You are on page 1of 12

WATER RESOURCES RESEARCH, VOL. 43, W11419, doi:10.

1029/2007WR005890, 2007

A two-parameter climate elasticity of streamflow index to assess


climate change effects on annual streamflow
Guobin Fu,1 Stephen P. Charles,1 and Francis H. S. Chiew2
Received 16 January 2007; revised 24 July 2007; accepted 20 August 2007; published 24 November 2007.
[1] This study extends the single parameter precipitation elasticity of streamflow index
into a two parameter climate elasticity index, as a function of both precipitation and
temperature, in order to assess climatic effects on annual streamflow. Application of the
proposed index to two basins indicates that the single parameter precipitation elasticity
index may give a general relationship between precipitation and streamflow in some cases,
but that it cannot reflect the complicated non-linear relationship among streamflow,
precipitation and temperature. For example, for the Spokane River basin the climate
elasticity of streamflow index varies from 2.4 to 0.2, for a precipitation increase of 20%, as
temperature varies from 1°C lower to 1.8°C higher than the long term mean. Thus a
20% precipitation increase may result in a streamflow increase of 48% if the temperature
is 1°C lower but only a 4% increase if the temperature is 1.8°C higher than the long-term
mean. The proposed method can be applied to other basins to assess potential climate
change effects on annual streamflow. The results of the two case studies can inform
planning of long-term basin water management strategies taking into account global
change scenarios.
Citation: Fu, G., S. P. Charles, and F. H. S. Chiew (2007), A two-parameter climate elasticity of streamflow index to assess climate
change effects on annual streamflow, Water Resour. Res., 43, W11419, doi:10.1029/2007WR005890.

1. Introduction validation of such a model remains a fundamental challenge


[2] Water resource planning and management increasingly [Sankarasubramanian et al., 2001]. For example, Nash and
need to incorporate the impacts of global climate change in Gleick [1991] and Schaake [1990] used the National
order to accurately predict future water supplies. Numerous Weather Service River Forecasting System (NWSRFS) to
studies have documented the sensitivity of streamflow to perform climate sensitivity analyses on the Animas River at
climate changes for basins across the world [Fu and Liu, Durango, Colorado. When precipitation was increased by
1991; Yates and Strzepek, 1998; Sankarasubramanian et 10%, holding temperature and potential evapotranspiration
al., 2001; McCarthy et al., 2001; Arnell, 2002; Chiew, constant, Nash and Gleick [1991] and Schaake [1990]
2006; Fu et al., 2007a]. Most of these studies involve reported an 11% and 20% increase, respectively, in annual
estimating the precipitation elasticity of streamflow, which streamflow. These are rather remarkable differences, consid-
was introduced by Schaake [1990] for evaluating the ering the same model was applied to the same basin in both
sensitivity of streamflow to changes in climate. Dooge instances. These are likely due to differences in model
[1992] and Dooge et al. [1999] termed it a sensitivity factor calibrations leading to differences in model parameter esti-
and Kuhnel et al. [1991] termed it a magnification factor. mates which result in differences in the models’ sensitivity to
[3] Climate elasticity of streamflow may be defined as the climatic variations [Sankarasubramanian et al., 2001].
proportional change in streamflow, Q, to the change in a [5] Vogel et al. [1999] use a regional multivariate regres-
climatic variable such as precipitation, P. Thus precipitation sion model to document that a 10% increase in precipitation
elasticity of streamflow is defined as [Schaake, 1990; leads, on average, to a 19% increase in annual streamflow
Sankarasubramanian et al., 2001]: for the entire upper Colorado River system, of which the
Animas River is only a small subbasin. Treating the entire
dQ=Q dQ P upper Colorado River as a single basin, Revelle and
ep ð P; QÞ ¼ ¼ ð1Þ Waggoner [1983] report an at-site multivariate regression
dP=P dP Q
relationship among streamflow, precipitation and tempera-
ture such that a 10% increase in annual precipitation leads to
[4] One difficulty with estimation of elasticity is that it is an 11% increase in streamflow.
often estimated from a hydrological model and, of course, [6] These four estimates of the sensitivity of streamflow
the form of the hydrological model is always unknown and to precipitation for the Animas basin are quite varied and
cannot all be correct. In order to solve this paradox,
Sankarasubramanian et al. [2001] introduced a specific
1
CSIRO Land and Water, Wembley, WA, Australia. case of (1) at the mean value of the climatic variable:
2
CSIRO Land and Water, Canberra, ACT, Australia.
  dQ mP
Copyright 2007 by the American Geophysical Union. ep mp ; mQ ¼ j ð2Þ
0043-1397/07/2007WR005890 dP P¼mP mQ

