Introduction:
A
critical step in applying a hydrologic model to a watershed
or a land surface parameterization scheme (LSPS) of an
atmospheric model to a specific grid element is to estimate
the coefficients or constants in the model or LSPS known
as parameters. These parameters are inherent in every
model. They vary spatially so they are unique to each
watershed or a grid point. Some model parameters may also
vary seasonally as well as spatially. How to estimate
model parameters has been receiving increasing attention
from the hydrology and land surface modeling community.
Presently
a priori relationships linking model parameters and land
surface characteristics such as soil and vegetation classes
are available for many hydrologic models and LSPSs. But
these relationships have not been fully validated through
rigorous testing using retrospective hydrometeorological
data and corresponding land surface characteristics data.
This is partly because the necessary database needed for
such testing has not been available. Moreover, there still
exist a gap in our understanding of the links between
model parameters and the land surface characteristics.
Generally available information about soils (e.g., texture)
and vegetation (e.g., type or vegetation index) only indirectly
relates to model parameters such as hydraulic properties
of soils and rooting depths of vegetation. Also it is
not clear how heterogeneity associated with spatial land
surface characteristics data affects those characteristics
at the scale of a basin or a grid cell. Consequently,
there is a considerable degree of uncertainty associated
with the parameters given by existing a priori procedures.
Recent
studies have illustrated clearly that existing a priori
parameter estimation procedures do not produce proper
parameter values and that improper model parameters result
in poor model performance. Figure
1 shows modeled partition of annual runoff and evapotranspiration
for many different land surface models (LSMs) participated
in the Project for Intercomparison of Land-surface Parameterization
Schemes Phase 2c (PILPS-2c). These LSMs were driven by
the same meteorological forcing data. More interestingly,
they were required to prescribe the same values for commonly
named parameters such as soil hydraulic properties and
vegetation phenology parameters. The large scattering
of model performance can be partly be explained by the
uncertainty in the values of the best parameters to use
in each model. Figure 2 is a scattergram
of Nash-Sutcliffe efficiency of daily streamflow simulation
from different models using a priori parameters and using
calibrated parameters for one of the MOPEX test basins.
The efficiency for calibrated cases is generally much
higher than the a priori cases, indicating a priori parameters
are problematic.
Clearly
there is a need to improve the existing a priori parameter
estimation procedure. A project known as Model Parameter
Estimation Experiment (MOPEX) has been funded by NOAA
Office of Global Programs to investigate techniques for
the a priori estimation of the parameters used in land
surface parameterization schemes of atmospheric models
and in hydrological models. A first major step by MOPEX
project is the development of a comprehensive database
that contain many years of historical hydrometeorological
time series data and land surface
characteristics data for many basins in the United States
and from other countries. MOPEX project has been truly
an international collaborative effort with involvement
of international scientists and data assembled from different
countries.
MOPEX
Goals and Objectives:
MOPEX
has the following science goals and objectives
1:
To develop improved a priori model parameter estimation
techniques for large scale modeling applications and
for ungaged basins
2: To develop an international database of retrospective
hydrometeorological data and basin characteristics data
for a wide of climate and geophysiological conditions
3: To develop objective measures to evaluate the parameter
estimate techniques and to understand parameter uncertainty
4: To develop diagnostic tools to foster improved understanding
of natural hydrologic processes at basin scales and related
behaviour of hydrologic models
5: To promote and facilitate the exchange of ideas and
experiences on approaches to model parameter estimation
for different climatic regimes
MOPEX
Science Strategy:
The
overall strategy of the land surface model experiments
is illustrated in Figure 3. The
first step of the MOPEX strategy is to develop the necessary
data sets. These data are then used to study individual
models using three parallel paths. The first path is
to make control runs with model parameters estimated
using existing a priori parameter estimation procedures.
The second path is to make model runs using calibrated
or tuned values of selected model parameters. Then,
relationships would be developed between the calibrated
parameters and basin climate, soils, vegetation and
topographic characteristics. These relationships are
used to define the new a priori parameters. The third
path is to make new model runs using the new a priori
parameter estimates. Achievement of the parameter estimation
goal is then established in two steps. The first is
to measure how much of the potential improvement in
model performance when operated in calibration
path is obtained when the model is operated using new
a priori parameters. This step uses the same data sets
as were used to develop the new a priori parameter estimates.
The second step is to demonstrate that new a priori
techniques produce better model results than existing
a priori techniques for basins not used to develop the
new a priori techniques. The outcome of this step has
very strong implications for the Prediction in Ungaged
Basins (PUBs), a major initiative undertaken in International
Association of Hydrological Sciences (IAHS).
www.nws.noaa.gov 
