Plain Language Summary
A large number of wetlands exist in the Prairie Pothole Region (PPR)
across U.S. and Canada. These wetlands are important to our environment
as they can provide flood control and cool the temperature, but they are
poorly represented in previous land surface model studies. In this
study, we updated a dynamic wetland module in the Noah-MP land surface
model to reasonably estimate wetland extent and seasonal variation in
the PPR. This wetland module shows significant impacts to the surface
energy and water balance and, hence, regional temperature. The results
show that wetland features would effectively cool the air temperature
1~3 in summer, especially for regions with high
wetland coverage. The implication of this study is very useful for
wetland conservation agencies and climate scientists, as this cooling
effect could potentially mitigate heat stress under climate change.
1 Introduction
Wetlands are important and unique ecosystems that play vital roles in
Earth’s ecosystem balance and biodiversity. Although wetlands occupy a
small portion of the global land surface (~6%), they
store about one third of terrestrial carbon (Lehner and D\(\ddot{o}\)ll,
2004; Mitra et al., 2005; Mitsch and Gosselink, 2007). Moreover, due to
their unique productivity, wetlands support a wide variety of plants,
birds, and amphibians, and are areas of high biodiversity (The Ramsar
Convention, 2007). Wetlands are natural reservoirs to prevent flooding,
especially in high latitude and mountainous regions (Hayashi et al.,
2016; Pattison-Williams et al., 2018). After springtime snowmelt or
heavy rainfall, surface runoff can be stored in wetlands, effectively
reducing the peak flow and delaying the peak time of flooding, hence,
mitigating flooding impacts. From a climate regulation perspective, the
presence of surface water and the moisture of wetland soils can
effectively store surface energy and favor energy partitioning to latent
heat flux over sensible heat. Specifically, greater partitioning of
latent heat flux over sensible heat flux in wetland water bodies
decreases summer temperature (Bonan, 1995) and reduces daily air
temperature variability (Hostetler et al. 1993). This land-atmosphere
interaction is analogous to the soil moisture-temperature feedback
(Seneviratnes et al., 2010), inducing a cooling effect to surrounding
environments.
The North American Prairie Pothole Region (PPR contains millions of
small wetlands, known as “potholes”, due to its unique geology,
hydrology, and climate conditions. The retreat of continental ice sheets
over 11, 000 years ago left glacial deposition upon the landscape,
forming millions of depressions. These depressions are isolated from
large river networks and are poorly hydraulically connected. The cold
winters allow snow to accumulate over cold seasons, and springtime
runoff and seasonal rainfall are major water inputs to these wetlands.
Over the warm season, evaporation exceeds precipitation, drying surface
water and exposing the underlying soils. The persistence and storage of
wetland ponds depend on receiving seasonal rainfall and connection with
shallow groundwater. Under extremely wet conditions, strong rainfall or
sudden snowmelt increases the water level of wetlands, exceeding the
maximum capacity. Several filled wetlands will spill water to other
surrounding wetlands, a “fill-and-spill” process, and form a largely
connected wetland complex (van der Kamp and Hayashi, 2009; Mekonnen et
al., 2014; Vanderhoof et al., 2018). These complex interactions between
climate, wetland, and groundwater make it challenging to simulate in
traditional hydrological models and land surface models (LSMs).
Given their importance to global and regional environments, the need to
represent wetland physics in earth system models (ESMs) and LSMs has
emerged in recent decades. In the Community Land Model (Oleson et al.,
2008) and Noah-MP LSM (Niu et al., 2011; Yang et al., 2011), a
relationship has been established between grid cell saturated fraction
and the depth of groundwater, based on the TOPMODEL hydrological model
(Beven and Kirkby, 1979) and its application in LSMs (Famiglietti and
Wood, 1991, 1994a). This method assumes the sub-grid representation of
grid cell saturation is based on a redistribution of water table depth,
given the variation of slope and contributing areas in the grid cell. A
sub-grid saturated fraction \(F_{\text{sat}}\) is defined for the local
water table at the surface and can be used for runoff generation as
saturationed excess runoff. While this may be sufficient estimation over
a large grid resolution in many GCM models (~50-100 km),
it is not sufficiently detailed for high-resolution regional simulation
(~5-10 km). Despite its limitations, TOPMODEL- based\(F_{\text{sat}}\) is widely used in many LSMs and ESMs, particularly in
representing global wetland extents. The discrepancies in projecting
wetland extents have significant implications for modeled
CH4 emissions, as summarized in a wetland
CH4 inter-comparison modeling project (WETCHIMP, Wania
et al., 2013, Melton et al., 2013).
