xxxxxmodeing the eco hydrologic response of a mediterian type ecosystem to the combined impacts of proects cc and altered fire frequencies

Climatic Change (2009) 93:137–155
DOI 10.1007/s10584-008-9497-7

Modeling the eco-hydrologic response
of a Mediterranean type ecosystem to the combined
impacts of projected climate change and altered fire
frequencies
C. Tague · L. Seaby · A. Hope

Received: 9 May 2006 / Accepted: 19 August 2008 / Published online: 3 October 2008
© Springer Science + Business Media B.V. 2008

Abstract Global Climate Models (GCMs) project moderate warming along with increases in atmospheric CO2 for California Mediterranean type ecosystems (MTEs).
In water-limited ecosystems, vegetation acts as an important control on streamflow
and responds to soil moisture availability. Fires are also key disturbances in semiarid environments, and few studies have explored the potential interactions among
changes in climate, vegetation dynamics, hydrology, elevated atmospheric CO2
concentrations and fire. We model ecosystem productivity, evapotranspiration, and
summer streamflow under a range of temperature and precipitation scenarios using
RHESSys, a spatially distributed model of carbon–water interactions. We examine
the direct impacts of temperature and precipitation on vegetation productivity and
impacts associated with higher water-use efficiency under elevated atmospheric CO2.
Results suggest that for most climate scenarios, biomass in chaparral-dominated
systems is likely to increase, leading to reductions in summer streamflow. However,
within the range of GCM predictions, there are some scenarios in which vegetation
may decrease, leading to higher summer streamflows. Changes due to increases in
fire frequency will also impact summer streamflow but these will be small relative to
changes due to vegetation productivity. Results suggest that monitoring vegetation

C. Tague (B)
Bren School of Environmental Science and Management,
University of California at Santa Barbara, Santa Barbara,

CA 93106, USA
e-mail:
L. Seaby
Sher Leff, 450 Mission Street, Suite 400, San Francisco,
CA 94105, USA
e-mail:
A. Hope
Department of Geography, San Diego State University,
San Diego, CA 92182, USA
e-mail:


138

Climatic Change (2009) 93:137–155

responses to a changing climate should be a focus of climate change assessment for
California MTEs.

1 Introduction
Recent summaries of GCM climate scenarios predict that the average temperature
in Southern California will increase by 1.5–5◦ C in the next century, with small
to moderate changes in annual precipitation (Cayan et al. 2006; Wilkinson 2002;
Goodrich et al. 2000). In this paper, we address the potential responses of chaparral
ecosystems to these projected changes, focusing on changes to vegetation productivity, carbon cycling and hydrology. Chaparral dominated, Mediterranean type
ecosystems (MTEs) of southern California are expected to be highly sensitive to
climate variability (Moreno and Oechel 1995). MTEs are strongly water limited,
and thus potential changes to chaparral ecosystem productivity are tightly linked
to vegetation water use. Substantial changes in vegetation production in these
ecosystems may therefore have important implications for local water resources.

Changes in vegetation productivity would also have implications for global carbon
budgets, with feedbacks to climate change, and may alter ecosystem health and
vulnerability to disease, fire and species change (Moreno and Oechel 1995).
Models can be useful tools in understanding spatial-temporal controls on ecohydrologic processes (Wigmosta and Lettenmaier 1999). There are numerous modelbased studies that examine the hydrologic impacts of projected climate change in
California (e.g. Knowles and Cayan 2002) as well as several studies on potential
changes in vegetation type (e.g. Lenihan et al. 2003). Few studies, however, have
considered the potential feedbacks between vegetation carbon cycling responses and
hydrologic stores and fluxes. In general, the contributions of vegetation dynamics
to hydrologic sensitivity to climate change have not been well studied in current
models of either carbon cycling or hydrologic behaviors in MTEs (Breshears and
Allen 2002). Interactions between vegetation and hydrology can be particularly
important in the semi-arid ecosystems, such as those in southern California, where
ET is a significant component of the water budget and vegetation dynamics are water
limited. In this paper, we use a process-based model to develop quantitative estimates
of chaparral ecosystem responses to climate, that include explicit representation of
soil moisture controls on vegetation carbon cycling and growth and concurrently,
vegetation controls on ET, soil moisture and ultimately streamflow.
Given a rise in atmospheric CO2 and associated changes in temperature and
precipitation, we are interested how summer streamflow might change. We focus
in this paper on summer streamflow because it is likely to be highly sensitive to
change in vegetation water use and climate forcing in semi arid ecosystems, and
because changes to summer streamflow regime are often used as indicators of aquatic
ecosystem stress (Poff et al. 1997).
Hydrologic models typically estimate streamflow as follows:
Q = f (ET, P, S)

(1)

ET = f (V, A, P, S)

(2)

And


Climatic Change (2009) 93:137–155

139

where Q is streamflow, ET is evapotranspiration, P is precipitation. V represents
vegetation characteristics often specified by LAI (leaf area index) and species type
parameters that control vegetation water use. LAI is a commonly used surrogate
for vegetation biomass and its influence on transpiration, evaporation from canopy
interception, and maximum canopy conductance. S represents soil characteristics
that influence drainage rates and storage and ultimately water availability for ET.
Atmospheric controls (A) include air temperature (T), radiation, windspeed, and
humidity. The impact of climate change in hydrologic models is often included by
changing temperature—and accounting for corresponding changes in humidity. Most
hydrologic models used in climate change assessment account for changes in A and
P as controls on ET and Q, while holding S and V constant.
Carbon cycling models typically assume that:
V = f (T, CO2, P)

(3)

