Ensemble reforecasts of recent warm-season weather:

by user

Category: Documents





Ensemble reforecasts of recent warm-season weather:
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, D24116, doi:10.1029/2007JD009480, 2008
Ensemble reforecasts of recent warm-season weather:
Impacts of a dynamic vegetation parameterization
Adriana Beltrán-Przekurat,1 Curtis H. Marshall,2 and Roger A. Pielke Sr.1
Received 9 October 2007; revised 9 March 2008; accepted 10 October 2008; published 27 December 2008.
[1] The impact of dynamic vegetation on ensemble re-forecasts of recent warm-season
weather over the continental U.S. was assessed using the Regional Atmospheric
Modeling System (RAMS) and a fully coupled dynamic vegetation version of RAMS,
the General Energy and Mass Transfer–RAMS (GEMRAMS). Two 10-member
ensembles were produced for the June-August periods of 2000 and 2001. For each
period, one of the members used the standard RAMS, and the other the GEMRAMS
version. Initial and lateral boundary conditions were provided by a re-forecast produced
with the NCEP Seasonal Forecast Model (SFM). In addition, a pair of ‘‘baseline’’
simulations was produced using the NCEP Reanalysis, the ‘‘perfect’’ global forecast, as
initial and lateral boundary conditions. Precipitation in the regional ensembles was
largely controlled by the driving large-scale forcing. A large precipitation bias exists
over the regional domain in the SFM itself that is amplified in the simulations. For the
time periods and model set-up considered in this work, under an explicitly predictive
model configuration, the use of a more complex parameterization of land-surface
processes with dynamic vegetation added little value to the skill of the seasonal forecast
over the regional domain. This is a consequence of the strong dependence of the
regional model results on the lateral boundary conditions provided by the parent global
model. Even the use of an ensemble of predictions does not remove all of the biases that
are inherent in the parent global model.
Citation: Beltrán-Przekurat, A., C. H. Marshall, and R. A. Pielke Sr. (2008), Ensemble reforecasts of recent warm-season weather:
Impacts of a dynamic vegetation parameterization, J. Geophys. Res., 113, D24116, doi:10.1029/2007JD009480.
1. Introduction
[2] Remotely sensed-derived datasets of Leaf Area Index
(LAI) have been incorporated into numerical simulations and
‘‘hindcasts’’ of seasonal weather [e.g., Lu and Shuttleworth,
2002], general circulation models [e.g., Buermann et al.,
2001] or mesoscale models [e.g., Lacaze et al., 2003]
replacing the default multi-year climatology used in many
global and regional climate modeling systems. Results (e.g.,
near-surface temperature and energy fluxes and precipitation)
were significantly different from simulation results for the
same periods when default vegetation fields were used
instead of the observed LAI data. As discussed in Pielke et
al. [1999], vegetation processes are an integral part of the
coupled climate system. On the seasonal time scale, the
phenology of vegetation related parameters (e.g., LAI)
evolves in response to the two-way fluxes of water, energy,
and carbon between the land surface and the atmosphere.
This phenology feeds back to influence the weather because
Department of Atmospheric and Oceanic Sciences, Cooperative
Institute for Research in Environmental Sciences, University of Colorado,
Boulder, Colorado, USA.
Board on Atmospheric Sciences and Climate, National Academy of
Sciences, Washington, DC, USA.
Copyright 2008 by the American Geophysical Union.
of the inherent control exerted by vegetation on the fluxes.
Thus, when multiyear averages of phenology are applied in
climate or seasonal weather modeling applications, an artificial constraint is imposed on the interactive dynamic climate
[3] In the ‘‘hindcast’’ or ‘‘simulation’’ framework, numerical modeling experiments use past observed LAI and lateral
boundary atmospheric conditions (e.g., the NCEP/NCAR
Reanalysis, Kalnay et al. [1996]; ERA40 Reanalysis,
Uppala et al. [2005]). Vegetation data like remotely-sensed
LAI or fractional vegetation cover are assimilated into the
modeling system as valid for the same time period as the
observed lateral boundary condition data. For example,
Lu and Shuttleworth [2002] studied the impact of incorporating Normalized Difference Vegetation Index (NDVI)derived LAI into a regional climate model (ClimRAMS)
on near-surface seasonal simulations. Matsui et al. [2005]
incorporated NDVI-derived vegetation cover data into the
Pennsylvania State University– National Center for Atmospheric Research Mesoscale Model (MM5) – Oregon State
University (OSU) Land Surface Model to show its influence
on North American monsoon simulations. On the other
hand, in a ‘‘predictive’’ framework, assimilation of vegetation data may be feasible in applications to short time
scales (e.g., mesoscale numerical weather prediction), because the vegetation fields do not change appreciably over
the course of a short-term model integration. The results of
1 of 18
short-term forecasts could potentially be improved by
assimilation of near real-time vegetation data when compared to short-term forecasts that use default, climatological vegetation parameters. For example, Kurkowski et al.
[2003] found an improvement in both 2 m and dewpoint
temperatures when NDVI-derived vegetation fraction was
used to initialize short-term (48 h) ETA forecasts. However,
in a predictive framework with applications to longer time
scales (e.g., for an operational application to the seasonal
weather forecast process), it is not possible to obtain a priori
the observed vegetation data that are valid over the long-term
integration. In other words, when using a regional model for
seasonal weather or regional climate model (RCM) for
forecasting applications, the vegetation parameters, much
like the lateral boundary conditions (LBCs), must be a part
of the predictive framework if one wishes to incorporate
season-specific (as opposed to default climatological) vegetation information.
[4] In the context of regional climate modeling, the outputs of the RCM experiments on the modeling domain
(generally with a relatively small grid-cell size) represent a
dynamical downscaling of the coarse data used as initial and
lateral boundary conditions, from global weather forecasts,
global reanalysis, or global general circulation models [e.g.
Castro et al., 2007a]. Castro et al. [2005] classified
dynamical downscaling into four types. In Type 1, the
RCM is forced by a global numerical prediction model at
short regular time intervals (i.e., every 6 or 12 h) and the
initial conditions are retained. In Type 2 the large-scale
forcing for the RCM is provided by reanalysis and observed
sea surface temperatures (SST), which can be considered as
‘‘perfect’’ lower and lateral boundary conditions. When the
LBCs for a RCM are given by a global prediction model that
contains some prescribed data such as observed SST, then the
dynamical downscaling is referred to as Type 3. In this Type 3
dynamic downscaling, the RCM is used in a long-term
‘‘predictive’’ framework, as in seasonal weather predictions.
On the other hand, in Type 4, the LBCs correspond to a fully
coupled climate modeling system, in which all the components of atmosphere, biosphere, ocean, and ice are interactive
and their behavior predicted within the modeling system.