W11419 1 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

Where mP and mQ denote mean values of precipitation and associated with the elasticity estimation procedure. Section 5
streamflow. The definition ep in (2) is the metric employed draws some conclusions and highlights future research
by Sankarasubramanian et al. [2001] for comparing the priorities.
climate response of river basins and has been suggested by
Chiew [2006] as an intrinsic property of catchments. 2. Two-Parameter Climate Elasticity of
[7] Sankarasubramanian et al. [2001] further verified
Streamflow Index
that the non-parametric estimator:
2.1. Streamflow-Precipitation-Temperature
  Relationship
Qt  Q P
eP ¼ median ð3Þ [11] Many investigations have used empirical rainfall-
Pt  P Q
runoff models to study the impacts of climate change on
is a robust estimator of the precipitation elasticity of hydrology. For example, the relationships between mean
streamflow for a wide class of hydrological models that annual precipitation, temperature, and runoff developed by
does not depend on the form of the hydrological model. t is Langbein [1949], based on 22 drainage basins in the
time in equation (3). contiguous United States, were used by Stockton and
[8] Several studies have used (3) to investigate the impacts of Boggess [1979] to estimate changes in the average annual
climate change on annual streamflow [Sankarasubramanian runoff of 18 designated regions throughout the United
and Vogel, 2003; Niemann and Eltahir, 2005; Chiew, 2006; States for different climate scenarios. Revelle and Waggoner
Novotny and Stefan, 2007]. For example, Sankarasubramanian [1983] used the same model as the basis for investigating
and Vogel [2003] used it to document the precipitation elasticity the effects of climate change on runoff in the western United
of streamflow for 1337 basins in the USA showing that a States [Leavesley, 1994]. Mirza [1997] also used the
1% change in precipitation results in a 1.5– 2.5% change in Langbein model for a similar study on the Ganges River
basin runoff, depending upon the degree of buffering by basin. However, applications of these empirical relation-
storage processes and other factors. ships to climate or basin conditions different from those
[9] The critical concern for (3) is that it is only a function used in the original development of these functions are
of precipitation, but not a function of evapotranspiration questionable [Leavesley, 1994].
(temperature). This limits the applicability of (3) when [12] Risbey and Entekhabi [1996] avoided this problem
assessing the climate effects on streamflow of future global by using the observed data from a single basin and pre-
warming scenarios. Arora [2002] introduced the use of the sented their results in contour format by using the adjustable
aridity index, which is a function of both precipitation and tension continuous curvature surface grid algorithm pro-
energy, to assess climate change effects on annual runoff, posed by Smith and Wessel [1990].
 The annual
 percentage
Qt Q
but Sankarasubramanian and Vogel [2003] reported that the departures for streamflow  100% , precipitation
  Q
aridity index is not an adequate predictor, by itself, of either Pt P
 100% , and temperature (Tt  T ) for a specific
the long-term water balance or the variability of annual P
runoff. Limbrunner [1998] and Chiew [2006] separated the basin were calculated and plotted on a precipitation-
precipitation elasticity of streamflow (eP) and the potential temperature plane based on the methodology of Risbey
evaporation elasticity of streamflow (ePET) and estimated and Entekhabi [1996], such that each point in the plane
the eP and ePET, for the United States and Australia represented one year of observed  data. The contours of
respectively. However, these elasticities should not be streamflow percentage change QtQQ  100% were then
considered separately, but rather as a joint precipitation interpolated from these points and transferred to a regular
and evaporation (temperature) elasticity of streamflow. grid for contouring using the adjustable tension continuous
[10] The primary goal of this study is to extend (3) into a curvature surface gridding algorithm proposed by Smith and
two-parameter index, i.e., climate elasticity of streamflow as Wessel [1990]. However, the interpolation algorithm of
a function of both precipitation and temperature. Section 2 Smith and Wessel [1990] is just one of many interpolation
describes the detailed development of this index in two steps: methods and may not be the ‘‘optimal’’ interpolation
(1) constructing the streamflow-precipitation-temperature method for a specific basin. Fu et al. [2007a] modified
relationship using the ArcGIS Geostatistical Analyst pack- the procedure by using ArcGIS Geostatistical Analyst,
age and historical records; and (2) computing the climate which provides a comprehensive set of tools for creating
elasticity of streamflow for different precipitation and tem- surfaces from measured sample points and interpolation
perature scenarios. Section 3 applies the proposed two- results can subsequently be used in GIS models for visual-
parameter climate elasticity of streamflow index to two ization, analyses, and understanding spatial phenomena.
basins, the Spokane River basin in the USA and Yellow There are at least two distinct advantages of this modified
River basin in China. The results indicate that the new two- approach compared to the Risbey and Entekhabi [1996]
parameter climate elasticity of streamflow can reflect the procedure. First, ArcGIS Geostatistical Analyst provides a
complicated non-linear relationship among streamflow, pre- comprehensive set of tools for creating surfaces from
cipitation, and temperature and is appropriate to assess the measured sample points compared to the single algorithm
climatic effects on annual streamflow of future climate of Risbey and Entekhabi [1996]. This allows users to
change scenarios. Section 4 compares the new two-parameter efficiently compare different interpolation techniques in
climate elasticity index with single parameter precipitation order to produce the ‘‘optimal’’ solution. Second, the
elasticity of streamflow results, analyses of the non-liner methodology can easily be applied and expanded to differ-
streamflow responses to precipitation, compares the differ- ent basins where the results can subsequently be used in
ent interpolation techniques, and discusses the uncertainties GIS models for visualization and analysis.
2 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

2.2. Climate Elasticity of Streamflow supplies. Climate change predictions need to be incorporat-
ed into the process of planning for future growth and
[13] The streamflow-precipitation-temperature relation- sustainable water management.
ship is converted into a climate elasticity of streamflow [18] On the basis of the estimates of Fu et al. [2007a], the
index: long-term (1917 – 2001) water year (October to next Sep-
  tember) average annual precipitation for Spokane River
QP;dT  Q P Basin is 797 mm. About 55.3% of precipitation, or
eP;dT ¼ ð4Þ
PP;dT  P Q 441 mm, becomes streamflow sometime during the water
year, a runoff coefficient of 0.553.
where dT = (T  T ) is the temperature departure. 3.1.2. Yellow River Basin
[14] This climate elasticity of streamflow index is a [19] The Yellow River is the 6th longest river in the world
conditional precipitation elasticity of streamflow index and 2nd longest river in China. The catchment area is as
accounting for the effects of temperature from the temper- large as 794,712 km2 if the Erdos inner flow area is
ature-precipitation plane, i.e., it is not an index of the full included [Fu et al., 2004, Fu and Chen, 2006]. The average
joint precipitation and temperature elasticity of streamflow. annual precipitation for 1957 –1997 was 455.8 mm for the
However, as a function of precipitation and temperature it entire basin, based on the 44 stations of Fu and Chen
can be used to assess the climatic effects of joint precipi- [2005]. The average annual streamflow at Hua-Yuan-Kou
tation and temperature changes on hydrological regimes at station is 5.747  1010 m3, a runoff depth of 78.7 mm. The
basin scales. Thus for a given precipitation and temperature runoff coefficient is thus only 0.173 [Fu and Chen, 2005,
change scenario (if the scenario is in the range of past 2006].
observed climate), this index can predict the annual stream- [20] With population growth and regional economic de-
flow response. velopment, the observed streamflow of the Yellow River
[15] An example serves to highlight how (4) is applied. In has decreased significantly during the last 50 a. In the
the Spokane River basin, USA, the long-term mean annual drought year 1997, the river no-flow period lasted 227 days
precipitation, streamflow and temperature are 797 mm (P), in Li-Jin station, and for 330 days there was no water
441 mm (Q), and 7.07°C (T), respectively. For an arbitrary discharged to the sea. Human activities are the leading cause
year, say the annual precipitation and temperature are 1036 of the water resources crisis in the Yellow River basin
mm (a 30% increase over the long-term mean) and 8.57°C during the last five decades [Chen et al., 2003; Fu et al.,
(1.5°C higher than the long term mean). From the stream- 2004; Liu and Xia, 2004; Fu and Chen, 2006]. However,
flow-precipitation-temperature interpolated surface (pro- the additional impact of future climate change on the
duced using historical data and ArcGIS Geostatistical regional hydrological regime and water resource availability
Analyst, as described later in the paper) we can estimate is critical for water resource planning and management [Fu
that the annual streamflow is about 551 mm for this et al., 2004; 2007b].
precipitation and temperature combination. We can then 3.2. Data Sets
use (4) to compute the climate elasticity of streamflow for
this scenario to be about 0.83. Thus for a future climatic 3.2.1. Spokane River Basin
change scenario, given the precipitation and temperature [21] Within or near the Spokane River basin there are
anomalies, the climate elasticity of streamflow index can be 51 climate stations whose data are available from the US
used to predict the annual streamflow. National Climatic Data Center (NCDC). The longest run-
ning stations go back to 1917 although most of these
stations do not have continuously observed data. Fifteen
3. Climate Elasticity of Streamflow Case Studies stations with records longer than 60 a were finally chosen
3.1. Study Basins for use in this study. Although a relatively dense network of
[16] This investigation applies the two-parameter climate climate observation was used, the NCDC climate stations in
elasticity of streamflow index to two basins: the Spokane this analysis are all located below 900 m elevation [Fu et
River basin in the USA and the Yellow River basin in al., 2007a]. About 57% of basin area is located above 900 m
China. and the average elevation for the basin is about 963 m. This
3.1.1. Spokane River Basin means that the arithmetical mean of the 15 stations would
[17] The Spokane River basin has a catchment area of underestimate the basin precipitation and overestimate the
17,200 km2 and is located in Eastern Washington and basin temperature. The results of the Parameter-elevation
Northern Idaho, USA. Spokane River, a tributary to the Regressions on Independent Slopes Model (PRISM) were
Columbia River above Grand Coulee Dam, is a vital part of therefore used to account for the dramatic spatial physical
the quality of life in Spokane and Coeur d’Alene cities, heterogeneity [Spatial Climate Analysis Service, 2004]. A
offering riverfront trails and parks, a prized trout fishery, previous study indicates that the 15 station averages under-
whitewater recreation, and dramatic natural scenery. How- estimated areal precipitation by 27.4% and overestimated
ever, the Spokane River was ranked No. 6 on the most areal temperature by 0.93°C [Fu et al., 2007a]. The stream-
endangered rivers in America list in 2005 by American flow station used in this study is USGS Station 12433000,
Rivers and its Partners [American Rivers, 2005] due to ‘‘too Spokane River at Long Lake, with a catchment area of
little water, too much pollution, and an uncertain future.’’ 15,592 km2.
Continued population growth and regional development 3.2.2. Yellow River Basin
will require innovative solutions to allocation of water, as [22] Forty-four standard meteorological stations in the
managers begin to understand the finite nature of local Yellow River basin, giving monthly precipitation and
monthly means of daily mean temperature, were used for
3 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

