Next Article in Journal
A Robust InSAR-DEM Block Adjustment Method Based on Affine and Polynomial Models for Geometric Distortion
Previous Article in Journal
A Channel-Adaptive Range-Doppler Domain Filtering Serial BAQ Algorithm and Comparative Analysis
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Revisiting the Role of SMAP Soil Moisture Retrievals in WRF-Chem Dust Emission Simulations over the Western U.S.

NSF National Center for Atmospheric Research, Boulder, CO 80301, USA
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(8), 1345; https://doi.org/10.3390/rs17081345
Submission received: 1 March 2025 / Revised: 3 April 2025 / Accepted: 7 April 2025 / Published: 10 April 2025

Abstract

:
Having good replication of the soil moisture evolution is desirable to properly simulate the dust emissions and atmospheric dust load because soil moisture increases the cohesive forces of soil particles, modulating the wind erosion threshold above which emissions occur. To reduce errors, one can use soil moisture retrievals from space-borne microwave radiometers. Here, we explore the potential of inserting soil moisture retrievals from the Soil Moisture Active Passive (SMAP) satellite into the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) to improve dust simulations. We focus our analysis on the contiguous U.S. due to the presence of important dust sources and good observational networks. Our analysis extends over the first year of SMAP retrievals (1 April 2015–31 March 2016) to cover the annual soil moisture variability and go beyond extreme events, such as dust storms, in order to provide a statistically robust characterization of the potential added value of the soil moisture retrievals. We focus on the Goddard Chemistry Aerosol Radiation and Transport (GOCART) model from the Air Force Weather Agency (GOCART-AFWA) dust emission parameterization that represents soil moisture modulations of the wind erosion threshold with a parameterization developed by fitting observations. The dust emissions are overestimated by the GOCART-AFWA parameterization and result in an overestimation of the aerosol optical depth (AOD). Sensitivity experiments show that emissions reduced to 25% in the GOCART-AFWA simulations largely reduced the AOD bias over the Southwest and lead to better agreement with the standard WRF-Chem parameterization of dust emissions (GOCART) and with observations. Comparisons of GOCART-AFWA simulations with emissions reduced to 25% with and without SMAP soil moisture insertion show added value of the retrievals, albeit small, over the dust sources. These results highlight the importance of accurate dust emission parameterizations when evaluating the impact of remotely sensed soil moisture data on numerical weather prediction models.

1. Introduction

Mineral dust is one of the most abundant atmospheric aerosols. To properly simulate its transport and dispersion and its interactions with radiation and microphysical processes, one needs to account for the emissions on the Earth’s surface. Emissions occur when the wind carries enough energy to overcome the forces that keep the particles on the ground (e.g., [1,2,3]) in what is known as the wind erosion threshold. The threshold is affected by cohesive forces between soil particles that are strengthened by soil moisture. Hence, the general effect of increasing soil moisture is to increase the wind erosion threshold.
The wind erosion threshold is affected by a range of physical processes (e.g., [2,3,4]). It is usually quantified as a friction velocity threshold. A first component is introduced to account for the effects of the particle size. This usually includes the effects both of gravity and electrostatic forces between particles (e.g., [2,5,6,7]). For example, Marticorena and Bergametti [7] removed the dependence on the Reynolds number in the formulation of Iversen and White [6] in order to rely only on the air density and particle properties. Other multiplicative factors are typically added to account for the modulations exerted by other physical processes. This is the case for soil moisture. The wind erosion threshold increases when there are increases in soil moisture content above a certain minimum amount in the top soil layer (e.g., [8,9]), as is generally found in observations [3]. The increased wind erosion threshold can be explained in terms of molecular adsorption of water on the grain surface and capillary effects due to the moisture tension that contribute to hold the soil grains together [10]. Fécan et al. [8] realized that the soil moisture value at which the friction velocity threshold increases is a function of the soil texture, and proposed to use the clay content to parameterize the increase in the erosion threshold associated with soil moisture, fitting parameters to match existing observations.
Having good replication of the soil moisture content in numerical weather prediction (NWP) models is desirable to properly simulate the modulations of the wind erosion threshold. To this end, one can use available observations to constrain the NWP models (e.g., [11]). Unfortunately, in situ soil moisture observations are not typically available in real time, and often lack sufficient density to characterize the soil moisture spatial variability. Therefore, alternate sources of soil moisture information to in situ observations are desirable. Space-borne microwave radiometers such as the Soil Moisture Active Passive (SMAP, [12]) mission routinely retrieve the soil moisture content. These retrievals, and those from other microwave radiometers, have been used to constrain the simulated soil moisture (e.g., [13,14,15,16]), focusing on the evaluation of standard meteorological variables. Recently, we examined the value of SMAP for simulations performed with the Weather Research and Forecasting (WRF) model coupled with Chemistry (WRF-Chem, [17,18]) for eight case studies of dust outbreaks within the contiguous U.S. (CONUS). We found inconclusive results regarding the benefits of the soil moisture retrievals for these extreme events, which we attributed to misrepresentations of larger-order effects in the Goddard Chemistry Aerosol Radiation and Transport (GOCART) model from the Air Force Weather Agency (GOCART-AFWA, [19]) dust emission parameterization that we used in these case studies [20].
In this work, we analyze for the first time the potential of soil moisture retrievals to improve dust emissions in a 1-year period in order to go beyond case studies of extreme blowing dust events to provide a statistically robust characterization with conclusive results regarding the added value of the retrievals. We focus on CONUS, which has important dust source regions in the deserts in the Southwest, and we use the WRF-Chem model with aerosols, including dust, represented using the GOCART module [21,22]. We analyzed two dust emission parameterizations: the original GOCART [23] and its evolution GOCART-AFWA. Major emphasis is given to the GOCART-AFWA dust emission parameterization because it uses the standard formulation of the effects of soil moisture on the emissions [8]. However, we also run simulations with the original GOCART in order to provide a reference for the amount of dust emitted since it is the default dust emission parameterization in WRF-Chem. For each scheme, we run WRF-Chem with and without direct insertion of SMAP soil moisture content retrievals to isolate the effects of soil moisture constraints. Our approach consists of direct insertion of soil moisture retrievals from SMAP to blend that information into WRF-Chem, as was performed in [20,24,25]. By performing this replacement consistently during the complete simulation period, we expect the soils and the atmosphere to adjust to the retrievals. While more sophisticated data assimilation (DA) methods exist that account for covariances between different variables, such as those used in operational NWP systems, a direct insertion approach is expected to yield larger increments to the soil moisture content in WRF-Chem than would occur with sophisticated DA systems. Thus, we expect that the approach we use here is providing an upper bound on the impact of improved soil moisture representation on the direct radiative effect in WRF-Chem.
This paper is organized as follows: Section 2 presents the in situ and remote observations used and describes the methods. The results are shown in Section 3, and the discussion is presented in Section 4. Finally, the conclusions are presented in Section 5.

