6. ROARING FORK RIVER STREAMFLOW: IMPACTS,
ADAPTATION, & VULNERABILITIES
scenarios selected for analysis in this report (see climate
modeling, chapter 2). This assessment of streamflows builds
on the analysis of snowpack conditions on Aspen Mountain,
This chapter is the result of work added to the project after
which is also included in this report (see snowpack analysis,
the work represented in Chapter 1 through 5 was completed.
chapter 3). The snowpack analysis focused on estimating
The additional work was made possible by support from
winter (October through March) snowpack conditions on
the Environmental Protection Agency and was produced by
Aspen Mountain. The snowmelt runoff analysis was designed
Stratus Consulting and the Aspen Global Change Institute
to provide insight into the annual runoff patterns, such as the
as deliverables under contract to TN & Associates, for
timing of peak flows
contract
WA3-1,
and relative changes
“Climate Impacts
in average monthly
and
Adaptation
Comparison of Water Years
streamflow.
Opportunities
Independence Pass
for Surface Water
Runoff modeling
Resources in the
Created 05/09.21:19 GMT
Colorado
Basin
River
Forecast
Center
results show that,
Roaring
Fork
20
based
on
the
Watershed.” The
To Date: 61% (8.3 / 13.6)
Average
18
simulations
for
first, section (6.2),
Seasonal: 47% (8.3 / 17.6)
1992
16
2001
the
A1B
emission
Melt
rate
-0.4
in/day
models snowmelt
averaged over last 3 days
scenario
natural
runoff in the upper
14
streamflow in the
Roaring Fork River
12
Roaring Fork River
to the confluence
10
in the year 2030
with Woody Creek.
should retain its
The
modeling
8
characteristic pattern
examines
runoff
6
of low winter flow
with the main
4
with late spring to
climate
scenarios
summer peak flow,
2
described
in
although peak flow
Chapter 2; Section
0
10-1 10-31 11-30 12-30 1-29 2-28 3-30 4-29 5-29 6-28 7-28 8-27 9-26
is predicted to shift
6.3 describes uses,
Date
from June to May in
rights, diversions,
2030. Predictions
and
impacts
for the year 2100,
from a historical
under the B1, A1B,
perspective,
and;
FIGURE 6.1: A comparison of water years for 1992 , 2001, and the historical average from 1981 to
2005. (
http://www.
and A1FI emission
Section 6.4 describes
cbrfc.noaa.gov/snow/station/sweplot/sweplot.cgi?IDPC2?avg.1992.2001??0???0?s))
scenarios,
also
the approach and
indicate that peak
results of interviews
flow will shift from June to May. However, increased winter
with stakeholders representing physical appropriators and inflow is predicted due to a mid-winter melt that does not occur
stream users. The stakeholder interviews explored how climate
under current conditions, with a corresponding reduction in
change (in the form of different runoff patterns) could affect
early summer flow due to a depleted snowpack. These results
those uses.
and the methods used to derive them are discussed in more
detail below.
SWE (inches)

6.1 INTRODUCTION

6.2 RUNOFF MODELING FOR THE
UPPER ROARING FORK RIVER

6.2.1 METHODS

We developed and applied a snowmelt runoff model to analyze
how streamflow in the Roaring Fork River at the Woody
Creek confluence might change under the future climate

Stratus Consulting used the Snowmelt Runoff Model
(SRM), developed and maintained by the U.S. Department
of Agriculture’s Agricultural Research Service (Martinec,

 1975; Martinec et al., 1994; http://hydrolab.arsusda.gov/cgibin/srmhome), to estimate runoff volume and timing in the
Roaring Fork River at the Woody Creek confluence under
selected future climate scenarios. The SRM simulates surface
processes, and is specifically designed to assess snow coverage
and snowmelt runoff patterns. The SRM uses a temperatureindex method, which is based on the concept that changes in
air temperature provide an index for snowmelt.

streamflow measured in the Roaring Fork River above Aspen
was similar during that year.

