A 85-year Retrospective Hydrologic Analysis for the Western US
Nathalie Voisin, Hyo-Seok Park, Alan F. Hamlet, Andrew W. Wood, Ned Guttman# and Dennis P. Lettenmaier
#National Climatic Data Center
Department of Civil and Environmental Engineering
University of Washington
Boulder, CO
1
3
Introduction
H22B-0913
Validation of VIC Water Balance
A frequently encountered difficulty in assessing model-predicted land–atmosphere exchanges of moisture and energy is the absence
of comprehensive observations to which model predictions can be compared at the spatial and temporal resolutions at which the
models operate. Various methods have been used to evaluate the land surface schemes in coupled models, including:
• comparison of model-predicted energy and radiative fluxes with tower measurements during periods of intensive observations
• comparison of model-predicted evapotranspiration with values derived from atmospheric balances over large river basins,
• comparison of (routed) model-predicted runoff with observed streamflow,
• comparison of model predictions of soil moisture with spatial averages of point observations.
While these approaches have provided useful model diagnostic information, the observation-based products used in the comparisons
typically are inconsistent with the model variables with which they are compared, due to differences in spatial and temporal resolution.
In a previous study (Maurer et al. 2002), a model-derived data set of land surface states and fluxes was derived for the conterminous
United States and portions of Canada and Mexico. The data set spans the period 1950–2000, and is at a 3-h time step with a spatial
resolution of 1/8 degree. These data allow the evaluation of the interaction of the water balance components over large regions for
long periods.
The National Climatic Data Center (NCDC) has recently created digital archives of daily climatological data for the continental U.S.
going back to the beginning of the period of instrumental records. We describe an extension of the Maurer et al data back to 1916 and
forward to 2003, a period for which adequate station density exists to perform hydrologic simulations with the Variable Infiltration
Capacity (VIC) model. A particular advantage of this 85-year period is that it includes the droughts of 1930s, which facilitates
comparative evaluations with more recent events.
Columbia River at the Dalles, OR
The preliminary VIC simulations for the Shasta reservoir, CA, the Dalles, OR and
Hoover Dam, NV show a fair agreement with observations. There is a noticeable
peak underestimation and phase lag at the Dalles, OR but the interannual
variability seems to be well preserved (see e.g. circled drought periods). The
selected drought periods of the 1930’s and 1970’s are analyzed below so as to
assess the severity of recent western U.S. drought in a centennial context.
2 Pre-processing of surface data input
Sacramento River at Shasta reservoir, CA
Colorado River at Hoover Dam, NV
The National Climatic Data Center (NCDC) has recently created digital archives of daily climatological data (primarily precipitation
and daily temperature maxima and minima) for the continental U.S. going back to the beginning of the period of instrumental
records. Previous electronic archives were typically available only back to about 1948, with a few stations digitized back to the
1930's. Using the newly available data merged with the previous archive (1948-2000), we have created a 1/8 degree data set of
precipitation, temperature, and derived radiative forcings and other surface variables needed to drive the Variable Infiltration
Capacity (VIC) macroscale hydrology model over the western U.S. (soon to be extended to the entire conterminous U.S.).
The first step though consists of developing a method, based an carefully quality controlled HCN (Historical Climatology Network;
see Easterling et al, “United States Historical Climatology Network (U.S. HCN) Monthly Temperature and Precipitation Data”,
ORNL/CDIAC-87, NDP-019/R3. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge,
Tennessee) stations to control for drift in the gridded data that otherwise results from changes in the stations included over time. We
proceeded as follows:
Schematic Diagram for Data Processing of VIC Meteorological Driving Data
PRISM Monthly
HCN/HCCD
Precipitation
Monthly Data
Maps
Preprocessing
Regridding
Lapse Temperatures
Correction to
Remove
Temporal
Inhomogeneities
Coop Daily Data
Topographic
Correction for
Precipitation
• Using the station meta data for indexing, the data were gridded to a
uniform 1/8 degree grid over the study domain.
• Precipitation data were then rescaled for each month and each grid
location by comparing the long term mean of the raw data from 1961-1990
to the PRISM (Daly et al. , 1984, J. of Applied Met. (33) pp 140-158)
means for the same location and time period.
•The temporal variability of the precipitation data was directly derived from
the station data, but the spatial distribution was forced to match the PRISM
results in the long-term mean for each month at each location. Earlier
records were corrected by the same scaling fraction, even though the
exact group of stations that define the raw precipitation values are not
guaranteed to be the same in different periods. No attempt was made in
this preliminary analysis to remove temporal inhomogeneities in the
precipitation time series.
(http://www.wrcc.dri.edu/pcpn/westus_precip.gif)
4
Hydrologic simulations – severity of ongoing extreme events relative to the past
1929 Drought
1930 Drought
1931 Drought
1976 Drought
( in California )
1977 Drought
2002 Drought
April 1st Snow
Anomaly
( 10 mm
threshold)
1915-1997
climatology
April 1st Snow
( > 50 mm)
(a)
August 1st
Soil Moisture
Anomaly,
3 layers
(mm/day)
August 1st Soil
Moisture
( > 0 mm/day )
(b)
• Wind data were gathered from NCAR reanalysis, and the daily
climatological mean value for each grid cell was used to define the daily
mean wind speed. This approach was used because reanalysis data are
not available prior to 1948.
June
Runoff
Anomaly
(mm/day)
June
Runoff
( > 0 mm/day )
(c)
Hydrologic Model
The Variable Infiltration Capacity (VIC) hydrologic model
(schematic diagram below) has been implemented at 1/8 degree
spatial resolution over the study domain. Each grid cell is about
12 km by 12 km. Runs were made in water balance mode using
a time step of 24 hours and a snow model time step of 1 hour.
(a)
Black values are under -300 mm snow anomalies,
(b)
Black values are under -200 mm/day anomalies ,
(c)
Black values are under -1mm/day anomalies
5 Conclusion
1/8 Degree VIC Hydrologic Model and Simulated Channel Network
The derived data (soil moisture, snow water content, runoff) are distinct from reanalysis products in that precipitation is a gridded product derived directly
from observations, and both the land surface water and energy budgets balance at every time step. Simulated runoff match observations fairly well, and
inter annual variability is well preserved over large river basins. On this basis, and given the physically based model parameterizations, we argue on the
basis of closure that other terms in the surface water balance (e.g., soil moisture and evapotranspiration) must be reasonably well represented, at least
for the purposes of diagnostic studies such as those in which atmospheric model reanalysis products have been widely used. These characteristics make
this dataset useful for a variety of studies, especially where ground observations are lacking.