2. Methods

2.1. Retrievals and In Situ Observations

We used soil moisture retrievals from SMAP to constrain some WRF-Chem experiments. More precisely, we used the SMAP SPL2SMP_E.004 product available at 9 km grid spacing. The data covers the 1-year period from 1 April 2015, when the SMAP mission started, until 31 March 2016. Focusing on the first year of SMAP operation reduces the impacts of instrument degradation in the study. Only retrievals flagged as recommended quality were used.
For model evaluation, we used data from three observational networks. The first one is the U.S. Climate Reference Network (USCRN, [26]). We selected this network because it covers CONUS relatively homogeneously, and provides hourly soil moisture averages with the first sensor at a 0.05 m depth. These data were used to evaluate the potential added value of the SMAP retrievals with respect to the soil moisture simulations from WRF-Chem. The evaluation is adequate because USCRN observations are independent of SMAP and WRF data. The second set of observations comes from Meteorological Aerodrome Reports (METARs). These are hourly reports of standard meteorological observations that we used to ensure a good replication of the atmospheric evolution, and, in particular, 10 m winds given their importance for dust emissions. The third dataset is the Aerosol Robotic Network (AERONET, [27]). We used the aerosol optical depth (AOD) at 500 nm as our primary variable to quantify the performance of the simulations and the added value of imposing the soil moisture retrievals. The original data were averaged every 15 min centered on:00, :15, :30, and :45 of every hour to match the output times from our WRF-Chem experiments. Additionally, a minimum of 1000 valid observations were required at each station to be included in the analysis.
In addition to the ground observations, we also used AOD retrievals at 550 nm from the Moderate Resolution Imaging Spectroradiometer (MODIS) instruments on board Terra and Aqua [28]. The retrievals cover the whole CONUS and are used to complement AERONET ground observations in our efforts to quantify the performance of the simulations. The MODIS level 2 data within a 15 min time window of each WRF-Chem output time point are extracted and then are co-located and compared with WRF results by bilinearly mapping WRF grids into MODIS pixels spatially.