cfs

Model calibration was conducted by first calculating the
2001 water year, monthly average runoff from daily estimates
generated by the SRM, then comparing those to the simulated
1992 natural streamflows. The SRM code simulates runoff
only, and does not estimate streamflow contributions from
groundwater (i.e., base flow). We adjusted the runoff model
Using the SRM, we simulated runoff in the 2001 water year
parameters using the correlation coefficient and root mean
(the 12 month period from October through September) for
square difference between the model output and the data,
calibration purposes. Model inputs were daily temperature,
to determine best fit. We achieved a correlation coefficient
precipitation, and snow covered area data from 2001. The
of 0.98, and a root mean square error value of 87 cubic feet
2001 water year was originally selected for the Aspen snowpack
per second (cfs) (2.5 cubic meters per second [m3s-1]) (Figure
study because it
6.2), compared to
is
representative
an average peak
of average winter
streamflow
of
Projected Runoff in the Roaring Fork River
(October through
approximately 1980
March) snowpack
cfs (56 m3s-1).
conditions on Aspen
1400
Mountain,
and
It should be noted
not because of its
that the predicted
CWCB_1992
1200
annual precipitation
natural streamflows
SRM_2001
or
streamflow
(CWCB,
2006)
1000
characteristics.
include both runoff
Streamflow
in
and base flow.
800
the Roaring Fork
However,
base
River is influenced
flow
is
only
a
small
CORR = 0.98
600
by
upstream
component of the
RMS = 87.5
diversions,
dams,
total annual flow in
400
and
withdrawals,
the Roaring Fork at
and thus may not
the Woody Creek
200
completely reflect
confluence.
We
natural changes in
estimated
base
flow
0
OCT NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP
streamflow due to
to be less than 5%
Month
runoff. We therefore
of peak flow from
used
simulated
graphs of predicted
average
monthly
natural streamflow
streamflow instead of
from the CWCB
FIGURE 6.2:
measured streamflow
2006
modeling
to calibrate the
study. Thus, runoff
snowmelt model. To
predictions could be
calibrate the runoff model for the 2001 water year, modeled
compared directly to predicted natural streamflows without
average monthly natural streamflows in the Roaring Fork
introducing significant error.
River at the Woody Creek confluence were obtained from a
watershed modeling study conducted by the Colorado Water
Once the model was calibrated to the 1992 streamflow
Conservation Board (CWCB, 2006). The modeled natural
data, we simulated runoff using selected climate models and
streamflows are only available for the years 1909 through
emission scenarios for the years 2030 and 2100 by scaling
1996. Snow covered area analysis results, as dictated by the
observed temperature and precipitation records by the changes
Aspen snowpack study, were only available for 2001. We
in the various scenarios. We applied the monthly changes in
temperature and precipitation from the climate scenarios to
therefore selected 1992 as a surrogate calibration water year
each day of the month in the daily data series for 2001.
for 2001 because it was the year that most closely resembled
the annual snow water equivalent accumulation and depletion
The climate scenarios span a range of different estimates of
at the Independence Pass SNOTEL site (Figure 6.1). Also,
84

© 2006 Aspen Global Change Institute

 cfs

greenhouse gas emissions and climate sensitivity, and produce
runoff model predictions can only be made monthly with
a range of potential regional changes in temperature and
confidence, at the same temporal resolution as the calibration
precipitation. We used the same climate scenarios as those
dataset. Based on visual examination of the projected runoff
used in the snowpack analysis. As is summarized in Chapter
output, the shift in peak runoff could be somewhat less than
2, Table 1 of this report, those scenarios include increases in
a month.
temperature in both 2030 and 2100, with a greater increase in
2100. Average precipitation is predicted to decrease in both
Changes in total annual runoff volume in 2030 reflect the
2030 and 2100, with the decrease being greater in 2030 than
climate scenario predictions regarding decreases in annual
in 2100, although there is high variance amongst the climate
precipitation. Under the A1B average and wet scenarios, total
models. For the year 2030, we simulated runoff using the wettest
annual runoff volume in 2030 is predicted to be approximately
and driest climate model predictions for the A1B emissions
0 to 5% less than runoff volume in 2001. Under the dry
scenario, as well as the average of all the climate models. There
scenario, total annual runoff in 2030 is predicted to be
is very little divergence between the emissions scenarios by
approximately 10% less than runoff volume in 2001.
2030, so we consider the A1B scenario to be indicative of the
other emission scenarios. For 2100, we simulated runoff using
In 2100, the seasonal pattern of runoff is predicted to be
the average of all the
dramatically
climate models for
different than the
emissions scenarios
seasonal
pattern
2030 Projected Runoff in the Roaring Fork River
B1, A1B, and A1FI.
observed in 2001
and that predicted
1400
for 2030 (Figure
simulated 2001
6.4). As in 2030,
6 . 2 . 2 R E S U LT S
1200
A1B_3 2030_avg
the timing of peak
A1B_3 2030_WET
The 2030 runoff
runoff is predicted
A1B_3 2030_DRY
1000
modeling predicts
to shift from June
that the general
to May, but the
800
seasonal pattern from
more substantial
the simulated 2001
warming in 2100
600
data is shifted to an
will
result
in
earlier peak runoff,
increased
winter
400
but most significant
flow caused by
snowmelt
doesn’t
mid-winter melt,
200
occur until April as
particularly
in
is historical. (Figure
0
February.
This
OCT NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP
6.3). The historical
mid-winter melt
Month
seasonal pattern of
subsequently
low winter flow with
will
cause
a
a later spring to early
corresponding
summer peak flow
FIGURE 6.3:
reduction in June
is retained under all
flows
because the
for 5.4
climate models for
winter snowpack
2030.
Therefore,
will no longer exist
temperatures used as input to the runoff modeling for 2030
in June. The rebound in total runoff volume in July will be the
are not warm enough to cause mid-winter melting of the
result of summer monsoons predicted by the climate models
snowpack, and the seasonal pattern is dominated by the late
in 2100. Therefore, by 2100, the historic runoff seasonal
spring-early summer melt of the winter snowpack.
pattern will be substantially altered. Not only will the timing
of peak runoff shift, but the pattern of low mid- to late winter
However, in the year 2030, peak runoff is predicted to occur in
flow will no longer be evident.
May rather than June (Figure 6.3). Since runoff is estimated in
the model on a monthly basis, it is not possible to determine
Under all climate scenarios, the total annual runoff volume
how many days or weeks earlier the runoff would occur.
in 2100 is predicted to be approximately 5 to 15% greater
Although the SRM code does predict flows on a daily basis,
than in 2001, and slightly greater than predicted for 2030.
only monthly natural streamflow predictions (CWCB, 2006)
These differences are a result of the monthly patterns of
were available for calibration of the runoff model. Thus, the
change in precipitation predicted for the year 2100 versus