2.2.1.1 Evidence of and Mechanisms for Seasonal Predictability
GCIP provides an excellent rationale and data source for
investigating the
hypothesis, in the context of North America, that (globally
determined) soil moisture anomalies at the beginning of the warm
season influence the regional precipitation in the subsequent
months. Atmospheric general circulation model runs with improved
representation of interactive moist processes along with
diagnostic studies are needed to test this hypothesis and
determine the conditions and limitations of its applicability.
These would involve comprehensive analyses to explore the lagged
correlation, both locally and perhaps downwind, between all the
relevant data (on rainfall, evaporation, temperature, clouds,
radiation, vegetative state, etc.) now available within GCIP. In
addition, it is now time to undertake experimental, free running
seasonal to interannual simulations with coupled models of the
land-atmosphere-ocean system to give global and regional
forecasts. Realistically, early expectations of the skill of such
forecasts should be limited to capturing modest indications of
regional-scale monthly or seasonal anomalies of precipitation and
temperature.
2.2.1.2 Coupling importance in the annual cycle
The strength and influence of the hydrologic-atmospheric
coupling varies between cool and warm seasons, which leads to
seasonal differences in the importance of land-atmosphere
coupling relative to other regional-scale and global-scale
influences. An explicit understanding of the seasonal variation
of relative coupling strength is necessary to define the relative
prediction. Land-atmosphere coupling processes which are
important in seasons when local controls are more important
likely need more precise representation than those which are
important in seasons when global-scale influences dominate.
The required studies will involve a combination of
measurement and modeling activities. Observations would likely
include atmospheric profiles of moisture, temperature, and wind
during both warm and cool seasons and during the transition from
cold to warm season, together with simultaneous measurements of
the surface fluxes of water and energy. Modeling studies could
include sensitivity studies using validated coupled models
applied in different seasons and at different spatial scales.
2.2.1.3 Significance of diurnal variations in surface energy
fluxes
Based on the results from a coupled land-atmosphere model,
Koster and Suarez (1995) suggested
that large scale circulation is
affected by short-term variability in the surface energy balance.
Hence a land surface scheme that realistically reproduces the mean
diurnal cycle of the surface energy balance may nonetheless be
inadequate for coupled modeling purposes. The scheme might also
need to reproduce the short-term variations in the balance of
energy.
The extent to which short-term variations in surface energy
balance require representation in predictive models when applied at
seasonal-to-interannual time scales merits more detailed
investigation. Modeling experiments are required to explore this
limit on the complexity of the representation of hydrologic-atmospheric processes.
2.2.2 Hydrological interpretation of meteorological predictions
The nature of the meteorological predictions calculated by
global-scale models of the ocean-atmosphere-land system is likely
to be profoundly different from actual meteorological observations
in terms of their spatial and temporal precision and accuracy, even
when those predictions have been down-scaled through mesoscale,
regional models. Existing hydrological models are designed to work
from observations, and their form and function reflect the nature
of these observations. Research is required to determine what type
of hydrological prediction is possible from seasonal-to-interannual
meteorological predictions and at what spatial and temporal scales
hydrological interpretation can have worthwhile credibility and
utility. Handling uncertainty in meteorological predictions is not
a resolved issue in hydrological models, even for short-term
forecasts, and reservoir management practice will always need to be
incorporated into the hydrological interpretation for North
American water resource issues.
There is opportunity to improve communication between
atmospheric scientists and hydrologists on this issue, because
neither of these two groups have hitherto had opportunity to fully
appreciate the relevant capabilities of the other. Hydrologists do
not yet appreciate what the nature and form of seasonal-to-interannual
meteorological predictions might be, and there is some
lack of clarity on this issue. Equally, meteorologists do not yet
have an appreciation of what type of seasonal-to-interannual
prediction might have practical value to hydrologists. At this
time, therefore, the need is to provide better definition of these
issues in order to establish a means of interaction between the two
communities.
2.2.2.1 Evaluation of seasonal-to-interannual predictions
As noted above, not only is GCIP in a strong position to
foster experimental seasonal-to-interannual forecasts focused on
the North American continent, it is also uniquely able to provide
effective validation of such forecasts by virtue of the existing
and new data that are being collected for the U.S. in general, and
for the Mississippi River basin in particular. However, some
redefinition of GCIP data products will be required. Specifically,
once the form, nature, and spatial and temporal scale of
seasonal-to-interannual prediction products are defined, it will be
necessary to synthesize equivalent observational products from
GCIP's precipitation and temperature measuring networks. Future
westerly extension of the GCIP study area also seems essential if
there is to be a better match between areas in the U.S., where
seasonal-to-interannual prediction is most feasible, and areas in
which data collection within GCIP has priority. Arguably, the
single-most challenging technical problem will be providing a
credible regional measurement of cold-season precipitation for the
purposes of comparison with seasonal-to-interannual predictions.
2.2.2.2 Definition of the predictive products required by
hydrologists
Resource managers within the hydrological community might be
able to make use of a range of predicted outputs from coupled
land-atmosphere-ocean models, but hitherto they have tended to rely on
traditional meteorological and hydrological measurements applied to
conventional hydrological models for streamflow predictions.
