Hydrologic Modeling Studies
DRAINMOD
The computer model DRAINMOD (Skaggs, 1978) was selected to simulate and
analyze the long-term effectiveness of restoration efforts on site hydrology.
The model was developed for the study and design of water management systems
on poorly drained, shallow water table soils. A combination of methods
are used in the model to simulate infiltration, drainage, surface runoff,
evapotranspiration, and seepage processes on an hour-by-hour, day-by-day
basis (Skaggs et al., 1991a). Skaggs et al. (1991b) modified DRAINMOD for
application to wetland determinations by adding a counter which accumulates
the number of times that the water table rises above a specified depth
and remains there for a given duration during the growing season.
DRAINMOD has been shown to reliably predict
water table hydrology in poorly drained, shallow water table soils for
a wide range of geographic locations and soil types (Skaggs, 1982; Rogers,
1985; Chang et al., 1983; Susanto et al., 1987). It has been used to simulate
the hydrology of pocosin and wetland areas in North Carolina (Skaggs et
al., 1991a; Chescheir et al.,1992; Skaggs et al., 1994; Richardson and
McCarthy, 1994), and has been found to compare closely with the results
obtained from other simulation models such as SWATREN and PREFLO (Workman
and Skaggs, 1989; Workman and Skaggs, 1991; Desmond et al., 1996).
Model
Calibration and Verification
DRAINMOD inputs were compiled and initial simulations were run to simulate
the hydrologic conditions observed at the restoration sites from January
1996 through December 1997. Water table depth was chosen as the comparison
criteria for defining the "goodness of fit" relationships between simulations
and observed conditions. Since soil upward flux and volume drained relationships
were approximated from soil water characteristic data, these parameters
were adjusted within acceptable limits to improve simulation results. Values
for surface storage were also altered within the range of values measured
in the field to more closely simulate observed depths and durations of
ponding.
At the Craven County site, lateral seepage
and seepage around the water control structures was observed in the field.
DRAINMOD allows for inputs to approximate lateral seepage losses to a nearby
outlet. The lateral seepage inputs which resulted in simulated water table
depths most closely approximating observed water table depths are given
in Table 2. The hydraulic head of the receiving waters was chosen to represent
the head at the bottom of the channelized stream on the eastern side of
the site. DRAINMOD computes seepage assuming the hydraulic head of the
receiving waters is constant. The large channelized stream, however, is
an intermittent stream and does not flow during dry periods. Therefore,
the model simulations represent only estimates of the seepage conditions
that actually affected the site. To more closely simulate the effect of
seepage around the water control structures at the Craven County site,
DRAINMOD inputs for the elevations of the weirs were adjusted over a range
of values to determine the setting which most closely simulated observed
conditions.
Simulated water table depths were found to
compare favorably with observed data for both restoration sites. Average
error (AE), average absolute error (AAE), and root mean squared error (RMSE)
were calculated using the methods described by Evans (1991) for each hydrologic
treatment and study year, as well as for pooled data from both study years.
DRAINMOD tended to over-estimate water table depths during 1996, a wet
year, and under-estimate depths during 1997, a dry year. For the pooled
data from both sites, AAE ranged from 9.57 to 15.55 cm and RMSE ranged
from 14.35 to 23.71 cm. These values compare favorably with Broadhead and
Skaggs (1989) who reported predicted water table depths within 10 cm of
observed values on both natural and drained pocosins in eastern North Carolina.
Differences between simulations and observed
field data are expected since model inputs rarely describe the conditions
of a site fully. Although some errors were incurred, there was generally
good agreement between observed and simulated hydrology. Furthermore, differences
between observed and predicted conditions were similar to the differences
between observed replicate values for the same treatment. It was concluded
that DRAINMOD could be used to adequately model the restoration sites.
Modeling
the Pre-Drained, Natural Wetland Conditions
On unaltered wetland systems with little slope, natural outlets for surface
water are provided by a network of meandering streams. It is difficult
to determine an accurate spacing for these natural stream systems, since
many areas have been channelized for well over 100 years. However some
published data has been presented. Richardson and McCarthy (1994) used
drain spacings of 1200 m to simulate natural pre-mining conditions on pocosin
peatlands in North Carolina. Lilly (1980) mentions the installation of
drainage canals in the 1800's for harvesting timber at a spacing of 1600
m. However, the two restoration sites modeled in this study were originally
located relatively close to natural streams. The natural streams were channelized
in the past to improve drainage, but the positions of the streams relative
to the restoration sites were changed very little. Therefore, two modeling
scenarios were used to evaluate the pre-drainage hydrologic conditions
of the restoration sites: 1) a drain spacing of 150 m for the Craven site
and a spacing of 200 m for the Beaufort site, based on measured distances
from instrumentation to the channelized stream, were used to simulate the
pre-drainage conditions at the location on the sites where field measurements
were recorded, and 2) a drain spacing of 1000 m was used for both sites
to simulate pre-drainage conditions on an area representative of an interstream
divide. Drain depths were set at 60 cm for the Beaufort County site and
40 cm for the Craven County site, based on field observations, for both
modeling scenarios.