In coupled models, V,A,P and CO2 in Eqs. 2 and 3 are dynamic. It could be argued
that there is greater uncertainty in coupled models given the need for additional
algorithms and parameters. Nonetheless these models support the exploration relationships among different variables and thus may be key tools in the development
hypothesis about which interactions or forcing conditions are likely to be important
for climate change assessment and monitoring. In this study we use a coupled model

to examine the relative sensitivity to ET, and ultimately Q to changes in V, T, P, CO2
and interactions among them for chaparral dominated MTEs.
Wildfire in MTEs is another important driver of land cover change. The fire return
interval is three to five decades in southern California chaparral (Keeley et al. 1999).
Wildfire directly alters ecosystem carbon cycling in these watershed and also changes
hydrologic response. Numerous studies have found that fire in chaparral causes an
increase in streamflow, sediment load, and peak discharges (Loaiciga et al. 2001;
Florsheim et al. 1991). Wildfire is likely to become more frequent under a warmer
climate. Further, frequency and severity of fire can be linked to climate driven
changes in both vegetation productivity and hydrology. Westerling et al. (2003) show
that in the Western US, fire severity increased in dry years, and was also higher
when the previous year was wet leading to higher biomass and greater fuels. Thus, in
estimating responses of MTEs to climate change, fire must be included as a driver
of vegetation change in Eq. (3). We recognize that vegetation composition may
change given time and fire frequency but we ignore it here for shorter-term (decadal)
analysis.
In this paper, we use RHESSys (Regional Hydro-Ecologic Simulation System)
(Tague and Band 2004), a spatially distributed model of carbon–water interactions,
to investigate how vegetation responses to climate change might alter the relationship between watershed hydrology and carbon cycling. We use the model to
estimate changes in annual productivity, ET, vegetation LAI and summer streamflow
under different climate scenarios, and to explore how including fire may alter
the resulting patterns of behavior. The goal is not necessarily to provide precise
quantitative estimates of water and carbon fluxes. Instead, this work compares how
different drivers of change (temperature, precipitation, atmospheric CO2, within
ranges provided by current GCM projections for southern California) contribute to
interactions between hydrology and vegetation carbon cycling and to offer insight


140

Climatic Change (2009) 93:137–155

into how important these interactions may be for quantifying future water availability
and ecosystem vulnerability.

2 Methods
2.1 Study site
We focus on chaparral ecosystems in the Santa Ynez mountains of Southern
California. Modeling scenarios are developed for the Jameson Creek watershed,
a 34 km2 watershed in the Santa Ynez Mountains in southern California (Fig. 1).

Fig. 1 Study site


Climatic Change (2009) 93:137–155

141

Average annual rainfall is 780mm with most of the rainfall occurring between
November and May. Elevations range from 677 meters at the watershed outlet
to 1771 meters at the highest point. The steeper hillslopes have rocky, nutrient
poor, sandy-loam soils; while the gentler slopes have deeper more developed sandyloam and loam soils. Vegetation cover is predominately evergreen chaparral (e.g.
Adenostoma fasiculatum, Ceanothus luecodermis, Arctostaphylos glauca) intermixed
with summer-deciduous sub-shrubs (e.g. Salvia mellifera, Artemisia californica, Eriogonum fasiculatum), oak woodland (e.g. Quercus spp.), grass and winter-deciduous
riparian trees (e.g. Salix spp. and Populus spp.) (Stephenson and Calcarone 1999).
Daily precipitation and temperature are available from 1952 to 2002 for a nearby
National Climate Data Center monitoring site. Details on processing and spatial
interpolation of precipitation data are provided in Tague et al. (2004). Streamflow
is recorded at U.S Geolological Survey Gage (no. 11121010) at the Jameson Lake
Reservoir.

2.2 RHESSys
RHESSys is a spatially distributed model of watershed scale linkages among water,
carbon and nitrogen. RHESSys sub-models can be used to estimate the impact of
air temperature, humidity and soil moisture on a number of ecosystem processes
including evaporation, transpiration, stomatal conductance, photosynthesis and respiration. RHESSys models both vertical hydrologic processes (ET, canopy and litter
interception, infiltration, drainage) and lateral routing between terrestrial patches.
The RHESSys carbon cycling sub-model is similar to that used by BIOME_BGC
(Thornton et al. 2002) and includes estimates of carbon assimilation, respiration
and allocation of net photosynthate to leaves, stems and roots as well as soil and
litter decomposition. Tague and Band (2004) provide a complete description of
the RHESSys model. For this study, the Jarvis (1976) based stomatal conductance
computation in RHESSys was modified following Medlyn et al. (2001) to include
the impact of increases in atmospheric CO2 . RHESSys uses the Farquhar and
vonCaemmerer (1982) approach to estimate net photosynthesis, which also responds
to increases in atmospheric CO2 concentration.
Tague et al. (2004) describe previous applications of RHESSys to the Jameson
watershed that evaluated model predictions of historic streamflow patterns and postfire LAI recovery trajectories, respectively. Calibration of RHESSys soil hydrologic
parameters for the Jameson watershed is summarized in Tague et al. (2004), and
show that the model obtained Nash-Sutcliffe efficiency (Nash and Sutcliffe 1970)
values of greater than 0.9 for monthly streamflow, and percent error in total flow
over a 5-year calibration period of less than 5% for a pareto optimal parameter
set. For this paper, a single parameter set is selected from optimal parameter
space. Calibration in this previous work was based on the correspondence between
model and observed streamflow at a monthly time step. The model is also able
to capture inter-annual variation in streamflow, giving an R2 of 0.96 for modeled
versus observed annual flow of a 30-year period of record. Strong correspondence
between observed and modeled flow at the annual time step suggests that the model
provides reasonable estimates of ecosystem ET, which is approximately 80% of the
mean annual water budget. Previous work has also evaluated the carbon cycling