Although hydrologists have a good capability for using statistical
forecast information, so far the coupled modeling community has not
given priority to providing this type of information. However,
research into the possible hydrological interpretation of these
predictions cannot begin until the nature and form of such
predictions are better defined.
There is a need to develop better understanding of the
requirements of the hydrological community so that any predictive
meteorological products provided at the seasonal-to-interannual
time scale can be tailored more precisely and the opportunities for
timely application of GCIP research within hydrology thus enhanced.
2.2.3 Improved coupling processes - issues and actions
Accepting the hypothesis that better representation of
processes in coupled atmospheric-hydrologic models will yield
improved meteorological prediction at all time scales, research is
required to determine, understand, and model such coupling
processes. The focus of research into several coupling processes
might evolve in response to better specification. However,
initially, research will address improved representation of
precipitation, soil moisture and biospheric processes.
2.2.3.1 Precipitation processes
Clouds and their associated precipitation are important in
the global water and energy cycle and their accurate
representation in atmospheric models is crucial. However,
incorporating moisture processes is difficult because cloud and
precipitation physics is poorly understood, and because the
horizontal resolution of large-scale models is much larger than
the scales at which clouds are formed --hence cloud-precipitation
processes are subgrid-scale mechanisms which must be
parameterized.
(1) Improved parameterization of convective precipitation in
atmospheric models
Focused smaller-scale modeling studies are needed to
investigate how to improve the parameterization of convective
precipitation within regional-scale atmospheric models. To have
credibility, such studies require experimental validation. Such
experiments would involve simultaneous measurements in the
atmosphere and at the surface, and would need to be framed in a
proper regional context by specification of the atmospheric flow
fields through the study area. GCIP has already begun planning the
provision of some of the required observations, in the form of a
Near-Surface Observation Data Set described in
Section 10. GCIP is
also fostering opportunities to validate regional models of
precipitation within the Mississippi River basin through
collaboration with other observational programs such as ARM, the US
Weather Research Program, and the GEWEX Cloud System Study as
described in Section 8.
(2) Statistical analyses of sub-grid scale precipitation
Studies are needed to characterize the true variability of
precipitation in space and time and its relation with the state of
the overlying atmosphere. Understanding the relationship between
actual continental precipitation and that predicted by atmospheric
models is a very high priority for GCIP. Such studies are
especially important at hourly to daily time scales and at spatial
scales up to the area covered by a few grid intervals in mesoscale
and large-scale atmospheric models.
The accuracy with which precipitation can be measured (by
gauges, radar, or both) is likely to be an issue in such studies.
Recognizing this last point, the LSA-SW would be the appropriate
initial focus for such studies since the stage 3, gauge-calibrated
radar precipitation products provided by the Arkansas-Red Basin
River Forecast Center (ABRFC) now have established value for
comparison against modeled estimates using the Eta, MAPS and RFE
regional NWP models.
(3) Research into cold season precipitation issues
Snow is an important component of precipitation, particularly
so in the northern and western regions of the U.S., where it
provides an important component of the available surface-water
resource. Many of the basic atmospheric parameterization issues
are similar for warm and cold season precipitation, though
parameter values are likely to change between seasons. However,
there are additional important research issues related to
quantifying cold season precipitation and its partition into runoff
or soil moisture which must be addressed. Such questions will be
priority issues in the scientific agenda for GCIP studies in the
LSA-NC.
The central question is how to develop precipitation volumes
that give an accurate measure of the temporal and spatial
distribution of snowfall. Associated with this question is the
need to determine how representative are rain gauge measurements of
snowfall and how to combine surface observations of snow depth and
with remote-sensing estimates from aircraft and satellites. These
questions of snowfall measurements are discussed further in
Section 6.1.
(4) Improved understanding of topographic influences on
precipitation
Water is a critical resource in the western U.S. It occurs
mainly through the winter season and to a great extent depends on
the total water vapor flux across the mountains and, hence, on
large-scale circulation in the atmosphere in winter. However, it
is strongly influenced by orography, and GCIP has the potential to
make an important contribution to the improved seasonal-to-interannual
prediction of water resources in the western U.S. by
improving the predictability of orographic precipitation. Accurate
forecasting of water resources requires better definition of the
location of precipitation than is possible with current weather
forecast models. The optimal spatial scale for these forecasts is
around 2-3 km, but to achieve this would require a nested modeling
approach as an extension of presently available systems.
Exploratory research is required to evaluate the value of
successively nested forecast models as a possible mechanism for
applying seasonal-to-interannual forecasts to water resource
issues.
2.2.3.2 Soil Moisture Processes
Soil moisture possesses a memory during its seasonal evolution,
and is determined as the residual between precipitation on the one
hand and evaporation and surface and subsurface runoff on the
other. Many of the modeling studies which have provided evidence
that seasonal predictions show sensitivity to hydrologic-atmospheric
coupling have in fact been framed in terms of
sensitivity to modeled or prescribed soil moisture. There is,
therefore, a clear understanding of the importance of soil moisture
for climate prediction at the seasonal time scale.