flooding events if any. The detailed methods and formula


for estimating each of these components are described and
discussed by YRCC [1997].
[25] Several studies have been undertaken to analyze the
annual variations of the estimated naturalized streamflow
and the water resources variability during the last half
century [Li et al., 2001; Wang et al., 2002; Mu et al.,
2003; Yang et al., 2004; Zhang et al., 2005]. There are no
obvious biases (i.e., overestimation and/or underestimation)
having been reported in these studies. Furthermore, a
bivariate test [Potter, 1981] shows that there is no system-
atic change in mean annual streamflow conditional on the
annual precipitation series. Our computed bivariate test
statistic T0 is 3.87 (n = 41), compared with the critical
value of 8.7 (a = 0.05) and 11.6 (a = 0.01) for n = 40
[Potter, 1981]. Therefore this naturalized streamflow data
Figure 1. Contour plot of percentage annual streamflow set is used in this study. The annual average naturalized
change as a function of annual percentage precipitation flow for the study period (1957– 1997) is 5.747  1010 m3,
change and temperature change for the Spokane River compared to the annual average observed flow of 4.028 
basin. 1010 m3, a difference of 29.9%.
3.3. Streamflow-Precipitation-Temperature
Relationship
this study. These stations were among 160 first-class mete-
orological stations in China with high quality data. The [26] Although the magnitudes and patterns of streamflow
1957-1997 data were used to ensure that each year had at response to changes in precipitation and temperature differ
least 41 stations. These stations were maintained according somewhat between the basins, the general relationship is
to standard methods with the data being quality control by clear from Figures 1 and 2. Streamflow is positively related
the National Meteorological Administration of China. The to precipitation but negatively related to temperature, with
details for each of these stations including location, eleva- the precipitation-streamflow sensitivity greater than the
tion, data period etc. are described by Fu et al. [2004]. The streamflow-temperature sensitivity.
streamflow station used in this study is Hua-Yuan-Kou [27] The impacts of temperature on streamflow change
station, with a catchment area of 730,036 km2. It is the are clear in Figures 1 and 2. For example, a 30% precipi-
most important hydrological station at the main channel of tation increase in the Spokane River basin results in a 50%
the Yellow River and is located where the middle and lower increase in streamflow at mean temperature, but only a 20–
reaches divide. Both ‘‘naturalized’’ and observed stream- 30% increase in streamflow for a temperature increase of
flow usually reach the maximum volume at this station 1.5°C (Figure 1). A 20% precipitation decrease results in a
because there is limited water flowing into the river channel 25 –30% decrease in streamflow at mean temperature but a
downstream of this point. This is because the riverbed is 50% decrease in streamflow if the temperature increases
higher than the land outside the dikes. The hydrologic 1.5°C (Figure 1). For the Yellow River basin a 30%
regime at this station thus represents an integration of the precipitation increase results in a 45% increase in stream-
hydrologic regime of the entire river basin [Fu and Chen, flow at mean temperature but only a 20% increase if the
2006]. temperature increases 0.8°C (Figure 2). A 20% precipitation
[23] Considering the influence of human activities such
as irrigation, industry, and domestic uses, and the role of
dams to control the streamflow, a calculated ‘‘naturalized
runoff’’ amount, instead of observed streamflow, is used in
this study. The monthly naturalized runoff data were pro-
vided by the Yellow River Conservancy Commission
(YRCC), which is the only authority in China to conduct
water data quality control of the Yellow River and to release
these data.
[24] The difference between observed runoff and natural-
ized runoff generally results from the following factors
[YRCC, 1997; Li et al., 2001]: (a) the amount of water
directly abstracted from the river channel for irrigation,
industry, and domestic use, and the amount returning to
the downstream river channel after use; (b) the amount of
water controlled by dams, including extra water losses
through evaporation and seepage due to dams; (c) the
amount of water transported into and out of the basin; Figure 2. Contour plot of percentage annual runoff
(d) the amount of water used for soil erosion conservation; change as a function of percentage precipitation change
and (e) the water amount taken via ‘‘bypass’’ channel during and temperature change for the Yellow River basin.