[5] Prescribed LAI seasonal distribution in Type 2 dynamic downscaling based on climatology can result in
strong atmospheric biases in atmospheric variables and
surface fluxes [Xue et al., 1996; Lu and Shuttleworth,
2002]. A predictive seasonal LAI scheme is needed in Type
3 dynamical downscaling, i.e., in seasonal weather prediction experiments, to account for seasonal and year-to-year
variations in vegetation status.
[6] Recently, dynamic vegetation parameterizations (i.e.,
predictive LAI schemes) have been incorporated into
regional climate modeling systems [Eastman et al., 2001a,
2001b; Lu et al., 2001; Tsvetsinskaya et al., 2001a, 2001b;
Narisma et al., 2003; Beltrán, 2005] and coupled to landsurface models (for example, the Community Land Model,
Kim and Wang [2005]; Gulden et al. [2007]). These schemes
do not include long-term plant competition or changes in
vegetation type. LAI is predicted at a daily or weekly basis
based on temperature, soil moisture, and shortwave radiation values provided by the model. These studies demonstrated that significant feedbacks occur on monthly to
seasonal time scales when vegetation is allowed to evolve
as part of the dynamic modeling system. However, the
experiments using the coupled vegetation-atmosphere
RCM still comprise ‘‘simulations’’ rather than explicit
‘‘predictions’’, because the RCM lateral boundary conditions were provided by reanalysis, such as the NCEP/NCAR
or ERA-40 reanalysis.
[7] The main purpose of this work is to investigate the
utility of using a dynamic vegetation parameterization
within a RCM in a truly explicitly predictive framework
(i.e., type 3 downscaling). Specifically, initial and lateral
boundary conditions for a RCM are provided by a 10member global ensemble reforecast produced with the
NCEP Seasonal Forecast Model (SFM) [Kanamitsu et al.,
2002], which was the operational global dynamical forecast
system in use by the Climate Prediction Center (CPC)
during 2000 –2001. In this paper we address two questions.
First, is there a ‘‘value added’’ by incorporating a dynamic
vegetation scheme to represent a season-specific, dynamically interactive phenology in the context of operational
seasonal forecasts? The second question is related to the
statistical nature of the ensemble: how does the dynamic
vegetation parameterization affect the statistical nature of
the ensemble information?
2. Model Description and Experimental Design
2.1. RAMS and GEMRAMS Descriptions
[8] GEMRAMS, comprised of the Colorado State University version of the Regional Atmospheric Modeling System
4.3 [RAMS; Pielke et al., 1992; Cotton et al., 2003] and the
General Energy and Mass Transport Model [GEMTM; Chen
and Coughenour, 1994, 2004], was the RCM used in this
study. GEMRAMS has been used to study the effects of land
cover and CO2 changes on weather and climate [Eastman et
al., 2001a, 2001b; Narisma et al., 2003; Narisma and
Pitman, 2004; Pitman et al., 2004; Beltrán, 2005]. The fully
coupled GEMRAMS contains several options for the typical
physical parameterizations of atmospheric modeling systems
including radiation, convection, and turbulence.
[9] GEMTM is an ecophysiological process-based modeling system that includes explicit C3 and C4 photosynthesis
pathways to determine the assimilation of carbon for sunlit
and shaded leaves. Assimilated carbon is allocated among
dynamically evolving plant biomass (roots, leaves, stems).
A new total LAI value is estimated daily from the leaf
biomass growth, using a vegetation-prescribed specific leaf
area. GEMTM serves in conjunction with the RAMS soilvegetation-atmosphere transfer scheme LEAF-2 [Walko et
al., 2000] to determine the canopy resistances, and ultimately the fluxes of heat and water from the model landsurface. Meteorological inputs to LEAF-2/GEMTM are
provided by the atmospheric outputs from RAMS such as
solar radiation, temperature, and rainfall.
[10] The models were integrated over a domain covering
the contiguous U.S., using a 120 72 grid at 40 km grid
spacing (Figure 1a). There were 32 vertical levels with a
thickness of 120 m at the surface, stretching to 1 km from
approximately 5.2 km to the domain top at 23 km. The soil
model had 8 soil layers, with the bottom layer at 3.0 m.
[11] Initial and LBC for the simulation experiments were
provided by the 2.5 latitude by 2.5 longitude NCEP global
reanalysis [Kalnay et al., 1996] and by global forecasts
2 of 18
Figure 1. (a) Simulation domain in the regional model integrations. (b) Vegetation distribution in the
simulation domain.
produced by the NCEP SFM (see section 2.2). In all the
experiments, lateral boundary nudging was performed every
6 h. Three lateral grid points were used for lateral nudging.
For internal nudging, a 24 h timescale was applied. This is a
relatively ‘‘weak’’ internal nudging, according to RAMS
Users Model Input Namelist Parameters (http://www.atmet.
com/html/docs/rams/ug44-mod-namelists.pdf). This value
allows RAMS to keep the large-scale variability and at
the same time letting RAMS capture small-scale features
[Castro et al., 2005; Saleeby and Cotton, 2004]. The use of
internal nudging vs. no nudging improved the model
representation of the large-scale flow and spatial distribution and variability of precipitation when the Kain-Fritsch
convective precipitation scheme was used [Castro et al.,
2005]. We performed several tests with no nudging to a
48 h timescale, using a 12 h interval (results not shown).
We decided to use the 24 h timescale value based on our
tests results and those of Castro et al. [2005].
[12] The Mellor and Yamada [1982] parameterization was
used for vertical diffusion and the modified Smagorinsky
[1963] scheme for horizontal diffusion. The lateral boundary conditions were those of Klemp and Wilhelmson [1978].
The short- and longwave radiative fluxes were parameterized by the Chen and Cotton [1987] radiation scheme.
Large-scale precipitation processes were simulated with a
‘‘dump-bucket’’ parameterization scheme [Cotton et al.,
1995]. The convective precipitation parameterization
employed was a modified Kain-Fritsch scheme [Castro et
al., 2002; Kain, 2004] which replaced the standard Kuo
scheme in this RAMS version. The incorporation of the
3 of 18
Kain-Fritsch scheme to RAMS greatly improves the amount
and spatial distribution of precipitation in the simulations
[Castro et al., 2002, 2005].