Heterogeneity in the spatial distribution of soil moisture is
an inevitable consequence of uneven precipitation, and this can be
exacerbated by the subsequent flow of surface and subsurface water
across uneven topography. Modeling investigations (e.g.
Avissar and
Liu 1996) indicate that naturally occurring soil moisture
heterogeneity (acting through land-atmosphere coupling process)
significantly influences the behavior of the overlying atmosphere.
Progress in understanding the effect of area-average soil moisture,
heterogeneity in soil moisture fields, and in validating models
which describe the seasonal evolution of soil moisture in space and
time have all been curtailed by the historic (and still current)
lack of soil moisture measurements.
(1) Improved and extended soil moisture measurements
The growing deployment within GCIP of arrays of field systems
capable of routine measurement of soil moisture and the prospect of
future deployment of aircraft- and space-borne sensors capable of
providing indirect measurements of near-surface soil wetness
promise relief from observational limits on understanding for soil
moisture processes in coupled models.
Installation of automated soil moisture sensors within the
ARM-CART, Little Washita Watershed and the Oklahoma Mesonet are in
place or underway, and plans are being made to extend deployment in
the Oklahoma Mesonet to include all 114 sites and further
extensions to similar distributed data collection networks
elsewhere in the Mississippi River basin. GCIP is coordinating the
collection of a set of soil moisture (and temperature) profile
measurements along a north-south transect to make observations over
the annual cycle, but with emphasis on documenting freezing and
thawing episodes during the cold season. This transect from
Plainview, TX (about 30N) to Bemidji, MN (about 47N) in the
vicinity of the 96W longitude. In addition to these new data
sources, the Illinois state water survey soil moisture data
(Hollinger and Icard 1994)
remain a valuable data resource for
GCIP. The distribution of soil moisture data from these new arrays
of soil moisture sensors to the GCIP coupled modeling community is
a high priority, as is their synthesis into regional products for
model initiation and calibration purposes. A more detailed
description of the soil moisture measurement and analysis is given
in Section 6.2.
The GCIP community strongly supports the proposal to provide
routine remotely sensed measurements of soil moisture using a
satellite L-band microwave radiometer. The community understands
that such observations can only provide indirect estimates of
near-surface soil wetness for certain vegetation covers, but also
recognizes that these data are most reliable for short-rooted and
sparse vegetation where soil moisture control is most important.
Routinely provided soil wetness estimates from satellites could be
exploited for coupled model initiation and validation using
four-dimensional data assimilation techniques to improve the prospect of
better seasonal climate predictions for North America. Moreover,
GCIP provides a unique opportunity to validate and calibrate
remote-sensing soil moisture data because of the richness of other
data fields, such as WSR-88D and gauged rainfall, runoff, and
modeled evaporation, from which alternative area-average soil
wetness estimates can be made. Calibration of remotely sensed soil
wetness data within the GCIP region could thus be the basis for
their application elsewhere in the world.
The potential availability of new sources of soil moisture data
gives rise to the need to determine how these data can best be used
to initiate and validate coupled models. Research is required to
investigate how to use sample data from arrays of surface
measurements and exploratory remote-sensing data from airborne
radiometers. Some modeling studies have been done, but with very
limited field validation. Properly conceived combined field and
modeling studies such as the recent CASES-97 should greatly
illuminate this issue. The coupled modeling community is aware of
and applauds the GCIP efforts during the last four years and the
DOE, NSF, and NOAA sponsored CASES-97 and the NASA sponsored SGP97
field studies within the ARM-CART study area in the Mississippi
River basin that fulfill some of these observational needs, and
look forward to working with the data that will result .
(2) Coupled modeling of the effect of soil moisture on the
atmosphere.
Investigations on the effect of soil moisture and its
heterogeneity in a fully coupled 3-D, atmospheric-hydrological
system are required. The opportunity exists to run fine-scale,
nested grid microscale (large-eddy simulation) models that can
resolve clouds and the resulting precipitation fields in the
context of the upcoming observational studies just described. These
model results (considered in a statistical sense) can be compared
with the airborne sensor and ground soil moisture observations and
with radar and gauged rainfall measurements to determine the
quality of the model simulation.
An alternative approach to coupled modeling is to assume that
precipitation and other atmosphere processes cannot be predicted
deterministically and to conceive models that provide statistical
representation of these processes. The challenge is then to develop
complementary hydrological models that can be forced with
statistical distributions of meteorological variables such as
precipitation, solar radiation, etc., and to use these to calculate
statistical estimates of the feedback to the atmosphere in the form
of sensible-heat fluxes, etc. Statistical models of this type would
also benefit from validation against the statistical distributions
of precipitation and soil moisture observed in the upcoming
observational studies discussed above.
A much greater understanding is needed on how coupled models
represent and would utilize the soil moisture observations for
testing and validation, both in a spatial and temporal context.
Efforts are needed to bridge the gap between the disparate scales
of the point measurement with the simulation model grid box.
Further, basic research is needed to determine to what extent
downscaling of remotely sensed soil moisture is required in order
to be used in coupled models.
