Volume 58, Issue 3 e2021WR030573
Research Article
Free Access

Cooling Effects Revealed by Modeling of Wetlands and Land-Atmosphere Interactions

Z. Zhang

Z. Zhang

School of Environment and Sustainability, University of Saskatchewan, Saskatchewan, SK, Canada

Global Institute for Water Security, University of Saskatchewan, Saskatchewan, SK, Canada

Contribution: Conceptualization, Methodology, Software, Validation, Formal analysis, Data curation, Writing - original draft, Visualization

Search for more papers by this author
F. Chen

F. Chen

Research Application Laboratory, National Center for Atmospheric Research, Boulder, CO, USA

Contribution: Conceptualization, Methodology, Supervision

Search for more papers by this author
M. Barlage

M. Barlage

Research Application Laboratory, National Center for Atmospheric Research, Boulder, CO, USA

Search for more papers by this author
L. E. Bortolotti

L. E. Bortolotti

School of Environment and Sustainability, University of Saskatchewan, Saskatchewan, SK, Canada

Institute for Wetland and Waterfowl Research, Ducks Unlimited Canada, Stonewall, MB, Canada

Contribution: Conceptualization, ​Investigation, Resources, Supervision

Search for more papers by this author
J. Famiglietti

J. Famiglietti

School of Environment and Sustainability, University of Saskatchewan, Saskatchewan, SK, Canada

Global Institute for Water Security, University of Saskatchewan, Saskatchewan, SK, Canada

Contribution: Validation, ​Investigation, Supervision

Search for more papers by this author
Z. Li

Z. Li

Global Institute for Water Security, University of Saskatchewan, Saskatchewan, SK, Canada

Contribution: ​Investigation, Resources, Writing - review & editing

Search for more papers by this author
X. Ma

X. Ma

School of Environment and Sustainability, University of Saskatchewan, Saskatchewan, SK, Canada

Global Institute for Water Security, University of Saskatchewan, Saskatchewan, SK, Canada

Contribution: Resources, Data curation

Search for more papers by this author
Y. Li

Corresponding Author

Y. Li

School of Environment and Sustainability, University of Saskatchewan, Saskatchewan, SK, Canada

Global Institute for Water Security, University of Saskatchewan, Saskatchewan, SK, Canada

Correspondence to:

Y. Li,

[email protected]

Contribution: Conceptualization, Methodology, Writing - review & editing, Supervision, Project administration, Funding acquisition

Search for more papers by this author
First published: 11 February 2022
Citations: 4


Wetlands are important ecosystems—they provide vital hydrological and ecological services such as regulating floods, storing carbon, and providing wildlife habitat. The ability to simulate their spatial extents and hydrological processes is important for valuing wetlands' function. The purpose of this study is to dynamically represent the spatial extents and hydrological processes of wetlands and investigate their feedback to regional climate in the Prairie Pothole Region (PPR) of North America, where a large number of wetlands exist. In this study, we incorporated a wetland scheme into the Noah-MP land surface model with two major modifications: (a) modifying the subgrid saturation fraction for spatial wetland extent and (b) incorporating a dynamic wetland storage to simulate hydrological processes. This scheme was evaluated at a fen site in central Saskatchewan, Canada and applied regionally in the PPR with 13-year climate forcing produced by a high-resolution convection-permitting model. The differences between wetland and no-wetland simulations are significant, with increasing latent heat and evapotranspiration while suppressing sensible heat and runoff in the wetland scheme. Finally, the dynamic wetland scheme was applied in the Weather Research and Forecasting (WRF) model. The wetlands scheme not only modifies the surface energy balance but also interacts with the lower atmosphere, shallowing the planetary boundary layer height and promoting cloud formation. A cooling effect of 1–3°C in summer temperature is evident where wetlands are abundant. In particular, the wetland simulation shows reduction in the number of hot days for >10 days over the summer of 2006, when a long-lasting heatwave occurred. This research has great implications for land surface/regional climate modeling and wetland conservation, especially in mitigating extreme heatwaves under climate change.

Key Points

  • An updated dynamic wetland scheme reasonably captures spatial extents and seasonal variation in the Prairie Pothole Region

  • Implementing this wetland scheme in Noah-MP LSM shows strong impacts on surface energy and water budget

  • Wetlands' effects on regional climate are evident, especially in cooling summer temperatures and mitigating heat stress from heatwaves

Plain Language Summary

A large number of wetlands exist in the Prairie Pothole Region (PPR) across the US and Canada. These wetlands are important to our environment as they can provide flood control and wildlife habitat and may cool the temperature, but they are poorly represented in previous land surface model (LSM) studies. In this study, we updated a dynamic wetland module in the Noah-MP LSM to reasonably estimate wetland extent and seasonal variation in the PPR. This wetland module shows significant impacts to the surface environmental conditions and interactions with regional climate. The results show that wetlands would effectively cool the air temperature 1–3 °C 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 & Döll, 2004; Mitra et al., 2005; Mitsch & Gosselink, 2007). Their unique productivity supports a wide variety of plants, birds, and amphibians (Gardner & Connolly, 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.

In particular, wetlands may influence the regional climate through changing the partition of turbulent energy fluxes of sensible and latent heat. These land-atmosphere interactions are analogous to soil moisture-temperature and the soil moisture-precipitation feedbacks (Seneviratnes et al., 2010). Greater partitioning of latent heat flux over sensible heat flux in wetlands has been shown to induce a cooling effect on summer temperature (Bonan, 1995) and reduce daily air temperature variability (Hostetler et al., 1993; Houspanossian et al., 2018). Wetlands also provide a moisture source for the formation of clouds, reducing solar radiation and atmospheric upwards motion, and thus resulting in a shallower planetary boundary layer height (PBLH; Pal et al., 2020). In African wetlands, wetland inundation may suppress local rainfall over wetlands, but increase the initiation of convective storms in the upwind direction (the “wetland breeze” effect; Taylor, 2010; Taylor et al., 2018). This suppression of precipitation has also been demonstrated in a model sensitivity study over the Canadian Prairies, where the accumulated and peak precipitation amount decreased with an increase in open-water bodies (Joshi et al., 2017).

