Assessment of Climate Change Impacts over Oregon’s Willamette River Basin Using NARCCAP Datasets
Andrew Halmstad (halmstad@cecs.pdx.edu), Mohammad Reza Najafi (najafim@cecs.pdx.edu),
Hamid Moradkhani (hamidm@cecs.pdx.edu )
Department of Civil and Environmental Engineering, Portland State University
Abstract
Observed Data- UW Gridded Precipitation
One important aspect related to the management of water resources under future
climate variation is the occurrence of extreme precipitation events. In order to
prepare for extreme events, namely floods and droughts, it is important to
understand how future climate variability will influence the occurrence of such
events. Recent advancements in regional climate modeling efforts provide
additional resources for investigating the occurrence of these extreme events at
scales that may be useful for regional watershed modeling. This study utilizes data
provided by the North American Regional Climate Change Assessment Program
(NARCCAP) in order to investigate the occurrence of extreme precipitation events.
A comparison between observed historical precipitation events and NARCCAP
modeled historical conditions was performed in order to investigate the reliability of
the regional modeling efforts. Future scenarios provided by NARCCAP, forced
using the A2 SRES scenario, were also investigated in order to capture the
expected variation of these events under future climates.
The UW gridded dataset, described by Maurer et al., (2002), was used in this study as an
estimate of observed precipitation. The observed precipitation in the UW dataset is based on
processed National Oceanic and Atmospheric Administration (NOAA) Cooperative Observer
(Co-op) records (Maurer et al., 2002). The spatial resolution of the UW dataset is 1/8 degree
and the temporal range, used for this study, covered the period 1979-1999.
Study Area
The Willamette River Basin, located in western Oregon, encompasses roughly
12,000 square miles (~30,000 square kilometers) and lies in a temperate marine
climate region. The basin supports a variety of land uses (including agriculture,
timber, recreation and urban) and is home to roughly 2.8 million people . The majority
of precipitation falls during the winter months (approximately 80% from October to
May according to Chang and Jung, 2010) where as the summer months are
predominantly dry.
Figure 1: Location of Willamette River Basin, the
study region, within the state of Oregon
Historical vs Future RCM EV
Historical 1979-1999 vs Future 2038-2069
CRCM
WRFG
Visualizing Extreme Value (EV) Precipitation Events
Historic
Historic
Historic
Figure 3. Location of grid points within the
Willamette River Basin for the three
RCMs. This difference is due to the RCM
output grid.
An initial comparison between the UW observed and RCM modeled extreme precipitation
events over the historical period 1979-1999 reveals the bias in each RCM model. Figure 3,
below, shows the UW observed extreme precipitation event average compared to the results
from each of the RCM models forced using the NCEP reanalysis data. The bottom row of
figure 4 displays the bias between the UW dataset and each RCM.
Future
Future
Future
UW "obs" vs RCM (NCEP forcing): Average of EVs
UW "Observed"
Future with Delta B.C.
Future with Delta B.C.
Future with Delta B.C.
Avg EV Precip [mm/day]
33
35
37
39
41
43
45
47
34
36
38
40
42
44
46
48
31 - 32
Figure 4: Comparison between
UW observed and RCM
modeled historic averages of
extreme value events.
25 Year Return Levels [mm/day]
WRFG
MM5I
CRCM
WRFG
MM5I
CRCM
Historic
(1979-1999)
Avg EV Precip [mm/day]
111 - 115
121 - 125
131 - 135
141 - 145
116 - 120
126 - 130
136 - 140
146 - 150
109 - 110
31-32 32-34 34-36 36-38 38-40 40-42 42-44 44-46 46-48
<65
66 - 70
81 - 85
71 - 75
86 - 90
76 - 80
91 - 95
Future
(2038-2069)
Precip Bias (RCM-UW) [mm/day]
WRFG
-11- -10 -9- -8
-8- -6
-6- -4
-4- -2
-2 - 0
0-2
2-4
4-6
6-8
In this study we fit a Generalized Extreme
Value Distribution (GEV) to the climate
model precipitation outputs at each grid
cell for the historical and future periods.
The parameters of the distribution were
calculated based on the maximum
likelihood estimation. The GEV combines
three types of extreme value distributions,
namely the Gumbel, Frechet and Weibull
families. The shape parameter determines
the type of the distribution. The changes
of this parameter are shown in figure 5.
The positive values of the shape
parameter imply heavy tail distributions
and higher probability of the extreme
occurrences.
Figure 5. EV distribution shape parameter, historic
versus future EV distribution shape parameters.
MM5I
CRCM
The return level “f” is a level that
is expected to be exceeded once
every T years where T is the
return period (25 years). The
return level can be calculated by
setting the cumulative GEV
distribution to the desired nonexceedance probability (1-1/T).