Many of the wetland areas surrounding the restoration
sites are dominated by hardwood and softwood species. In modeling natural
systems, rooting depths ranging from 40 to 60 cm have been cited (Richardson
and McCarthy, 1994; Skaggs et al., 1991a; Skaggs et al., 1994). A constant
rooting depth of 45 cm was chosen to represent the natural conditions of
the restoration sites.
Surface storage values on undisturbed wetland
areas are generally higher than those of agricultural fields. Gayle and
Skaggs (1978) reported maximum surface storage values ranging from 0.25
to 3 cm on agricultural row crops, while Skaggs et al. (1994) gave surface
storage values ranging from 1.5 to 5.0 cm for native forests with hydric
soils in the coastal plain of North Carolina. Based on these values, a
surface storage value of 2.5 cm was chosen for modeling the pre-drainage
conditions of the restoration sites.
Saturated hydraulic conductivity values are
typically much higher for soils in natural wetland systems than for those
in agricultural production. The rooting action of forest species increases
the number of macropores, allowing for increased lateral flow. Hydraulic
conductivity values ranging from 43 to 156 cm/hr in the top 30 to 60 cm
of the soil profile have been documented by Chescheir et al. (1995) and
Amatya et al. (1995) for an undrained, natural forested wetland site located
near Plymouth, North Carolina. Chescheir et al. (1995) found the average
hydraulic conductivity of these soils to be 80 cm/hr. The Plymouth site
is similar in hydrology to the condition of the restoration sites in their
undrained state. Hydraulic conductivity inputs to the model were therefore
adjusted to a value of 80 cm/hr in the surface soil horizons of the soil
profiles.
Modeling
the Long-Term Hydrology of the Restoration Sites
Simulations were performed to evaluate the long-term effectiveness of the
restoration techniques used at the Beaufort and Craven sites. Model inputs
used to model the existing, observed conditions were adjusted to values
representative of a time when a mature forest stand would be present on
the sites for a long enough duration such that the soil properties and
forest stand would stabilize.
For reasons previously discussed, a rooting
depth of 45 cm was used to represent conditions in which a mature stand
of trees had developed on the restoration sites. Rooting action over time
would be expected to influence the soil structure and number of macropores
in the soils of the restoration sites in the same manner as observed on
natural forested wetlands. Therefore, a hydraulic conductivity value of
80 cm/hr was assigned to the surface soil horizons to simulate the long-term
hydrology of mature restoration sites. Reductions in surface storage would
be observed on the restoration sites over time, due to erosional degradation
of the berms surrounding the treatment plots. To reflect this change in
the simulations for the long-term hydrology of the sites, maximum surface
storage inputs were reduced on the smooth microtopography treatments to
2.5 cm, a value typical of natural wetland systems. For the rough microtopography
treatments, maximum surface storage values were reduced to the field measured
amount of surface storage that must be filled before movement of trapped
water between depressions could occur.
Results
DRAINMOD was found to adequately simulate the observed
water table hydrology of the four hydrologic treatments at the Beaufort
and Craven County restoration sites. Water table control was the most important
factor influencing restoration hydrology. In addition to water table control,
surface microtopography also strongly influenced simulation results. Model
predictions for the pre-drained, natural wetland conditions of the restoration
sites were sensitive to changes in surface storage and, to a lesser degree,
saturated hydraulic conductivity in the surface soil horizons. This implies
that the degree of soil roughness incorporated during the initial preparation
of a restoration site can greatly influence the restoration hydrology.
The use of simulation models, such as DRAINMOD, which incorporate inputs
for outlet control and surface storage provides a powerful tool for evaluating
the future hydrologic conditions of a potential restoration site.
Some restoration efforts on prior converted
lands have focused on the hydrologic objective of establishing a water
table depth of 30 cm or less during 12% of the growing season for at least
50% of all years. This criteria was evaluated for the restoration treatments
at the Beaufort and Craven sites, and for the predicted pre- drainage conditions.
For both sites, the predicted pre-drainage conditions were drier than the
conditions necessary to meet the stated criteria. This implies that sites
being restored to meet this criteria may exhibit hydrologic conditions
wetter than the natural, pre-drainage conditions of the sites. Wetland
hydrology influences wetland function, therefore sites restored to conditions
wetter than they were originally cannot be assumed to support the same
wetland functions as they did originally.
The model simulations conducted for the Craven
County site demonstrated that the restoration of prior converted lands
located near deeply incised streams may be difficult. Assuming no seepage
around the water control structures, the Craven County site was predicted
to exhibit much drier conditions than the Beaufort County site, due to
lateral seepage and drainage caused by its close proximity to the nearby
channelized stream. Potential restoration sites located near streams or
drainage canals should be carefully evaluated to determine the effect of
the nearby stream on restoration hydrology.
This research has shown that simulation models
can be used to evaluate the hydrology of restored prior converted lands.
Modeling allows for evaluation of appropriate techniques to achieve restoration
objectives, and the degree of hydrologic wetness which can be achieved
at a proposed restoration site. It is recommended that modeling be incorporated
into the selection process used to determine which prior converted lands
will be chosen for future restoration.
More Information about:
References
Back to Index