4 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

precipitation change while the Yellow River has a smaller


percentage streamflow decrease than the precipitation
change when the precipitation decreases more than 15–
20%. (2) The streamflow in the Yellow River is more
sensitive to temperature change than that in the Spokane
River. If we compare Figure 2 with Figure 1, it can be seen
that there are parallel ‘‘vertical’’ lines for the Spokane River
basin but no such parallel lines for the Yellow River basin.
Simple linear regression analyses verify this observation.
The coefficient of determination (R2) improved from 0.649
for the streamflow-precipitation regression to 0.730 for the
streamflow, precipitation, and temperature regression for the
Yellow River basin [Fu et al., 2007b], whereas it only
increases from 0.707 to 0.759 for the Spokane River basin
[Fu et al., 2007a].
3.4. Climate Elasticity of Streamflow
[29] The quantified streamflow-precipitation-temperature
Figure 3. Annual streamflow change minus precipitation relationship (Figures 1 and 2) is used as input to equation
change as a function of precipitation change for the two (4) to calculate the climate elasticity of streamflow index in
basins. the precipitation-temperature plane (Figures 4 and 5).
3.4.1. Non-Constant Climate Elasticity of Streamflow
decrease results in a 15% decrease in streamflow at mean [30] The climate elasticity of streamflow is not a fixed
temperature but a 30% decrease if the temperature increases value for a specified basin. This contradicts the conclusion
0.8°C (Figure 2). of Sankarasubramanian et al. [2001] that precipitation
[28] Differences between the Yellow River and Spokane elasticity is fixed for each individual basin. Figures 4 and
River basins are evident when comparing Figures 1 and 2. 5 clearly show that the climate elasticity of streamflow
(1) The percentage streamflow change as a function of varies with precipitation and temperature: it usually
percentage precipitation and temperature change is not increases with precipitation if the temperature increases,
symmetric for the Yellow River basin with respect to the but decreases with precipitation if the temperature
direction of precipitation change (Figure 2). This implies decreases; it increases with temperature if the precipitation
that the magnitude of streamflow response to precipitation decreases, but decreases with temperature if the precipita-
change will be different for precipitation increase or de- tion increases. The general trend is summarized in Figure 6.
crease scenarios. This is not the case for the Spokane River 3.4.2. Numerical Instability Near P
basin where the streamflow response to the changes in [ 3 1 ] Figures 4 and 5 confirm the conclusion of
precipitation is symmetric (Figure 1). This difference is Sankarasubramanian et al. [2001] that there is a numerical
highlighted in Figure 3. As precipitation increases, the two instability with equations (3) or (4) when Pt approaches P,
basins show similar patterns of streamflow increases that are causing the elasticity to approach infinity. This is evident
proportionally greater than the precipitation changes. How- from the relationship between climate elasticity of stream-
ever, when the precipitation decreases, the streamflow flow and precipitation (Figure 7). Because of this numeric
responses go in different directions: the Spokane River instability, all discussions of precipitation change herein will
basin has a larger percentage streamflow decrease than the focus on precipitation changes of more than 10%. This

Figure 4. Climate elasticity of streamflow as a function of percentage precipitation change and


temperature change for the Spokane River basin.
5 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

Figure 5. Climate elasticity of streamflow as a function of percentage precipitation change and


temperature change for the Yellow River basin.

index cannot be used for a scenario of precipitation change 3.4.4. Precipitation Sensitivity
less than 5%. [33] The slope of Figure 3 indicates that a 1% precipita-
3.4.3. Negative Climate Elasticity of Streamflow tion change in the Spokane River basin results in a 1.5%
[32] Note that the climate elasticity of streamflow can be streamflow change. This is consistent with the equation (3)
negative, i.e., a precipitation increase resulting in a decrease estimation that the precipitation elasticity of streamflow is
in annual streamflow or vice versa. This happens in two 1.67, i.e., a 1% precipitation change results in a 1.67%
cases: a small precipitation increase combined with a large streamflow change. However, Figure 7 indicates that the
temperature increase, and a small precipitation decrease climate elasticity of streamflow may be different depending
combined with a large temperature decrease. The former on whether precipitation increases or decreases, even if
is more significant for both basins. For example, a 10% temperature is unchanged. The climate elasticity of stream-
precipitation increase in the Spokane River basin could still flow is 1.1 – 1.4 for precipitation increases and 1.6 for
lead to an annual streamflow decrease if the basin temper- precipitation decreases for the Spokane River basin (Figure 7).
ature increases by 1.0°C or more (Figure 4). This critical This indicates that a 1% precipitation increase generally
temperature reduces into about 0.4°C for the Yellow River results in a 1.1 – 1.4% increase in streamflow and a 1%
basin, where the annual streamflow may still reduce even if precipitation decrease generally results in 1.6% decrease in
precipitation increases by 20% (Figure 5). Conversely, an streamflow. Thus streamflow is more sensitive to decreasing
increase in annual streamflow could occur when precipita- precipitation.
tion decreases less than 5% and temperature decreases. [34] The slope of Figure 3 indicates that that a 1%
However, as mentioned earlier, the climate elasticity of precipitation increase in the Yellow River basin results in
streamflow is not stable for such a small precipitation a 1.4% streamflow increase, approximately consistent with
change. the equation (3) estimation that the precipitation elasticity of

Figure 6. Climate elasticity of streamflow changes with precipitation and temperature.


6 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

temperature variation is small. Therefore for small pertur-


bations in temperature, these methods would be expected to
work reasonably well. However, for larger temperature
changes (e.g., >2°C) there are thresholds in the response
that may not be well represented in the observed data.
Snowmelt dominant basins in particular have topographi-
cally driven thresholds related to snow loss that are rela-
tively weak in 20th century records, but grow much stronger
(in simulations) as the temperature continues to increase.
[38] However, these methods and results are still useful
for water resource management and planning as (1) for the
next two decades, a warming of about 0.2°C per decade is
projected for a range of IPCC SRES emission scenarios
[IPCC, 2007]; (2) a relatively small temperature increase
already has serious implications on water supply and a
Figure 7. Climate elasticity of streamflow as a function of larger temperature increase would amplify them based on
precipitation change (%) for the two basins. the temperature-streamflow relationships explored in this
study.