An important aspect of coupled modeling research concerns the
possible importance of soil moisture on the formation and evolution
of mesoscale convective complexes (MCCs) and mesoscale convective
systems (MCSs). Such large mesoscale systems are often initiated
over mountainous terrain and move eastward, and they produce a
significant portion of warm season precipitation in the Mississippi
River basin. Current studies in the western Mississippi River basin
need to take account of these mesoscale systems because they play
a major role in the warm
season hydrological cycle in the southeastern Mississippi River
basin. Fine-scale modeling studies are required to ensure adequate
simulation of MCCs and to investigate their relation to the
underlying soil moisture fields in the regional NWP models. Again
these studies would be best linked to upcoming observational
initiatives. After accurate trial simulation of MCCs is
accomplished in these particular situations model tests of the
effect of MCCs on the regional hydrology can be made under varying
soil moisture conditions. A further description of this topic in
connection with research relative to the GEWEX Cloud Systems Study
is given in Section 8.
2.2.3.3 Snowcover and Other Cold-Season Processes
(1) Snow cover
With its high albedo, low thermal conductivity, and
considerable spatial and temporal variability, the seasonal snow
cover overlying land plays a key role in governing the Earth's
global radiation balance; this balance is the primary driver of
the Earth's atmospheric circulation system and associated
climate. Of the various surface radiation balance components,
the location and duration of snow cover comprises one of the most
important seasonal variables. In the northern hemisphere, the
mean monthly land area covered by snow ranges from 7% to 40%
during the annual cycle, making snow cover the most
rapidly-varying surface-feature on Earth. In light of the role that snow
plays in determining weather and climate, it is essential that
regional and global models used to simulate weather and climate
be capable of accurately describing the evolution of seasonal
snow covers. In past years, significant strides have been made
to better represent snow cover in climate models, but there are
still indications that current representations of seasonal snow
in these models are plagued by significant deviations from
observed snow-related fields.
The timing of snow deposition and melt is one of the most
important climatic and hydrologic influences affecting
agriculture and water resources. A detailed understanding of
snow pack processes, including thermal properties, meltwater
percolation, density and albedo evolution (especially during the
melt season) is necessary to understand the relationships between
snow cover, atmospheric processes, and surface hydrology during
'normal' and 'anomalous' snow
cover regimes. Lagged relationships between snow cover and other
parameters are also expected to be important. A main goal of
GCIP should be to improve our understanding of these processes,
and to build better representations of snow cover in the global
and regional climate models that are used for GCIP diagnostic and
prediction studies. Accomplishing this will require studies
which 1) evaluate sub-models using GCIP snow and meteorological
field data; 2) develop parameterizations of snow-cover extent for
atmospheric models; 3) evaluate coupled simulations using
satellite data and other gridded data sets; and 4) identify
snow-atmosphere feedback-processes operating within the GCIP regions.
(2) Subgrid Interactions
The interactions between wind, topography, vegetation, and
snowfall produce snow covers of non-uniform snow-water-equivalent.
During these blowing and drifting snow events, a
largely unknown amount of moisture is returned to the atmosphere
as the snow grains sublimate during transport. When the snow
melts, the variable snow depth leads to a patchy mosaic of
vegetation and snow that evolves as the snowmelt progresses.
From the perspective of a surface energy
balance, the interactions between land and atmosphere are
particularly complex during this period. Subgrid-scale
heterogeneity in land surface use/cover is likely to be an
important factor in determining the time dependence of snow cover
fraction and albedo during melt episodes. Field experiments need
to be conducted to determine the magnitude and principal sources
of subgrid-scale heterogeneity in albedo and snow cover for
agricultural and mountainous landscapes
typical of the Mississippi River Basin. This should be done
using both ground-based and airborne measurements. These
analyses should also focus on developing methods to incorporate
heterogeneity effects into land-surface models using land-cover
data. The estimation of snow-cover extent and winter surface
albedo also needs to be addressed from the perspective of knowing
the fractional snow-covered area within the model domain.
Because of importance of snow in simulating weather and
climate, the level of snow-model complexity for use in these
models needs to be addressed. Several features of process-level
snowmelt models might be used to help capture subgrid-scale
variability and improve snowmelt
simulations. In particular, important tasks relevant to cold-season
processes include: quantifying the feedbacks between
exposed vegetation and the melt of adjacent snow covers,
demonstrating the
importance of fractional snow-covered area in surface-energy
partitioning during snowmelt, and identifying significant
interrelationships between non-uniform snow distributions, melt
rates, and exposure of vegetation.
(3) Hydrologic interactions
Important cold-season hydrologic processes include frozen
soil, the infiltration of frozen soil, frozen lakes, and
relationships between snowmelt and stream discharge. Frozen
soils have, until
recently, been ignored by most land-surface macroscale models,
but they can play an important role in increasing runoff during
snowmelt events. In particular, the interactions between
snowmelt, frozen soil, ponding, and infiltration need to be
addressed. As part of studies focusing on the amount and timing
of water resources and the soil moisture available for subsequent
evaporation, it is also necessary to document, understand, and
model how the water is partitioned into runoff and infiltration
when snow and ice melts. Land-surface hydrology models developed
for GCIP must include these cold-season-related interactions.