[13] In this GEMRAMS version, several grid cells of the
simulation domain were reclassified into wheat as the main
winter crop, and corn and soybean as the main summer
crops based on the standard Global Land Cover Characteristics database version 1.2 Global Ecosystems framework
[Olson, 1994] and the 0.5° 0.5° Historical Croplands
Dataset, 1700 – 1992 [Ramankutty and Foley, 1999; see
Figure 1b]. Spatially variable soil types were provided by
the Food and Agriculture Organization (FAO) of the United
Nations [FAO, 1997]. Initial LAI values were estimated
using the 8 km 8 km Global Inventory Modeling and
Mapping Studies Satellite Drift Corrected and NOAA-16
incorporated Normalized Difference Vegetation Index
(GIMMS-NDVI) [Pinzon et al., 2005; Tucker et al.,
2005], available bimonthly from July 1981 to December
2005. The algorithm proposed by Sellers et al. [1996] was
applied on the May and June GIMMS-NDVI data of 2000
and 2001. The average value of May and June LAI for each
year was used to initialize the experiments. Climatological
sea surface temperature from NCEP global 1° 1° grid data
base were used on a daily-basis update [Reynolds and
Smith, 1994]. Soil moisture initial conditions were provided
by the NCEP-Land Data Assimilation System model. Observed daily precipitation from the CPC U.S. Unified
Precipitation [Higgins et al., 1996] was used to evaluate
the performance of the modeling experiments.
2.2. NCEP Seasonal Forecast Model
[14] For the explicitly predictive framework experiments,
initial and lateral boundary conditions were provided by
global ensemble reforecasts produced with the NCEP SFM
[Kanamitsu et al., 2002]. This model was the operational
seasonal prediction model at NCEP during the 2000/2001
time of interest for this paper.
[15] The dynamical core of the SFM is based on updated
physics from the NCEP/Department of Energy (DOE)
reanalysis II. The model resolution is T62 spectral truncation, approximately 1.9° 1.9° grid spacing, with 28 layers
in the vertical sigma coordinate system. The model physics
include a relaxed Arakawa – Schubert convective parameterization [Moorthi and Suarez, 1992], shortwave and longwave radiation schemes by Chou [1992] and Chou and
Suarez [1994], respectively, and a cloud scheme by Slingo
[1987]. The land processes parameterization is based on the
OSU two-layer soil model [Pan and Mahrt, 1987]. The
vegetation type and cover and soil type used by SFM
correspond to the Simple Biosphere model climatology
[Dorman and Sellers, 1989].
[16] Until recently, SFM was used at NCEP to produce
global ensemble seasonal forecasts by staggering initial
conditions every 12 h, at 0000 and 1200 UTC of days 1 –
5 of each month, exactly as the ensembles used in this
paper. The SFM has now been replaced by the Climate
Forecast System [Saha et al., 2006].
2.3. Experimental Design
[17] The experiments covered a three month period, from
June 1 to September 1 for 2000 and 2001. The 2000 and
2001 periods were chosen because they are relatively close
to the present time, and remote sea surface temperature
anomalies (El Niño-Southern Oscillation, for example)
were not particularly strong during these periods, such
that the seasonal weather may have been more strongly
influenced by local land-atmosphere interaction than in
other years. In order to assess the ‘‘value added’’ by the
dynamic vegetation in an explicitly predictive framework, a
first set of ‘‘baseline’’ experiments for the periods of
interest were carried out using the NCEP/NCAR Reanalysis
as initial and lateral atmospheric boundary conditions.
These experiments included two runs for each period.
The difference among them is that one encompassed the
‘‘control’’ scenario, wherein RAMS is integrated without
the dynamic vegetation option (i.e., with the default vegetation phenology); these are called RAMS_NCEP simulations.
The second used the dynamic vegetation option GEMRAMS, in which LAI varies according to simulated temperature, radiation and soil moisture conditions; these are called
GEM_NCEP simulations.
[18] In a second set of experiments, RAMS and GEMRAMS runs are performed in a ‘‘predictive’’ mode, by using
the NCEP SFM as forecast atmospheric initial and lateral
boundary conditions throughout the model integration; the
runs are called RAMS_SFM and GEM_SFM, respectively.
Two (i.e., for 2000 and 2001) 10-member ensemble forecasts were obtained from the SFM. These integrations of the
SFM begin at 0000 UTC on 1 June of each year, with
successive integrations initialized every 12 hours, out to
1200 UTC on 5 June. Each of these 10 forecast runs of the
SFM, with their initial conditions staggered in time, were
used to provide initial and lateral boundary conditions to
RAMS and GEMRAMS. A total of 40 3-months runs are
generated in this second experiment.
[19] These two sets of experiments can be used to address
the ‘‘value added’’ by incorporating a dynamically interactive phenology scheme within a perfect global (SFM)
forecast (i.e., Type 3 dynamic downscaling, GEM_SFM
vs. RAMS_SFM comparisons) and the utility of the
dynamic interactive phenology in a ‘‘simulation’’ framework (i.e., Type 2 dynamic downscaling, GEM_NCEP vs.
RAMS_NCEP comparisons).
3. Results
3.1. Baseline Simulations: RAMS_NCEP
3.1.1. Leaf Area Index
[20] Prognosed LAI from GEMRAMS (Figure 2b) shows
higher spatial variability than the prescribed LAI used in
RAMS_NCEP runs (Figure 2a). Higher values are found on
the northeast (deciduous broadleaf trees), northwest and
south areas (evergreen needleleaf trees). The default-RAMS
LAI for those vegetation types do not present spatial
variation because the algorithm used has a slight dependence of latitude. The center and western portion of the
simulation domain, corresponding to shortgrass and semidesert areas, have the lowest LAI values. The August
domain-averaged LAI is also higher in RAMS_NCEP than
in GEM_NCEP simulations, 4 m2m 2 vs. 3 m2m 2 for
2000 and 2001.
[21] Simulated LAI from satellite estimates from GIMMSNDVI and MODIS are shown in Figures 2c and 2d. MODIS-
4 of 18
Figure 2. Mean August LAI from (a) RAMS_NCEP, (b) GEM_NCEP, (c) ‘‘Observed’’ GIMMS-NDVI
derived, (d) ‘‘Observed’’ MODIS, for 2000 (left) and 2001 (right), for (b) to (d).
derived LAI have a higher spatial variability than LAI
derived from GIMMS-NDVI. Largest differences between
these two LAI satellite estimations appear in the maximum
LAI areas of the northeast and northwest. Crop areas in the
Midwest have LAI values mostly between 2 and 3 m2m 2 in
the MODIS dataset, but mostly between 4 and 5 m2m 2
in GEM_NCEP and GIMMS-NDVI LAI. Prognosed
GEM_NCEP LAI values (Figure 2b) are in closer agreement
with MODIS LAI values and spatial distribution (Figure 2d)
than with GIMMS NDVI-derived LAI values.
5 of 18
[22] Interannual variability in LAI can be slightly seen in
the lower Mississippi basin in the MODIS estimates and
GEM_NCEP LAI, and also in the upper Midwest in the
MODIS estimates. Differences between LAI in August 2000
and 2001 were small in the GIMMS-NDVI LAI estimates.