streamflow is 1.69, i.e., a 1% precipitation change results in 4. Discussion


a 1.69% streamflow change. The climate elasticity of
streamflow in the Yellow River basin, as shown in Figure 7, 4.1. Comparisons With Single Parameter Precipitation
varies from 0.76 to 1.78 when precipitation decreases Elasticity of Streamflow
(excluding the non-stable values within ±10% precipitation [39] The single parameter precipitation elasticity index
change). This indicates that the single parameter elasticity (3) provides a general relationship between precipitation
(3) may give the general relation between precipitation and and streamflow for the Spokane River basin. However, it
streamflow in some basins (e.g the Spokane River basin), does not provide a general relationship between precipita-
but it cannot reflect the complicated non-linear relation tion and streamflow for the Yellow River basin as stream-
between streamflow and precipitation in some basins, such
as the Yellow River basin.
3.4.5. Temperature Sensitivity
[35] Figures 4 and 5 clearly indicate that climate elasticity
of streamflow varies with temperature. It usually increases
with temperature as precipitation decreases, and decreases
with temperature as precipitation increases (Figure 6).
This sensitivity is clearer in Figures 8 – 11, which are
several cross sections of Figures 4 and 5. Increases in
temperature always reduce streamflow for a fixed precipi-
tation scenario, increase the climate elasticity of streamflow
for decreasing precipitation, and decrease the climate elas-
ticity of streamflow for increasing precipitation scenarios
(Figures 9 and 11).
[36] This result means that water availability in both
basins is likely to be more critical in the future if current
global warming projections are correct. The Intergovern-
mental Panel on Climate Change (IPCC) in its Fourth
Assessment Report [IPCC, 2007] stated that for the next
two decades, a warming of about 0.2°C per decade is
projected for a range of IPCC Special Report on Emission
Scenarios (SRES). Even if the concentrations of all green-
house gases and aerosols had been kept constant at year
2000 levels, a further warming of about 0.1°C per decade
would be expected. Furthermore, based on recent global
model simulations, nearly all land areas will warm more
rapidly than the global average, and warming is expected to
be greatest over land for most high northern latitudes [IPCC,
2007]. This will have serious consequences for urban water
supply, agricultural production, industry development, and
ecological systems in both basins, based on our climate
elasticity of streamflow results.
Figure 8. The streamflow change (a) and climate elasticity
[37] One critical caveat of this climate elasticity index is
of streamflow (b) as function of precipitation change at
that it is derived from the observed data, in which the annual
different temperature scenarios for the Spokane River basin.

7 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

[43] Figure 8 is constructed from several cross sections of


Figures 1 and 4 in the temperature plane and shows the
streamflow change and climate elasticity of streamflow as a
function of precipitation change at three fixed temperature
scenarios for the Spokane River basin. The streamflow-
precipitation relationship is close to linear for a fixed
temperature scenario, but the climate elasticity of stream-
flow - precipitation relationship is not close to linear
even for a fixed temperature scenario. For example, for
the temperature increase of 1.5°C scenario, the climate
elasticity of streamflow is close or larger than 2.0 for a
decrease in precipitation, but is less than 1.0 for an increase
in precipitation.
[44] Figure 9, several cross sections of Figures 1 and 4 in
the precipitation plane, shows the streamflow change and
climate elasticity of streamflow as a function of temperature
change for different precipitation scenarios for the Spokane
River basin. The streamflow-temperature relationship is
close to linear for a given precipitation change scenario,
although the slopes are quite different for the different
precipitation change scenarios. For the precipitation de-
crease of 20% scenario, the slope of streamflow-temperature
line is about 9.0%/°C, i.e., one degree temperature increase
usually results in a streamflow decrease of 9.0%. However,
for the precipitation increase of 20% scenario, the slope of
streamflow-temperature line is about 16.6%/°C, i.e., one
degree temperature increase usually results in a streamflow
decrease of 16.6%. Thus streamflow is more sensitive to
temperature change when precipitation increases. If the
Figure 9. The streamflow change (a) and climate elasticity
of streamflow (b) as function of temperature change at
different precipitation scenarios for the Spokane River
basin.

flow was shown to have different responses depending on


whether precipitation increases or decreases (Figures 2, 3
and 7).
[40] In addition, it is clear that the single parameter
precipitation elasticity index cannot reflect the complicated
non-linear relationship among streamflow, precipitation and
temperature as it is not a function of temperature. The case
studies indicate that temperature, as a proxy for evapotrans-
piration, is a critical factor for runoff generation. Therefore
the current precipitation elasticity of streamflow index
should not be used to assess the effects of future climate
change on annual streamflow.
[41] The modified two-parameter climate elasticity of
streamflow index, a function of both precipitation and
temperature, can be readily used to assess the climate
change effect on streamflow at the basin-scale, especially
the impacts of temperature on annual streamflow.
4.2. Non-Linear Streamflow Responses to Precipitation
[42] If we observe Figure 3 alone, we may conclude that
the streamflow-precipitation relation is linear for the Spo-
kane River basin and is non-linear for the Yellow River
basin. This might be true if the temperature impact is
ignored. An obvious feature of Figures 1, 2, 4 and 5 is that
the response of streamflow to temperature is nonlinear, Figure 10. The streamflow change (a) and climate
which results in the non-linear streamflow-precipitation- elasticity of streamflow (b) as function of precipitation
temperature relationship for both basins. change at different temperature scenarios for the Yellow
River basin.
8 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