2.2.3.4 Biospheric processes
Vegetation type and amount influence various aspects of the
hydrologic cycle, from
interception of rainfall to active control on the transpiration
process through stomatal
regulation. This has direct impact on the partitioning of
available energy (net radiation) into sensible, latent, and
ground heat fluxes. Development of a vegetated canopy also
affects the absorption of momentum from the local wind and is
manifested in the local friction velocity and corresponding
surface roughness. Although the absorption of momentum by the
canopy is determined in part by the total amount of leaf
biomass, the energy partitioning is closely coupled to the amount
of active or green leaf biomass. Land-surface models must be
able to account for this difference not only during periods of
drought or water stress, but also during the later stages of the
yearly growth cycle when many plant species begin senescence.
These various stages of the growth cycle are likely to contain a
spectral signature that may be used not only to assess the
total biomass and percent green leaf area for model evaluation
but for future applications in near real-time model data
assimilation mode for both short and long range forecasts. Much
of this research will be addressed through the International
Satellite Land Surface Climatology Project (ISLSCP) from joint
NOAA/NASA sponsored observations and modelling studies. In
addition to the
NOAA sponsored long-term flux measurements, additional sites
within the GCIP domain where observations of mass and energy
fluxes and information on the planetary boundary layer are
proposed. These measurements will be used to not only document
the energy balance over the annual cycle, but to test many of
the land-surface parameterizations currently used in the short,
medium and long range forecasts. It is also anticipated that
short-term field campaigns will be conducted in order to
evaluate how the energy fluxes and forcing variables vary
spatially. This
will be accomplished with instrumented aircraft capable of
similar energy flux and radiative measurements found on the
surface flux systems.
Measurements of soil moisture will play an important role in
assessing the role of water stress of plant response and
biological processes and the resulting impacts on the surface
energy balance. Water stress can play an important role in
regulating the surface energy balance. Although an ample amount
of precipitation can replenish the available soil water for plant
processes, plant recovery is not immediate and may not always be
entirely reversible. This "timescale" of recovery can be
different depending on the plant species considered. Attempts
will be made to evaluate these water related impacts and its
impact on the local surface energy balance.
2.2.4 Model Parameter Estimation
A key step in applying land surface parameterization schemes
is to estimate model parameters that vary spatially and are
unique to each grid point. Local model parameters are estimated
on the basis of information about vegetation, soils and geology,
so gridded fields of these characteristics are needed at various
scales to provide such estimates. It has been shown (e.g. In
PILPS2c) that existing a priori parameter estimation techniques
may produce large errors and biases in the water balance,
especially in mean annual runoff. Improved methods for parameter
estimation are needed.
2.2.4.1 Parameter Estimation Techniques
Existing parameter estimation techniques generally assume
that land surface model parameters are related to various land
surface characteristics according to simple a priori
relationships. These relationships are assumed to apply
universally without regional or climatic variation. For example,
the rooting depth of a given type of vegetation is assumed to be
the same regardless of the climate where the vegetation is
located. Existing parameter estimation schemes are largely
untested. It remains to be seen if improved universal parameter
estimation techniques can be found for at least some models or if
regional relationships may be required.
There already is a wide range of models available. The
Project for Intercomparison of Land surface Parameterization
Schemes (PILPS) has shown that the available schemes can indeed
produce a wide range of results given the same
hydrometeorological forcing, land surface
characteristics and common rules for parameter estimation. No
doubt part of this is due to model
structure differences, but much is due to the way model
parameters are estimated One approach to developing improved
parameter estimation techniques would be to use historical
observations of runoff response to observed hydrometeorological
forcing of many watersheds over a wide range of climatic and land
surface conditions. The steps, which are illustrated in
Figure 2-1 are:
1. Develop historical hydrometeorological data sets (model
forcing and output) and basin characteristics data
(soils, vegetation, topography and climate).
2. Calibrate model parameters for a large number of basins
(for a given model).
3. Relate calibrated model parameters to basin
characteristics to develop regionalized a priori
parameter estimation techniques for selected parameters
(for a given model)
4. Use the regionalized parameter estimates for a large
number of basins. Evaluate the results in terms of
model performance when parameters are estimated by:
A. Initial a priori parameter estimation techniques
B. Model calibration
C. Regionalized a priori techniques derived
using calibrated parameters
5. Test the transferability of the results to other basins
not used in the above analysis. These basins may
be in the same region or in other continents.
6. Expand the available data sets to include
representation from all climate regimes of the
earth and to achieve the best possible global coverage.
7. Assess whether parameters for some models are easier to
estimate than parameters of others and modify
models to have more "observable" parameters.
Figure 2-1 Steps in Parameter Estimation for the Model Parameter Estimation
Experiment (MOPEX).
An important step in achieving this goal is to assemble
historical hydrometeorological data and river basin
characteristics for about 200 intermediate scale river basins
(500 - 10,000 km2) from a range of climates throughout the world.
The data sets to be developed would not be model
specific and would be appropriate for developing parameter
estimation schemes for most, if not
all, land surface parameterization schemes. A Model Parameter
estimation EXperiment task
(MOPEX) has been initiated by GCIP to begin to assemble the
required data sets and to organize
parameter estimation experiments among model developers and
users.
