plant physiological ecology. Springer, Berlin.

Laughrin, L. (2005) Santa Cruz Island rainfall record, 1904–

2005. Santa Cruz Island Reserve, UC Natural Reserve Sys-

tem, Santa Barbara, CA.

Leipper, D.F. (1994) Fog on the U.S. west coast: a review.

Bulletin of the American Meteorological Society, 75, 229–

240.

Leyton, L. & Armitage, I.P. (1968) Cuticle structure and water

relations of needles of Pinus radiata (D. Don). New Phy-

tologist, 67, 31–38.

Marotz, G.A. & Lahey, J.F. (1975) Some stratus/fog statistics in

contrasting coastal plant communities of California. Journal

of Biogeography, 2, 289–295.

Mason, H.L. (1934) Pleistocene flora of the Tomales forma-

tion. Studies of the Pleistocene palaeobotany of California,

pp. 81–179. Carnegie Institution, Washington, DC.

McGuirk, J.P. (1982) A century of precipitation variability

along the Pacific coast of North America and its impact.

Climatic Change, 4, 41–56.

McKenney, M.S. & Rosenberg, N.J. (1993) Sensitivity of

some potential evapotranspiration estimation methods to

climate change. Agricultural and Forest Meteorology, 64,

81–110.

Mendelssohn, R. & Schwing, F.B. (2002) Common and

uncommon trends in SST and wind stress in the California

and Peru–Chile current systems. Progress in Oceanography,

53, 141–162.

Michaelsen, J., Haston, L. & Davis, F.W. (1987) 400 years of

central California precipitation variability reconstructed

from tree rings. Water Resources Bulletin, 23, 809–818.

Millar, C.I. (1986) The Californian closed cone pines (sub-

section Oocarpae Little and Critchfield): a taxonomic history

and review. Taxon, 35, 657–670.

Millar, C.I. (1999) Evolution and biogeography of Pinus rad-

iata, with a proposed revision of its Quaternary history. New

Zealand Journal of Forestry Science, 29, 335–365.

National Weather Service (2005) Santa Barbara temperature

records. National Climatic Data Center, Boulder, CO.

National Weather Service (2007) NWS Detroit/Pontiac weather

glossary. Available at: http://www.crh.noaa.gov/dtx/glossary.

php (last accessed 10 January 2007).

Olson, D.M. & Dinerstein, E. (1998) The global 200: a

representation approach to conserving the Earth’s most

biologically valuable ecoregions. Conservation Biology, 12,

502–515.

Pounds, J.A., Fogden, M.P.L. & Campbell, J.H. (1999) Bio-

logical response to climate change on a tropical mountain.

Nature, 398, 611–615.

Pounds, J.A., Bustamante, M.R., Coloma, L.A., Consuegra,

J.A., Fogden, M.P.L., Foster, P.N., La Marca, E., Masters,

K.L., Merino-Viteri, A., Puschendorf, R., Ron, S.R., San-

chez-Azofeifa, G.A., Still, C.J. & Young, B.E. (2006) Wide-

spread amphibian extinctions from epidemic disease driven

by global warming. Nature, 439, 161–167.

Raven, P.H. & Axelrod, D.I. (1978) Origin and relationships of the

California flora. University of California Press, Berkeley, CA.

Richardson, D.M., ed. (1998) Ecology and biogeography of

Pinus. Cambridge University Press, Cambridge.

Ruiz, G. (2005) Characterization of fog water collection

potential and quality on California State University Mon-

terey Bay and Glen Deven ranch near Big Sur. EOS,

Transactions, American Geophysical Union, 86. Fall Meeting

Supplement, Abstract A33B-0891.

Rundel, P.W., Dillon, M.O., Palma, B., Mooney, H.A., Gulmon,

S.L. & Ehleringer, J.R. (1991) The phytogeography and ecology

of the coastal Atacama and Peruvian deserts. Aliso, 13, 1–49.

Schonher, T. & Nicholson, S.E. (1989) The relationship

between California rainfall and ENSO events. Journal of

Climate, 2, 1258–1269.

Schulman, E. (1947) Tree-ring hydrology in southern California.

Bulletin No. 4, Laboratory of Tree-Ring Research, University

of Arizona, Tucson, AZ.

Schwing, F.B. & Mendelssohn, R. (1997) Increased coastal

upwelling in the California current system. Journal of Geo-

physical Research, 102, 3421–3438.

Snyder, R.L. & Eching, S. (2004) PMhrXLS: Penman-Monteith

hourly ETref for short and tall canopies. University of

California, Davis. Available at: http://biomet.ucdavis.edu/

evapotranspiration.html (last accessed 22 October 2008).

Snyder, M.A., Sloan, L.C., Diffenbaugh, N.S. & Bell, J.L. (2003)

Future climate change and upwelling in the California

current. Geophysical Research Letters, 30, 1823–1826.

Stephenson, N.L. (1998) Actual evapotranspiration and defi-

cit: biologically meaningful correlates of vegetation distri-

bution across spatial scales. Journal of Biogeography, 25,

855–870.

