1192 13 MARCH 2020 • VOL 367 ISSUE 6483 sciencemag.org SCIENCE
PHOTO: CHRIS WINSOR/GETTY IMAGESBy Mike Hobbins and Joseph Barsugli
A
s the beating heart of the American
Southwest, the Colorado River (CR)
serves ~40 million people from Denver
to Los Angeles and supports 16 million
jobs in a $1.4 trillion regional economy
( 1 ). Most of the river’s streamflow orig-
inates as snowpack in the mountains of the
Upper Colorado River Basin (UCRB). Flow
is threatened by increasing demands from
a growing population, extended drought,
and climate change. A 2018 workshop ( 2 )
convened experts in climate and hydrologic
modeling, observational analysis, and paleo-
climatology to address the causes of the ob-
served declines in streamflow. All agreed that
rising local temperatures were associated
with drying of the basin, but varied methods
yielded large differences in the magnitude of
this effect. Workshop participants urged the
scientific community to identify sources of
these differences by deciphering the contri-
butions of various processes to hydroclimatic
changes. On page 1252 of this issue, Milly and
Dunne address this challenge ( 3 ).
The CR is a well-studied exemplar of
snowmelt-dominated rivers rising in water-
rich mountains in otherwise water-limited
regions. Water sources for ~10% of the
global population depend on high moun-
tain regions that are especially vulnerable
to climate change ( 4 ). Projected reductions
in CR streamflow related to regional warm-
ing range from <10 to 45% by 2050 ( 5 ).
The higher end of projected losses would
be calamitous. The design of new policies
and adaptive plans requires process un-
derstanding, rigorous synthesis of observa-
tions, and modeling.
The surface water balance and surface en-
ergy balance must be resolved for accurate
streamflow estimation and projection. At cli-
mate scales, these balances are affected by a
tangled web of competing physical processes,
but they intersect in evapotranspiration (ET).
Across the UCRB streamflow-producing
mountains, an uncertainty of just 5 W/m^2
of equivalent ET represents ~4 × 10^9 m^3 per
year, or close to 30% of mean annual basin
streamflow. In energy-limited (water-rich) ar-
eas, evaporative demand (E 0 ) drives ET. Thus,
proper estimation of both ET and E 0 are cru-
cial. Data for observation-based assessments
of long-term hydrology are limited largely
to precipitation and temperature, with tem-
perature serving as a proxy for E 0 and the
complexities of the energy balance. This poor
process representation has hobbled observed
analyses of trends in streamflow and land-
surface aridity in two ways ( 6 , 7 ). A full rep-
resentation of E 0 also requires data on wind
speed, humidity, and solar radiation—varia-
tions in all of which drive long-term E 0 trends
across the UCRB ( 8 ) and globally ( 9 ). Also, E 0
and ET should covary in a complementary
fashion in water-limited areas and in paral-
lel in energy-limited areas ( 8 )—both of which
are represented in the basin’s complex hy-
droclimate. Although these issues may seem
obvious, temperature-based E 0 parameteriza-
tions remain obscured in many analyses.
The relationship between long-term
streamflow and regional warming drivesmuch of the uncertainty in flow projections.
Milly and Dunne call this sensitivity b (per-
cent streamflow change per degree Celsius of
warming). The analogous sensitivity to pre-
cipitation is less in dispute. The sensitivity
approach provides a simple metric for com-
paring different analyses and models ( 5 , 10 );
further, temperature and precipitation are
the only meteorological elements for which
century-long, station-based observations are
widely available in the basin. Yet, the over-
all sensitivity metrics can obscure important
details. Large ensembles of climate-model
analyses indicate that meteorological vari-
ability alone can lead to a large range in
regression-based estimates ( 10 , 11 ), even if
the underlying basin hydrologic processes
remain the same. To counter this, basin sen-
sitivity can be estimated from controlled ex-
periments with uniform warming applied to
a hydrologic model.
Milly and Dunne focused on reduced
snowpack and the resulting changes in the
proportion of surface-reflected sunlight (al-
bedo). They adopted an innovative approach
for measuring the energy budget with re-
mote sensing that short-circuits the explicit
calculation of various feedbacks from albedo
changes. They then used their empirically
calibrated model of radiative balance to de-
rive net surface radiation from snowpack for
use in the Priestley-Taylor formulation of E 0.
This permitted an estimate of radiative ef-
fects of snowpack variability using tempera-
ture and precipitation records from 1920 to
the present and incorporated the estimates
into a surface water balance model. They
calculated an overall temperature sensitivity
(b) of –9.3% °C–1 for the basin streamflow. By
examining the seasonality of precipitation inCooperative Institute for Research in Environmental
Sciences, University of Colorado, Boulder, CO 80309,
USA, and Physical Sciences Division, Earth System
Research Laboratory, National Oceanic and Atmospheric
Administration, Boulder, CO 80305, USA.
Email: [email protected]
CLIMATE CHANGE
Threatening the vigor of the Colorado River
Loss of sunlight-reflecting snow spurs evaporation and ebbs river flow
PERSPECTIVES
Published by AAAS