On the other hand, many models have incorporated surface water storage
schemes to represent the dynamics of lakes and wetlands to investigate
their impacts on the energy and water cycles. For example, Pitman (1991)
incorporated a sub-grid scheme for water surfaces and their contribution
to latent and sensible heat as the weighted average over the fraction of
water, vegetated and bare ground surface in a coarse resolution
(~2°) GCM. The Variable Infiltration Capacity model
(VIC, Liang et al., 1994) has developed a dynamic lake and wetland
scheme to study the impacts of surface water heterogeneity on energy and
water balance (Bowling and Lettenmaier, 2010). Results show that
incorporating wetlands increases the annual ET by 5% and decreases
runoff by ~ 12% in the U.S. Midwest region. Latent heat
fluxes also increase, with corresponding decreases in sensible heat
fluxes . Despite robust results in surface energy and water balance,
this research is not coupled with regional climate models, therefore
omitting the feedback from wetlands to temperature and precipitation.
The purpose of this study is to quantify the impacts of wetlands on the
surface energy and water balances, as well as their feedback to regional
climate in a high-resolution convection-permitting regional climate
model (CPRCM, Prein et al., 2015). For this purpose, we have established
three steps: (1) Develop a physical process-based parameterization of
sub-grid wetland extent and a dynamic wetland storage scheme; (2)
Explore the impacts of inclusion of this wetland parameterization on the
surface energy and water balance in offline regional land-surface
hydrology simulations using Noah-MP; (3) Investigate the interactions
between the wetland hydrological cycle and its feedback to regional
climate using a coupled Weather Research & Forecasting (WRF, Skamarock
et al., 2019) and Noah-MP model system. In particular, we want to
investigate the potential cooling effect of surface wetlands in
mitigating summertime heat stress, especially during the widespread
high-intensity heatwave of 2006 in Southern Canada and the U.S.
2 Materials and Methods
2.1 Global Inundation Extent from Multiple Satellites (GIEMS-2)
The 1993-2007 Global Inundation Extent from Multiple Satellites
(GIEMS-2) is a unique dataset that provides estimates of surface water
extent and dynamics, based on a collection of satellite observations
(https://lerma.obspm.fr/spip.php?article91&lang=en). The satellite data
are used to calculate monthly-mean inundated fractions of equal-area
grid cells (0.25°x0.25° at the equator), taking into account the
contribution of vegetation (Prigent et al., 2001, 2007, 2012; Papa et
al., 2010). Such estimates use both passive and active microwave
measurements, along with visible and near-infrared reflectance to
capitalize on their complementary strengths, to extract maximum
information about inundation characteristics, and to minimize problems
related to one instrument only. The technique is globally applicable
without any tuning for particular environments. The GIEMS data have been
widely used to evaluate surface wetland extents in multiple GCM
intercomparison studies for simulating wetland extents (Wania et al.,
2012; Melton et al., 2012).
2.2 Convection-permitting regional climate simulation
Convection-permitting models (CPMs) are atmospheric models whose grid
spacing is fine enough (usually < 5-km) to permit convection
and resolve mesoscale orography (Rasmussen et al., 2011; Prein et al.,
2015; Liu et al., 2017). Long-term high-resolution climate downscaling
using CPMs provides important added value to improve precipitation
forecasts, which is critical to surface wetland hydrology, as well as
for resolving fine-scale land surface heterogeneity (Kenden et al.,
2017).