2.2. Experimental Setup

This section describes the WRF-Chem experiments performed. In each experiment, we ran WRF version 4.2.1 with its chemistry extensions, WRF-Chem, continuously from 1 January 2015 until 31 March 2016. This provides a conservative 3-month spin-up period, since the evaluation period starts with the insertion of SMAP retrievals that became available in April 2015 (see Section 2.1). The model domain covers CONUS at 9 km grid spacing and used 45 levels with the top at 50 hPa. The model time step is 30 s, and the output was saved every 15 min. The simulations did not explicitly resolve atmospheric chemistry but accounted for aerosol processes.
Initial and lateral boundary conditions for the atmospheric and aerosol variables were obtained every 6 h from the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2, [29]). The resolution of MERRA-2 is 0.625° longitude by 0.5° latitude. Spectral nudging was applied in order to better represent the synoptic forcing. Every 3 min, we nudged the wind components, geopotential height, and temperature above model level 21 (around 1600 m). Only wave numbers below 6 and 4 were nudged in the west–east and north–south directions, respectively.
The aerosol evolution was simulated with the GOCART model implemented in WRF-Chem. The GOCART model accounts for black carbon, organic carbon, sulfates, sea salt, and dust. Both sea salt and dust are represented using size distributions. The dust aerosol has five bins with effective radii ranging from 0.73 µm to 8.0 µm. Organic carbon and black carbon are originally hydrophobic but age to hydrophilic with 2.5 days’ e-folding time. No atmospheric chemistry is explicitly resolved. This made necessary the use of gas climatologies (OH, NO 3 , and H 2 O 2 ) based on a global model to produce sulfate aerosols.
The aerosol emissions at the surface were parameterized. Hourly anthropogenic emissions were based on the U.S. Environmental Protection Agency (EPA)-provided National Emissions Inventory (NEI) for 2017 at 12 km grid spacing. The NEI emissions were mapped to our WRF-Chem domain using a mass-conserving emission preprocessor for 2017. Since the NEI emissions have day-of-the-week variability, we developed a mapping to conserve the day of the week when using NEI 2017 emissions for other years (e.g., 2018). In this mapping, the first Monday of 2017 was mapped to the first Monday of the target year and so on. Derived emissions for years other than 2017 were also adjusted via the application of EPA-reported annual state-wise trends to the NEI 2017 emissions. Emissions from open biomass burning are represented using the Fire Inventory from NCAR version 2.5 [30]. Fire emissions were distributed vertically online within the model using a plume rise parameterization [31]. Fire emissions were assumed to have a diurnal variation with a daytime peak. The sea-salt emissions were also calculated online within the model following [32]. The dust emissions, the main focus of this work, were parameterized based on either the GOCART or the GOCART-AFWA parameterization. Both parameterizations used the same dust source, or erodibility, dataset, which defines the locations that can emit dust. The erodibility ranges from 0 (no dust emissions) to 1 (maximum strength) and thus also modulates the strength of the dust emissions.
The dust emission parameterization from GOCART was used because it is the default option in WRF-Chem. The wind erosion threshold is defined in terms of the 10 m wind speed. The threshold, U t , consists of two parts, one dependent on the particle size and density [7], the threshold for dry conditions ( U t _ d r y ), and another factor, f ( θ s ) , that includes the effects of soil moisture [23], as follows:
U t = U t _ d r y f ( θ s )
f ( θ s ) = 1.2 + 0.2 log 10 θ s , if θ s < 0.5 . , otherwise .
where θ s is the degree of saturation defined as the volumetric soil moisture over the porosity of the soil, which is land use dependent. This formulation of the effects of the soil moisture, f ( θ s ) , does not depart much from unity, as is shown in the example in Figure 1 (dashed line). Hence, the modulations to the wind erosion threshold exerted by the soil moisture in the GOCART dust emission parameterization are going to be small. In addition, the values differ from the more standard approach of Fécan et al. [8] that is fitted to observations (solid line).
The dust emission parameterization from GOCART-AFWA is an evolution of GOCART and uses the Fécan et al. [8] formulation to incorporate the effects of soil moisture in the wind erosion threshold, which is why we used it in this work. The wind erosion threshold is defined in terms of the friction velocity. The threshold, U * t , has also two parts, one dealing with the threshold for dry conditions, U * t _ d r y , that is the same as GOCART’s U t _ d r y , and the other being the factor from [8], f ( θ g ) , with the effects of soil moisture, as follows:
U * t = U * t _ d r y f ( θ g )
f ( θ g ) = 1 + 1.21 ( θ g θ g ) 0.68 , if θ g > θ g . 1 , otherwise .
where θ g is the gravimetric soil moisture content that is a function of the clay content, soil porosity, and volumetric soil moisture; and θ g is the soil moisture fraction that can be absorbed before inter-particle capillary forces begin to influence soil particle detachment, and is a function of the clay content of the soil. Figure 1 shows the effects of soil moisture in the wind erosion threshold, f ( θ g ) (solid line). For this case, soil moisture values larger than around 0.07 [ m 3   m 3 ] are needed for soil moisture modulations of the wind threshold to take effect. Quickly, at around 0.11 [ m 3   m 3 ], the wind erosion threshold is doubled. The impact continues to grow at a slower rate for larger values of soil moisture, increasing the threshold by a factor of three at around 0.25 [ m 3   m 3 ]. One can understand the impact of these modulations on dust emissions, assuming a neutral atmosphere where the logarithmic wind profile applies. Under these conditions, typical of strong winds, the friction velocity is proportional to the surface winds, and therefore, any increase in the wind erosion threshold (e.g., fuel moisture modulations shown in Figure 1) will require a proportional increase in the surface winds for dust emissions to occur. For the example shown in Figure 1, increasing the soil moisture from 0.07 [ m 3   m 3 ] to around 0.11 [ m 3   m 3 ] would require doubling the wind speed threshold for dust emissions to occur. The reader is referred to [19] for more specific details of the GOCART-AFWA dust emission parameterization.
Other physical processes were also parameterized. The surface layer parameterization followed [33], while the atmospheric turbulent mixing was represented with the Mellor–Yamada–Nakanishi–Niino (MYNN) planetary boundary layer parameterization [34,35]. The longwave and shortwave radiation used the Rapid Radiative Transfer Model for Global Circulation Models (RRTMG, [36]). The microphysical processes were based on the Thompson–Eidhammer aerosol-aware microphysics [37] coupled to the GOCART model. To perform this coupling, we converted the mass mixing ratio to aerosol number, joined together the water-friendly aerosols (organic carbon, sulfates, and sea salt) and the ice-friendly aerosols (dust), and passed the information to the microphysics. Once the microphysics was completed, the revised aerosol number concentration was converted back to mixing ratios assuming the same proportions as there were before entering the microphysics. Cumulus processes were simulated using the [38] parameterization. Finally, we used the Noah-MP model [39,40] to represent land-surface processes [41]. The default Noah-MP soil layers were reduced from 0.1 m, 0.3 m, 0.6 m, and 1 m to 0.05 m, 0.1 m, 0.25 m, and 0.6 m in order to have a first layer of 0.05 m, which is the region that the SMAP instrument samples.
Thirteen WRF-Chem experiments were run (Table 1). The experiments differ in the dust emission model (GOCART, GOCART-AFWA, or no dust emissions), the insertion or not of SMAP soil moisture retrievals, and, for the case of GOCART-AFWA, a constant multiplicative factor to modulate the dust emissions. For the GOCART parameterization, we ran a pair of simulations with and without SMAP soil moisture constraints (experiments 1 and 2 in Table 1). The GOCART dust parameterization has small sensitivity to the soil moisture (Figure 1), but we ran these experiments since GOCART is the default emission scheme in WRF-Chem and will serve as a reference against which to compare the dust load from GOCART-AFWA. With GOCART-AFWA, we ran five pairs of simulations, with and without SMAP, for five multiplicative factors of the dust emissions, 1.0, 0.5, 0.25, 0.1, and 0.05 (experiments 3 through 12 in Table 1). The run with a factor of 1.0 is the standard GOCART-AFWA simulation, whereas the runs with smaller factors (0.5, 0.25, 0.1, and 0.05) reduce the dust emissions (to 50%, 25%, 10%, and 5% of the standard GOCART-AFWA). These experiments are motivated by identifying an overestimation of dust production by GOCART-AFWA, as will be shown. In addition to the two GOCART experiments and 10 GOCART-AFWA experiments, we ran a final experiment using SMAP retrievals but without dust emissions to isolate the impact of the dust emissions within CONUS (experiment 13 in Table 1).
The experiments using SMAP soil moisture constraints included a direct insertion of the soil moisture retrievals. Our approach consists of replacing the top soil layer moisture content with the SMAP soil moisture content retrievals when available, as previously demonstrated [20,25]. To this end, as mentioned above, we modified the thickness of the top soil layer in the model to match the soil depth that the SMAP instrument samples, i.e., 5 cm. We admit that modifying the top soil layer without adjusting lower ones can potentially lead to imbalances in the surface fluxes. However, this method has been demonstrated before for a range of variables in NWP or land surface models [20,24,25,42,43,44,45], and it is an acceptable approach to start exploring model sensitivities before considering a more sophisticated method. More sophisticated DA methods will provide better consistency of the soil state, but the analysis increments would likely be smaller, leading to a reduced impact than would likely be obtained from direct insertion. The soil moisture constraints started on 1 April 2015 when SMAP retrievals became available and concluded at the end of the simulation on 31 March 2016. During this period, the SMAP retrievals, the SPL2SMP_E.004 product at 9 km grid spacing, for each day were interpolated to the WRF grid by assigning the SMAP data to the nearest grid point in the WRF grid, and interpolating data as a function of the distance to grid points within a radius of 2 grid points. We speculate that results are not going to be very sensitive to the interpolation method, considering that we are using the same grid spacing as the retrievals, 9 km, and there is no large variability between nearby retrievals. Then we grouped the retrievals at 00 and 12 UTC in order to replace the simulated soil moisture in the top level of the soil every 12 h. The retrievals over WRF grid cells classified as water were not used. No attempt was made to apply a bias correction to the retrievals. The evaluation period does not include a spin-up period after SMAP data were inserted because the length of this period (days) is much smaller than the evaluation period (1 year). A year-long evaluation period allows us to represent the annual evolution of soil moisture and different ranges of wind speed, to cover a wide range of conditions, in order to provide a statistically robust characterization of the impact of soil moisture insertion. The number of retrievals inserted at each location during the evaluation period is shown in Figure 2. Although the sampling of the retrievals has gaps in parts of eastern CONUS and the Pacific Northwest, the retrievals provide good sampling in the southwestern CONUS where the main dust sources are located. By comparing experiments with/without the SMAP retrievals (see Table 1), we are able to quantify the added value of SMAP retrievals to improve the dust emission in WRF-Chem, which is our primary objective.