The values of the 25 year return
level are estimated for both
historical and future time periods.
It is important to note that the
uncertainties due to different
RCMs are significant. As well, the
changes of the 25 year return
level vary in different locations.
Figure 7 . Comparison between
historic and future extreme events
with a 25 year return period.
Conclusions
Extreme Value Analysis: Distribution Shape Parameter
Figure 2: Comparison between NCEP and GCM forcing data. Comparison between the NCEP driven runs
and the CCSM, a GCM, driven runs displays the relative differences between each RCM as well as the
influence of the initial forcing conditions. The CRCM model displays noticeably lower average values
compared to both MM5I and WRFG.
Figure 6. Historic versus future average
precipitation EV comparison: This figure
displays the average of extreme value
precipitation events, using GCM forcing
data and all three RCMs, over both the
historic (1979-1999) [row 1] and future
(2038-2069) [row 2] periods. Row 3
displays the future results after the bias
correction.
Extreme Value Analysis: 25 Year Return Levels
NARCCAP Project Models
The NARCCAP program utilizes several regional climate models (RCMs) driven by
five atmosphere-ocean general circulation models (AOGCMs) and is focused on
providing data over the majority of the North American continent. The spatial
resolution of the RCM datasets is ~50km; temporal resolution is generally 3hrs. For
this study three RCMs were selected: CRCM (Canadian Regional Climate Model by
OURANOS/UQAM), WRFG (Weather Research & Forecasting Model by Pacific
Northwest National Lab) and MM5I (MM5-PSU/NCAR Mesoscale Model by Iowa
State University). Initial conditions for these three RCMs was provided by the same
GCM driver, CCSM (Community Climate System Model) as well as a reanalysis
dataset called NCEP (NCEP/DOE AMIP-II Reanalysis). In choosing three RCMs that
were driven by the same GCM forcing data this study investigates the relative
differences between the RCMS alone.
MM5I
In order to provide a more accurate
depiction
of
the
relationship
between
historic
and
future
scenarios, a “delta change” method,
similar to the approach taken by
Mote and Salathe (2010) and
Johnson and Sharma (2011) was
implemented. The delta change
factor was determined using the
ratio between UW observed and
RCM modeled historical conditions.
This factor was then used to adjust
the future scenario results. Figure 6
graphically depicts the relationship
between historical and future model
results for each of the RCMs.
• As expected, the RCM models display noticeable differences even when the same
forcing data/initial conditions are used to drive the models.
• The WRFG and MM5I RCMs consistently display higher precipitation values
compared to the CRCM model.
• Bias correction of RCM outputs has a noticeable impact on resulting data.
• As indicated by the shape parameter of the extreme values distributions, the
occurrence of extreme precipitation events over the Willamette River Basin is likely
to vary. The future variation depends heavily on the RCM.
References
•
•
•
•
•
•
Gao, Y., Vano, J.A., Zhu, C., Lettenmaier, D.P., Evaluating climate change over the Colorado River Basin using regional
climate models, Submitted to Journal of Geophysical Research-Atmosphere Revised 3/25/11, retrieved from
http://www.narccap.ucar.edu/contrib/papers.html
Johnson, F., and A. Sharma (2011), Accounting for interannual variability: A comparison of options for water resources
climate change impact assessments, Water Resour. Res., 47, W04508, doi:10.1029/2010WR009272
Maurer, E.P., Wood, A.W., Adam, J.C., Lettenmaier, D.P., and Nijssen, B., 2002, A Long-Term Hydrologically Based
Dataset of Land Surface Fluxes and States for the Conterminous United States, Journal of Climate, 3237-3251. 15
November 2002
Mearns, L.O., et al., 2007, updated 2011, The North American Regional Climate Change Assessment Program dataset,
National Center for Atmospheric Research Earth System Grid data portal, Boulder, CO. Data downloaded Jan-April
2011
Mearns, L. O., W. J. Gutowski, R. Jones, L.-Y. Leung, S. McGinnis, A. M. B. Nunes, and Y. Qian: A regional climate
change assessment program for North America, EOS, Vol. 90, No. 36, 8 September 2009, pp. 311-312.
Mote, P.W., Salathe, E.P., 2010, Future Climate in the Pacific Northwest, Climate Change, 102:29-50,
DOI:10.1007/s10584-010-9848-z
Acknowledgements: We wish to thank the North American Regional Climate Change Assessment Program (NARCCAP) for providing the data used in this paper. NARCCAP is funded by the National Science Foundation (NSF),
the U.S. Department of Energy (DoE), the National Oceanic and Atmospheric Administration (NOAA), and the U.S. Environmental Protection Agency Office of Research and Development (EPA).