[47] Figure 11, several cross sections of Figures 2 and 5


in the precipitation plane, shows the streamflow change and
climate elasticity of streamflow as a function of temperature
change for several fixed precipitation scenarios for the
Yellow River basin. Unlike the linear relationship between
the streamflow and temperature for a fixed precipitation
scenario in the Spokane River basin (Figure 9), the stream-
flow-temperature relationship is non-linear. The streamflow
change is almost independent of temperature if the temper-
ature decreases, but is sensitive to increases in temperature,
especially for the precipitation increase 20% scenario. For
example, for the precipitation increase of 20% scenario, a
0.5°C temperature increase usually results in a streamflow
decrease of 17%, which is almost four times that of the
Spokane River basin. For the precipitation decrease 20%
scenario, every 0.5°C temperature increase results in a
streamflow decrease of 11%, which is only about 1.3 times
that of the Spokane River basin. Like the Spokane River
basin, streamflow is more sensitive to temperature change
for the precipitation increase scenarios. If precipitation
remains the same, the streamflow varies within ±10% if
the temperature change ranges from 0.9°C to +0.4°C, but
a 0.8°C temperature increase alone results in a streamflow
decrease of 20% (Figure 11).
[48] Similar to the climate elasticity of streamflow-
temperature relationship in the Spokane River basin
(Figure 9), the climate elasticity of streamflow in the Yellow
Figure 11. The streamflow change (a) and climate
River basin increases with temperature as precipitation
elasticity of streamflow (b) as function of temperature
decreases, but decreases with temperature as precipitation
change at different precipitation scenarios for the Yellow
increases (Figure 11). However, there are two distinct
River basin.
differences: (1) the range of climate elasticity of streamflow
in the Yellow River basin is smaller than that in the Spokane
precipitation remains the same, the streamflow will not River basin. This might be due to the fact that the area of the
respond to temperature change within a ±1°C range. How- Yellow River basin is much larger than that of the Spokane
ever, a 1.5°C temperature increase alone could result in a River basin and the temperature change range is narrower in
streamflow decrease of 15%. the Yellow River basin; (2) the climate elasticity of stream-
[45] Correspondingly, the climate elasticity of streamflow flow in the Yellow River basin is less sensitive to decreasing
increases with temperature in the precipitation decrease temperature compared to the Spokane River basin, where
scenario, but decrease with temperature in the precipitation the climate elasticity of streamflow is sensitive to both
increase scenario. For example, the climate elasticity of temperature increases and decreases (Figure 9). This might
streamflow varies from 1.0 to 2.4 as temperature varies from be related to the snow accumulation and snowmelt processes
1°C lower to 1.8°C higher than the long-term mean for the of the Spokane River basin [Fu et al., 2007a], the spatial
precipitation decrease 20% scenario. In the precipitation inhomogeneous temperature increase trend and magnitude
increase 20% scenario, the climate elasticity of streamflow in the Yellow River [Fu et al., 2004], and the arid or semi-
decreases from 2.4 to 0.2 for the same temperature variation arid climate of the Yellow River [Fu and Chen, 2006], but
(Figure 9). needs further investigation. In addition, the impacts of
[46] Figure 10, several cross sections of Figures 2 and 5 El Niño and La Niña on precipitation and streamflow show
in the temperature plane, shows the streamflow change and a corresponding response, i.e., streamflow in the Yellow
climate elasticity of streamflow as a function of precipita- River is more sensitive to temperature increases rather than
tion change for three fixed temperature scenarios for the temperature decreases [Fu et al., 2007b].
Yellow River basin. The streamflow-precipitation relation- 4.3. Interpolation Technique Comparisons
ship is not linear even for a fixed temperature scenario. This
[49] The streamflow-precipitation-temperature relation-
is in contrast to the Spokane River basin (Figure 8), but
ships and climate elasticity of streamflow index reported
consistent with Figure 3, reflecting the differences in basin
in this study (Figures 1, 2, 4 and 5) depend on the geo-
runoff generation processes. The climate elasticity of
statistical interpolation techniques used to interpolate the
streamflow - precipitation relationship is accordingly not
streamflow change at each precipitation change and tem-
linear even for a fixed temperature scenario. For example,
perature change point based on the observed data. Mathe-
for the temperature increase 0.8°C scenario the climate
matically, the best geostatistical model for spatial
elasticity of streamflow varies from 1.4 to 2.3 when the
interpolation is the one that has the standardized mean
precipitation decrease changes from 26% to 10%, but
(dimensionless) nearest to zero, the smallest root-mean
varies from 1.0 to 0.8 when precipitation increases from
square prediction error (streamflow change in percentage),
10% to 41% (Figure 10).
the average standard error (streamflow change in percent-
9 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

Table 1. The Prediction Errors of Different Interpolation Methods in ArcGIS Geostatistical Analyst for the
Spokane River basin
Standardized Root Mean Average Standard Standardized Root
Methods Mean Mean Square Error Mean Square

Inverse distance weighted 0.0735 16.03


Global polynomial 0.09825 15.68
Local polynomial 0.1031 15.75
Radial basis functions 0.1758 15.18
Ordinary Kriging 0.02854 0.004384 16.11 13.06 1.309
Ordinary Kriging with first 0.3208 0.01975 15.76 14.74 1.064
order trend removal
Simple Kriging 0.3608 0.0113 17.44 24.51 0.7015
Universal Kriging with 0.02854 0.004384 16.11 13.06 1.309
constant order of trend
Universal Kriging with 0.2584 0.01748 17.19 15.19 1.107
first order of trend
Disjunctive Kriging 0.2329 0.00623 17.73 23.22 0.7516
Disjunctive Kriging with 1.927 0.1452 15.04 13.37 1.125
first order trend removal

age) nearest the root-mean square prediction error, and the Spokane River basin (Table 1) and the ‘‘Ordinary Kriging’’
standardized root-mean square prediction error (dimension- model produced the ‘‘optimal’’ fit with observed data for the
less) nearest to one [Johnston et al., 2001]. Yellow River basin (Table 2). The global polynomial
[50] ArcGIS Geostatistical Analyst provides two groups method gave the simplest form and best visualization of
of interpolation techniques: deterministic and geostatistical the relationship among streamflow, precipitation, and tem-
models. Deterministic techniques use mathematical func- perature. One advantage of ArcGIS Geostatistical Analyst,
tions for interpolation whereas geostatistical methods rely compared to the adjustable tension continuous curvature
on both statistical and mathematical methods and can be surface gridding used by Risbey and Entekhabi [1996], is
used to create surfaces and assess the uncertainty of that it automates the application of the different interpola-
predictions [Johnston et al., 2001]. tion models so users can quickly compare results to obtain
[51] Of the eleven interpolation models (four determinis- the ‘‘optimal’’ interpolation. Regarding the climate elasticity
tic interpolation models: Inverse Distance Weighted (IDW), of streamflow, although the eleven models give different
global polynomial, local polynomial, and radial basis func- results (Table 3) they give consistent results for different
tions (RBFs) and seven geostatistical models: Ordinary basins, i.e., if one model produces a ‘‘high’’ elasticity value
Kriging, Ordinary Kriging with first order trend removal, for a basin, then it is likely this model also produces a
Simple Kriging, Universal Kriging with constant order of ‘‘high’’ value for another basin (Table 3).
trend, Universal Kriging with first order of trend, Disjunc- [52] Compared with linear regression analysis, equation
tive Kriging, and Disjunctive Kriging with first order (3) seems to overestimate the precipitation elasticity of
removal) available in ArcGIS Geostatistical Analyst, the streamflow and our ‘‘optimal’’ interpolation models seem
‘‘Ordinary Kriging with first order trend removal’’ model to give closer precipitation elasticity of streamflow values
produced the ‘‘optimal’’ fit with observed data for the (Table 3). Although the climate elasticity of streamflow has

Table 2. The Prediction Errors of Different Interpolation Methods in ArcGIS Geostatistical Analyst for the
Yellow River basin
Standardized Root Mean Average Standard Standardized Root
Methods Mean Mean Square Error Mean Square

Inverse distance weighted 1.276 15.91


Global polynomial 0.1542 12.91
Local polynomial 0.1226 12.4
Radial basis functions 1.011 15.17
Ordinary Kriging 0.229 0.006707 14.32 14.0 1.108
Ordinary Kriging with 0.1516 0.01148 12.33 14.95 0.8328
first order trend removal
Simple Kriging 1.175 0.05006 17.98 21.49 0.8183
Universal Kriging with 0.229 0.006707 14.32 14.0 1.108
constant order of trend
Universal Kriging with 0.1037 0.03441 13.98 15.56 0.9913
first order of trend
Disjunctive Kriging 1.548 0.07559 18.3 18.93 0.9421
Disjunctive Kriging with 0.201 0.03559 12.79 11.2 1.156
first order trend removal