3. Results

3.1. Added Value of SMAP Soil Moisture

A first indication of the added value of SMAP with respect to WRF-Chem is shown in Table 2, which shows the mean bias error (MBE), mean absolute error (MAE), root mean square error (RMSE), and correlation of SMAP soil moisture retrievals and the two GOCART experiments versus the USCRN observations. Only those times with SMAP retrievals available at the USCRN sites when the USCRN site also had a valid observation are selected for validation. As mentioned previously, GOCART is used because it is the default dust emission parameterization in WRF-Chem, and there are only small differences in soil moisture content compared with GOCART-AFWA because we impose the SMAP soil moisture retrievals regularly. The SMAP retrievals have a slightly negative MBE (−0.5%), paired with MAE and RMSE values of 7.2% and 9.2%, respectively. The correlation is 0.65. The results for the GOCART experiment are similar except for a larger positive MBE (2.0%). When we insert SMAP retrievals into WRF-Chem in the GOCART_SMAP experiment, we obtain the best scores, although only slightly better than the GOCART experiment or for the SMAP retrievals directly. The MBE is −0.5%; the MAE and RMSE are 6.5% and 8.3%, respectively (10% improvement with respect to SMAP); and the correlation is 0.71. The similarities between the retrieved soil moisture and GOCART_SMAP are expected. The differences are a result of evaluating the simulated soil moisture just before performing the insertion of SMAP into WRF-Chem.
A more site-specific evaluation is provided in Figure 3, which shows the scatter plot between the mean retrieved soil moisture (SMAP, Figure 3a) or mean simulated soil moisture (GOCART experiments, Figure 3b,c) and the observed mean soil moisture at the USCRN sites. For SMAP (Figure 3a), there is scatter around the diagonal line, consistent with the nearly unbiased retrievals (Table 2). The correlation coefficient is r = 0.60 . For the GOCART experiments (Figure 3b,c), the correlation is higher, r = 0.75 in both cases. This indicates a better ability of WRF-Chem in reproducing the spatial pattern of the soil moisture than SMAP. However, there is a tendency in the scatter of the GOCART experiment to be above the diagonal, indicating a positive MBE in WRF-Chem (see also Table 2). As expected by the previous results (Table 2), the GOCART_SMAP experiment shows the scatter around the diagonal line, confirming the value of SMAP to reduce a bias in the model.
The value of the retrievals is better appreciated in the spatial distribution of the MBE and MAE for SMAP retrievals and the GOCART experiment (Figure 4). SMAP retrievals have lower-magnitude MBE in western CONUS, whereas larger negative values can be seen in the central U.S. and larger positive values in the southeastern U.S. (Figure 4a). This pattern resembles the MAE (Figure 4c), which reveals an increasing gradient in the MAE from west to east. On the other hand, the GOCART experiment shows a rather systematic overestimation of the soil moisture in western CONUS and a less clear pattern in the eastern part (Figure 4b). The MAE for the GOCART experiment (Figure 4d) has a less clear spatial structure. By comparing the MAE from SMAP (Figure 4c) and the GOCART experiment (Figure 4d), it is apparent that SMAP will provide an added value over the western part of CONUS because the MAE is lower than GOCART there.
To quantify the added value of SMAP retrievals versus WRF-Chem simulations, Figure 5 shows the skill score (SS) of GOCART_SMAP with respect to the GOCART experiment, defined as
SS = 1 RMSE GOCART _ SMAP RMSE GOCART
With this definition, SS values larger than zero will indicate added value, with a value of one being a perfect match with the USCRN observations used to calculate the RMSE. In other words, SS larger than zero indicates better performance. As expected, the added value of SMAP is clearly evident in the western part of CONUS. In the eastern part, the advantage is less clear. However, the major sources of dust are located in the Southwest, the region where there is a clear added value. In the case of WRF-Chem, the dust source regions are defined by the erodibility, which is also shown in Figure 5. With the exception of one site in southeastern Colorado, all the USCRN locations over regions with erodibility larger than zero show an added value of SMAP with respect to WRF-Chem. The improvements are around 20–40%. Hence, there is potential value from inserting the SMAP soil moisture to improve the calculation of dust emissions and the dust simulation in WRF-Chem.

3.2. Near-Surface Meteorological Variables

Before analyzing the added value of SMAP in terms of dust load, we inspected the ability of the simulations to reproduce standard near-surface variables. Among these variables, near-surface wind speed is the most relevant for dust emissions. Statistics summarizing the ability of the WRF-Chem GOCART-AFWA experiment to reproduce the 10 m wind speed and wind direction are shown in Table 3. The statistics are calculated for each of the 10 regions defined by the U.S. Environmental Protection Agency (EPA). The wind speed MBE absolute value is lower than 0.8 m s 1 across all regions, and for the wind direction, the MBE is smaller than 10 degrees. The wind speed MAE and RMSE are also lower than 1 m s 1 (except for the RMSE of 1.01 m s 1 in the Mid-Atlantic region, R3). For the wind direction, the MAE and RMSE are no larger than 26° and 38°, respectively. Over the most relevant regions for dust sources, Pacific Southwest and South Central, the wind speed MAE is around 0.65 m s 1 , and the wind speed RMSE is around 0.75 m s 1 , while the wind direction MAE/RMSE is around 18°/23°. The correlation is high for wind speed and wind direction, between 0.8 and 0.9 for most of the regions. For the 2 m temperature, the statistics are also good, with the MAE and RMSE across regions ranging from 1 to 3 °C and the MBE between –3 and 0 °C. The correlation is almost 1 in all regions. The 2 m relative humidity also compares well with observations. The MAE and RMSE across regions range from 6% to 8% and 7% to 13%, respectively, and the MBE is between −5% and 4%. The correlation of the relative humidity is between 0.76 and 0.95 across the regions. The GOCART-AFWA simulation with SMAP retrievals inserted shows similar performance. Hence, WRF-Chem is able to reproduce the near-surface variables, and especially the wind, with sufficient accuracy to encourage inspecting the added value of inserting SMAP soil moisture retrievals, which is analyzed in the next section.

3.3. Aerosol Optical Depth

The evaluation against AERONET and MODIS observations will quantify the value of inserting SMAP retrievals in WRF-Chem.