[23] A strict comparison of the simulated LAI with
satellite estimates is a difficult task. Uncertainties associated
with LAI comparisons include NDVI to LAI conversion
algorithms for the satellite data, and parameters related to
green leaf biomass to LAI conversion in the model results,
NDVI data set used [Buermann et al., 2002], different
spatial scales associated with the different estimations
(i.e., satellite pixel: from 105 to 108 m2; model grid-cell:
109 m2), and different spatial vegetation datasets. More
work needs to be done in the assessment of modeling LAI
at this regional scale, with respect to modeling parameters
and datasets used in the comparisons. The ‘‘ground truth’’
could be provided by the few plot level LAI measurement,
but their comparisons with modeling and/or satellite estimates are also affected by scaling issues as well as experimental settings and methodology used in the measurements
[Scurlock et al., 2001].
3.1.2. Surface Energy Fluxes
[24] August mean latent heat fluxes (LH) in both experiments present similar patterns, with the highest values on
the eastern portion of the domain (Figure 3a). Sensible heat
fluxes (SH) have the opposite spatial pattern, with high
values on the east and low values on the south and west
(Figure 3b). In both cases interannual differences are
noticeable on the east portion of the domain, with higher
LH and lower SH values, respectively, in 2001 than in 2000,
reflecting the simulated precipitation interannual differences
in that region (see Figures 4b and 4c).
[25] No major differences in domain-averaged values are
found in LH and SH between RAMS_NCEP and GEM_
NCEP experiments, but large spatial variability can be seen
in Figures 3c and 3d. Latent heat values from the fully
coupled model GEMRAMS (GEM_NCEP) tends to be
lower than from RAMS (RAMS_NCEP) in the center-north
part of the domain consistently in both years (Figure 3c).
The opposite pattern is found for SH values (Figure 3d). In
the southern part of the domain, LH values are higher in the
GEM_NCEP experiments in 2000, but lower in 2001. SH
differences in that region are larger in 2000 than in 2001. In
general, the location of the largest differences in the fluxes
tended to be related with the location of the largest LAI
(Figure 3e), precipitation (see Figure 4d) and soil moisture
differences (not shown) in both experiments. Although LAI
differences are important in the central part of the domain
(Figure 3e), this area corresponds to the semiarid to arid
western region (see precipitation in Figures 4b and 4c), and
then LH values are water-limited.
3.1.3. Precipitation
[26] Both GEMRAMS and RAMS, using the NCEP/
NCAR Reanalysis, captured the general observed spatial
and temporal precipitation patterns (Figure 4). Only results
for August 2000 and 2001 are shown. Observed precipitation indicates a relatively drier August 2000 compared to
2001, with area averages of 48 mm and 58 mm, respectively.
Modelled precipitation showed areas of maximum precipitation along the east coast and Gulf of Mexico, and
minimum precipitation in the center and west of the domain,
in agreement to observations. Over the western U.S. small
features in simulated precipitation appeared closely related
to topography. The overall simulated values indicate a
relatively dry August 2000, like in the observations,
although the RAMS_NCEP and GEM_NCEP area-average
values were lower, 33 mm and 31 mm respectively. During
the relatively wet August 2001 area-averaged simulated
precipitation from RAMS and GEMRAMS was higher than
the observations, 68 mm and 64 mm, respectively, with
precipitation particularly overestimated over southeast U.S.
Although the simulated area of maximum precipitation shift
towards the Gulf of Mexico in agreement with observations,
a maximum of precipitation greater than 300 mm was
located SE of the Appalachians where observations showed
a relative precipitation minimum. The location of this
simulated maximum might be related with combined effects
of the convection parameterization scheme, a RAMS-overestimated strength of the diurnal cycle [Castro et al., 2007a]
and topography.
[27] Several modeling studies have used RAMS [Liston
and Pielke, 2001; Eastman et al., 2001a; Adegoke et al.,
2003; Marshall et al., 2004; Saleeby and Cotton, 2004;
Castro et al.,. 2005, 2007a, 2007b] and GEMRAMS
[Narisma et al., 2003; Beltrán, 2005], with different experimental set-up for warm season simulations. All of them
showed that simulated precipitation amounts and their
spatial and temporal distribution were satisfactory. Castro
et al. [2005, 2007a, 2007b] also found that precipitation in
RAMS is sensitive to simulation size and location, grid
spacing, and convective parameterization used, consistent
with other studies [e.g., Seth and Giorgi, 1998; Xue et al.,
2001, 2007]. In particular for RAMS, the Kuo scheme (the
original convective scheme in RAMS) underestimates precipitation and although the Kain-Fritsch convective scheme
tends to produce excessive precipitation, gives a better
overall performance.
[28] The use of a more complex parameterization of landsurface processes, including the prognosis of LAI, did not
produce major differences in precipitation (Figure 4d).
Domain average differences GEM_NCEP – RAMS_NCEP
were – 2 mm and – 3 mm for August 2000 and 2001
respectively. The differences were located approximately
in the same areas in both years, with larger values in 2001
than in 2000.
3.2. ‘‘Predictive Framework’’ Simulations:
[29] In the second set of experiments, 10 forecast runs of
the SFM for 2000 and 2001 were used as forecast
atmospheric initial and lateral boundary conditions
throughout the integration of RAMS and GEMRAMS.
These 10-member ensemble simulation experiments allow
us to take into account the uncertainties in the SFM runs
themselves and how they downscale using a RCM.
3.2.1. Leaf Area Index
[30] The general pattern of mean prognosed LAI using
SFM (Figure 5a) is similar to the one using NCEP reanalysis (Figure 2b): high values on the east, north and
northwest, decreasing to the center and south. The differences NCEP vs. SFM expressing the effect of the LBC on
prognosed LAI can be seen in Figure 5b (comparison of
Figure 5a and Figure 2b). The domain-average differences
6 of 18
Figure 3. GEM_NCEP simulated latent (a) and sensible (b) heat fluxes. Difference GEM_NCEP –
RAMS_NCEP latent (c) and sensible (d) heat fluxes. (e) LAI differences between GEM_NCEP and
RAMS_NCEP. For all panels, 2000 (left) and 2001 (right).
7 of 18
Figure 4. Monthly observed (b), RAMS_NCEP (b), GEM_NCEP (c) precipitation, and differences
GEM_NCEP – RAMS_NCEP (d) for August 2000 (left) and 2001 (right).
8 of 18
Figure 5. (a) Prognosed mean August LAI for GEM_SFM experiments; LAI differences between
GEM_NCEP and GEM_SFM (b) and GEM_SFM – RAMS_SFM (c); spread in LAI from GEM_SFM
simulations of 10 member-ensemble (d); 2000 (left) and 2001 (right) for all panels.