10 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

Table 3. The Climate Elasticity of Streamflow From Different temperature scenarios where there is limited or no sample
Interpolation Methods data.
Spokane River Yellow River
Methods Basin Basin 5. Conclusions
Inverse distance weighted 0.87 0.72 [57] The single parameter precipitation elasticity method
Global polynomial 1.45 1.17 may represent the general relationship between precipitation
Local polynomial 1.35 1.13
Radial basis functions 1.06 0.94 and streamflow in some cases, such as the Spokane River
Ordinary Kriging 1.43 1.20 basin, but it cannot reflect the complicated non-linear
Ordinary Kriging with 1.23 1.34 relationship among streamflow, precipitation and tempera-
first order trend removal ture. The modified model introduced here extends the
Simple Kriging 1.42 1.18
Universal Kriging with 1.22 1.11
precipitation elasticity of streamflow into a two-parameter
constant order of trend index, which depends on both precipitation and tempera-
Universal Kriging with 0.88 0.57 ture. This index can then be used to assess the sensitivity of
first order of trend annual streamflow to both precipitation and evapotranspi-
Disjunctive Kriging 1.23 1.34 ration (temperature) changes. The application of the mod-
Disjunctive Kriging with 1.30 1.52
first order trend removal ified method for two basins indicates that temperature is a
Equation (3) 1.67 1.69 critical factor controlling the regional runoff generation and
Regression analysis 1.49 1.27 that streamflow response to precipitation and temperature
exhibits a non-linear relationship.

different values from different interpolation models, the


climate elasticity of streamflow-precipitation relation seems
consistent among the different models (Figure 12).
4.4. Uncertainties
[53] There are several aspects to the uncertainties of the
climate change effects on streamflow obtained using elas-
ticity estimates derived from interpolation models. First, this
model is developed and applied using historical observed
data for the last 40– 60 a. It assumes that the streamflow-
precipitation-temperature relationship is independent of
land surface change, such as land use/land cover change,
which could change the characteristics of the hydrological
cycle, including infiltration, evapotranspiration, and runoff
generation processes. When it is used to assess the climatic
effects on streamflow, it again assumes that the current
streamflow-precipitation-temperature relationship is still
valid under the future climate change scenarios.
[54] Second, this model only applies to the annual stream-
flow response to climate change. It cannot be used to
analyze response to the sub-annual (e.g., monthly to sea-
sonal) time step. As both basins have large catchment areas
and mountain regions, the monthly streamflow-precipitation
relationship is not highly correlated due to the greater than
one month runoff routing time (runoff lag), snow accumu-
lation and snowmelt processes, and groundwater-surface
water interaction.
[55] Third, the effect of climate change on streamflow
depends on regional precipitation and temperature change.
The current GCM models have high confidence about future
temperature change scenarios, but have very low confidence
in projections for precipitation, especially at regional or Figure 12. Climate elasticity of streamflow as a function
basin scales because different models produce different of precipitation change produced from different interpola-
regional results [Houghton et al., 2001]. The uncertainty tion techniques (a: Yellow River; b: Spokane River) (Dis-
in future projected precipitation produces uncertainty in Disjunctive Kriging; Dis2- Disjunctive Kriging with first
future streamflow for the studied basins. order removal; Global- global polynomial; idw- Inverse
[56] Fourthly, the accuracy of the interpolation of stream- Distance Weighted; Local- local polynomial; Ord- Ordinary
flow in the precipitation-temperature plane depends on the Kriging; Ord2- Ordinary Kriging with fist order trend
available sample points. We have high confidence for the removal; Rbf- radial basis functions; Sim- Simple Kriging;
precipitation-temperature scenarios where there are many Uni- Universal Kriging with constant order of trend; and
sample data and lower confidence for the precipitation- Uni2- Universal Kriging with first order of trend).
11 of 12
W11419 FU ET AL.: CLIMATE CHANGE EFFECTS ON STREAMFLOW W11419