Given their importance to global and regional environments, the need to represent wetland spatial extents and hydrological processes in Earth system models (ESMs) and land surface models (LSMs) has emerged in recent decades. From a modeling perspective, wetlands are defined as grid cells, or fractions, where the land surface is inundated or saturated. These are usually associated with a shallow water table depth. Previous studies have used prescribed wetland maps from remote sensing products, e.g., the Global Inundation Extent from Multiple Satellite (GIEMS; Prigent et al., 2007), or land survey data, such as Matthews and Fung (1987) and the Global Lake and Wetland Database (GLWD; Lehner & Döll, 2004). Other modeling studies applied parameterization schemes to estimate wetland extents. For example, in the Community Land Model (Oleson et al., 2008) and Noah-MP LSM (Niu et al., 2011; Yang et al., 2011), grid cell saturated fraction was determined by the depth of groundwater, based on the TOPMODEL hydrological model (Beven & Kirkby, 1979) and its application in LSMs (Famiglietti & Wood, 19911994a1994b). In Fan and Miguez-Macho (2011), the authors used a threshold 0.25 m of groundwater depth to determine the spatial extent of wetlands.

On the other hand, large-scale hydrological models and ESMs have incorporated surface water inundation schemes to represent the dynamics of lakes, wetlands, and floodplains to investigate their impacts on the water cycle and climate system. For example, Yamazaki et al. (2011) developed a new global river routing model, CaMa-Flood, to explicitly represent floodplain inundation dynamics, based on subgrid topographic parameters. Dadson et al. (2010) employed an overbank flow parameterization to the Joint UK Land-Environment Simulator (JULES) LSM to simulate wetland inundation dynamics and evaporation loss from the Niger inland delta. The model reproduces spatial and seasonal wetland inundation dynamics and river flow and shows the inundation scheme doubling the water vapor fluxes to the atmosphere. The Variable Infiltration Capacity (VIC) model (Liang et al., 1994) has a dynamic lake and wetland scheme to study the impacts of surface water heterogeneity on energy and water balance (Bowling & Lettenmaier, 2010). Results show that incorporating wetlands increases the annual evapotranspiration (ET) and latent heat fluxes while decreasing runoff and sensible heat fluxes in the US. Pitman (1991) incorporated a subgrid 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.

Despite progress in developing wetland schemes in LSMs, the wetland physics in the Noah-MP LSM and its coupled regional climate model (RCM), Weather Research and Forecasting (WRF, Skamarock et al., 2008), are still crude. As in many RCMs and operational weather models, wetlands are treated as a land cover type with static parameters in WRF. Moreover, there is no wetland storage in Noah-MP, so that the simulated surface runoff will leave the model grid instantly. In reality, wetland depressions actively collect surface runoff from snowmelt/rainfall and allow interaction with the atmosphere. Therefore, a dynamic wetland scheme, incorporating both subgrid energy and water balance, is needed to represent the complex hydrological processes in prairie wetlands and their potential feedback to the atmosphere.

This study was conducted over the Prairie Pothole Region (PPR) in North America where numerous small wetlands exist over a large spatial extent and play important roles in regional hydrology, ecology, and climate. There were two objectives: (a) improve the representation of wetland extents and hydrological processes in the Noah-MP LSM and (b) explore the impacts of wetlands on the regional climate, especially the wetland-temperature feedback, in a high-resolution convection-permitting regional climate model (CPRCM; Prein et al., 2015). For the above purposes, this paper is organized into the following sections: The study region, data materials, and newly proposed wetland scheme are introduced in Section 2. Three sets of simulations are conducted with the new wetland schemes in Noah-MP, including single-point simulations at a fen site, regional offline simulations, and coupled WRF (Skamarock et al., 2008) simulations in the PPR, and their results are shown in Section 3. Then, these results are discussed with other similar model studies in Section 4, focusing on the cooling effect of wetlands. Section 5 provides a summary of the study and ending remarks.

2 Data and Methods

2.1 Study Region—Prairie Pothole Region

The PPR is in the center of North America, covering about 770,000 km2 across Canada and the US. Figure 1 presents the topography of the PPR, whose boundary is outlined by a solid black line. Millions of small wetland depressions, also known as “potholes,” exist in the PPR, as a result of continental ice sheets retreating over 11,000 years ago, which left behind uneven glacial deposition upon the landscape (La Baugh et al., 1998; Pomeroy et al., 2005). These depressions are isolated from large river networks and are poorly hydraulically connected. The cold winters allow snow to accumulate over cold seasons, accounting for about one-third of annual precipitation, and snowmelt runoff is a major water input to these wetlands (Dumanski et al., 2015). 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 (Hayashi et al., 2016). Under extremely wet conditions, surface runoff by strong rainfall or sudden snowmelt exceeds the maximum capacity, spilling water to other surrounding wetlands, and form a largely connected wetland complex through the “fill-and-spill” process (Vanderhoof et al., 2018; van der Kamp & Hayashi, 2009).

Details are in the caption following the image

Topography of the Prairie Pothole Region (PPR) in North America. The black line shows the boundary of the PPR. The red triangle in Central Saskatchewan represents the location of the fen validation site. Black dots show the locations of 3,095 stations from the Global Historical Climate Network (GHCN).