2.2.4.2 Potential Improvements for Model Parameterization
Some factors which need to be considered with regard to
critical variables in land surface modeling are summarized below:
1) Soil evaporation. In many cases, the soil evaporation
efficiency parameter (i.e. the so-called beta function) is
determined empirically as a function of top-layer soil
moisture. It should be thermodynamically based. Most models
use very thick layers; this should be adequately thin to
represent the diurnal variation of soil evaporation.
Therefore, more layers may be needed to overcome the numeric
problem. Most models neglect the flux of water vapor in the
soil, which may limit their performance over semi-arid land
surfaces.
2) Canopy transpiration. Most models use the Penman-Monteith
approach, and parameterize the stomatal resistance as a
function of environmental conditions. The same equations are
assumed to apply for different types of vegetation, while
the only difference is to adjust values for some parameters.
More work is needed to test the universality of these
equations.
3) Canopy evaporation. In most models, the evaporation from
the intercepted water on canopy surface are essentially a
"bucket-type" model, in which a constant interception
capacity is assumed, and it yields the canopy interception
capacity by multiplying a leaf-area index (LAI). The canopy
drip would occur only when the canopy interception capacity
is reached.
4) Runoff. The land-surface models used in atmospheric models
have not explicitly considered the effects of the subgrid
features of topography on runoff generation on scales from
10km x 10km to 300km x 300km, typical resolutions of weather
forecasting and climate prediction models.
5) Routing. The horizontal water transport from neighboring
grid-boxes are generally neglected in the land-surface models.
6) Snow and Ice. Variations of snow- and ice-related
features within the GCIP domain frequently occur at subgrid-scales to
the regional and global weather and climate models
being applied to the area. As a result, these features,
which include snow-water-equivalent distributions, snow
thermal properties, frozen lakes and soils, and patchy
configurations of snow and vegetation during melt, and their
associated influence on subgid-scale energy and moisture
fluxes, must be accounted for (or parameterized) within these
models.
2.3 Improvements to regional mesoscale models
For the past four years there has been an extensive effort to
acquire the model output from several operational/experimental
centers from a range of operational models of varying resolution,
physics and data assimilation systems. GCIP is concentrating on
three regional mesoscale models (IGPO 1995):
The participation by the operational centers in providing
regional model output for GCIP leads to a mutually beneficial
relationship. The principal benefit to GCIP is to provide a
measure of the inter-model variability of the outputs from the
different regional models which can also be related to the global
model output from the operational centers. GCIP can provide
benefit to the operational centers by enabling them to make use
of the enhanced data sets to calibrate and validate the model
data assimilation and forecast systems.
The regional mesoscale models are supporting GCIP research in
the following manner:
The regional models now running operationally (NCEP/Eta and
CMC/GEM) will be upgraded with numerous improvements during the
next several years. The GCIP investigators need to be aware of
these plans for improvements and the schedule being followed to
incorporate these improvements into the operational models. The
experimental MAPS model will also be upgraded during the next
several years. Projected improvements to each of the regional
models is described in the remainder of this section.
2.3.1 The NCEP Mesoscale Eta Model and Eta Data Assimilation
System (EDAS)
Since April 1, 1995, output from the NCEP Eta model
(Black 1994) and its
associated Eta-based 4-D Data Assimilation System
known as EDAS (Rogers 1995)
have been routinely archived for
GCIP. In conjunction with this milestone, NCEP implemented for
GCIP an extensive expansion of the routine ETA/EDAS output
products, including a vast suite of surface and near-surface
products that encompass all the surface energy and water fluxes,
soil moisture and temperature, snowpack and snowmelt, and surface
and subsurface runoff. These output products include a) 3-hourly
analysis and 6-hourly forecast horizontal gridded fields (known
in GCIP as MORDS) and b) hourly station time series output (known
is GCIP as MOLTS) at nearly 300 sites. A number of GCIP
investigators have completed and published assessments of the
coupled ETA/EDAS land-surface and/or water budget performance,
including Berbery et al. 1996;
Yarosh et al. 1996;
Betts et al. 1997; and
Yucel et al. 1997. The
assessment being done by
Curtis Marshall at the University of Oklahoma is making use of
the Oklahoma Mesonet, including the newly installed soil
moisture profile sensors.
With GCIP support, and in collaboration with GCIP
investigators including the NWS Office of Hydrology (OH), NCEP
over the last four years has accelerated ETA/EDAS development and
improvement in the following three key areas:
1) Coupled land-surface/hydrology model
2) 4DDA assimilation techniques and data sources
3) Precipitation, cloud, and radiation physics
These three GCIP-supported NCEP development areas have
resulted in the following GCIP-related improvements being
implemented operationally into the ETA/EDAS system (including the
GCIP model output data sets):
- New NESDIS green vegetation fraction replaces ISLSCP;
- New bare soil evaporation function, improved snowmelt
physics;
- Improved snow albedo, improved solar insolation via
improved cloud and ozone effects;
- Improved numerical advection scheme for water vapor and
cloud water.