The WRF convection-permitting regional climate simulation over the
Contiguous U.S. (CONUS WRF, Liu et al., 2017) provides the opportunity
for long-term (13-year), high-resolution (4-km) land surface modeling
(Zhang et al., 2020). The CONUS WRF consists of simulations for the
current climate and for future climate using the Pseudo Global Warming
(PGW) method (Sch\(\ddot{a}\)r et al., 1996, Rasmussen et al. 2011). The
current climate simulation is a retrospective run from 2000-10-01 to
2013-10-01, forced by ERA-Interim (Dee et al., 2011) as boundary and
initial conditions. For the future simulations, a delta climate
perturbation, derived from the 19-model ensemble in the CMIP5 project
under RCP8.5 scenario at the end of the 21st century, is added upon the
ERA-Interim forcing. The future simulation represents an equivalent
13-year period at the end of the 21st century. The CONUS WRF forcing has
been used in multiple climate, hydrology, and land surface studies
(Zhang et al., 2020; Fang et al., 2021). In this study, we use CONUS WRF
forcing in the PPR for offline land-surface model regional simulations
to study the impacts of incorporating a surface wetland scheme on the
regional energy and water balance in the PPR.
2.3 Application of TOPMODEL in LSMs
TOPMODEL (TOPography based hydrological MODEL) is a rainfall-runoff
model that uses topography data to reflect dynamic process response in
downslope hydrology, especially in runoff generation on variable
contributing areas (Beven and Kirkby, 1979; Beven et al., 2020). Its
basic assumption is that the runoff generation response to steady state
rainfall is proportional to the spatial variation of moisture content in
a drainage basin and can be characterized by its topographic variation,
characterized by digital topography analysis. In the model, a
topographic index is defined, \(\Lambda=ln(\frac{a}{\text{tanβ}})\),
where \(a\) is the area draining through a point from upslope andtanβ is the local slope angle. High index values are likely
to saturate first, hence, they indicate potential subsurface or surface
contributing areas (Beven, 1997).
The simplicity of the model comes from the assumption that all the
points of the same value of the index respond similarly in the
catchment. Therefore, it is not necessary to calculate all the points in
a catchment, but rather to integrate the hydrologic response of each
interval of index values in a representative distribution function. At
steady state, a critical threshold value for the local topographic index
(\(\Lambda_{\text{cri}}\)) can be obtained when local water table depth
is at the surface, compared to the grid cell mean water table depth.
Hence, a subgrid fraction \(F_{\text{sat}}\) can be defined by
integrating the topographic index interval from this critical value to
the maximum, following its probability distribution function:
\(F_{\text{sat}}=\int_{\Lambda_{\text{cri}}}^{\infty}{pdf(\Lambda)d\Lambda\text{\ \ }}\)(1)
This probability distribution function was assumed to be a
three-parameter gamma distribution by Sivapalan et al. (1987).
This \(F_{\text{sat}}\) fraction is an important parameter in
partitioning surface water using the saturation runoff mechanism, i.e.,
the \(F_{\text{sat}}\) portion of the surface water from rainfall and
snowmelt becomes surface runoff and the remaining (1-\(F_{\text{sat}}\))
becomes infiltration. The sub-grid \(F_{\text{sat}}\) is also critical
in controlling surface energy balance and land-atmosphere interactions
(Famiglietti and Wood, 1994a&b). In Famiglietti and Wood (1994a&b), a
Soil-Vegetation-Atmosphere Transfer Scheme (SVATS) is applied at local-,
catchment- and macro-scales to demonstrate the sub-grid soil moisture
heterogeneity in controlling both evapotranspiration and runoff. The
total evapotranspiration over the sub-grid topographic index in a grid
cell is the integration of the potential evaporation from saturated
portion to drier land surface outside the transitional region, where
evapotranspiration is restricted by active vegetation and soil moisture
(Famiglietti and Wood 1994a). This framework for incorporating TOPMODEL
into LSMs (TOPLATS) was utilized in the NASA GISS land surface model
(Stieglitz et al., 1997) and the NASA Catchment Land Surface Model
(CLSM, Koster et al., 2000; Bechtold et al., 2018) among others.
Due to its computational simplicity, the \(F_{\text{sat}}\) fraction is
also very popular to represent surface wetland extent in large-scale
global models (Gedney and Cox, 2003; Ringeval er al., 2011). The
temporal and spatial variation of \(F_{\text{sat}}\) is based on
groundwater dynamics interacting with soil moisture, simulating the
expansion and shrinkage of surface wetlands. Although the meaning of
saturation is not necessarily the same as inundation of wetland soils,
this fractional area to some degree reflects the wetness conditions in a
given grid cell, as well as its function partitioning surface water in
“saturation excess” runoff generation. Thus, it has been widely
applied in various LSMs and multiple modeling studies simulating wetland
extents (WETCHIMP, Wania et al., 2013; Melton et al., 2013).