142

Climatic Change (2009) 93:137–155

component of RHESSys by comparing RHESSys modeled and Thematic Mapper
remote sensing derived LAI post-fire recovery trajectories and show that the model
is able to reproduce post-fire regrowth as well as mature LAI values within ±20%
for this watershed (Seaby et al. 2006).
2.3 Climate change scenarios
Cayan et al. (2006) summarize recent GCM climate projections for California. While
there is consensus that temperatures will increase in California during the next
100 years, the magnitude of this increase depends upon emission scenarios and
varies to some extent between different GCMs. For Southern California, predicted
temperature increases relative to historic (1961–1990 period) conditions range from
0.8–2.3◦ C for the 2035–2064 period and 1.6 to 4.4 for the 2070–2099 period. Changes
to precipitation show even greater variability across emission scenarios and GCM
models, and range from predicted decreases of up to 30% to smaller increases (up to
10%).
A key challenge in applying GCM model climate projections is downscaling
to local scales, particularly in the complex topography of mountain environments
(Ghan et al. 2006; Wood et al. 2004). Given the uncertainties involved in downscaling
GCM data, we choose to develop scenarios based on historic meteorologic records
for our site. One of the advantages of this approach is that we can consider both
the separate and synergistic effects of changes in temperature and precipitation.
We applied a 2◦ C and 4◦ C degree temperature increases to existing meteorologic
station records to generate scenarios consistent with moderate and more extreme
warming projected by GCMs. Temperature increases assume a uniform temperature
increase throughout the year. To simulate changes in precipitation, we generated
50-year time series by randomly selecting water years from the historic meteorologic

record. Note that for meteorologic data used in the Jameson study site, there is
no statistically significant temporal correlation in annual precipitation for successive
water years. We randomly selected water years from the 50-year time series, allowing
for repetition of water years, to generate new 50-year climate records that vary
in terms of decadal statistics. Using this method we were able to generate 50-year
time series with mean annual precipitation greater or less than current mean annual
precipitation. We chose scenarios with ±30, ±10 and no change in precipitation and
applied the baseline, 2◦ C and 4◦ C warming to these scenarios to generate the set of
9 climate scenarios for our analysis.
2.4 Fire frequency
In addition to comparing model predictions of ecosystem function across climate
scenarios, we also contrast simulations with and without fire. Current fire return
interval ranges from 30 to 50 years for southern California chaparral ecosystems
(Keeley et al. 1999). Fire return intervals of less than 15 years are unlikely given that
chaparral reaches reproductive maturity in 5–10 years and becomes most flammable
after 15–20 years (Radtke et al. 1982; Haidinger and Keeley 1993). We compare
simulations run assuming no fire over the 50-year simulation period and simulations
with a moderate-high level of fire frequency (return period of 30 years). Because the


Climatic Change (2009) 93:137–155

143

region is prone to large fires greater than 120 km2 (Mensing et al. 1999; Radtke et al.
1982), it is assumed that the entire 34 km2 watershed will burn as a result of each
ignition. To simulate fire, we set all above-ground carbon and nitrogen model stores,
as well as fine root stores to zero. Soil carbon and nitrogen stores, however, are not
altered.
2.5 Scenario analysis

RHESSys was run for combinations of the 3 temperature (no increase, +2◦ C,
+4◦ C) and 5 precipitation scenarios (−30, −10, 0, +10, +30), with and without fire to
produce 30 model realizations. In addition, we also consider the sensitivity of model
prediction to assumptions made about atmospheric CO2 concentrations. For initial
runs we assume a relatively modest, CO2 concentration of 400 ppm and compare
results with simulations run using 600 and 800 ppm. We also perform several addition
model runs using only the hydrologic component of RHESSys. These simulations are
similar to standard hydrologic models that do not account for changes in vegetation
with a changing climate. These static simulations allow us to explore the impact
of coupling a hydrologic model with the dynamic ecosystem model and answer the
question: Is the additional model complexity warranted?
Model outputs for comparison include August streamflow, leaf area index (LAI),
ET (ET) and net primary productivity (NPP). Given the Mediterranean climate of
our study site, August streamflow typically has the lowest monthly streamflow, and it
is likely to be highly sensitive to vegetation water-use. As a summer month with the
lowest streamflow, August streamflows are also likely to be critical from an aquatic
ecosystem perspective. For example, maintenance of low flows to support Steelhead
habitat in the Santa Ynez region is an ecosystem management goal (EIR Lower
Santa Ynez River Fish Management Plan 2004). We examine the impact of climate
change on summer streamflow through the probability of obtaining average monthly
streamflow values below a threshold. We use the low quartile yearly mean August
streamflows from baseline climate scenario to define this threshold. ET estimates
show the direct impacts of climate variability on vegetation water use. NPP and
LAI demonstrate interactions with chaparral carbon cycling. For each scenario, we
examine annual mean and inter-annual variation of these 4 ecosystem response
variables.

3 Results and discussion
Figure 2 summarizes annual ET for all climate scenarios. Results are shown for
dynamic simulations, in which carbon-cycling driven changes in vegetation are

included, and static simulations in which vegetation biomass does not change in
response to climate forcing. For dynamic simulations, we begin by examining results
obtained by using atmospheric CO2 concentration of 400 ppm. As expected ET
increases with increasing precipitation, for all temperature scenarios. In this semiarid Mediterranean climate, precipitation increases associated within climate change
are likely to be small relative to inter-annual variation precipitation. Modeled
ET reflects this high inter-annual variation in precipitation such that inter-annual


144

Climatic Change (2009) 93:137–155

0 400 1000

ET mm/year

ET Static

dc30

dc10

cczero

in10

in30

dc30T2 dc10T2 cczeroT2 in10T2

in30T2 dc30T4 dc10T4 cczeroT4 in10T4

in30T4

in30T2 dc30T4 dc10T4 cczeroT4 in10T4

in30T4

0 400 1000

ET mm/year

ET Dynamic

dc30

dc10

cczero

in10

in30

dc30T2 dc10T2 cczeroT2 in10T2

0 400 1000

ET mm/year

ET with Fire

●
●

●
●

●

●
●

dc30

dc10

cczero

in10

●

in30

●

●
●

dc30T2 dc10T2 cczeroT2 in10T2

●

●

●
●

in30T2 dc30T4 dc10T4 cczeroT4 in10T4

in30T4

Fig. 2 Annual ET (mm/year) across precipitation and climate scenarios. Scenario key is provided in
Table 1. Variance within each 50-year climate scenario reflects year to year differences in ET. Results
are show for a static model (vegetation does not change), a dynamic model in which vegetation
responds to climate and a dynamic model that also includes vegetation losses due to fire (30-year
return interval)