9 of 18
Figure 6. (a) August mean latent heat (LH) fluxes GEM_SFM; LH differences between GEM_NCEP
and GEM_SFM (b) and GEM_SFM – RAMS_SFM (c); LH spread of 10 member-ensemble of
GEM_SFM; 2000 (left) and 2001 (right) for all panels.
10 of 18
Figure 7. Same as Figure 6 for sensible heat fluxes.
11 of 18
Figure 8. Differences between GEM_SFM and RAMS_SFM for 2 m temperature (°C) (a), and soil
moisture content (m3m 3) for 2000 (left) and 2001 (right).
were –1.5 m2m 2 for both years. Over the Midwest for
both 2000 and 2001 simulations, LAI is higher when using
NCEP reanalysis as LBC. The opposite pattern is found
west and south of this area.
[31] Figure 5c shows the effect of the different landsurface parameterization, with the quantification of LAI
differences between GEM_SFM and RAMS_SFM. The
RAMS_ experiments have all the same LAI values, i.e.,
they only depend on vegetation type, time of year, and
latitude. This means that Figure 5c is reflecting the effect of
the LBC (i.e., NCEP vs. SFM, Figure 5b) on the baseline
simulations (see Figure 3e). LAI differences are reinforced
over the Midwest and suppressed in the south and west.
[32] The spread of LAI for the 10 member-ensemble, as
measured by the root mean square error (RMSE), presents
the highest values in the center and southern part of the
domain (Figure 5d). Part of that area coincides with the area
of the largest differences between GEM_NCEP and
GEM_SFM (Figure 5b). This can be seen for both years,
2000 and 2001, indicating a preferred area where the
differences in these two LBC are likely to affect the RCM
simulations, at least for LAI.
3.2.2. Surface Energy Fluxes and Temperature
[33] Surface fluxes were affected by the different lateral
boundary conditions and the new land-surface scheme. The
August average LH for 2000 and 2001 simulated by
GEM_SFM are shown in Figure 6a. In both GEM_SFM
and RAMS_SFM (not shown) experiments, highest and
lowest LH values are found on the southeastern, and central
and western part of the domain, respectively (Figure 6a).
Overall, LH values are higher when SFM was used as LBC
compared to NCEP Reanalysis, for both RAMS and GEMRAMS experiments (compare Figures 3a and 6a). In the
case of the GEM_ experiments, the effects of the LBC
(NCEP vs. SFM) can be seen in Figure 6b: LH increased in
the southeast and decreased in the northeast when SFM was
used. Those differences tend to be collocated with the areas
of largest differences in LAI, in the case of GEM_NCEP
and GEM_SFM experiments (see Figures 5b and 7).
[34] The effect of the new land-surface scheme on LH
under an explicitly predictive framework (i.e., when SFM is
used as LBC) is shown in Figure 6c. The spatial pattern of the
differences is very similar to the ‘‘simulation’’ experiments,
when Reanalysis were used as LBC (Figure 3c). Differences
tended to be slightly larger in the _SFM experiments than in
the _NCEP experiments (Figure 3c): domain-averaged
values for 2000 and 2001 are 7 Wm 2 for GEM_SFM –
RAMS_SFM, and 4 Wm 2 and 5 Wm 2 for GEM_
NCEP – RAMS_NCEP, respectively. In both experiments,
the location of the larger LH differences tends to be related to
LAI and soil moisture differences between the experiments.
[35] The highest values of LH spread of the 10-member
ensemble (Figure 6d) comprised a larger area than the one
for LAI (Figure 5d), in particular in the center-southern part
of the domain. In addition, an area related to topography
appeared on the west.
[36] Sensible heat flux (SH) had approximately the opposite pattern than LH (Figure 7a) with the largest values over
the west coast and central U.S. and smallest values over the
southeastern United States. Also, opposite to the LH behavior,
12 of 18
Figure 9. Scatterplots of latent heat differences GEM_SFM- RAMS_SFM, for different vegetation
types for August 2000, as a function of: (a) the relative difference of LAI; (b) the relative difference of
soil moisture content; (c) Idem a for GEM_NCEP – RAMS_NCEP.
overall SH values are higher in GEM_NCEP experiments
than in GEM_SFM ones (Figure 7b). SH differences are
collocated with LAI differences and are of opposite sign
(Figure 5b).
[37] Figure 7c shows the effect of the different landsurface schemes on SH. The differences were mostly
positive, indicating larger SH in GEM_SFM experiments
than in RAMS_SFM ones. The areas with large differences
13 of 18
Figure 10. Simulated August precipitation from RAMS_ SFM (a), and GEM_SFM (b) experiments;
(c) difference GEM_SFM – RAMS_SFM. In all panels 2000 (left) and 2001 (right).
are approximately collocated with the LH ones, but tend to
be bigger and of opposite sign (Figure 6c).
[38] The SH spread of the 10-member ensemble (Figure 7d)
is collocated with the LH spread (Figure 6d), but comprised a
smaller area. The SH spread values were lower values than
the ones for LH (Figure 6d).
[39] Near-surface temperature also shows the effects of a
more complex land-surface scheme on the downscaling of
reforecast (Figure 8a). GEM_SFM 2 m temperatures were
predominantly lower (higher) than RAMS_SFM temperatures in the eastern (western) part of the domain. Largest
positive differences were higher than 2.0°C in north-central
U.S. and collocated with the SH differences (Figure 7c).
Two areas with negative differences, one in the center of the
domain and another one on the west coast, corresponded to
negative and positive SH and LH values, respectively.
[40] Another variable that was affected by the different
land-surface parameterization was soil moisture. Figure 8b
shows the average of the first top soil layers. Lower soil
moisture in GEM_SFM was found related to high and low
values of LH and SH respectively.
[ 41 ] Figure 9a shows the differences in LH from
GEM_SFM and RAMS_SFM (Figure 6c) experiments as
a function of LAI (Figure 5c) and soil moisture content
(Figure 8b) differences for August 2000. A priori, no
relationship can be seen if all the grid points are considered.
If the grid points are then discriminated by vegetation types,
a clearer behavior appears for different values of LAI. For
example, that is the case for grasses (tall and short) and
14 of 18
Figure 11. Spread of the 10 ensemble members from (a) SFM simulations themselves and (b)
GEM_SFM experiments for August 2000 (left) and 2001 (right).
semidesert, with a decreasing in LH differences if LAI
differences increase for mean LAI above a certain value,
and the opposite behavior or no change for mean LAI below
a certain level. Instead, in crops and mixed woodland no
relationship is present. But for the latter vegetation types,
differences in LH are better related to soil moisture differences, i.e., LH differences increase with an increase in soil
moisture content, (Figure 9b), but no clear relationship
appears with either LH or soil moisture for the rest of the
vegetation types. This indicates that complex interactions
exists among the variables affected by the land-surface
scheme and time-varying LAI when GEM and RAMS are
being used to downscale reforecasts. Moreover, similar
relationships can be seen when NCEP reanalysis are used
as LBC (Figure 9c).