In the Noah-MP LSM, the energy balance is calculated separately for two
sub-grid semitiles: a fractional vegetated area (\(F_{\text{veg}}\)) and
a fraction bare ground area (1-\(F_{\text{veg}}\)). In this semitile
scheme, shortwave radiation transfer is computed over the entire grid,
while longwave radiation, sensible and latent heat flux, and ground heat
flux are computed separately over these two tiles. As such, these two
tiles in a Noah-MP grid neglect the large extent and seasonal
variability of open-water wetlands. The total latent (LH) and sensible
heat (SH) of these two semitiles are aggregated in a weighted function:
\(LH=F_{\text{veg}}\left(\text{LE}_{\text{gv}}+\text{LE}_{v}\right)+(1-F_{\text{veg}})\text{LE}_{\text{gb}}\)(2)
\(SH=F_{\text{veg}}\left(\text{SH}_{\text{gv}}+\text{SH}_{v}\right)+(1-F_{\text{veg}})\text{SH}_{\text{gb}}\)(3)
Where the subscript v represents the vegetation canopy, gvis ground under canopy and gb is the bare ground flux.
Additionally, the TOPMODEL-based runoff generation model is utilized for
surface water partitioning: \(F_{\text{sat}}\) portion of the surface
available water (\(Q_{\text{insurf}}\)) from snowmelt or precipitation
becomes surface runoff (\(R_{\text{srf}}\)) and (1-\(F_{\text{sat}}\))
portion becomes infiltration (\(Q_{\text{infil}}\)). In Niu and Yang
(2005), the probability distribution function of \(F_{\text{sat}}\) in
equation (1) is replaced by an exponential function of the water table
depth (\(Z_{\nabla}\), equation (6)) and has been utilized in both CLM
(Oleson et al., 2008) and Noah-MP LSM (Niu et al., 2011; Yang et al.,
2011). \(F_{\text{satmx}}\) is the maximum saturated fraction in a grid
cell derived from digital elevation model (DEM).
\(R_{\text{srf}}=Q_{\text{insurf}}*F_{\text{sat}}\) (4)
\(Q_{\text{infil}}=Q_{\text{insurf}}*\left(1-F_{\text{sat}}\right)\)(5)
\(F_{\text{sat}}=F_{\text{satmx}}*exp(-0.5*f*(Z_{\nabla}-2))\) (6)
However, the above water balance setting does not reflect dynamic water
movement in prairie wetlands. These wetland depressions actively receive
surface water from snowmelt and rainfall, but there is no surface water
storage process in Noah-MP, so that the simulated surface runoff
component will leave the model grid. Additionally, this setting further
neglects evaporation from the wetland surface to the atmosphere and
discharge to surrounding wetlands in the fill-and-spill process.
Therefore, a dynamic surface wetland storage scheme, incorporating both
sub-grid energy and water balance, is needed to represent the complex
hydrological processes in the prairie wetland landscape and their
potential feedback to the atmosphere.
2.4 Modifying \(F_{\text{sat}}\) fraction to represent wetlands
The original TOPMODEL-based \(F_{\text{sat}}\), based on an exponential
function of water table depth, does not reasonably reflect the magnitude
and seasonal variation of wetland extent in the Prairies. Figures 1 and
2 show the spatial distribution and temporal evolution of the inundation
fraction from GIEMS and Noah-MP simulated \(F_{\text{sat}}\) fraction in
the PPR region from 2000 to 2014. It is clear that the modeled\(F_{\text{sat}}\) has underestimated the maximum extent while
overestimating the minimum extent. This is because of two reasons: (1)
the parameter \(F_{\text{satmx}}\) is a fixed value (0.38) for the
global mean; and, (2) the seasonally frozen soil and glacial till with
low hydraulic conductivity prevent direct groundwater connection with
surface water, hence the water table dynamic is not a good indicator of
surface water extent in the PPR. Detailed reasons for this discrepancy
are provided in the discussion section.