variation in ET within each climate scenario is large relative to differences in mean
ET between scenarios (Fig. 2).
While ET increases with precipitation (for all scenarios), this increase is typically
non-linear. Figure 3 shows a leveling off of ET with higher annual precipitation
for the dynamic simulation with baseline and 4◦ C temperature increase scenario.
A similar pattern is found for all climate scenarios. RHESSys model estimates
of mean potential ET for this site are approximately 1,300 mm/year. Similarly,
Hidalgo et al. (2005) estimate potential ET ranging from 1,300 to 1,700 mm/year
using data from CIMIS (California Irrigation Management System) stations and pan
evaporation from local NCDC (National Climate Data Center) stations within the
Santa Ynez region. We note that leveling off of ET estimates for years with high

annual precipitation occurs at values significantly below these potential ET estimates.
Thus there is still unmet ET potential even in wet years. The ET-precipitation
relationship also reflects the within year temporal distribution of precipitation. Years
with high annual precipitation are typically dominated by one or more large storm
events, where most of the additional water is lost as runoff; thus these precipitation
increases do not lead to large gains in ET. The greatest increases in ET are seen in
the shift from years with low to moderate precipitation.
When averaged across precipitation scenarios, the dynamic model predicts decreases in ET with warming, while the static model (no vegetation change) show

Table 1 Key for climate change scenario names
Precip/temperature

30% decrease

10% decrease

No change

10% increase

30% increase

Baseline
Increase 2◦ C
Increase 4◦ C

dc30
dc30T2
dc30T4

dc10
dc10T2
dc10T4

cczero
cczeroT2
cczeroT4

in10
in10T2
in10T4

in30
in30T2
in30T4


**

ET mm/day
600
800
1000

Fig. 3 Annual ET (mm/year)
as a function of annual
precipitation. Results are
shown for the baseline climate
scenario and a scenario with a
4-degree increase in

temperature

145
1200

Climatic Change (2009) 93:137–155

*

●

●
●
●●

●

●
●

*
***

●

* *
**
●

●

400

**
** *
* *
*
**
*
*
*
*
* ***
*****
●

● ●
●
●
●
●
●
●
●
●●●
●

●●
●

● ●
●

●

●

* ***
*
●●
●

●

**
●
●

●
●

*
●
●

*
**
*
●

●
●

●

●

●

o

●●
●

*

0.5

1.0

baseline
T4

1.5

Pcp mm/yr

negligible changes in ET (Fig. 4). For example mean annual ET for the 4◦ C warming
scenario, using the dynamic model, is 10% lower relative to baseline. This counter
intuitive decrease in ET with warmer temperature using the dynamic model reflects

the impact of changes in vegetation. Mean LAI for the dynamic model reduces by
more than 30% with a 4◦ C warming. Mean annual NPP for the dynamic model
reduces from 111 gC/m2/year under baseline scenarios to 83 and 60 gC/m2/year with
2 and 4 degree warming respectively. Similar reductions in NPP were found in a fieldbased warming experiment in Mediterranean shrublands of Spain, where moderate
warming of approximately 3◦ C in a single year, reduced above ground NPP from
approximately 160 to 130 gC/m2/year (Penuelas et al. 2007). Lower LAI is caused
by the higher respiration costs under the warmer temperature. The magnitude of
changes in LAI for the dynamic model are sensitive to assumptions made about
atmospheric CO2 and will be discussed below. Results here, however, demonstrate
that changes in vegetation can impact model estimates of vegetation water use and
its response to climate.
Including fire in dynamic simulations does reduce ET in specific years, but does
not alter overall relationships between decadal average ET and decadal mean climate
variables (precipitation and temperature). An overall reduction in ET due to lower
LAI with warmer scenarios is also evident and is of similar magnitude to results from
the dynamic simulation when fire is not included (Fig. 4a).
Modeled changes in August streamflow parallel these changes in ET across
climate scenarios, such that for dynamic simulations (with and without fire) there
is a moderate increase in mean August streamflow with warming (Fig. 4b). The
frequency of low flow years shows slightly greater sensitivity to climate scenarios
(Fig. 5). For scenarios with static vegetation, a 30% decrease in decadal precipitation
means results in only minor increases in the frequency of low flow conditions (or
years with August flow below the baseline scenario lower quartile). Low flows are
more sensitive to increases in precipitation. With a 10% increase in precipitation,
the frequency of low flow (below baseline quartile) reduces by more than 50%. For
the dynamic model patterns are dominated by the hydrologic effects of vegetation
responses to temperature. With the dynamic model, baseline climate scenarios show
a greater frequency of low flow years relative to results from the static model. This

146

Climatic Change (2009) 93:137–155

a

700
600
500

ET (mm/year)

800

baseCO2

base

T2

T4

700
600
500

ET (mm/year)

800

with Fire

base

T2

T4

700
600
500

ET (mm/year)

800

static

base

T2

T4

Fig. 4 Annual ET (a) and Mean August Streamflow (b) for baseline, 2◦ C and 4◦ C warming
scenarios. Each box-whisker plot shows mean and variance across precipitation scenarios. Results
are show for a static model (vegetation does not change), a dynamic model in which vegetation
responds to climate and a dynamic model that also includes vegetation losses due to fire (30-year
return interval)

difference in low flow year frequency reflects the hydrologic impact of year-to-year
(within the baseline climate scenario) variation in LAI associated with the dynamic
simulations. For the dynamic model, decreases in vegetation biomass with warmer
temperatures (assuming baseline CO2 conditions) leads to reductions in water use
and a dramatic decrease in the frequency of low flow years under a warmer climate.
In contrast to static model, the dynamic model results suggest that variation in the


Climatic Change (2009) 93:137–155

b

147

0.6
0.5
0.4
0.3

Aug Str (mm/year)

