ABSTRACT

1
LDAS Long-Term Retrospective Land Surface Data
Set 1950-2000 Edwin P. Maurer, Andrew W. Wood,
and Dennis P. Lettenmaier aDepartment of Civil
and Environmental Engineering, Box 352700,
University of Washington, Seattle, WA 98195
2
Water Budget Evaluation Over Mississippi River
Basin
ABSTRACT The objective of the Land Data
Assimilation System (LDAS) project is to improve
the initialization of land surface state
variables for numerical weather prediction, and
eventually to provide the basis for assimilation
of land surface state variables, such as soil
moisture and snow water storage, into weather and
climate forecasts. In addition to a real-time
pathway, LDAS has utilized multi-decadal
retrospective studies to aid in estimation of the
parameters for land surface models needed for
real-time applications. The retrospective
studies are also intended to provide diagnostic
information that can aid in improvement of land
surface process representations. The
retrospective pathway, in addition to its roles
in parameter estimation and model diagnostics,
affords the opportunity to evaluate the
climatology of various surface energy and
moisture fluxes, and their spatial patterns and
variability. We summarize results from
implementation of LDAS-North America
retrospective results using the Variable
Infiltration Capacity (VIC) model, run in an
off-line mode at a 3-hour time step for 10 years
(1988-97) for the Mississippi River basin at 1/8
degree spatial resolution, as well as preliminary
results for a 50-year simulation (1950-2000) over
the entire continental U.S. The model is
constrained by observed precipitation and
temperature, and is calibrated against observed
streamflow at key points throughout the region.
For the Mississippi River basin, comparisons are
made between predicted water balance fluxes,
storages, and soil moisture persistence using
both the 10- and 50-year simulations. Because the
model is driven by observed (or derived from
observed) surface meteorology and radiative
forcings, and is validated with streamflow, the
model is constrained to produce long-term mean
evapotranspiration that closely balances the
precipitation and runoff. The observed and
simulated fluxes, therefore, can in a sense be
considered analogous to a "land surface
reanalysis" which can be used to help diagnose
coupled model results.
Use of the Derived Data Set as a Benchmark in
Coupled Model Diagnosis In contrast to the VIC
model, NCEP/NCAR Reanalysis (Reanalysis-1,
Kalnay, et al., 1996) and NCEP/DOE AMIP II
(Reanalysis-2, Kanamitsu, et al., 2000) include
a simplified land surface representation, with
many parameters set as global constants (e.g.,
wilting point0.12, critical pt0.25,
porosity0.47, canopy coverage70, soil depth2
m). They both include a soil moisture adjustment
(nudging) term, and do not require closure of the
surface water budget. They do not include any
precipitation observations, but are driven by
that simulated by the reanalysis model. The major
difference at the land surface between the two
reanalyses is that Reanalysis-2 corrects some of
the known biases in Reanalysis-1, and replaces
the soil moisture nudging to climatology in
Reanalysis-1 with limits on infiltration based on
observed precipitation.
Use of GCIP Results for Reanalysis-1 and -2 Water
Budget Diagnosis
1
Hydrologic Model Validation
The Hydrologic Model The hydrologic model is the
VIC macroscale land surface model (see Liang et
al., 1994 and http//www.hydro.washington.edu/
for model details), which has been applied
off-line at 1/8 resolution. Forcing variables
are daily precipitation, maximum and minimum
temperatures (from NCDC cooperative observer
stations in the U.S, and from Mexican and
Canadian government sources), wind from NCEP
Reanalysis, and humidity and incoming shortwave
and longwave radiation (derived from temperature
and precipitation using established
relationships). Soil parameters are taken from
the Penn State State Soil Geographic Database
(STATSGO) database for the continental U.S., and
the FAO global soil map for areas outside the
U.S. Land cover is from the University of
Maryland 1-km Global Land Cover product (derived
from AVHRR). VIC has been shown to reproduce
observed streamflow when forced with observed
precipitation, and closes its water budget by
construct. This, and the use of physical
representations of the soil moisture and runoff
producing processes, suggests that other surface
flux and state variables can serve as reliable
surrogates for observations that are not
otherwise available. These derived fluxes can
arguably be used as a benchmark against which to
compare coupled models, and to investigate land
surface responses to climatic forcing.
VIC MODEL VALIDATION The VIC model is calibrated
by adjusting the soil depths, baseflow parameters
and infiltration capacity curve parameter to
reproduce observed streamflow. Runoff from each
1/8 degree grid cell is routed to points with
estimated naturalized flows (or if not available,
measured flows), where the hydrographs are
compared. The VIC model is also validated using
the Illinois soil moisture data set of Hollinger
and Isard (1994), which includes observations
from 1981-1996, using soil moistures in the top 1
meter of the soil column. Although the
measurements are not directly comparable (due to
different the use of 19 observing stations are
compared to the 17 1/8º modeled grid cells that
contain the observation points), good agreement
can be seen of average fluxes, interannual
variation, and persistence characteristics.