3.3.1. Effects of the Dust Emission Parameterization

The MBE for the 500 nm AOD at the AERONET sites for three of our WRF-Chem simulations is shown in Figure 6. GOCART_SMAP simulations show a tendency to underestimate the AOD (Figure 6a). This is particularly evident in the eastern half of the domain. This underestimation might be a result of missing nitrate and secondary organic aerosol formation in GOCART because aerosol thermodynamics was not included in the original design of GOCART for computational reasons [21,22]. Over the Southwest, where the main dust sources are located (see Figure 5), there is no clear pattern except for a slight overestimation of the AOD. The GOCART_SMAP MBE pattern is also evident in the GOCART-AFWA_SMAP simulation, except that in this case, there is a clear overestimation in the Southwest (Figure 6b). The overestimation is a result of the dust emissions since the simulation without dust emissions, NODUST, shows a negative MBE across CONUS (Figure 6c), which means that other aerosol sources are also uncertain.
As expected, analysis of the MBE calculated with MODIS reveals the same patterns. Figure 7 shows the MBE calculated using the retrievals from the MODIS-Terra instrument for the same WRF-Chem simulations evaluated at the AERONET sites (Figure 6). Similar results are found for the MODIS-Aqua instrument (not shown in image). The wider coverage of the retrievals more clearly reveals the GOCART_SMAP positive bias in the Southwest (Figure 7a). This overestimation is larger for the GOCART-AFWA_SMAP simulation, which shows MBE values larger than 0.3 over land and even an overestimation over the ocean (Figure 7b). The NODUST simulation does not show the overestimation (Figure 7c), which points to too-strong dust emissions in the GOCART-AFWA_SMAP experiment (and to a lesser extent in GOCART_SMAP). The location of maximum overestimation, around the junction of California, Arizona, and Mexico, coincides with the maximum in the erodibility (Figure 5), which suggests that the erodibility is too high in this region and/or that some other factor in the GOCART-AFWA scheme is causing too much dust to be lofted in areas with high erodibility. The agreement between the MODIS results and the ground observations from AERONET is remarkable, which confers more robustness to the patterns identified.
In order to explore any potential added value of inserting the SMAP retrievals in WRF-Chem, one needs to reduce the systematic dust overestimation by GOCART-AFWA dust emission parameterization. To this end, we further show in Figure 8 the MBE calculated with the MODIS-Terra retrievals and the GOCART-AFWA_SMAP simulations with dust emissions reduced to 50%, 25%, 10%, and 5%. Reducing the emissions to 50% still shows an AOD overestimation in the Southwest (Figure 8a). Reducing the emissions to 25% shows an MBE pattern (Figure 8b) similar to the one obtained with the standard dust emission parameterization in WRF-Chem (GOCART_SMAP, Figure 7a). Reducing the emissions to 10% largely alleviates the overestimation (Figure 8a,c). This pattern is similar to the one obtained by reducing the dust emissions to 5% (Figure 8d). These results suggest that reducing the emissions in the GOCART-AFWA_SMAP experiment to at least 25% is necessary to obtain results similar to the default GOCART_SMAP simulation.
A better understanding of the benefits of reducing the dust emissions is shown in Figure 9, which shows the SS comparing GOCART-AFWA_SMAP AOD errors (using AERONET observations) between two dust emission factors. We use the RMSE in spite of the systematic errors (Figure 8) because we want to identify any improvement regardless if it is systematic or not. The SS comparing reducing the emissions to 50% against the standard GOCART-AFWA emissions clearly shows an added value over the Southwest (Figure 9a). This is also the case when one compares the 25% and the 50% emission experiments (Figure 9b), confirming that reducing the GOCART-AFWA emissions to at least 25% is desirable. The SS comparing the emissions reduced to 10% with the emissions reduced to 25% also shows in general added value at the locations sampled by AERONET in the Southwest, but the pattern is no longer systematic, and some sites show a negative SS, indicating a degradation (Figure 9c). This degradation becomes even more evident in the SS when we reduce the emissions to 5% and compare against the simulation with emissions set to 10% (Figure 9d).
Reproducing the previous analysis with the MODIS retrievals further confirms the results obtained with AERONET observations. The SS obtained with the MODIS-Terra retrievals is shown in Figure 10. Reducing the emissions to 50% is clearly advantageous over the Southwest, as indicated by the positive SS (Figure 10a). Reducing the emissions from 50% to 25% also shows regions with large positive SS, although some slightly negative values are also apparent in the Southwest (Figure 10b). The SS quantifying the benefit of reducing the emissions from 25% to 10% also shows regions of positive values in the Southwest (Figure 10c), but the regions with negative values have increased considerably in comparison with the benefit of the 25% emissions (Figure 10b). Finally, there appears to be very little added value as a result of reducing the emissions from 10% to 5% (Figure 10d). These results based on MODIS are in agreement with the AERONET analysis (Figure 9) and further suggest that reducing the GOCART-AFWA emissions to 25%, and perhaps even to 10%, is desirable.

3.3.2. Effects of SMAP Data Insertion

To quantify the added value of inserting SMAP retrievals, we use the GOCART-AFWA simulations with emissions reduced to 25% and 10%. To illustrate the impact of the SMAP insertion in the dust emissions, we show in Figure 11 the histogram of mean dust emissions (ratio of emissions with SMAP insertion to emissions without SMAP insertion) using data at every grid point that is a dust source for the GOCART-AFWA experiment with the emissions reduced to 25%. When inserting SMAP soil moisture retrievals, the emissions are enhanced by a factor ranging from 1 to 3 for most of the dust sources. This enhancement was expected as a result of the WRF-Chem soil moisture positive bias in the western U.S. and the tendency of inserting SMAP to alleviate it (recall Figure 4 and related discussion). Similar results are obtained for the GOCART-AFWA experiments with the emissions reduced to 10%.
To quantify the added value of the retrievals, we calculate the AOD SS with GOCART-AFWA_SMAP experiments using the GOCART-AFWA (no SMAP) experiment as a reference for the 25% and 10% emissions reductions (Figure 12). The SS calculated with AERONET observations and the emissions reduced to 25% (Figure 12a) and 10% (Figure 12b) do not show a clear positive or negative pattern over the Southwest, making it difficult to extract conclusions. The analysis of the MODIS-Terra retrievals provides more information as a result of the larger spatial coverage. The simulation with the emissions reduced to 25% appears to show overall slightly positive values in the Southwest (Figure 12c), with a less clear pattern or even negative SS for the 10% emissions (Figure 12d). The SS shows a less clear structure outside the southwestern CONUS where there are no large dust sources, and thus, a weaker signal is expected. To better analyze the SS over the dust sources, we also show the MODIS-Terra SS only over the regions with erodibility larger than zero (Figure 12e,f). The SS over the dust sources in the Southwest is positive overall with values of up to 0.2, but some regions show negative values for the emissions reduced to 25% (Figure 12e). The negative values dominate in the SS for the emissions reduced to 10% (Figure 12f). Very similar results are obtained with the MODIS-Aqua retrievals.
For a better quantification of the SS, the histogram of the SS for the MODIS retrievals calculated with the GOCART-AFWA_SMAP and GOCART-AFWA with emissions reduced to 25% and 10% is shown in Figure 13. We show the results for MODIS-Terra and MODIS-Aqua separately because the satellites Terra and Aqua have different overpass times. Terra crosses the equator at around 10:30 AM local time, and Aqua at around 01:30 PM local time. Both MODIS-Terra and MODIS-Aqua show a median SS around zero, with the MODIS-Aqua histogram a bit wider than the one from MODIS-Terra. The histograms calculated with the SS over the complete simulated region (Figure 13a,b) show a frequency of 60–70% for the values in the zero SS bin. The bin for the 5% SS is larger than the −5% SS bin, which, to a lesser extent, can also be seen for the 10% and −10% bins. These differences are larger for the SS calculated with the simulations using 25% emissions. For example, for the 25% emissions and the 5% SS bin for the Terra retrievals, there is a frequency of 17%, compared with 8% for the −5% SS bin. Aqua retrievals show similar values with a frequency of 18% for the 5% SS bin and 10% for the −5% SS bin. Results for the 10% emissions show a smaller added value because a portion of the values in the 5% SS bin are shifted to the zero bin. These results show that the distribution of the SS is asymmetrical, with more frequent values in the positive part of the distribution. The histograms of the SS calculated over the dust sources (Figure 13c,d) show a similar picture. The frequency of values in the zero SS bin is smaller than for the CONUS results (Figure 13a,b) for the 25% emissions. The same is not true for the 10% emission experiment, which shows larger frequencies in the zero SS bin. The SS distribution for the 25% emission experiment is more asymmetrical for Terra, with a frequency of 23% in the 5% SS bin and 8% in the −5% SS bin. This is also the case for Aqua retrievals, with a frequency of 23% in the 5% SS bin and 10% in the −5% SS bin. These differences are much smaller for the 10% emissions as the frequency of the positive bins is reduced and shifted to the zero SS bin. We found equivalent results for the histograms calculated with AOD values larger than the 75 th percentile at each grid point acting as a dust source, with the histograms skewed towards positive skill scores. For the emissions reduced to 25%, the frequency of the 5% bin for Terra (Aqua) is 14% (10%) versus 3% (3%) in the −5% bin. For the 10% emissions, the frequency of the 5% bin and the −5% bin is 7% and 3% for both Terra and Aqua. Hence, there is added value, albeit small, in imposing the SMAP soil moisture retrievals, provided that the GOCART-AFWA emissions are reduced to 25%.