0.7

baseCO2

base

T2

T4

0.6
0.5
0.4
0.3

Aug Str (mm/day)

0.7

with Fire

base

T2

T4

0.6
0.5
0.4
0.3

Aug Str (mm/day)

0.7

static

base

T2

T4

Fig. 4 (continued)

frequency of low flow years is greater across temperature scenarios than across
precipitation scenarios. Dynamic simulation results, therefore, highlight vegetation
responses as a first order control on hydrologic responses.
Inclusion of fire, in general reduces the frequency of low flow years. This is
expected, given the reduction in vegetation and associated ET losses. The small
reduction in the frequency of low flow years, however, suggests that even with
the fairly high fire return interval used here, the impact of fire on low flows is
relatively minor at decadal time scales. Rapid recovery of vegetation, within several
years following fire, and high inter-annual variation in precipitation, means that
the likelihood of low flow years occurring during the short hydrologically relevant
post fire period is small. Rapid recovery of chaparral estimated by the model is


148

Climatic Change (2009) 93:137–155
Static Vegetation

Relative Frequency
0.10 0.20 0.30

30% dec
10% dec

baseline
10% inc
30% inc

base

T2

T4

base

T2

T4

base

T2

T4

base

T2

T4

base

T2

T4

base

T2

T4

base

T2

T4

T2

T4

T2

T4

Relative Frequency
0.10 0.20 0.30

Dynamic Vegetation

base

T2

T4

base

T2

T4

base

T2

T4

Relative Frequency
0.10 0.20 0.30

Dynamic Vegetation with fire

base

T2

T4

base

T2

T4

base

T2

T4

base

base

Fig. 5 Frequency of years with August streamflow below a threshold value. Threshold is defined as
the lower quartile flow from the baseline climate scenario (0.24 mm/day). Relative frequency refers
to the proportion of years within a 50-year climate scenario that have August streamflow below
the threshold value. Results are show for all temperature and precipitation scenarios and for static
model (vegetation does not change), a dynamic model in which vegetation responds to climate and a
dynamic model that also includes vegetation losses due to fire (30-year return interval)

consistent with remote-sensing based estimates of recovery trajectories for this
region (McMichael et al. 2004).
In addition to simulations with baseline atmospheric CO2 concentrations, we also
compare dynamic model results across low (400 ppm), moderate (600 ppm) and high
(800 ppm) CO2. Increasing CO2 increases plant water use efficiency leading to higher
estimates of LAI (Fig. 6). Variation in LAI within a given temperature scenario
(shown on each box-whisker plot) reflect the range of LAI estimates across different
precipitation scenarios, averaged over the 50-year simulation period. Variation in
LAI due to a change in decadal precipitation means is small relative to variation due

to temperature. For all scenarios LAI decreases linearly with increasing temperature.
Higher levels of atmospheric CO2, however, support higher LAI overall, such that
the gains due to increased CO2 outweigh losses due to temperature (up to a 4◦ C
temperature increase). Differences between scenarios using 600 ppm and 800 ppm
are smaller than those between the baseline (400 ppm) and 600 ppm concentration,


Climatic Change (2009) 93:137–155

149
with Fire
LAI
1.0 1.5 2.0 2.5 3.0 3.5 4.0

LAI
1.0 1.5 2.0 2.5 3.0 3.5 4.0

baseCO2

T2
CO2 600

T4

base

T2
CO2 800

T4

LAI
1.0 1.5 2.0 2.5 3.0 3.5 4.0

LAI
1.0 1.5 2.0 2.5 3.0 3.5 4.0

base

base

T2

T4

base

T2

T4

Fig. 6 Mean LAI for baseline, 2◦ C and 4◦ C warming scenarios. Each box-whisker plot shows mean
and variance across precipitation scenarios. Results are show for dynamic model runs with baseline
(400 ppm), moderate (600 ppm) and high (800 ppm) CO2 atmospheric concentrations. Results for
dynamic model with fire assume a 400 ppm atmospheric CO2 concentration

suggesting that gains due to increased water use efficiency begin to level off at the
higher atmospheric CO2 concentrations. LAI for scenarios with fire are similar to
those without fire. The small impact of fire on decadal means reflects rapid post-fire
re-growth and associated recovery of vegetation water use.

Differences in NPP reflect these differences in LAI, and water use efficiency
(Fig. 7) While NPP and LAI are higher for elevated CO2 scenarios, year-to-year
variation in NPP is also greater in elevated CO2 scenarios and suggest greater
vulnerability to year-to-year variation in water availability. As noted with ET, within
scenario variation in NPP tends to be large relative to between scenario variation.
Summer water use tends to stabilize with increased CO2 concentration such that
mean August streamflow no longer varies across temperatures scenarios (Fig. 8).
Once the ecosystem reaches some threshold LAI, soil water is efficiently used by
the system in all years and water that remains for the summer baseflow is deep
groundwater bypass flow. Frequency of low flow years is substantially higher for
scenarios with greater atmospheric CO2 (Fig. 9). Changes in frequency of low flow
years across temperature are negligible for these higher CO2 concentrations and
are not shown. With 400 ppm scenario, LAI decreases under warmer years due to
increase in water stress, leading to fewer low flow years. With greater water use
efficiency, however, higher LAIs increase the likelihood of low flow years under a
warmer climate.