[58 ] The ArcGIS Geostatistical Analyst is a useful Langbein, W. B. (1949), Annual Runoff in the United State, U.S. Geological
Survey Circular 5. U.S. Dept. of Interior, Washington D. C.
tool to quantify potential impacts of climate change on Leavesley, G. H. (1994), Modeling the effects of climate change on water
regional hydrological regimes, as it provides a streamflow- resources - A review, Clim. Change, 28, 159 – 177.
precipitation-temperature relationship based on observed Li, D., X. Jiang, Y. Wang, and H. Li (2001), Analysis of calculation of
data. Compared to an earlier method [Risbey and Entekhabi, natural tunoff in the Yellow River, People’s Yellow River, 23(2), 35 – 37,
1996] using the adjustable tension continuous curvature (in Chinese).
Limbrunner, J. F. (1998), Climatic elasticity of streamflow in the United
surface gridding algorithm, it is easily applied to different States, M.S. thesis, 39 pp., Tufts Univ., Medford, Mass.
basins. Users can compare different interpolation results and Liu, C., and J. Xia (2004), Water problems and hydrological research in the
select an ‘‘optimal’’ interpolation technique for a specified Yellow River and the Huai and Hai basins of China, Hydrol. Processes,
basin. This work will be extended to assess the generality of 18, 2197 – 2210.
McCarthy, J. J., O. F. Canziani, N. A. Leary, D. J. Dokken, and K. S. White
the findings in future research. (2001), Climate Change 2001: Working Group II: Impacts, Adaptation
and Vulnerability, Contribution of Working Group II to the Third Assess-
[59] Acknowledgments. We wish to thank Stewart W. Franks, Nick ment Report of the Intergovernmental Panel on Climate Change, Cam-
J. Potter, Richard Silberstein, three anonymous reviewers, and Dennis bridge Univ. Press, Cambridge, UK.
Lettenmaier (Associate Editor) for their invaluable comments and construc- Mirza, M. Q. (1997), The runoff sensitivity of the Ganges River basin to
tive suggestions used to improve the quality of the manuscript. This climate change and its implications, J. Environ. Hydrol., 5, 1 – 13.
research was part funded by the Australian Greenhouse Office Climate Mu, X., J. Li, F. Wang, and X. Xu (2003), Analysis of the annual natural
Change Science Program, the Indian Ocean Climate Initiative, and Out- runoff variety process of the Yellow River, J. Arid Resour. Environ.,
standing Overseas Scholar Fund Program of the Institute of Geographical 23(2), 1 – 5, (in Chinese).
Sciences and Natural Resources Research of the Chinese Academy of Nash, L. L., and P. H. Gleick (1991), Sensitivity of streamflow in the
Sciences.
Colorado basin to climatic changes, J. Hydrol., 125, 221 – 241.
Niemann, J. D., and E. A. B. Eltahir (2005), Sensitivity of regional hydrol-
References ogy to climate changes, with application to the Illinois River basin, Water
American Rivers (2005), http://www.waterplanet.ws/ endangeredriver/#6. Resour. Res., 41, W07014, doi:10.1029/2004WR003893.
Arnell, N. (2002), Hydrology and Global Enviromental Change, Prentice Novotny, E. V., and H. G. Stefan (2007), Stream flow in Minnesota: In-
Hall. dicator of climate change, J. Hydrol., 334, 319 – 333.
Arora, V. K. (2002), The use of the aridity to assess climate change effects Potter, K. W. (1981), Illustration of a new test for detecting a shift in mean
on annual runoff, J Hydrol., 265, 164 – 177. in precipitation series, Mon. Weather, 109, 2040 – 2045.
Chen, J., D. He, and S. Cui (2003), The response of river water quality and Revelle, R. R., and P. E. Waggoner (1983), Effects of a carbon dioxide-
quantity to the development of irrigated agriculture in the last 4 decades induced climatic change on water supplies in the Western United States,
in the Yellow River basin, China, Water Resour. Res., 39(3), 1047, in Changing climate: Report of the carbon dioxide assessment commit-
doi:10.1029/2001WR001234. tee, National Academy Press, Washington D. C.
Chiew, F. H. S. (2006), Estimation of rainfall elasticity of streamflow in Risbey, J. S., and D. Entekhabi (1996), Observed Sacramento Basin stream-
Australia, Hydrol. Sci. J., 51, 613 – 625. flow response to precipitation and temperature changes and its relevance
Dooge, J. C. I. (1992), Sensitivity of runoff to climate change: A Hortonian to climate impacts studies, J. Hydrol., 184, 209 – 223.
approach, Bull. Am. Meteorol., 73, 2013 – 2024. Sankarasubramanian, A., and R. M. Vogel (2003), Hydroclimatology of the
Dooge, J. C. I., M. Bruen, and B. Parmentier (1999), A simple model for continental United States, Geophys. Res. Lett., 30(7), 1363, doi:10.1029/
estimating the sensitivity of runoff to long-term changes in precipitation 2002GL015937.
without a change in vegetation, Adv. Water Resour., 23, 153 – 163. Sankarasubramanian, A., R. M. Vogel, and J. F. Limbrunner (2001), Cli-
Fu, G., and S. Chen (2005), ArcGIS Geostatistical Analysis of Observed mate elasticity of streamflow in the United States, Water Resour. Res., 37,
streamflow and its response to precipitation and temperature changes in 1771 – 1781.
the Yellow River, in Regional Hydrological Impacts of Climatic Change Schaake, J. C. (1990), From climate to flow, in Climate Change and U.S.
- Hydroclimatic Variability, IAHS Publ., 296, 238 – 245. Water Resources, edited by P. E. Waggoner, chap. 8, pp. 177 – 206, John
Fu, G., and S. Chen (2006), Water crisis in the Yellow River: Facts, reasons, Wiley, New York.
impacts, and countermeasures, Water Practice Technol., doi:10.2166/ Smith, W. H. F., and P. Wessel (1990), Gridding with continuous curvature
wpt.2006.028. splines in tensions, Geophysics, 55, 293 – 305.
Fu, G., and C. Liu (1991), Impacts of global warming on regional hydro- Spatial Climate Analysis Service (2004), Oregon State Univ., http://
logical regime - A case study on Wanquanhe basin, ACTA Geogr. Sinica, www.ocs.oregonstate.edu/prism/.
46, 277 – 289, (in Chinese). Stockton, C. W., and W. R. Boggess (1979), Geohydrological implications
Fu, G., S. Chen, C. Liu, and D. Shepard (2004), Hydro-climatic trends of of climate change on water resource development, U.S. Army Coastal
the Yellow River basin for the last 50 years, Clim. Change, 65, 149 – 178. Research Engineering Center, Fort Belvoir, VA.
Fu, G., M. E. Barber, and S. Chen (2007a), The impacts of climate change Vogel, R. M., I. Wilson, and C. Daly (1999), Regional regression models of
on regional hydrological regimes in the Spokane River watershed, annual streamflow for the United States, J. Irrig. Drain. Eng., 125, 148 –
J. Hydrol. Eng., 12, 452 – 461. 157.
Fu, G., S. P. Charles, N. R. Viney, S. Chen, and J. Q. Wu (2007b), Impacts Wang, Y., X. Zhang, L. Wang, and J. Zhou (2002), Analysis on variation of
of climate variability on streamflow in the Yellow River, Hydrol. Pro- naturalized runoff during the 90s of twenty century in the Yellow River
cesses, 21, 3431 – 3439, doi:10.1002/hyp.6574. basin, People’s Yellow River, 24(3), 9 – 11, (in Chinese).
Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, Yang, D., C. Li, H. Hu, Z. Lei, S. Yang, T. Kusuda, T. Koike, and K. Musiake
X. Dai, K. Maskell, and C. A. Johnson (2001), Climate Change 2001: (2004), Analysis of water resources variability in the Yellow River of
The Scientific Basis: Contribution of Working Group I to the Third China during the last half century using historical data, Water Resour.
Assessment Report of the Intergovernmental Panel on Climate Change, Res., 40, W06502, doi:10.1029/2003WR002763.
Cambridge Univ. Press, Cambridge, UK. Yates, D. N., and K. M. Strzepek (1998), Modeling the Nile basin under
IPCC (2007), Summary for policymakers, in Climate Change 2007: The climatic change, J. Hydrol. Eng., 3, 98 – 108.
Physical Science Basis. Contribution of Working Group I to the Fourth YRCC (1997), Data Quality Control and Assessment of Hydrological Data
Assessment Report of the Intergovernmental Panel on Climate Change, in the Yellow River and Estimation of Naturalized Streamflow: 1950 –
edited by S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. 1990, Zhengzhou, Henan Province, China (in Chinese).
Averyt, M. Tignor, and H. L. Miller, Cambridge Univ. Press, Cambridge, Zhang, X., C. Liu, and D. Li (2005), Analysis on Yellow River surface
U.K. and New York. runoff consumption, Acta Geogr. Sinica, 60, 79 – 86, (in Chinese).
Johnston, K., J. M. Ver Hoef, K. Krivoruchko, and N. Lucas (2001), Using
ArcGIS Geostatistical Analyst, ESRI Press, Redlands, CA, USA. 

Kuhnel, V., J. C. I. Dooge, J. P. J. O’Kane, and R. J. Romanowicz (1991), S. P. Charles and G. Fu, CSIRO Land and Water, Private Bag 5,
Partial analysis applied to scale problems in surface moisture fluxes, Wembley, WA 6913, Australia. (guobin.fu@csiro.au)
Surv. Geophys., 12, 221 – 247. F. H. S. Chiew, CSIRO Land and Water, GPO Box 1666, Canberra, ACT
2601, Australia.

12 of 12

You might also like