3.2.3. Precipitation
[42] The mean simulated precipitation from the ensemble forecasts RAMS_SFM and GEM_SFM showed a large
positive bias with respect to observations over the southeastern U.S. and western mountain areas (Figures 10a,
10b). Simulated domain averages for RAMS_SFM and
GEM_SFM were 156 and 140 mm for 2000, and 130 and
127 mm for 2001, respectively. A large area of precipitation higher than 400 mm was simulated over the
southeastern U.S. for August 2000 and 2001, while
observed precipitation was less than 250 mm in the same
area. In addition, interannual precipitation differences
were hardly noticeable in both experiments compared to
the baseline simulations (Figures 4b, 4c). Small features
appeared over the western mountain areas again showing
a clear overestimation.
[43] Figure 10c shows the GEM_SFM – RAMS_SFM
values, indicating of the effect of land-surface parameterization including a dynamic vegetation scheme. No major
differences in domain-averaged precipitation were found
between the GEM_SFM and RAMS_SFM experiments,
16 and 3 mm for 2000 and 2001. Large spatial variability is found in both years with the largest negative
differences occurring over the western portion of the
domain, but very few grid cells present statistically significant differences (not shown).
[44] Figure 11 shows the spread of August 2000 and 2001
precipitation for the each of the SFM forecasts and the
spread of the GEM_SFM experiments. The areas with the
largest spread of the simulated ensemble members tended to
be collocated with the ones found in the SFM forecasts. A
large spread in the RAMS_SFM and GEM_SFM simulations was also found over the western mountain areas,
increasing and expanding the spread found in the SFM
forecasts in this area. Some areas on the west also coincide
with the largest values of the spread in LH (Figure 6d).
4. Discussion and Conclusions
[45] In this paper we addressed several questions. The
first was related to the value added by incorporating a
dynamic vegetation scheme to represent a season-specific
interactive phenology within a truly explicitly predictive
15 of 18
Figure 12. Precipitation from (a) NCEP reanalysis and (b) NCEP SFM for August 2000 (left) and 2001
framework, i.e., downscaling a ‘‘perfect’’ global (SFM)
forecast as an example of a type 3 downscaling based on
Castro et al. [2005]. A second question is associated with
the value added of using a RCM to dynamically downscale
from a global reanalysis, corresponding to a type 2 application of a RCM [Castro et al., 2005].
[46] Although large differences appeared between RAMS
default LAI and GEMRAMS simulated LAI, not major
differences were found in domain-averaged surface energy
fluxes, although high spatial variability and large local
differences were observed. For some vegetation types, the
differences in LH tended to decreased with an increase in
LAI differences between GEMRAMS and RAMS (Figure 9).
In some other cases, a direct relationship with soil moisture
content is found. This confirms that different LAI parameterizations were able to influence near-surface variables, both
when reanalysis and reforecasts are used. Other variables,
like green vegetation fraction, also impact the surface fluxes
and may also affect precipitation [e.g. Matsui et al., 2005].
Although both can be equally important in describing the
state of the vegetation our focused in LAI is related to the fact
that LAI is the predicted variable in the new coupled
GEMRAMS version.
[47] We found that precipitation in the regional simulations was largely dominated by driving large-scale forcing.
Both GEMRAMS and RAMS, using the NCEP/NCAR
Reanalysis, captured the general observed spatial and temporal precipitation patterns. The domain-averaged biases
with respect to observations ranged between 10 and 17 mm
for August. When GEMRAMS and RAMS were forced
with the SFM reforecasts, an example of type 3 downscaling, the domain-averaged precipitation was approximately
three times greater than the observations. The August
precipitation fields from NCEP Reanalysis and SFM are
shown in Figure 12. Large precipitation biases exist in the
Reanalysis and in SFM themselves that were amplified in
the RAMS and GEMRAMS simulations (Figures 4 and 10).
GEMRAMS, and RAMS forecasts tended to reproduce the
precipitation fields from the large-scale forcing, and did not
improve the large biases found in the SFM forecasts
themselves. In addition, the areas with largest spread in
the forecast tended to coincide with the areas with the
largest precipitation biases (Figure 11).
[48] These results are similar to the ones found by Castro et
al. [2005, 2007b]: the use of a RCM to dynamically downscale from a global reanalysis and/or climate model only adds
value when the global model accurately represents the
observed atmospheric conditions. Then, the value-added of
the RCM is to improve the representation of small-scale
features due to landscape heterogeneity. But, ultimately, the
accuracy of this improvement depends on the lateral boundary conditions provided by the large-scale forcing.
[49] In summary, for the time periods and model set-up
considered in this work, under an explicitly predictive
model configuration, the use of a more complex parameterization of land-surface processes with dynamic vegetation added little value to the skill of the seasonal forecast
over the regional domain since it was dominated by the
16 of 18
larger-scale model results. This conclusion reaffirms other
studies [see Gustafson and Leung, 2007; Castro et al.,
2007b], which demonstrates that the addition of improved
weather and climate processes within a regional domain,
cannot, unfortunately, correct for biases that exist in the
parent larger scale (global) model. Land-surface processes,
like dynamic vegetation effects [e.g., Pielke, 2001], have
important impacts on warm-season weather including precipitation forecasts. Our results suggests that in an operational-style ensemble forecast system for recent warm
seasons, the addition of a time-varying LAI may not
significantly impact the dynamical downscaling of the
operational forecast product.
[50] Acknowledgments. This work was supported by the National
Aeronautics and Space Administration under Grant NAG5-11370. We
thank Dr. Christopher Castro for valuable suggestions and discussions.
The CPC US Unified Precipitation and NCEP Reanalysis data provided by
the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site
at http://www.cdc.noaa.gov/. We thank Dr. Jae Schemm of NCEP/CPC,
who assisted the authors with the production of the global SFM forecasts
used in this study.
Adegoke, J. O., R. A. Pielke Sr., J. Eastman, R. Mahmood, and K. G.
Hubbard (2003), Impact of irrigation on midsummer surface fluxes and
temperature under dry synoptic conditions: A regional atmospheric model
study of the U.S. High Plains, Mon. Weather Rev., 131, 556 – 564.
Beltrán, A. B. (2005), Using a coupled atmospheric-biospheric modeling
system (GEMRAMS) to model the effects of land-use/land-cover
changes on the near-surface atmosphere, Ph.D. dissertation, 186 pp.,
Colorado State Univ., Fort Collins, Colo.
Buermann, W., J. Dong, X. Zeng, R. B. Myneni, and R. E. Dickinson
(2001), Evaluation of the utility of satellite based vegetation leaf area
index data for climate simulations, J. Clim., 14, 3536 – 3550.