150

Climatic Change (2009) 93:137–155

NPP gC/m2/yr
-1000 0
1000

CO2 400

●
●

●

dc30

dc10 cczero in10

in30 dc30T2

cczeroT2

●

in30T2 dc30T4

cczeroT4

in30T4

dc30T4c6

cczeroT4c6

in30T4c6

dc30T4c8

cczeroT4c8

in30T4c8

NPP gC/m2/yr
-1000 0
1000

CO2 600

●
●
●

dc30c6

cczeroc6

in30c6

dc10T2c6

in10T2c6

●

NPP gC/m2/yr
-1000 0
1000

CO2 800
●

●

dc30c8

cczeroc8

in30c8

dc10T2c8

in10T2c8

Fig. 7 Annual net primary productivity across precipitation and climate scenarios. Scenario key is
provided in Table 1. Variance within each 50-year climate scenario, reflects year-to-year differences
in NPP. Results are shown for dynamic model runs with baseline (400 ppm), moderate (600 ppm)
and high (800 ppm) CO2 atmospheric concentrations

4 Discussion
In most semi-arid ecosystems, ET varies strongly with annual precipitation—
although the relationship is a non-linear one. In years with high precipitation, much
of the additional water is lost as runoff and may not increase ET rates. Model
results for this chaparral dominated ecosystem follow this expected relationship
with precipitation. Responses to high inter annual variation in precipitation often
outweigh climate change effects on ET and NPP. Nonetheless, there are changes to
decadal mean ET and NPP with temperature increases that can have implications for
ecosystem function and water resources, particularly in terms changing the frequency
of years with low summer streamflow.
In many ecosystems, such as those in humid or snow-melt dominated regions,
changes in vegetation water use may be small relative to potential changes in water
inputs as precipitation or snow-melt. Further in many systems, the dominant impact

of climate change on ET may be due to changes in temperature or soil moisture
availability rather than change in vegetation biomass. Recent modeling of hydrologic
responses to projected climate warming in California has focused changes in snow
accumulation and melt and corresponding reductions in summer streamflow. These
hydrologic modeling studies do not typically incorporate changes to vegetation with
climate scenarios (Vicuna et al. 2007; Hayhoe et al. 2004; Dettinger et al. 2004). For
snow-dominated regions, vegetation dynamics may play a secondary role and the
added complexity of incorporating dynamic vegetation into a hydrologic model is
not warranted. For more semi-arid, warmer regions in California, however, changes
in vegetation may substantially alter annual water balances. In this study, modeled
changes in vegetation for a range of possible climate scenarios led to both significant
increases or decreases in the frequency of low flow conditions. Further changes in


Climatic Change (2009) 93:137–155

151

0.3

0.3

Aug Str (mm/day)
0.4
0.5
0.6
0.7

with Fire

Aug Str (mm/day)
0.4
0.5
0.6
0.7

baseCO2

base

T2

T4

base

T2

T4

0.3

0.3

Aug Str (mm/day)
0.4
0.5
0.6
0.7

CO2 800

Aug Str (mm/day)
0.4
0.5
0.6
0.7

CO2 600

base

T2

T4

base

T2

T4

Fig. 8 Mean August streamflow for baseline, 2◦ C and 4◦ C warming scenarios. Each box-whisker plot
shows mean and variance across precipitation scenarios. Results are show for dynamic model runs
with baseline (400 ppm), moderate (600 ppm) and high (800 ppm) CO2 atmospheric concentrations.
Results for dynamic model with fire assume a 400 ppm atmospheric CO2 concentration

summer streamflow due to changes in ecosystem production were as large as changes
due directly to longer term changes in decadal precipitation means. Our results argue
that for chaparral dominated systems, models that do no account for interactions

among vegetation growth, hydrology and climate ignore a first order control on
hydrologic response to climate change. In this environment, the added complexity
of a coupled ecohydrologic model is necessary. On the other hand, model results
also suggest that the effects of fire are small relative to these changes, at least in
terms of water availability and summer streamflow. For peak flows and associated
erosion, hydrologic responses may be more sensitive to fire, particularly given that
soils in chaparral environments show hydrophobicity in the first year following fire
(Hubbert and Oriol 2005; DeBano 2000). For low flows, however, fast recover of
chaparral in the context of high year to year variation in climate mean that the
effect of fire frequency on decadal streamflow behavior is small and modeling of post
fire recovery trajectories is not essential for estimating decadal patterns in summer
streamflow.
Results from this study also suggest that the response of chaparral ecosystem ET
will depend largely on how vegetation productivity responds to the combined effect
of CO2 and temperature. The current ecosystem is situated such that the magnitude
of atmospheric CO2 increases and associated changes in water use efficiency will
determine whether increases or decreases in vegetation biomass are likely to occur.


0.30
0.25
0.20
0.15
0.10
0.05
0.00

Relative Frequency

Fig. 9 Frequency of years

with August streamflow below
a threshold value. Threshold is
defined as the lower quartile
flow from the baseline climate
scenario (0.24 mm/day).
Relative frequency refers to
the proportion of years within
a 50-year climate scenario that
have August streamflow below
this threshold value. Results
are shown for baseline climate
scenarios, with an atmospheric
CO2 concentration of 400 ppm
and for moderate (2◦ C)
warming and baseline
(400 ppm), moderate
(600 ppm) and high (800 ppm)
atmospheric CO2
concentrations

Climatic Change (2009) 93:137–155
0.35

152

base

T2 C400

T2 C600

T2 C800

Results from this model based study suggest that a CO2 concentrations at the
lower end of recent GCM predictions would lead to reductions in LAI (due to
increasing temperatures). For higher levels of atmospheric CO2, productivity and
biomass would increase, however, these increases diminish for CO2 concentrations
at the upper end of GCM scenario ranges. Field and other model-based studies
have generally shown increases in ecosystem LAI and NPP, under elevated CO2
(Schimel et al. 2000; Antle et al. 2001). Reductions in LAI predicted here for low
levels of elevated CO2, reflect the importance of linking changes due to increased
CO2 with temperature impacts on respiration, particularly given the warm, semi-arid
Mediterranean climate. Both field and model based studies remain uncertain and
long-term physiological adaptation with higher levels of atmospheric CO2 may occur
and have not been account for. Nonetheless, this modeling study provides insight
into the balance between different and interacting controls for California chaparral.
Results here emphasize that CO2 concentrations will be major control on chaparral
hydrologic and carbon cycling responses, both in terms magnitude and direction.
Model results suggest that reducing uncertainty in projections of CO2 concentrations
and plant response may be more important than efforts to refine temperature and
precipitation estimates.
Changes in ecosystem productivity and biomass are also important indicators of
ecosystem heath. Although increases in vegetation biomass (shown here as LAI)
suggest a more productive, potentially healthier ecosystem under higher atmospheric
CO2 concentration, modeled NPP also show greater year-to-year variation under
higher levels of CO2, which may increase ecosystem vulnerability to insect infestation
and disease. Increases in LAI with elevated CO2 will increase the frequency of years
with very low summer streamflow, which may have negative impacts on aquatic
ecosystems. It is important to note that this study did not account for species
change but focused solely on changes in productivity. Increased fire frequency in

chaparral dominated systems may result in conversion to grass (Oechel et al. 1995).