4. Discussion

Directly inserting the SMAP retrievals in WRF-Chem shows added value in the soil moisture simulation in the Southwest, which is where the most important dust sources within CONUS are located. The improvements are about 20–40%. Hence, there is a potential benefit in assimilating the SMAP retrievals over this region. Considering that we are inserting the SMAP retrievals in the simulation every 12 h to avoid large drifts in the model, having better quality in the retrievals would be desirable to further improve soil moisture and its impacts in the dust emissions. It is also possible that the current resolution of the retrievals, 9 km, does not provide detailed-enough information on very small-scale heterogeneity in soil moisture, and increasing the resolution of the retrievals would become necessary. In this direction, it is also possible that some relevant dust outbreaks occur at shorter time scales not captured by the revisit frequency of the retrievals. If this is the case, it would be desirable to have more frequent retrievals, which may be available in the upcoming decades over CONUS from microwave instruments on board geostationary satellites. Extending the study to other regions is desirable to better characterize the value of inserting the SMAP retrievals, as the WRF-Chem performance in reproducing soil moisture and the quality of the retrievals could vary regionally.
From the modeling point of view, our analysis focuses on the GOCART-AFWA dust emission parameterization because it has a more standard representation of the soil moisture modulations of the wind erosion threshold than GOCART. Our results indicate that the emissions are overestimated in this scheme (WRF-Chem version 4.2.1). Reducing the emissions to 25% of the default emissions is necessary to have a similar dust load as the default emission scheme, GOCART, in WRF-Chem. However, the GOCART scheme does not allow for analyzing the impacts of soil moisture in dust emissions because the impact of soil moisture in the wind erosion threshold is minimal. Furthermore, reducing the emissions in GOCART-AFWA to 10% appears to produce more realistic dust loads, but it may be going to the limits of the benefits of reducing emissions in bulk (i.e., by a constant factor domain-wide).
Another modeling aspect controlling the emissions that deserves further attention is the definition of the dust sources. The dust sources are controlled by the erodibility dataset (Figure 5), which determines the locations that can emit dust and their strength. Our results suggest that the maximum values of the erodibility over CONUS appear to be too large. Hence, revisiting the definitions of the dust sources and their strength appears to be needed in WRF-Chem. It is possible that the 9 km grid spacing used in our simulations is unable to capture relevant small-scale details of the soil heterogeneity, as was also pointed out for the soil moisture above. Having temporally variable dust sources, as opposed to the current static values, may further improve results and perhaps is worth considering in future studies as well.
After improving the definition of the dust sources, it may be worth inspecting sensitivities to the soil layers. The thickness of the first layer in our study is 5 cm to match the sensible region of the SMAP retrievals, but soil moisture variability in a thinner layer near the surface may be needed for applications of dust emissions. Closely related to the soil layers is the assimilation strategy. In our study, we used direct insertion of the retrievals in the top soil layer, which, as indicated above, is the first step in understanding model sensitivities. Having better integration with deeper layers using a more comprehensive DA method should be considered in future studies.
A couple of observational aspects should be highlighted. First, it should be mentioned that we treated AOD observations as perfect, though they, of course, have uncertainties. For example, MODIS AOD retrievals may be affected by cloud contamination. This could also be the case with AERONET observations. However, we have found consistency between the model results obtained with both observational datasets. This seems to indicate that the role of observational uncertainties plays a secondary role in our evaluation. Second, in our analysis of the adequacy of the dust load, we attribute the biases to the dust emissions, implicitly assuming that the load of other aerosol species is adequate. There is some evidence supporting this interpretation as there is an overall positive bias when we suppress the dust emissions in the NODUST experiment. However, there may be smaller-magnitude impacts that are worth inspecting in future studies.

5. Conclusions

We analyzed the benefits of imposing SMAP soil moisture retrievals to improve dust emissions. We focused our study on CONUS due to the presence of important dust sources and good ground observational networks. Our analysis is composed of 1 year of simulations to cover the annual evolution of soil moisture and a range of wind speeds in order to obtain a statistically robust characterization of the impact of directly inserting the retrievals, going beyond the more typical analysis of extreme blowing dust events. Directly inserting the SMAP retrievals in WRF-Chem shows soil moisture improvements of about 20–40% in the Southwest, where the main dust sources are located. After reducing the emissions to 25% in GOCART-AFWA, this translates into added value with respect to the aerosol optical depth as a result of imposing the SMAP retrievals in WRF-Chem. The improvements are modest or even small, with around 10% larger frequencies of the 5% skill score improvements compared with the 5% skill score degradations. To further improve the value of the retrievals for dust applications, our work suggests that it is necessary to improve the accuracy of the retrievals, reduce the bulk amount of dust emitted by GOCART-AFWA, and revisit the definition and characteristics of the source regions. Other aspects of the dust emission parameterization may be exposed after these improvements. After addressing the major sources of model errors, DA methods that combine observations and simulations to go beyond a direct insertion of retrievals, resulting in smaller analysis increments but more consistent analyses of model variables, should also be considered to further contribute to improving the value of assimilating soil moisture retrievals for better representations of dust.

Author Contributions

Conceptualization, P.A.J.y.M. and R.K.; methodology, P.A.J.y.M., R.K., C.H. and J.A.L.; software, P.A.J.y.M., C.H. and R.K.; validation, P.A.J.y.M., C.H., R.K. and J.A.L.; formal analysis, P.A.J.y.M.; investigation, P.A.J.y.M., R.K., C.H. and J.A.L.; resources, P.A.J.y.M., R.K., J.A.L. and C.H.; data curation, P.A.J.y.M.; writing—original draft preparation, P.A.J.y.M.; writing—review and editing, J.A.L., R.K. and C.H.; visualization, P.A.J.y.M.; project administration, P.A.J.y.M.; funding acquisition, P.A.J.y.M. and R.K. All authors have read and agreed to the published version of the manuscript.

Funding

This material is based upon work supported by the NSF National Center for Atmospheric Research, which is a major facility sponsored by the U.S. National Science Foundation under Cooperative Agreement No. 1852977. The funding for this project for all authors was provided by NASA under Award No. 80NSSC20K1798. The WRF-Chem simulations were conducted on Cheyenne (Computational and Information Systems Laboratory 2019), which is provided by NSF NCAR’s Computational and Information Systems Laboratory (CISL, [46]), and sponsored by the U.S. National Science Foundation.