A fen site (53.802 N, 104.618°W; red triangle in Figure 1) from the Boreal Ecosystem Research and Monitoring Sites (BERMS) is selected to test the new wetland scheme in this study. Observation measurements of wind, temperature, humidity, pressure, precipitation, solar, and longwave radiation were used as meteorological forcing to drive the Noah-MP model in a single-point simulation. Latent heat (LH) and sensible heat (SH) fluxes measured by eddy covariance system are used to evaluate single-point model results. Moreover, to evaluate the wetlands' impacts on regional climate, in situ station measurements of daily temperature and precipitation data, in total 3,095 stations, from the Global Historical Climate Network (GHCN) have been obtained. Their locations are shown in black dots in Figure 1.

2.2 Data

In this study, the Global Inundation Extent from Multiple Satellites-2 (GIEMS-2; Prigent et al., 2020) data set is used to prescribe the wetland spatial extents and seasonal dynamics in the PPR. The GIEMS-2 data set is an extension of the unique GIEMS data set which uses a collection of satellite sensors to provide estimates of surface water extent and dynamics at the global scale (Prigent et al., 20072012). 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 GIEMS-2 data provide monthly mean inundated fractions of equal-area grid cells (0.25° × 0.25° at the equator) from 1992 to 2015, covering 24 years of global inundation dynamics. The GIEMS and GIEMS-2 remote sensing products have been evaluated extensively over the globe (Papa et al., 2010) and used to evaluate simulated wetland fraction in ESM intercomparison studies (Melton et al., 2013; Ringeval et al., 2012). Figure 2 presents the maximum inundation extent, seasonality (month of inundation), and month of maximum inundation from GIEMS-2 in PPR.

Details are in the caption following the image

Spatial distribution of (a) the maximum inundation extent, (b) seasonality (the number of months of inundation), and (c) month of maximum inundation in the Prairie Pothole Region (PPR; black outline) from the Global Inundation Extent from Multiple Satellites-2 (GIEMS-2) data.

A CPRCM simulation over the Contiguous U.S. (CONUS WRF; Liu et al., 2017) is used to provide long-term (13 year) high-resolution (4 km) meteorological forcing for regional offline simulations. Convection-permitting models (CPMs) are atmospheric models whose grid spacing is fine enough (usually <5 km) to permit convection and resolve mesoscale orography (Liu et al., 2017; Prein et al., 2015; Rasmussen et al., 2011). 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 (Kendon et al., 2017). The CONUS WRF data have been extensively evaluated and applied in multiple climate, hydrology, and land surface studies (Fang & Pomeroy, 2021; Zhang et al., 201820202021).

To evaluate the offline Noah-MP simulation at the regional scale, two remote sensing data sets from MODIS Terra satellite are used, including land surface temperature (LST, MOD11) and ET (MOD16). The Terra satellite passes above the PPR twice a day at about 4 and 18 hr UTC, representing the LST for nighttime and daytime, respectively. The MODIS satellite data sets were obtained from NASA Earthdata Search Engine (https://search.earthdata.nasa.gov/search). The Gravity Recovery and Climate Experiment (GRACE) satellite provides terrestrial water storage (TWS) for global coverage from 2002 to 2017, at monthly interval for 1urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0001 × 1urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0002 resolution. The GRACE TWS data were used to evaluate the Noah-MP-simulated water storage for the whole wetland-soil-groundwater column in the PPR.

2.3 Surface Water and Energy Partition Scheme in Noah-MP LSM

The Noah-MP LSM adopts a runoff scheme to estimate a subgrid saturated fraction and surface water partition based on the TOPMODEL (TOPography based hydrological MODEL; Beven & Kirkby, 1979; Beven et al., 2021). This method assumes the subgrid representation of grid cell saturation, urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0003, is based on a redistribution of water table depth, given the topographic variations in the grid cell. The urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0004 fraction is an important parameter in partitioning surface water using the saturation runoff mechanism and was first integrated into a Soil-Vegetation-Atmosphere Transfer Scheme (SVATS) at local-scale, catchment-scale, and large-scale model by Famiglietti and Wood (1994a1994b). In Noah-MP, the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0005 portion of the available surface water from rainfall and snowmelt (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0006) becomes surface runoff (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0007), which is a loss term leaving the grid cell, and the remaining (1 − urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0008) portion becomes infiltration (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0009). In Niu and Yang (2005), urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0010 is estimated by an exponential function of the water table depth (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0011, Equation 3) and has been utilized in the Noah-MP LSM (Niu et al., 2011; Yang et al., 2011). urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0012 is the maximum saturated fraction in a grid cell derived from digital elevation model (DEM)

This TOPMODEL-based urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0016 framework is also widely used in the NASA GISS land surface model (Stieglitz et al., 1997) and the NASA Catchment Land Surface Model (CLSM; Bechtold et al., 2019; Koster et al., 2000).

The energy balance in Noah-MP is calculated separately for two subgrid semitiles: a fractional vegetated area (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0017) and a fraction bare-ground area (1 − urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0018). 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 LH and SH of these two semitiles are aggregated in a weighted function
where the subscript v represents the vegetation canopy, gv is ground under canopy, and gb is the bare-ground flux.

However, the above water and energy 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 the latent heat flux/evaporation contribution from the wetland surface to the atmosphere. Therefore, the deficiency of current TOPMODEL-based urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0021 parameterization in estimating surface saturation extents will be demonstrated and an updated method will be proposed in the next section.

2.4 Modifying Fsat Fraction to Represent Wetlands

The original TOPMODEL-based urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0023, based on an exponential function of water table depth, does not reasonably reflect the magnitude and seasonal variation of wetland extent in the Prairies. Figure 3 shows the temporal evolution of the inundation fraction from GIEMS and Noah-MP simulated urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0024 fraction in the PPR from 2000 to 2014. It is clear that the modeled urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0025 has underestimated the maximum extent while overestimating the minimum extent. This is for two reasons: (a) the parameter urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0026 is a fixed value (0.38) for the global mean and (b) 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 Section 4.