Still, C.J., Foster, P.N. & Schneider, S.H. (1999) Simulating the

effects of climate change on tropical montane cloud forests.

Nature, 398, 608–610.



Thornthwaite, C.W. & Mather, J.R. (1955) The water balance.

Laboratory of Climatology, Centerton, NJ.

Thornthwaite, C.W. & Mather, J.R. (1957) Instructions and

tables for computing potential evapotranspiration and the

water balance. Laboratory of Climatology, Centerton, NJ.

Wall, J.B. (2005) Weather data, Santa Cruz Island, CA, 1995–

2005. Department of Geography, California State University,

Northridge, CA.

Walter, H.S. & Taha, L.A. (1999) Regeneration of bishop pine

(Pinus muricata) in the absence and presence of fire: a case

study from Santa Cruz Island, California. Fifth California

Islands Symposium. Santa Barbara Museum of Natural

History, US Department of the Interior, Minerals Manage-

ment Service, Pacific OCS Region, Santa Barbara, CA.

Wehtje, W. (1991) Response of bishop pine (Pinus muricata) to

cessation of browsing by feral sheep on Santa Cruz Island,

California. Master’s Thesis, Department of Geography,

University of California, Los Angeles.

White, A.B., Rich, P.M. & Pasqualini, D. (2006) Drought-in-

duced tree mortality in pinon-juniper woodlands. EOS,

Transactions, American Geophysical Union, 87. Fall Meeting

Supplement, Abstract B43C-02.

Williams, A.P. (2006) Teasing foggy memories out of pines on

the California Channel Islands using tree-ring width and

stable isotope approaches. Master’s Thesis, Department of

Geography, University of California, Santa Barbara.

Williams, A.P., Still, C.J., Fischer, D.T. & Leavitt, S.L. (2008)

The influence of summertime fog and overcast clouds on the

growth of a coastal Californian pine: a tree-ring study.

Oecologia, 156, 601–611.

BIOSKETCHES

Douglas T. Fischer is an assistant professor of geography at

California State University Northridge. This paper is derived

from his dissertation work at University of California, Santa

Barbara. His research focuses on microsite- to regional-scale

controls on species range limits, climate change and develop-

ment of conservation strategies that incorporate long-term

environmental variability.

Christopher J. Still is an associate professor of geography at

the University of California, Santa Barbara. His research

interests include ecosystem ecology, climate change, earth

system science, isotope biogeochemistry and biogeography.

A. Park Williams is a PhD student at University of

California, Santa Barbara. His research focuses on how large-

scale climatic variability affects plant species distribution and

on using plants, and tree rings in particular, to infer details

about past climate.

Editor: Glen MacDonald

APPENDIX

The standardized reference evapotranspiration equation (Allen

et al., 2005) follows:

ETrs ¼0:408DðRn � GÞ þ c Cn

Tþ273 u2ðes � eaÞDþ cð1þ Cdu2Þ

;

where Rn = calculated net radiation at the crop surface

(MJ m)2 h)1); G = soil heat flux density at the soil surface

(MJ m)2 h)1); T = mean hourly air temperature at 2-m height

(�C); u2 = mean hourly wind speed at 2-m height (m s)1);

es = mean hourly saturation vapour pressure at 2-m height

(kPa); ea = mean hourly actual vapour pressure at 2-m height

(kPa); D = slope of the saturation vapour pressure–tempera-

ture curve (kPa �C)1); c = psychrometric constant (kPa �C)1);

Cn=numerator constant (66 K mm s3 Mg)1 h)1 for tall veg-

etation, hourly time step); Cd = denominator constant

(0.25 s m)1 for tall vegetation, hourly time step, during the

day, 1.7 at night). Units for the 0.408 coefficient are

m2 mm MJ)1. Snyder & Eching’s (2004) spreadsheet calculates

these terms (following procedures in Allen et al., 2005) using

site parameters and hourly measurements of air temperature,

dew point, wind speed and incoming solar radiation.

The relevant terms of the monthly water balance model are

presented below (modified from Thornthwaite & Mather,

1955, 1957, and excluding terms for calculating run-off). Soil

moisture storage is in mm (or mm3 mm)2) and all fluxes are

in mm month)1.

Term Symbol Calculation

Precipitation Pt Estimated as f(t), after

excluding stormflow

Potential

evapotranspiration

PETt Estimated as f(t), from

modelled ETref

Field soil moisture

capacity

FSMC Estimated as 500 mm

Accumulated

potential water loss

APWLt If (Pt ) PETt) < 0,

APWLt = APWL(t–1)

+ (Pt ) PETt); else APWLt

= FSMC · ln(SMt/FSMC)

Soil moisture SMt If (Pt ) PETt) < 0, SMt

= FSMC · e(APWLt/FMSC);

else SMt = minimum of

[FSMC or SM(t–1) + (Pt ) PETt)]

Actual

evapotranspiration

AETt AETt = minimum of

[PETt or Pt ) (SMt ) SMt–1)]

Deficit Dt PETt ) AETt



significance of summer fog and overcast for drought stress and

Documents