Climatic Change (2009) 93:137–155

153

Incorporation of species change responses into model-based analysis will be the focus
of future work.

5 Conclusion
RHESSys, a coupled hydro-ecologic model, was used to tease apart the effect
of multiple, interacting controls that influence the sensitivity of chaparral systems
and their hydrology to climate change. We focus particularly on linkages among
ecosystem water use, vegetation growth and atmospheric drivers of hydrologic behavior. Simulation results demonstrate that vegetation responses are both uncertain
and likely to be particularly important, given our current understanding of system
dynamics. Results highlight the importance of changes in water-use efficiency in
MTEs for summer streamflow, NPP and ET estimates. Model results suggest that
future hydrologic behavior and ecosystem productivity will depends on the balance
between CO2 controls on vegetation water use efficiency and vegetation responses
to increasing temperatures. It is likely, however, that increases in vegetation biomass
will result in a greater frequency of low flow conditions. While fire frequency
can influence hydrology in individual years, it is less important for understanding
decadal hydrologic behavior. Resource managers in MTEs must therefore consider
ecosystem responses as a major component of changes in water resources within the
coming decade.
The interpretation of model results must be made in the context of model
uncertainty (Beven and Freer 2001). Sources of uncertainty in this study include both
uncertainty in climate scenarios predictions and uncertainty in RHESSys predictions
of hydrologic and ecosystem responses. Models such as RHESSys essentially offer

quantitative estimates of the implications current understanding of key controls on
water and carbon cycling. It is worth noting that recent reviews of field studies
of ecosystem response show responses that follow our current understanding of
physiological controls on plant function (Antle et al. 2001). Model estimates of
decadal behavior and dominant controls on climate change responses discussed in
this paper provide hypothesis that can be used to guide future research and focus
adaptive management, assessment and monitoring efforts. Results demonstrate the
importance of using coupled eco-hydrologic models in estimating climate change impacts when, as is the case with California chaparral, there are multiple, simultaneous
controls whose relative influence can vary dramatically from season to season and
year to year.

References
Antle J, Apps M, Beamish R (2001) Ecosystems and their goods and services. In: McCarthy JJ,
Canziani OF, Leary NA, Dokken DJ, White KS (eds) Climate change 2001: impacts, adaptation,
and vulnerability. Cambridge University Press, Cambridge, pp 235–342
Beven K, Freer J (2001) Equifinality, data assimilation, and uncertainty estimation in mechanistic
modeling of complex environmental systems using the GLUE methodology. J Hydrol 249:11–29
Breshears DD, Allen CD (2002) The importance of rapid, disturbance-induced losses in carbon
management and sequestration. Glob Ecol Biogeogr 11:1–5


154

Climatic Change (2009) 93:137–155

Cayan D, Maurer E, Dettinger M, Tyree M, Hayhoe K, Bonfils C, Duffy P, Santer B (2006)
Climate scenarios for California, white paper, California climate change center, CEC-500–2006–
203-SF
DeBano LF (2000) The role of fire and soil heating on water repellency in wildland environments: a
review. J Hydrol 231–232:195–206

Dettinger MD, Cayan DR, Meyer M, Jeton AE (2004) Simulated hydrologic responses to climate
variations and change in the Merced, Carson, and American River basins, Sierra Nevada,
California, 1900–2099. Clim Change 62:283–317
Environmental Impact Statement (EIS) Lower Santa Ynez River Fish Management Plan and
Cachuma Project Biological Opinion for Southern Steelhead Trout (2004) Prepared by Cachuma
Operation and Maintenance Board, Santa Barbara County, California and Depart of the Interior
Bureau of Reclamation
Farquhar G, vonCaemmerer S (1982) Modeling photosynthetic response to environmental conditions. Encyclopedia of Plant Physiology
Florsheim JL, Keller EA, Best DW (1991) Fluvial sediment transport in response to moderate storm
flows following chaparral wildfire, Ventura County, southern California. Geol Soc Amer Bull
103:504–511
Ghan SJ, Shippert T, Fox J (2006) Physically based global downscaling: regional evaluation. J Climate
19:429–445
Goodrich DC, Chehbouni A, Goff B, MacNish B, Maddock T, Moran S et al (2000) Preface paper to
the Semi-Arid Land-Surface-Atmosphere (SALSA) program special issue. Agric For Meteorol
105:3–20
Haidinger TL, Keeley JE (1993) Role of high fire frequency in destruction of mixed chaparral.
Madroño 40:141–147
Hayhoe K, Cayan D, Field C, Frumhoff P, Maurer E, Miller N, Moser S, Schneider S, Cahill K,
Cleland E, Dale L, Drapek R, Hanemann RM, Kalkstein L, Lenihan J, Lunch C, Neilson R,
Sheridan S, Verville J (2004) Emissions pathways, climate change, and impacts on California.
Proc Natl Acad Sci (PNAS) 101(34):12422–12427
Hidalgo HG, Cayan DR, Dettinger MD (2005) Sources of variability of ET in California. J
Hydrometeorol 6(1):3–19
Hubbert KR, Oriol V (2005) Temporal fluctuations in soil water repellency following wildfire in
chaparral steeplands, southern California. Int J Wildland Fire 14:439–447
Jarvis PG (1976) The interpretation of the variations in leaf water potential and stomatal conductance found in canopies in the field. Philos Trans R Soc Lond B 273:593–610
Keeley JE, Fotheringham CJ, Morias M (1999) Reexamining fire suppression impacts on brushland
fire regimes. Science 284(5421):1829–1832
Knowles N, Cayan D (2002) Potential effects of global warming on the Sacramento/San Joaquin