- Computational spatial resolution improved from 48-km
to 32-km horizontally and from 38-layers to 45-layers
vertically: Output grid for GCIP remains the same at 40-km;
- Number of soil layers increased from 2 to 4 (total soil
depth remains 2 meters);
- Soil moisture and temperature is continuously cycled
day-after-day in the regional Eta assimilation/EDAS
replacing the soil initialization from the Global Data Assimilation System;
- Longstanding optimal interpolation technique is
replaced with 3-D variational assimilation;
- Assimilation of hourly GOES-derived water vapor data.
The steady march of major land-surface implementations in
January 1996, February 1997, and January 1998 represent major
GCIP coupled model milestones, resulting from the focused and
concerted GCIP-funded efforts of both the GCIP Core Project
(NCEP, OH, NESDIS) and external GCIP researchers. These
operational milestones represent noteable success in the model
development strategy laid out in
Figure 1-3 of
Section 1.4.1, which
called for a clearly identifiable operational path to serve as a
demonstration and implementation environment and for the research
advancements in the research path.
In addition to the above ETA/EDAS model/assimilation changes,
production of the following realtime products were initiated for
GCIP:
The following is a list of ongoing GCIP-focused ETA/EDAS
developments now underway with a projected implementation within
the next 18 months:
The 3-D variational assimilation technique (3-D VAR)
implemented in January 1997 represents a major data assimilation
milestone. This advanced data assimilation technique allows
easier incorporation of non-traditional data sources, such as
direct use of satellite radiances, cloud cover, precipitation,
radar radial winds, radar reflectivities, wind profilers, WVSS,
ACARS, and ASOS.
The follow-on to 3-D VAR, namely 4-D variational assimilation
(4-D VAR), employing the linear adjoint of the Eta model, is well
advanced at NCEP/EMC and undergoing routine testing. The linear
adjoint has been extended to include the new land-surface/hydrology
physics, which provides an opportunity for the
assimilation of land-surface related information such as
streamflow and satellite-derived surface skin temperature. 4-D
VAR is computationally expensive and its operational
implementation awaits the next generation computer upgrade at NCEP.
2.3.2 Regional Model Upgrade at CMC
It is expected that the quality of the regional model outputs
at CMC will improve significantly during the coming two or three
years, especially in terms of the variables that are important
for the water and energy budgets which are of prime interest to
GCIP.
The developments during this period will stem from a major
change that was made in February 1997 when the Regional Finite
Element (RFE) was replaced by the Global Environmental Multiscale
(GEM) model. This is the result of a major project which has
been in progress at Recherche en prevision numerique (RPN) with
the goal of developing a non-hydrostatic variable-resolution
global model of the atmosphere. This model uses a finite-element
based spatial discretization and has been designed for efficiency
and flexibility to satisfy the requirements of
operational weather forecasting on a wide range of time and space
scales (see the description published earlier in the GCIP Major
Activities Plan for 1995, 1996 and Outlook for 1997;
IGPO 1994c).
This model uses a global variable-resolution strategy, permitting
the focusing of an arbitrarily-rotated high resolution latitude-
longitude mesh on any geographical area of interest, be it
tropical or extra-tropical, making it a more flexible strategy
than the former operational RFE model which is limited to
extra-tropical applications. The overhead associated with using a
model of global extent for short-range forecasting, even at the
meso-gamma scale, is relatively small: more than half of the
total number of meshpoints are on the uniform-resolution area of
interest, and the overhead of using variable resolution outside
this area is consequently comparable to that of the sponge
regions of one-way interacting models. To maintain the validity
of the model at the mesoscale, the formulation uses the
non-hydrostatic Euler equations, with a switch to revert to the
hydrostatic primitive equations for larger scale applications
where the hydrostatic assumption is valid. To this end a
pressure-type hybrid vertical coordinate is adopted. Prototype
tangent linear and adjoint versions have also been developed in
preparation for a 4-dimensional variational (4DVAR) analysis
system.
The first implementation of GEM was made with a configuration
giving a uniform fine mesh grid over North America and the
adjacent ocean areas at a resolution equivalent to the former 35
km RFE one. However, testing of an experimental version with a
15 km resolution over more limited regions of North America is
already in progress. This experimental version uses the
Fritsch-Chappell meso-scale convective parameterization as opposed to a
Kuo-based scheme in the currently operational version. The
recent installation of a 32 processor NEC SX-4 supercomputer,
with other upgrades to come during the three-year period, should
permit a continuing gradual increase in the horizontal
resolution, with corresponding increases in the number of
vertical levels. Progress with the 4DVAR analysis scheme should
also lead to better assimilation of meso-scale data, especially
related to moisture which is crucially important for GEWEX.
Energy budget calculations are also expected to benefit from more
sophisticated solar and radiation parameterizations.
A new concentrated activity on model coupling has been
initiated at RPN, to do research and development on the coupling
of atmospheric, land surface, hydrology, ocean, ice and wave
models in order to construct a more comprehensive environmental
prediction system. The construction of the baseline system is
expected to take about a year and a half, but it is likely that a
more interactive coupled data assimilation and prediction system
will become available to
CMC during the year 2000, with significant impact on the water and
energy components in the analyses and forecasts.
2.3.3 Improvements to MAPS
The gridded output from the Mesoscale Analysis and Prediction
System (MAPS) will improve over the next several years of the
GCIP EOP in different areas including model physics, data
assimilation, and spatial resolution.