Details are in the caption following the image

Temporal evolution of the inundation fraction from Global Inundation Extent from Multiple Satellite (GIEMS) and Noah-MP modeled urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0027 in the Prairie Pothole Region (PPR; regional average for urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0028).

Therefore, we propose a new formula for the saturated fraction urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0029, based on the first layer soil saturation, instead of water table depth

The first layer soil moisture (SH2O) responds more rapidly to surface hydrological processes, such as snowmelt infiltration and ET, than groundwater level. urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0031 is determined by the maximum saturated fraction (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0032) and a relative soil moisture saturation condition, normalized by the soil moisture wilting point (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0033) and field capacity (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0034). This method assumes the mean soil moisture saturation in the first layer soil can empirically represent spatial heterogeneity of soil saturation at the subgrid scale.

2.5 Implementing the Surface Wetland Storage Scheme

In this study, we incorporate a subgrid bucket-style surface water storage scheme to represent three important hydrological processes in Prairie Pothole wetlands: (a) The surface runoff from snowmelt and rainfall becomes the inflow to surface water storage (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0035). The water in surface wetlands evaporates to the atmosphere at the potential rate, calculated by the Priestley-Tayer equation (9). The outflow is a result of total water exceeding the maximum water storage (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0036), representing the “fill-and-spill” process. Note this wetland storage scheme is not connected to other wetland storage or a river network, so that the outflow term will leave the grid point and is lost to the water balance, similar to the runoff term in the default Noah-MP. The change of surface water storage urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0037 is calculated by the net balance of inflow, evaporation, and outflow in Equation 11. The contribution to the latent heat flux is calculated as a weighted average over all three subgrid types in Equation 12, similar to the treatment in Pitman (1991). The sensible heat flux is calculated as the residual term from the energy balance equation

Figure 4 illustrates the difference between the default Noah-MP and the new surface wetland scheme in this study. The left-hand side shows the default Noah-MP surface runoff scheme based on the TOPMODEL saturation-excess concept. The inflow from rain and snowmelt (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0044) will be partitioned into infiltration (in the 1 − urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0045 portion), which enters soil moisture, and to surface runoff (in the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0046 portion), which eventually leaves the grid cell. The right-hand side shows the two modifications in our study: (a) the modified urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0047 parameterization based on first layer soil saturation and (b) creating a surface water storage urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0048 representing surface wetland dynamics. The urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0049 portion of the inflow will now be collected within the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0050 storage and evaporate to the atmosphere with a weighted function. The water amount exceeding the maximum capacity will become the outflow from the wetland (also referred to as the new runoff term, urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0051).

Details are in the caption following the image

Simple diagram demonstrating the modifications in this study, which includes the modification of surface saturated fraction and incorporating a surface wetland storage scheme in the Noah-MP land surface model (LSM).

2.6 Simulation Design

Three sets of numerical simulations are conducted to study the impacts of representing wetlands on the simulated energy and water balance in the Noah-MP LSM, as well as feedback to the regional climate in the coupled WRF model. A summary of these three simulations is in Table 1.

Table 1. Summary of the Three Sets of Simulations Conducted in This Study
Simulation design Location Period Purpose
Single-point Noah-MP Fen site, SK 01 January 2003 to 31 December 2010 Examine the sensitivity of urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0052 formula and different levels of storage
Offline regional Noah-MP PPR region 01 October 2000 to 01 October 2013 Incorporate spatially varied urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0053 and urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0054 parameters in the PPR
Coupled regional WRF PPR region 2006, summer from Apr to Aug Conduct coupled WRF-Noah-MP-wetland simulation and study the feedback to temperature, driven by ERA-Interim reanalysis data (Dee et al., 2011)

The first set of simulations was conducted at the fen site in central Saskatchewan. The meteorological forcings were wind, temperature, humidity, pressure, precipitation, solar, and longwave radiation from a tower measurement. The purpose of this simulation was to evaluate the improved estimation on the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0055 fraction and ET and explore the sensitivity of maximum storage (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0056) in the wetland scheme. A variety of urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0057 levels were selected to demonstrate the impacts of different wetland capacities on simulated energy/water balance at the fen site.

The second set of simulations was at the regional scale in the PPR, driven by the 4-km WRF regional climate simulation (CONUS WRF; Liu et al., 2017). The purpose of this offline simulation was to investigate the wetland scheme over the PPR, focusing on its impacts on energy and water balance at a regional scale. In this regional simulation, we constrain the maximum urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0058 by the GIEMS-2 data and the maximum surface water storage capacity urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0059 by the 90-m DEM (MERIT; Yamazaki et al., 2017, http://hydro.iis.utokyo.ac.jp/∼yamadai/MERIT_DEM/). Figure 5 presents the spatial map of urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0060 from GIEMS-2 data and urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0061 derived from the MERIT 90-m DEM and aggregated to a 4-km resolution grid
urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0063 represents the 90-m elevation and urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0064 is the mean elevation for a 4-km grid, such that urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0065 represents the collective topographical variation in the depressional area from 90-m DEM and aggregated into the 4-km grid. The high urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0066 regions are located in the Northeast part of the domain, near Lake Winnipeg in Manitoba and the Red River Valley. These regions also correspond with the low urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0067 regions.
Details are in the caption following the image

Map of maximum saturation (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0068) and wetland storage capacity (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0069) in the Prairie Pothole Region, derived from the Global Inundation Extent from Multiple Satellites-2 (GIEMS-2) product and MERIT 90-m digital elevation model (DEM), respectively.