watershed and the San Francisco estuary. Geophys Res Lett 29:18
Lenihan JM, Drapek R, Bachelet D, Neilson RP (2003) Climate change effects on vegetation
distribution, carbon, and fire in Califronia. Ecol Appl 13(6):1667–1681
Loaiciga HA, Pedreros D, Roberts D (2001) Wildfire-streamflow interactions in a chaparral watershed. Adv Environ Res 5:295–305
McMichael CE, Hope AS, Roberts DA, Anaya M (2004) Post-fire recovery of leaf area index in
California chaparral: a remote sensing—chronosequence approach. Int J Remote Sens
25(21):4743–4760
Medlyn BE, Barton CVM, Broadmeadow MSJ, Ceulemans R, De Angelis P, Forstreuter M,
Freeman M, Jackson SB, Kellomaki S, Laitat E, Ray A, Roberntz P, Sigurdsson BD,
Strassemeyer J, Wang K, Curtis PS, Jarvis PG (2001) Stomatal conductance of forest species
after long-term exposure to elevated CO2 concentration: a synthesis. New Phytol 149:247–264
Mensing SA, Michaelsen J, Byrne R (1999) A 560-year record of Santa Ana fires reconstructed from
charcoal deposited in the Santa Ana Basin, California. Quat Res 51:295–305
Moreno JM, Oechel WC (1995) Global change and Mediterranean-type ecosystems. Springer,
New York
Nash J, Sutcliffe J (1970) River flow forecasting through conceptual models: Part I—a discussion of
principles. J Hydrol 10:282–290
Oechel WC, Hastings SJ, Vourlitis GL, Jenkins MA, Hinkson CL (1995) Direct effects of elevated
CO2 in Chaparral and Mediterranean-type ecosystem. In: Moreno J, Oechel W (eds) Global
change and Mediterranean-type ecosystems. Springer, New York, pp 58–75
Penuelas J, Prieto P, Beier C, Cesaraccio C, de Angelis P, de Dato G, Emmett BA, Estiarte M,
Garadnai J, Gorissen A, Lang EK, Kroel-Dulay G, Llorens L, Pellizzaro G, Riis-Nielsen T,


Climatic Change (2009) 93:137–155

155

Schmidt IK, Sirca C, Sowerby A, Spano D, Tietema A (2007) Response of plant species richness
and primary productivity in shrublands along a north-south gradient in Europe to seven years of

experimental warming and drought: reductions in primary productivity in the heat and drought
year of 2003. Glob Chang Biol 12(12):2563–2581. doi:10.1111/j.1365-2486.2007.01464.x
Poff NL, Allan JD, Bain MB, Karr JR, Prestegaard KL, Richter BD, Sparks RE, Stromberg JC
(1997) The natural flow regime. BioScience 47:769–784
Radtke KW-H, Arndt AM, Wakimoto R (1982) Fire history of the Santa Monica Mountains. General
Technical Report PSW-58: Pacific Southwest Forest and Range Experiment Station, Forest
Service: U.S. Department of Agriculture, Berkeley, CA, pp 438–443
Schimel DS, Melillo J, Tian H, McGuire AD, Kicklighter D, Kittel T, Rosenbloom N, Running S,
Thornton P, Ojima D, Parton W, Kelly R, Sykes M, Neilson R, Rizzo B (2000) Contribution
of increasing CO2 and climate to carbon storage by ecosystems in the United States. Science
287:2004–2006
Seaby LP, Tague CL, Hope A (2006) Post-fire recovery of eco-hydrologic behavior given historic and
projected climate variability in California Mediterranean type environments. Eos Transactions,
American Geophysical Union, 87(52) Fall Meeting. Suppl. Abstract H13B–1938
Stephenson JR, Calcarone GM (1999) Southern California Mountains and Foothills assessment:
habitat and species conservation issues. General Technical Report: PSW-GTR-172, Albany,
California: Pacific Southwest Research Station, Forest Service, U.S. Department of Agriculture,
402 pp
Tague C, Band LE (2004) RHESSys: Regional Hydro-Ecologic Simulation System—an objectoriented approach to spatially distributed modeling of carbon, water, and nutrient cycling. Earth
Interact 8(19):1–42
Tague C, McMichael C, Hope A, Choate J, Clark R (2004) Application of the RHESSys model to a
California semi-arid shrubland watershed. J Am Water Resour Assoc 40(3):575–589
Thornton PE, Law BE, Gholz HL, Clark KL, Falge E, Ellsworth DS, Goldstein AH, Monson
RK, Hollinger D, Falk M, Chen J, Sparks JP (2002) Modeling and measuring the effects of
disturbance history and climate on carbon and water budgets in evergreen needleleaf forests.
Agric For Meteorol 113:185–222
Vicuna S, Maurer E, Yoyce B, Dracup J, Purkey D (2007) The sensitivity of California
water resources to climate change scenarios1. J Am Water Resour Assoc 43(2):482–498.
doi:10.1111/j.1752-1688.2007.00038.x
Westerling AL, Gershunov A, Brown TJ, Cayan DR, Dettinger MD (2003) Climate and wildfire in

the Western United States. Bull Am Meteorol Soc 84(5):595–604
Wigmosta M, Lettenmaier D (1999) A comparison of simplified methods for routing topographically
driven subsurface flow. Water Resour Res 35(1):255–264
Wilkinson R (2002) The potential consequences of climate variability and change for California:
a report of the California Regional Assessment Group for the U.S. Global Change Research
Program, September pp 1–432
Wood AW, Leung LR, Sridhar V, Lettenmaier P (2004) Hydrologic implications of dynamical and
statistical climate model outputs. Clim Change 62:189–216