Data Availability Statement

The ground observations from AERONET, USCRN, and METAR are publicly available online, as are the SMAP soil moisture retrievals and the MODIS AOD retrievals. The MERRA-2 data used to create the initial and boundary conditions for WRF-Chem are also publicly available. The WRF-Chem source code changes described in Section 3 can be found in a public GitHub repository (https://github.com/NCAR/WRF_SMAP/tree/develop accessed on 6 April 2025). The WRF-Chem data are available from https://rda.ucar.edu/datasets/d010062 accessed on 6 April 2025.

Conflicts of Interest

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Abbreviations

The following abbreviations are used in this manuscript:
AERONETAerosol Robotic Network
AODaerosol optical depth
EPAU.S. Environmental Protection Agency
GOCARTGoddard Chemistry Aerosol Radiation and Transport
GOCART-AFWAGOCART from the Air Force Weather Agency
GOCART-AFWA_SMAPexperiment using GOCART-AFWA dust emissions and
SMAP retrievals
GOCART_SMAPexperiment using GOCART dust emissions and SMAP retrievals
NWPnumerical weather prediction
CONUScontiguous U.S.
MAEmean absolute error
MBEmean bias error
MERRA-2Modern-Era Retrospective analysis for Research and Applications,
Version 2
METARsMeteorological Aerodrome Reports
MODISModerate Resolution Imaging Spectroradiometer
NODUSTexperiment without dust emissions
NEINational Emissions Inventory
RMSEroot mean square error
SMAPSoil Moisture Active Passive
SSskill score
USCRNU.S. Climate Reference Network
EPAU.S. Environmental Protection Agency
WRFWeather Research and Forecasting
WRF-ChemWRF model coupled with Chemistry