The third set of simulations was the coupled WRF regional climate simulation for the summer of 2006. This was to investigate the impacts of wetland dynamics on regional climate, in particular under a high-resolution convection-permitting configuration. It is noteworthy that in the summer of 2006, an intense and prolonged heatwave occurred in the Central US and Southern Canada from mid-July to early August. Two simulations were conducted: the default simulation (DEF) uses the TOPMODEL-based runoff scheme (Niu et al., 2005) and the wetland simulation (WS) uses the updated wetland scheme in this study. Moreover, the model sensitivity to two groundwater schemes was also investigated for the Simple Groundwater Scheme (SIMGW; Niu et al., 2007) and the shallow groundwater scheme from Miguez-Macho and Fan (MMF-GW; Miguez-Macho et al., 2007). Please see Supporting Information S1 for detailed results for these two GW schemes.

3 Results

3.1 Implementation and Sensitivity Tests on a Single-Point LSM

We first performed a single-point LSM simulation at the fen site in Central Saskatchewan. A sensitivity test was performed with the updated wetland scheme with various storage capacities (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0070 = 0, 5, 50, 500 mm). Figure 5 shows the subgrid urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0071 fraction, energy, and water balance at the fen site simulated by Noah-MP, evaluated against the GIEMS-2 time series and in situ measurement. Due to the scale difference between GIEMS-2 data and the fen site observations, we selected the closest grid point from the GIEMS-2 data and surrounding eight grid points to the fen site for this comparison. In Figure 6a, the default urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0072 formula using the exponential function of the water table depth fails to represent the large magnitude and strong seasonal variation, as shown by the GIEMS-2 data. The modified formula using the first layer of soil moisture improves both the magnitude and seasonal cycle of the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0073, while different capacity levels have little influence on the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0074. By increasing the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0075 fraction and incorporating different storage capacities, the modeled ET increases, accounting for the evaporation contribution from wetland surface water (Figure 6c). Correspondingly, the modeled sensible heat fluxes decrease as water storage capacity increases (Figure 6d). As a result, the modeled Bowen ratio (SH/LH) also decreases, as more energy was partitioned into latent heat fluxes over sensible heat fluxes (Figure 6b).

Details are in the caption following the image

Single-point simulation at the fen site. (a) Surface saturated fraction from default (DEF) and wetland scheme with different capacities (WS = 0, 5, 50, 500 mm) and GIEMS-2 inundation extent; (b) mean Bowen ratio from observation, default model, and different WS simulations with one standard deviation error bar (c) scatter plot of monthly evapotranspiration (ET) from DEF and WS models against observation; (d) same as (c) but for sensible heat (SH). A black 1:1 line is shown in both (c) and (d) for reference.

Given different capacity levels, WS = 0 demonstrates the driest case with the smallest ET, highest SH, and largest Bowen ratio. This is because even with the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0076 increasing, a larger amount of surface water is partitioned into runoff, but not collected by the wetland (WS = 0), neglecting the surface water evaporation. By increasing the storage capacity, ET and latent heat fluxes are enhanced while sensible heat fluxes are suppressed (WS = 5 and WS = 50). With large storage capacities (WS = 50 and WS = 500), the summer evaporation demand cannot sufficiently dry out all water from the wetland, so the contribution to ET was the same for these two storage levels. Compared to the observations from the fen site, the WS = 5 simulation provided the best estimate of ET, and although slightly underestimating SH, had the best estimation of Bowen ratio for partitioning surface energy at the wetland surface.

3.2 Offline Simulation Over the PPR

Two 13-year offline Noah-MP simulations were conducted: one with the default setting and one with the new wetland scheme. Three aspects of the model simulations were evaluated, including the TWS against the GRACE satellite data, the LST and ET against the MODIS Terra satellite data. Figure 7 shows the TWS trend from the GRACE satellite and Noah-MP model in the PPR. Over the 15-year GRACE period, the PPR has experienced an increasing trend of around 6–20 mm/year (Figure 7a). This increasing trend is also shown by the time series in Figure 7b, together with results from two Noah-MP simulations. The annual cycle of the TWS anomalies is shown in Figure 7c. Both Noah-MP simulations reasonably capture the increasing trend and annual cycle of the TWS. However, the DEF simulation shows smaller seasonal variation than GRACE TWS. On the other hand, the WS simulation produces higher TWS from spring to summer than DEF, which matches better with the GRACE data. This increase in TWS can be attributed to the implementation of the wetland scheme, which collects inflow from snowmelt and rainfall, increasing the surface water storage, while decreasing surface runoff.

Details are in the caption following the image

(a) Spatial distribution of the anomaly trend for Gravity Recovery and Climate Experiment (GRACE) Terrestrial Water Storage (TWS) in the Prairie Pothole Region; (b) time series of TWS anomalies from GRACE, default (DEF) and wetland scheme (WS) simulations; (c) annual cycle of TWS anomalies.

Figure 8 demonstrates the effect of the wetland scheme on monthly ET compared to DEF and their evaluation against the MODIS Terra satellite over the PPR. The domain spatial plots are 13-year averages for May, June, July, and August. The scatter plot is for the spring and summer months for the whole 13-year period (78 months). The increased ET from WS simulation mostly concentrates on the Northeast domain, where urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0077 are the highest (Figure 5), by about 20–40 mm/mon. As compared to the MODIS ET data, both DEF and WS simulations present overestimation in spring months (MAM) and underestimation in summer months (JJA). The increased ET in WS simulation improves the underestimation of ET in summer months compared to DEF, while also contributes to much overestimation in spring months. Overall, the WS simulation is more comparable to (77.12 mm/mon) the MODIS ET data (80.01 mm/mon) than the DEF simulation (72.17 mm/mon).

Details are in the caption following the image

Differences in ET from wetland scheme (WS) and default (DEF) simulations (WS-DEF) from May to August (13-year average) and their scatter plot against the MODIS evapotranspiration (ET) data in spring and summer months. Two linear regression lines are fitted for the DEF (blue) and WS (red) simulation and a 1:1 line (black).