Some improvements related to GCIP have already been
implemented, including access to daily lake-surface temperatures
(from NOAA's Great Lakes Environmental Research Laboratory), snow
and ice cover (from NCEP and the US Air Force). MAPS is
currently using monthly climatological sea-surface temperature,
to be replaced by daily information from NCEP in the near future.
The most important of the GCIP-related changes has been the
implementation of a multi-level soil/vegetation model. This
model, currently running with 5 soil levels, is described by
Smirnova et al. (1997).
High-resolution data sets for fixed or
seasonally varying surface characteristics (soil type, vegetation
indices, albedo) made available by NCEP are being used now in MAPS.
Full atmospheric radiation has also been added to the MAPS
experimental 40-km model, substantially improving lower
troposphere temperature forecasts.
A number of changes implemented during WY'97 in the
experimental 40-km MAPS, include:
- 3-d variational analysis in the MAPS isentropic-sigma
coordinate, to replace the current optimal
interpolation scheme.
- explicit cloud microphysics in the MAPS model, with
forecasts for cloud water, rain water, snow, ice,
graupel, and the number concentration or ice particles.
This is the revised microphysics from the NCAR/Penn State MM5 model.
- an improved forward/backward digital filter initialization.
- an improved turbulence parameterization (Burk-Thompson
level-3.0) with explicit forecast of turbulent kinetic energy
- Addition of snow, frozen soil physics to
soil/vegetation package, including 1-d tests with PILPS 2-d data sets.
- Initial cloud/moisture analysis.
- Assimilation of precipitation data.
- Assimilation of water vapor data from the ACARS WVSS and GPS data
- Assimilation of GOES, SSM/I precipitable water and wind products.
- Use of improved covariances in 3-d variational analysis
allowing better representation of divergent wind component.
Plans for WY98 -
- Incorporation of GOES radiance/imager data in
cloud/moisture/temperature analysis.
- Assimilation of WSR-88D radar radial winds.
- Possible specification of soil moisture from an off-line data assimilation system.
Plans for WY99.
- Resolution at 15-20km range.
- Possible incorporation of a non-hydrostatic hybrid
isentropic-sigma model.
- Experiments with simplified Kalman filter or 4-d
variational techniques.
One of the principal functions of the regional mesoscale
models, as was noted in Section 2.3 is to produce the model
assimilated and forecast output products for GCIP research,
especially for energy and water budget studies. The production
of such data sets is designed to achieve the following objectives:
(i) To produce model assimilated and forecast data
products for GCIP investigators with an emphasis on
those variables needed to produce energy and water
budgets over a continental scale with detailed emphasis
in 1997 on the LSA-SW and the LSA-NC and beginning the
application of such detailed emphasis capability to the
LSA-E during 1998, and to the LSA-NW during 1999.
(ii) To produce a quantitative assessment of the accuracy
and reliability of the model assimilated and forecast
data products for applications to energy and water budgets.
(iii) To conduct the research needed to improve the time
and space distribution along with the accuracy and
reliability of the model assimilated and forecast data products.
The activities relevant to the third objective above were
described in the previous section. The activities relative to the
first two objectives are summarized in the remainder of this
section.
2.4.1 Regional Mesoscale Model Output
The list of model output fields needed by GCIP researchers
was given in Table 3, Volume I of the GCIP Implementation Plan
(IGPO 1993).
From the beginning of GCIP, it has been the intent
to acquire model output from several different models of varying
resolution, physics and data assimilation systems. The large
volume of data produced by the current generation of atmospheric
models has forced a number of compromises in order to achieve a
tractable data handling solution for model output data. The data
volume is further enlarged by the GCIP need to enhance the
traditional model output to include additional fields needed by
researchers to perform meaningful studies of the water and energy
cycles. The near-term GCIP needs for model output data will be
met by concentrating on three regional mesoscale models:
The model output is divided into three types:
(1) One-dimensional vertical profile and surface time
series at selected locations referred to as Model
Location Time Series (MOLTS)
(2) Gridded two-dimensional fields, especially ground
surface state fields, ground surface flux fields,
top-of-the-atmosphere (TOA) flux fields, and atmospheric
fields referred to as Model Output Reduced Data Sets (MORDS)
(3) Gridded three-dimensional atmospheric fields containing
all of the atmospheric variables produced by the
models.
Each model output type is described in more detail in
Appendix B.
2.4.2 Evaluation of Model Output
Objective: To produce a quantitative assessment of the
accuracy and reliability of the model assimilated and forecast
data products for applications to energy and water budgets.
All of the evaluations require a lengthy series of observed
data for the variables considered critical to achieve the
objective stated above. As a start on this evaluation effort ,
GCIP is compiling a composite data set for as many of the
variables as reasonably available. In order to maximize the
number of observed variables this composite is focused on the
LSA-SW with the initial emphasis on the ARM/CART site as
described in Section 10. Evaluations of the regional model
output is also part of the budget studies described in Section 5.
![[LSAs]](/gcip/map_98/images/fig2.1.gif)
2.4 Model Assimilated And Forecast Data Sets