Buermann, W., Y. Wang, J. Dong, L. Zhou, X. Zeng, R. E. Dickinson, C. S.
Potter, and R. B. Myneni (2002), Analysis of a multiyear global vegetation leaf area index data set, J. Geophys. Res., 107(D22), 4646,
Castro, C. L., W. Y. Y. Cheng, A. B. Beltrán, R. A. Pielke Sr., and W. R.
Cotton (2002), The incorporation of the Kain-Fritsch cumulus parameterization scheme in RAMS with a terrain-adjusted trigger function,
paper presented at Fifth RAMS users and related applications Workshop, Santorini, Greece. ATMET Inc.
Castro, C. L., R. A. Pielke Sr., and G. Leoncini (2005), Dynamical downscaling: Assessment of value restored and added using the Regional
Atmospheric Modeling System (RAMS), J. Geophys. Res., 110(D5),
D05108, doi:10.1029/2004JD004721.
Castro, C. L., R. A. Pielke Sr., and J. Adegoke (2007a), Investigation of the
summer climate of the contiguous U. S. and Mexico using the Regional
Atmospheric Modeling System (RAMS). Part I: Model climatology
(1950 – 2002), J. Clim., 20, 3844 – 3865.
Castro, C. L., R. A. Pielke Sr., J. Adegoke, S. D. Schubert, and P. J. Pegion
(2007b), Investigation of the summer climate of the contiguous U. S. and
Mexico using the Regional Atmospheric Modeling System (RAMS).
Part II: Model climate variability, J. Clim., 20, 3866 – 3887.
Chen, C., and W. R. Cotton (1987), The physics of the marine stratocumulus-capped mixed layer, J. Atmos. Sci., 44, 2951 – 2977.
Chen, D.-X., and M. B. Coughenour (1994), GEMTM: A general model for
energy and mass transfer of land surfaces and its application at the FIFE
sites, Agric. For. Meteorol., 68, 145 – 171.
Chen, D.-X., and M. B. Coughenour (2004), Photosynthesis, transpiration,
and primary productivity: Scaling up from leaves to canopies and regions
using process models and remotely sensed data, Global Biogeochem.
Cycles, 18, GB4033, doi:10.1029/2002GB001979.
Chou, M.-D. (1992), A solar radiation model for use in climate studies,
J. Atmos. Sci., 49, 762 – 772.
Chou, M.-D., and M. J. Suarez (1994), An efficient thermal infrared
radiation parameterization for use in general circulation models, NASA
Tech. Rep. TM-1994-104606, Series on Global Modeling and Data
Assimilation, 85 pp., NASA, Greenbelt, Md.
Cotton, W. R., J. F. Weaver, and B. A. Beitler (1995), An unusual summertime downslope wind event in Fort Collins, Colorado on 3 July 1993,
Weather Forecasting, 10, 786 – 797.
Cotton, W. R., et al. (2003), RAMS 2001: Current status and future directions, Meteorol. Atmos. Phys., 82, 5 – 29.
Dorman, J. L., and P. J. Sellers (1989), A global climatology of albedo,
roughness and stomatal resistance for atmospheric general circulation
models as represented by the Simple Biosphere Model (SiB), J. Appl.
Meteorol., 28, 833 – 855.
Eastman, J. L., M. B. Coughenour, and R. A. Pielke (2001a), The effects of
CO2 and landscape change using a coupled plant and meteorological
model, Global Change Biol., 7, 797 – 815.
Eastman, J. L., M. B. Coughenour, and R. A. Pielke Sr. (2001b), Does
grazing affect regional climate?, J. Hydrometeorol., 2, 243 – 253.
Food and Agriculture Organization (FAO) (1997), Digital soil map of the
world and derived soil properties, Version 3.5, Land and Water Digital
Media Series 1. [CDROM], FAO, Rome, Italy.
Gulden, L. E., Z.-L. Yang, and G.-Y. Niu (2007), Interannual variation in
biogenic emissions on a regional scale, J. Geophys. Res., 112, D14103,
Gustafson, W. I., Jr., and L. R. Leung (2007), Regional downscaling for air
quality assessment, Bull. Am. Meteorol. Soc., 88, 1215 – 1227,
Higgins, R. W., J. E. Janowiak, and Y.-P. Yao (1996), A Gridded Hourly
Precipitation Data Base for the United States (1963-1993), NCEP/Climate Prediction Center Atlas 1, 46 pp., Natl. Cent. for Environ. Predict.
Kain, J. S. (2004), The Kain-Fritsch convective parameterization: An
update, J. Appl. Meteorol., 43, 170 – 181.
Kalnay, E., et al. (1996), The NCEP/NCAR 40-Year reanalysis project,
Bull. Am. Meteorol. Soc., 77, 437 – 471.
Kanamitsu, M., et al. (2002), NCEP Dynamical seasonal forecast system
2000, Bull. Am. Meteorol. Soc., 83, 1019 – 1037.
Kim, Y., and G. Wang (2005), Modeling seasonal vegetation variation and
its validation against Moderate Resolution Imaging Spectroradiometer
(MODIS) observations over North America, J. Geophys. Res., 110,
D04106, doi:10.1029/2004JD005436.
Klemp, J. B., and R. B. Wilhelmson (1978), The simulation of three-dimensional convective storm dynamics, J. Atmos. Sci., 35, 1070 – 1096.
Kurkowski, N. P., D. J. Stensrud, and M. E. Baldwin (2003), Assessment of
implementing satellite-derived land cover data in the Eta model, Weather
Forecasting, 18, 404 – 416.
Lacaze, R., S. Donier, P. Lacarrere, and J.-L. Roujean (2003), AVHRRderived land surface conditions for flux simulations with a mesoscale
model over the Hapex-Sahel study area, J. Appl. Meteorol., 42, 686 – 700.
Liston, G. E., and R. A. Pielke (2001), A climate version of the Regional
Atmospheric Modeling System, Theor. Appl. Climatol., 68, 155 – 173.
Lu, L., and W. J. Shuttleworth (2002), Incorporating NDVI-Derived LAI
into the climate version of RAMS and its impact on regional climate,
J. Hydrometeorol., 3, 347 – 362.
Lu, L., R. A. Pielke, G. E. Liston, W. J. Parton, D. Ojima, and M. Hartman
(2001), Implementation of a two-way interactive atmospheric and ecological model and its application to the central United States, J. Clim., 14,
900 – 919.
Marshall, C. H., Jr., R. A. Pielke Sr., L. T. Steyaert, and D. A. Willard
(2004), The impact of anthropogenic land-cover change on the Florida
peninsula sea breezes and warm season sensible weather, Mon. Weather
Rev., 132, 28 – 52.