• Using as a benchmark observations of
  precipitation and the VIC-simulation results for
  other water balance components, a preliminary
  diagnostic evaluation can be made of the
  Reanalysis-1 and -2 models
• In Reanalysis-1, there is the well-documented
  overprediction of summer precipitation in the
  southeastern Mississippi River basin. Improvement
  is seen in Reanalysis-2, though an overprediction
  still exists.
• There is an overprediction of evapotranspiration
  in all months for both Reanalysis-1 and -2, and
  this occurs throughout the basin.
• Runoff patterns do not match VIC spatially or
  temporally for either Reanalysis-1 or -2.
• The large soil moisture cycle in Reanalysis-1 (5
  times the amplitude of VIC) is largely remedied
  in Reanalysis-2.
• In Reanalysis-1 the magnitude of the soil
  moisture adjustment is comparable (1.6 mm/d) to
  principal forcing term P (2.2 mm/d), showing the
  large non-closure of the surface water budget.
• Snow water equivalent is underpredicted in
  Reanalysis-1, and overpredicted in Reanalysis-2,
  and for both model formulations the melt occurs
  earlier than with VIC.
• See Maurer et al. (2001a 2001b) for a more
  detailed discussion.

Illinois Soil Moisture Comparison
Stream Flow Hydrograph Comparison
4
3
Summary of Archived Data Availability
Extension of Data Set to LDAS-North America Region
Illustration of Soil Moisture Persistence

• A comprehensive data set forcing data and
  hydrologic model output has been archived, with
  the following characteristics
• 1/8ºspatial resolution, for a total of 76,864
  grid cells
• 3-hour time step
• Continuous data from 1 January 1950 - 31 July
  2000 (to be periodically updated as observed
  forcing data becomes available)
• Domain is shown in charts below (Mexico is in
  process)

The land surface flux and state variables are
produced at a 3-hour timestep over North America
(between latitudes 25 and 53) for the period Jan.
1950-Jul. 2000. These are archived in netCDF
format, and are available to the public via ftp.
As the data are processed, they will be available
through the link at http//www.hydro.washington.e
du/ where announcements of the complete data set
availability and updates will also be posted.
As an example of exploratory analysis using this
data set, we show the persistence of soil
moisture anomalies over the domain. These are
based on the full 50 year timeseries at each
grid cell. These patterns are consistent with
those produced by Huang, et al (1996). The array
of water and energy variables presented with this
data set allows investigation of many
correlations between variables that has not
previously been possible at this scale.
Shown below is an example of the seasonal
variation of soil moisture, scaled by its dynamic
range for a 20-year period. The drying of the
soil through the year is especially evident in
the east coast and southwest regions.
Shown to the right is the 20-year soil moisture
range scaled by average annual precipitation.
Regions with high values have a soil column that
stores, and later releases, a greater proportion
of the annual precipitation, and therefore has
greater capability to exhibit persistence.
REFERENCES Hollinger, S.E. and S.A. Isard, A soil
moisture climatology of Illinois, J. Clim. 7,
822-833, 1994. Huang, J., H.M. van Den Dool,and
K.P. Georgakakos, Analysis of Model-calculated
soil moisture over the United States (1931-1993)
and applications to long-range temperature
forecasts, J. Clim 9, 1350-1362, 1996. Kalnay,
E., et al., The NCEP/NCAR 40-year reanalysis
project, Bull. Am. Meteorol. Soc., 77, 437-471,
1996 Kanamitsu, M., W. Ebisuzaki, J. Woolen, J.
Potter, and M. Fiorino, An overview of NCEP/DOE
Reanalysis-2, in Proceedings, Second WCRP
International Conference on Reanalyses, Rep.
WCRP-109, World Meteorol. Org., Geneva, 2000.
Liang, X., D. P. Lettenmaier, E. F. Wood, and S.
J. Burges, A Simple hydrologically Based Model of
Land Surface Water and Energy Fluxes for GCMs,
J.Geophys. Res., 99(D7), 14,415-14,428,
1994. Maurer, E.P., G.M. O'Donnell, D.P.
Lettenmaier, and J.O. Roads, 2001a, Evaluation of
NCEP/NCAR Reanalysis Water and Energy Budgets
using Macroscale Hydrologic Simulations, In Land
Surface Hydrology, Meteorology, and Climate
Observations and Modeling, AGU series in Water
Science and Applications, V. Lakshmi, J.
Albertson, and J. Schaake eds., pp.
137-15 Maurer, E.P., G.M. O'Donnell, D.P.
Lettenmaier, and J.O. Roads, 2001b, Evaluation of
the Land Surface Water Budget in NCEP/NCAR and
NCEP/DOE AMIP-II Reanalyses using an Off-line
Hydrologic Model. (J. Geophys. Res., in press)