Figure 9 shows the bar plot for regional average LST at daytime and nighttime from May to August in the PPR. The DEF simulation shows substantially warmer LST than the MODIS data, especially in July by >6°C in the daytime. The enhanced ET and suppressed SH effect from the wetland scheme creates a cooling effect on LST for both daytime and nighttime in all months compared to DEF. The cooling effect is stronger in the daytime (∼3 °C) than the nighttime (∼1°C), contributing to reduced warm biases in the WS simulation (day: 0.78°C; night: 2.16°C) than in DEF simulation (day: 3.57°C; night: 2.69°C) relative to MODIS.

Details are in the caption following the image

Bar plot of land surface temperature (LST) from MODIS data, default (DEF), and wetland scheme (WS) simulations from May to August in (a) daytime and (b) nighttime.

Details are in the caption following the image

Monthly temperature biases from default (DEF) and wetland scheme (WS) simulations against GHCN station observation, and the cooling effect (WS-DEF) in the summer (May-August) of 2006.

3.3 Regional Climate Simulation With Coupled Wetland Dynamics

To study the feedback from wetlands to regional climate, we performed two coupled WRF-WSs for the summer of 2006. Our analysis focused on wetlands' cooling effect on temperature from May to August, especially in 2006 when an intense summer heatwave occurred from mid-July to early August in the Central US and Southern Canada.

Figure 10 shows the monthly temperature biases from two simulations, and the cooling effect induced by the WS scheme in 2006. In the DEF simulation, it is clear that a warm bias exists in the southern part of the domain, ranging from >4°C in the Central US to 1°C in the Western Canadian Prairies. This warm bias is stronger in July and August. The WS simulation shows a significant cooling effect in the Northeast portion of the domain, where the saturated fraction is high. The cooling in temperature ranges from <1°C in May to about 1–2°C in July. This cooling signal is evident in high-urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0078 regions from May to August.

In the summer of 2006, a record-breaking heatwave hit much of the US and Southern Canada. The extreme heat conditions can be represented by the number of “hot days” during the summer, with the daily maximum temperature (Tmax) exceeding the 90th percentile (TX90) of the 30-year climatology. We calculated the number of hot days from May to August in 2006 from station observations and two simulations and the results are presented in Figure 11. Through these four months, the hottest region is in the southeast of the domain in South Dakota, Nebraska, and Wyoming—with >30 hot days—while in the Canadian Prairies, there are about 10–20 hot days. The DEF simulation demonstrates significant overestimation of hot days in the US Midwest and Southeast PPR, with 8–30 days more than in observation data. The WS simulation shows that wetlands could effectively reduce the hot days by about 10–15 days in the entire domain, and it reduces the overestimation of hot days in the southern domain while overcooling in the Canadian Prairies. Two regions, including southern Manitoba and the area between Nebraska and Iowa, receive greater impacts from wetlands. The time series of regional average Tmax also shows a consistent cooling of 1–2 °C through the summer period. These results demonstrate the important role of wetlands in mitigating climate change, especially in extreme heat events.

Details are in the caption following the image

Number of hot days in default (DEF), wetland scheme (WS) simulations, and the reduction in hot days from WS to DEF, as well as the time series of Tmax in the summer of 2006 from these two simulations and observation data. OBS is from the station temperature from GHCN network and TX90 represents the 90th percentile threshold to define extreme hot days.

Figure 12 shows the differences in model-simulated cloud fraction (WS-DEF) for 2006, two cross-sections are also provided at 52urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0079N and −95urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0080W. Through the summer of 2006, increased cloud fraction by the wetland scheme is evident for up to 0.1 in the Northeast of the PPR domain and is strongest in May. Vertically, this enhanced cloud fraction emerged below 1,000 m in the lower troposphere. Conversely, the wetland scheme produces less cloud in the middle troposphere, roughly 7,000 m in height. These results suggest that incorporating wetland storage in the WS simulation not only modifies surface energy and water balance, but also impacts cloud formation in the lower troposphere.

Details are in the caption following the image

Differences in cloud fraction from wetland scheme-default in May, June, July, and August in 2006. Two vertical cross-sections of cloud fraction are shown at 52urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0081N and −95urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0082W. Upper, middle, and lower roughly correspond to 15,000, 7,000, and 1,000 m in height.

Figure 13 shows the differences in midday boundary layer height (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0083PBLH) and soil moisture (urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0084SM) from the two simulations (WS-DEF). There is a significant negative correlation between urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0085PBLH and urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0086SM that is more evident in July and August. The WS simulation clearly reduces the midday PBLH occurring in these four months in 2006, ranging up to 800 m shallower than the DEF simulation. The greater reductions in PBLH are associated with wetter conditions in SM. This analysis further adds evidence to the wetland scheme modifying surface energy partitioning, suppressing boundary layer height at midday. On the other hand, the differences in urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0087SM between the two simulations are not as obvious as the differences in urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0088PBLH. Our study also found that precipitation differences in these two simulations are spatially heterogenous, with ∼2 mm/mon more precipitation in the WS regional average. This change is not large enough to conclude there is a positive SM-precipitation feedback for the region due to high variability of precipitation (see Supporting Information S1).

Details are in the caption following the image

Scatter plot of urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0089PBLH and urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0090SM are shown from the differences between wetland scheme (WS) and default (DEF) simulation for May, June, July, and August in 2006. Color represents the density of samples, red means data samples are more converged and blue means data samples are sparse. Linear repressions were performed with Pearson's correlation coefficient and Student's t-test P-value in the top right corner for each month.