Matsui, T., V. Lakshmi, and E. E. Small (2005), The effects of satellitederived vegetation cover variability on simulated land – atmosphere interactions in the NAMS, J. Clim., 18, 21 – 39.
Mellor, G. L., and T. Yamada (1982), Development of a turbulence closure
model for geophysical fluid problems, Rev. Geophys., 20, 851 – 875.
Moorthi, S., and M. J. Suarez (1992), Relaxed Arakawa – Schubert: A
parameterization of moist convection for general circulation models,
Mon. Weather Rev., 120, 978 – 1002.
Narisma, G. T., and A. J. Pitman (2004), The effect of including biospheric
responses to CO2 on the impact of land-cover change over Australia,
Earth Interact., 8, 1 – 28.
Narisma, G. T., A. J. Pitman, J. Eastman, I. G. Watterson, R. Pielke Sr., and
A. Beltrán-Przekurat (2003), The role of biospheric feedbacks in the simulation of the impact of historical land cover change on the Australian January
climate, Geophys. Res. Lett., 30(22), 2168, doi:10.1029/2003GL018261.
Olson, J. S. (1994), Global ecosystem framework-definitions, USGS EROS
Data Cent. Internal Rep., Sioux Falls, SD, 37 pp.
Pan, H.-L., and L. Mahrt (1987), Interaction between soil hydrology and
boundary layer developments, Boundary Layer Meteorol., 38, 185 – 202.
Pielke, R. A., Sr. (2001), Influence of the spatial distribution of vegetation
and soils on the prediction of cumulus convective rainfall, Rev. Geophys.,
39, 151 – 177.
Pielke, R. A., Sr., et al. (1992), A comprehensive meteorological modeling
system – RAMS, Meteorol. Atmos. Phys., 49, 69 – 91.
17 of 18
Pielke, R. A., G. E. Liston, J. L. Eastman, L. Lu, and M. Coughenour
(1999), Seasonal weather prediction as an initial value problem, J. Geophys. Res., 104, 19,463 – 19,479.
Pinzon, J. M., M. E. Brown, and C. J. Tucker (2005), EMD Correction of
Orbital Drift Artifacts in Satellite Data Stream, in Hilbert-Huang Transform and Its Applications, edited by N. E. Huang and S. S. P. Chen, chap.
8, pp. 167 – 186, World Sci., Hackensack, New Jersey.
Pitman, A. J., G. T. Narisma, R. A. Pielke Sr., and N. J. Holbrook (2004),
The impact of land cover change on the climate of southwest Western
Australia, J. Geophys. Res., 109, D18109, doi:10.1029/2003JD004347.
Ramankutty, N., and J. A. Foley (1999), Estimating historical changes in
global land cover: Croplands from 1700 to 1992, Global Biogeochem.
Cycles, 13, 997 – 1027.
Reynolds, R. W., and T. M. Smith (1994), Improved global sea surface
temperature analyses using optimum interpolation, J. Clim., 7, 929 –
Saha, S., et al. (2006), The NCEP Climate Forecast System, J. Clim., 19,
3483 – 3517.
Saleeby, S. M., and W. R. Cotton (2004), Simulations of the North American Monsoon System. Part I: Model Analysis of the 1993 Monsoon
Season, J. Clim., 17, 1997 – 2018.
Scurlock, J. M. O., G. P. Asner, and S. T. Gower (2001), Global Leaf Area
Index Data from field measurements, 1932 – 2000, Oak Ridge Natl. Lab.,
Oak Ridge, Tenn. (Available online at: http://daac.ornl.gov/daacpages/
Sellers, P. J., S. O. Los, C. J. Tucker, C. O. Justice, D. A. Dazlich, G. J.
Collatz, and D. A. Randall (1996), A revised land surface parameterization (SiB2) for atmospheric GCMs. Part II: the generation of global fields
of terrestrial biophysical parameters from satellite data, J. Clim., 9, 706 –
Seth, A., and F. Giorgi (1998), The effects of domain choice on summer
precipitation simulation and sensitivity in a regional climate model,
J. Clim., 11, 2698 – 2712.
Slingo, J. M. (1987), The development and verification of a cloud prediction model for the ECMWF model, Q. J. R. Meteorol. Soc., 113, 899 –
Smagorinsky, J. (1963), General circulation experiments with the primitive
equations. Part I: The basic experiment, Mon. Weather Rev., 91, 99 – 164.
Tsvetsinskaya, E., L. O. Mearns, and W. Easterling (2001a), Investigating
the effect of seasonal plant growth and development in 3-dimensional
atmospheric simulations. part I: Simulation of surface fluxes over the
growing season, J. Clim., 14, 692 – 709.
Tsvetsinskaya, E., L. O. Mearns, and W. Easterling (2001b), Investigating
the effect of seasonal plant growth and development in 3-dimensional
atmospheric simulations. Part II: Atmospheric response to crop growth
and development, J. Clim., 14, 711 – 729.
Tucker, C. J., J. E. Pinzon, M. E. Brown, D. Slayback, E. W. W. Pak,
R. Mahoney, E. Vermote, and N. El Saleous (2005), An extended AVHRR
8-km NDVI data set compatible with MODIS and SPOT vegetation NDVI
data, Int. J. Remote Sens., 26, 4485 – 4498.
Uppala, S. M., et al. (2005), The ERA-40 re-analysis, Q. J. R. Meteorol.
Soc., 131, 2961 – 3012.
Walko, R. L., et al. (2000), Coupled atmosphere-biophysics-hydrology
models for environmental modeling, J. Appl. Meteorol., 39, 931 – 944.
Xue, Y., M. J. Fennessy, and P. J. Sellers (1996), Impact of vegetation
properties on U.S. summer weather prediction, J. Geophys. Res., 101,
7419 – 7430.
Xue, Y., F. J. Zeng, K. E. Mitchell, Z. Janjic, and E. Rogers (2001), The
impact of land surface processes on simulations of the U. S. hydrological
cycle: A case study of the 1993 flood using the SSiB land surface model
in the NCEP Eta regional model, Mon. Weather Rev., 129, 2833 – 2860.
Xue, Y., R. Vasic, Z. Janjic, F. Mesinger, and K. E. Mitchell (2007),
Assessment of Dynamic Downscaling of the Continental U. S. Regional
Climate Using the Eta/SSiB Regional Climate Model, J. Clim., 20,
4172 – 4193.
A. Beltrán-Przekurat and R. A. Pielke Sr., Department of Atmospheric
and Oceanic Sciences, Cooperative Institute for Research in Environmental Sciences, University of Colorado, UCB 311, Boulder, CO 80523
1371, USA. ([email protected])
C. H. Marshall, Board on Atmospheric Sciences and Climate, National
Academy of Sciences, 500 Fifth Street, NW, Room 634, Washington, DC
20001, USA.
18 of 18
Fly UP