4 Discussion

Reasonable representations of subgrid saturated fraction for wetland extents and runoff generation for dynamic water storage are challenging in light of data scarcity, coarse model resolution, and insufficient understanding of the physical processes (Ringeval et al., 2012). Traditional TOPMODEL-based parameterization fails to represent highly variable urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0091 seasonal cycles, as we showed in Section 2.2. Here, we provide three possible reasons for the discrepancy between TOPMODEL urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0092 and the inundation dynamics from GIEMS-2 data. (a) The underlying assumption of the TOPMODEL method requires “steady state” precipitation and soil moisture heterogeneity, which is more likely in wet, relatively shallow soils on moderate slopes (Beven & Kirkby, 1979, 2021). However, this is not the case in the PPR, where the climate is usually semiarid and the large-scale topography is flat with small-scale variation. (b) Another possible reason for this discrepancy is that the TOPMODEL method calculates a critical topographic index value when the local water table is at the surface. However, in the PPR, frozen soils in wintertime prevent interaction between the soil moisture and groundwater (Ireson et al., 2013). (c) The inadequacy stems from the scale differences between regional-scale LSM and the catchment-scale hydrological study, where TOPMODEL was originally developed.

In our modification of the urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0093 formulation, we used the first layer of soil saturation to indicate the subgrid saturation fraction. This method assumes the grid cell mean soil moisture saturation can be translated into a spatial fraction for surface saturation, which further plays an important role in the saturation runoff mechanism. This positive correlation between soil moisture and surface runoff is highlighted in Ghajarnia et al. (20202021), in which the authors studied this close covariation from multicatchment data across Europe and the globe, respectively. In these studies, the authors found that there is a strong correlation between soil moisture and runoff exhibited in independent observations and reanalysis data, but that fails to manifest in ESM data. Moreover, we also incorporate a spatially varied maximum urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0094 map from the GIEMS-2 product to replace the default global mean value (0.38) in Noah-MP and WRF. Both modifications improve the spatial heterogeneity and the temporal dynamics of wetland extents in the PPR.

The wetland scheme in this study sufficiently modified the surface energy and water partition in prairie wetlands, showing increased ET with decreased surface runoff and an increase in LH with decreased SH. This finding aligns with our expectations, as well as with previous VIC model wetland and lake simulations in the US Midwest region (Mishra et al., 2010) and a floodplain modeling study in the Niger river delta (Dadson et al., 2010).

One highlight of this study is the wetland cooling effect on atmospheric temperature. Previous studies have documented this effect in detail, specific to different wetland characteristics and dominant vegetation types (Bonan, 1995; Pitman, 1991). The cooling effect shown in the WS is not only due to modified surface energy partition, enhancing latent heat while suppressing sensible heat, but also the result of land-atmosphere interactions, involving boundary layer dynamics and cloud formation, which is analogous to the soil moisture-temperature feedback (Seneviratne et al., 2010). That inundated conditions are associated with shallow boundary layer height and smaller daily temperature range has been also demonstrated in an observational study in Mississippi and Missouri River flooding (Pal et al., 2020) and the long-lasting floods in the Pampas (Houspanossian et al., 2018). A previous study using WRF with a prescribed soil moisture threshold to indicate wetlands in the Great Plains at coarser resolution (12 km) also showed a temperature cooling effect, but the precipitation effect was negligible (Capehart et al., 2011). The wetland cooling effect, especially during extreme heatwave events, echoes a previous study in the Central US where antecedent soil moisture could effectively reduce the frequency, intensity, and duration of extreme heatwaves (Zhang et al., 2018).

The simulated warm temperature biases associated with dry soil moisture is a long-standing issue in modeling the summer climate in Central US (Cheruy et al., 2014; Liu et al., 2017). Efforts have been dedicated to improve the representation of shallow groundwater in the MMF-GW scheme (Miguez-Macho et al., 2007), in which groundwater closely connects with soil moisture and provides a moisture source to ET and precipitation recycle, hence cooling the atmosphere (Barlage et al., 2021). Our study presents an alternative solution to mitigate the warm biases in the PPR, where wetlands' contribution to ET is nonnegligible. However, the combined cooling effect of the MMF-GW scheme and the WS scheme is too strong, inducing cool biases in the Canadian Prairies (see Supporting Information S1). Moreover, the wetlands' impacts on regional precipitation in the PPR (∼2 mm/mon increase) are not as strong as the precipitation recycle shown in applying the MMF-GW scheme in the Central US (∼10 mm/mon). This may be because the PPR is further north than the Central US, which is one of the coupling “hot spots” for soil moisture-precipitation feedback (Koster et al., 2004). Other observational studies in African wetlands (Taylor, 2010; Taylor et al., 2018) found convective initiation and cloud fraction is enhanced around the wetland edges but suppressed over the wetland itself, known as the “wetland breeze” effect. Our study found an increase of lower cloud fraction but decrease of upper cloud, this discrepancy could be due to the wetlands in the PPR are small in sizes and large in extents, so that it is difficult to draw clear boundaries near the edges, as compared to the several large wetlands selected in the African study.

This study approach has some limitations. For example, there is uncertainty in prescribed wetland maximum extents from GIEMS-2 data. Aires et al. (2017) compared three high-resolution global inundation data sets, including two from visible satellites at 30-m, Global Surface Water Explorer (GSWE) and Global 3 arc-second Water Body Map (G3WBM), and one downscaled product from GIEMS (GIEMS-D3) at 90 m. The advantages of the GIEMS method are that the multisensor technique minimizes limitations from single instruments, e.g., under vegetation canopy and cloudy conditions. The disadvantages of the GIEMS data set are its low original resolution (25 km) which may underperform at detecting small water bodies. Although it has been widely used in prescribing wetland extents (Ringeval et al., 2012) and analysis of wetland-precipitation feedbacks (Taylor et al., 2018) at regional and global scales, it was challenging to apply at the single-point fen site. We compared the discrepancies between inundation fraction from GIEMS-2 and GSWE (aggregated to different resolution) and showed that GIEMS-2 data may overestimate surface water extent on water-saturated soils compared to GSWE data, while their seasonality are similar (see Supporting Information S1). Although uncertainties exist for using different remote sensing product to prescribe maximum wetland extents, it would not change the main conclusions from this study. In addition, different types of wetlands have distinct characteristics due to their soil composition and dominant vegetation and, hence, exhibit a large range in partitioning of surface energy (e.g., drier in bogs and wetter in fens). In this study, wetlands are only represented as open-water storage to capture their contribution to evaporation, neglecting these detailed classifications. Moreover, it is admitted that the WS scheme in this study is insufficient to address blowing snow (Fang & Pomeroy, 20082009) and wetland fill-and-spill (Shook & Pomeroy, 2011; Shook et al., 2021) in the PPR. Previous studies showed that these processes are important to modeling wetlands' water balance and streamflow at basin scale; however, their hydrological cycle and atmosphere feedback at regional scale are unknown. Future studies are encouraged to include sophisticated hydrological processes to represent these two processes, though will need to overcome scarcity in spatial data, uncertainties in model parameters, and great computational cost.

5 Conclusions

Wetlands play a crucial role in Earth systems for their climatic and hydrological functions. However, reasonably representing the spatial extents and dynamics of small-scale wetlands has been challenging to LSMs and coupled ESMs. This is particularly important and urgent in the PPR as the wetlands are critical to the region's ecology and the hydrological conditions are complex. In this research, we developed a wetland scheme with two modifications to represent wetland dynamics in the Noah-MP LSM: (a) modification of the subgrid saturation fraction to represent spatial wetland extents based on grid cell soil moisture and (b) incorporation of a dynamic surface water storage scheme to represent the hydrological processes in wetlands. The new wetland scheme was incorporated in a single-point, offline regional simulation, and coupled WRF simulation in the PPR. The main findings are as follows:

The single-point simulation at the fen site showed that the modified subgrid urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0095 reasonably reproduces the magnitude and seasonality of surface inundation dynamics from the GIEMS-2 data, compared to the default TOPMODEL-based method. Incorporating the wetland scheme effectively modified the surface energy and water balance, enhancing latent heat fluxes and ET while suppressing sensible heat fluxes and surface runoff. This results in improved estimate of ET and the Bowen ratio but a slight underestimation of sensible heat fluxes. The modeled wetland's impacts on surface energy and water balance also depend on its maximum capacity, urn:x-wiley:00431397:media:wrcr25805:wrcr25805-math-0096, a parameter related to the shape of the wetland and its surrounding topography.

Incorporating the wetland scheme in the PPR demonstrates three improvements: (a) The simulated TWS increases from March to June, matching better results from the GRACE satellite. (b) The simulated ET also increases compared to the default simulation, reducing the underestimation of ET in summer months while overestimating ET in spring months. (c) These increases in ET and latent heat fluxes contribute to a cooling effect, reducing the warm biases in LST ∼3°C in the daytime and 1°C at nighttime.

Finally, the wetlands' feedback to regional climate was explored in the coupled WRF-Noah-MP-wetland model. The cooling effect induced by wetlands was evident in summer for about 2–3°C from May to August, significantly reducing warm biases from the default simulation. This cooling is the result of wetlands altering energy balance partitioning as well as interactions with atmosphere, shallowing the boundary layer height and promoting cloud formation. In the summer of 2006, when an extreme heatwave hit the Central US and Southern Canada, the presence of wetlands could profoundly reduce the number of extreme hot days by >10 during the summer period, effectively relieving the heat stress to human comfort.

In recent years, the tradeoffs between agriculture and wetland conservation have been a serious topic of discussion among the public, universities, and government agencies. Agricultural land expansion at the cost of wetland drainage increases the risk of emerging flooding in springtime (Dumanski et al., 2015; Pattison-Williams et al., 2018). Wetland drainage also results in increased nutrient export and carbon release to the atmosphere (Badiou et al., 2011), reducing resilience to drought and high temperature, which leads to crop failures (Hatifield, 2016). However, the loss of wetlands to agricultural, industrial, and residential land development is not confined to the PPR but rather is a common problem worldwide and requires greater attention (Gardner & Connolly, 2007; Nature Geoscience, 2021). Our results show that the presence of wetlands could be beneficial to many sectors by cooling atmospheric temperatures during heatwaves. These highlights should inspire future studies to understand wetlands' value in regional environments and the Earth system, especially those that have been neglected at the cost of human expansion.


The authors gratefully acknowledge valuable suggestions and discussions with Dr. John Pomeroy, which is constructive to improve this study. Z. Zhang, Z. Li, and Y. Li acknowledge the financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant, and Global Water Futures Program, Canada First Research Excellence Fund and Global Institute for Water Security (GIWS). Z. Zhang was funded by a Mitacs Accelerate Fellowship funded by Ducks Unlimited Canada's Institute for Wetland and Waterfowl Research. This project was supported by grants from Wildlife Habitat Canada, Bass Pro Shops Cabela's Outdoor Fund, and the Alberta NAWMP Partnership.

    Data Availability Statement

    The CONUS WRF simulation over the contiguous US (Liu et al., 2017) can be accessed at https://rda.ucar.edu/datasets/ds612.0/TS1. The simulation data in this study, including the single-point, offline, and coupled WRF simulations for the Prairie Pothole Region, can be accessed in a FAIR compliant repository at osf.io:https://osf.io/nckxy/?view_only=3fa18c1a466a46f1a414ecdaa0c24d67. The Noah-MP model is driven by the NCAR high-resolution land data assimilation system (Chen et al., 2007) and can be downloaded from https://github.com/NCAR/hrldas/. The Noah-MP LSM can be accessed from https://github.com/NCAR/noahmp. We appreciate Dr. Catherine Prigent for her help and support in this study and for providing the GIEMS-2 data for surface water inundation. The GIEMS-2 data can be obtained by personal communication with Dr. Prigent: [email protected].