References

  1. Bagnold, R. The Physics of Blown Sand and Desert Dunes; Chapmann and Hall: London, UK, 1941; p. 265. [Google Scholar]
  2. Shao, Y. Physics and Modelling of Wind Erosion; Springer: Berlin/Heidelberg, Germany, 2008; p. 452. [Google Scholar]
  3. Knippertz, P.; Stuut, J.B.W. Mineral Dust: A Key Player in the Earth System; Springer: Berlin/Heidelberg, Germany, 2014; p. 509. [Google Scholar]
  4. AlNasser, F.; Chehbouni, A.; Entekhabi, D. Influences of soil moisture and vegetation cover on dust emission using satellite observations. Aeolian Res. 2025, 72, 100961. [Google Scholar] [CrossRef]
  5. Iversen, J.; Pollack, J.; Greeley, R.; White, B. Saltation threshold on Mars: The effect on interparticle force, surface roughness, and low atmospheric density. Icarus 1976, 29, 381–393. [Google Scholar] [CrossRef]
  6. Iversen, J.; White, B. Saltation threshold on Earth, Mars, and Venus. Sedimentology 1982, 29, 111–119. [Google Scholar] [CrossRef]
  7. Marticorena, B.; Bergametti, G. Modeling the atmospheric dust cycle: 1. Design of a soil-dreived dust emission scheme. J. Geophys. Res. 1995, 100, 16415–16430. [Google Scholar] [CrossRef]
  8. Fécan, F.; Marticorena, B.; Bergametti, G. Parameterization of the increase of the aeolian erosion threshold wind friction velocity due to soil moisture for arid and semi-arid areas. Ann. Geophys. 1999, 17, 149–157. [Google Scholar] [CrossRef]
  9. Cornelis, W.; Gabriels, D.; Hartmann, R. A parameterisation for the threshold shear velocity to initiate deflation of dry and wet sediment. Geomorphology 2004, 59, 43–51. [Google Scholar] [CrossRef]
  10. McKenna-Neuman, C.; Nickling, W.G. A theoretical and wind tunnel investigation of the effect of capillarity water on the entrainment of sediment by wind. Can. J. Soil Sci. 1989, 69, 79–96. [Google Scholar] [CrossRef]
  11. Lin, L.F.; Pu, Z. Improving near-surface short-range weather forecasts using strongly coupled land-atmosphere data assimilation with GSI-EnKF. Mon. Weather Rev. 2021, 148, 2863–2888. [Google Scholar] [CrossRef]
  12. Entekhabi, D.; Njoku, E.; OŃeill, P.; Kellogg, K.H.; Crow, W.; Edelstein, W.; Entin, J.; Goodman, S.; Jackson, T.; Johnson, J.; et al. The Soil Moisture Active Passive (SMAP) mission. Proc. IEEE 2010, 98, 704–716. [Google Scholar] [CrossRef]
  13. Lin, L.F.; Pu, Z. Examining the impact of SMAP soil moisture retrievals on short-range weather prediction under weakly and strongly coupled data assimilation with WRF-Noah. Mon. Weather Rev. 2019, 147, 4345–4366. [Google Scholar] [CrossRef]
  14. Carrera, M.L.; Bilodeau, B.; Bélair, S.; Abrahamowicz, M.; Russell, A.; Wang, X. Assimilation of passive L-band microwave brightness temperatures in the Canadian Land Data Assimilation System: Impacts on short-range warm season numerical weather prediction. J. Hydrol. 2019, 20, 1053–1079. [Google Scholar] [CrossRef]
  15. Muñoz-Sabater, J.; Lawrence, H.; Albergel, C.; de Rosmay, P.; Isasken, L.; Mesklenburg, S.; Kerr, Y.; Drusch, M. Assimilation of SMOS brightness temperatures in the ECMWF Integrated Forecasting System. Q. J. R. Met. Soc. 2019, 145, 2524–2548. [Google Scholar] [CrossRef]
  16. Ferguson, C.R.; Agrawal, S.; Beauharnois, M.C.; Xia, G.; Burrows, D.A.; Bosart, L.F. Assimilation of satellite-derived soil moisture for improved forecasts of the Great Plains low-level jet. Mon. Weather Rev. 2021, 148, 4607–4627. [Google Scholar] [CrossRef]
  17. Grell, G.A.; Peckham, S.E.; Schmitz, R.; McKeen, S.A.; Frost, G.; Skamarock, W.C.; Eder, B. Fully coupled “’online” chemistry within the WRF model. Atmos. Environ. 2005, 39, 6957–6975. [Google Scholar] [CrossRef]
  18. Fast, J.D.; Gustafson, W.I., Jr.; Easter, R.C.; Zaveri, R.A.; Barnnard, J.C.; Chapman, E.G.; Grell, G.A. Evolution of ozone, particulates, and aerosol direct forcing in an urban area using a new fully-coupled meteorology, chemistry, and aerosol model. J. Geophys. Res. 2006, 111, D21305. [Google Scholar] [CrossRef]
  19. LeGrand, S.; Polashenski, C.; Letcher, T.; Creighton, G.; Peckham, S.; Cetola, J. The AFWA dust emission scheme for the GOCART aerosol model in WRF-Chem v3.8.1. Geoesci. Model Dev. 2019, 12, 131–166. [Google Scholar] [CrossRef]
  20. Lee, J.A.; Jiménez, P.A.; Kumar, R.; He, C. Impact of direct insertion of SMAP soil moisture retrievals in WRF-Chem for dust storm events in western U.S. Atmos. Environ. 2024, 321, 120348. [Google Scholar] [CrossRef]
  21. Chin, M.; Ginoux, P.; Kinne, S.; Torres, O.; Holben, B.N.; Duncan, B.N.; Martin, R.; Logan, J.; Higurashi, A.; Nakajima, T. Tropospheric aerosol optical thickness from the GOCART model and comparisons with satellite and sun photometer measurements. J. Atmos. Sci. 2002, 59, 461–483. [Google Scholar] [CrossRef]
  22. Colarco, P.; da Silva, A.; Chin, M.; Diehl, T. Online simulations of global aerosol distributions in the NASA GEOS-4 model and comparisons to satellite and ground-based aerosol optical depth. J. Geophys. Res. 2010, 115, D14207. [Google Scholar] [CrossRef]
  23. Ginoux, P.; Chin, M.; Tegen, I.; Prospero, J.M.; Holben, B.; Dubovik, O.; Lin, S.J. Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res. 2001, 106, 20255–20273. [Google Scholar] [CrossRef]
  24. Tang, Y.; Pagowski, M.; Chai, T.; Pan, L.; Lee, P.; Baker, B.; Kumar, R.; Monache, L.D.; Tong, D.; Kim, H.C. A case study of aerosol data assimilation with the Community Multi-scale Air Quality Model over the contiguous United States using 3D-Var and optimal interpolation methods. Geoesci. Model Dev. 2017, 10, 4743–4758. [Google Scholar] [CrossRef]
  25. Santanello, J.; Lawston, P.; Kumar, S.; Dennis, E. Understanding the impacts of soil moisture initial conditions on NWP in the context of land-atmosphere coupling. J. Hydrometeor. 2019, 20, 793–819. [Google Scholar] [CrossRef]
  26. Diamond, H.J.; Karl, T.R.; Palecki, M.A.; Baker, C.B.; Bell, J.E.; Leeper, R.D.; Easterling, D.R.; Lawrimore, J.H.; Meyers, T.P.; Helfert, M.R.; et al. U.S. Climate Reference Network after one decade of operations: Status and assessment. Bull. Amer. Met. Soc. 2013, 94, 489–498. [Google Scholar] [CrossRef]
  27. Holben, B.; Eck, T.; Slutsker, I.; Tanré, D.; Buis, J.; Setzer, A.; Vermote, E.; Reagan, J.; Kaufman, Y.; Nakajima, T.; et al. AERONET—A federated instrument network and data archive for aerosol characterization. Remote Sens. Environ. 1998, 66, 1–16. [Google Scholar] [CrossRef]
  28. Levy, R.C.; Matto, S.; Munchak, L.A.; Remer, L.A.; Sayer, A.M.; Patadia, F.; Hsu, N.C. The collection 6 MODIS aerosol products over land and ocean. Atmos. Meas. Tech. 2013, 6, 2989–3034. [Google Scholar] [CrossRef]
  29. Gelaro, R.; McCarty, W.; Suárez, M.J.; Todling, R.; Molod, A.; Takacs, L.; Randles, C.A.; Darmenov, A.; Reichle, R.; Wargan, K.; et al. The modern-era retrospective analysis for research and applications, version 2 (MERRA-2). J. Clim. 2017, 30, 5419–5454. [Google Scholar] [CrossRef]
  30. Wiedinmyer, C.; Kimura, Y.; McDonald-Buller, E.C.; Emmons, L.E.; Buchholz, R.; Tang, W.; Seto, K.; Joseph, M.; Barsanti, K.; Carlton, A.; et al. The Fire Inventory from NCAR version 2.5: An updated global fire emissions model for climate and chemistry applications. Geoesci. Model Dev. 2023, 16, 3873–3891. [Google Scholar] [CrossRef]
  31. Freitas, S.; Longo, K.M.; Chatfield, R.; Latham, D.; Dias, M.S.; Andreae, M.; Prins, E.; Santos, J.; Gielow, R.; Carvalho, J., Jr. Including the sub-grid scale plume rise of vegetation fires in low resolution atmospheric transport models. Atm. Chem. Phys. 2007, 7, 3385–3398. [Google Scholar] [CrossRef]
  32. Gong, S.; Barrie, L.; Blanchet, J.P. Modeling sea-salt aerosols in the atmosphere: 1. Model development. J. Geophys. Res. 1997, 102, 3805–3818. [Google Scholar] [CrossRef]
  33. Jiménez, P.A.; Dudhia, J.; González-Rouco, J.F.; Navarro, J.; Montávez, J.P.; García-Bustamante, E. A revised scheme for the WRF surface layer formulation. Mon. Weather Rev. 2012, 140, 898–918. [Google Scholar] [CrossRef]
  34. Nakanishi, M.; Niino, H. Development of an improved turbulence closure model for the atmospheric boundary layer. J. Meteorol. Soc. Jpn. 2009, 87, 895–912. [Google Scholar] [CrossRef]
  35. Olson, J.B.; Kenyon, J.S.; Djalalova, I.; Bianco, L.; Turner, D.D.; Pichugina, Y.; Choukulkar, A.; Toy, M.D.; Brown, J.M.; Angevine, J.; et al. Improving wind energy forecasting through numerical weather prediction model development. Bull. Amer. Met. Soc. 2019, 100, 2201–2220. [Google Scholar] [CrossRef]
  36. Iacono, M.J.; Delamere, J.S.; Mlawer, E.J.; Shephard, M.W.; Clough, S.A.; Collins, W.D. Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res. 2008, 113, D13103. [Google Scholar] [CrossRef]
  37. Thompson, G.; Eidhammer, T. A study of aerosol impacts on clouds and precipitation development in a large winter cyclone. J. Atmos. Sci. 2014, 71, 3636–3658. [Google Scholar] [CrossRef]
  38. Grell, G.A.; Freitas, S. A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling. Atm. Chem. Phys. 2014, 14, 5233–5250. [Google Scholar] [CrossRef]
  39. Niu, G.Y.; Yang, Z.L.; Mitchell, K.; Chen, F.; Ek, M.; Barlage, M.; Kumar, A.; Niyogi, K.M.D.; Rosero, E.; Tewari, M.; et al. The community Noah land surface model with multiparameterization options (Noah-MP) 1. Model description and evaluation with local-scale measurements. J. Geophys. Res. 2011, 116, D12109. [Google Scholar] [CrossRef]
  40. He, C.; Valayamkunnath, P.; Barlage, M.; Chen, F.; Gochis, D.; Cabell, R.; Schneider, T.; Rasmussen, R.; Niu, G.Y.; Yang, Z.L.; et al. The Community Noah-MP Land Surface Modeling System Technical Description Version 5.0; Technical Report TN-575+STR; NCAR: Boulder, CO, USA, 2023. [Google Scholar]
  41. Yangl, Z.L.; Niu, G.Y.; Mitchell, K.; Chen, F.; Ek, M.; Barlage, M.; Longuevergne, L.; Manning, K.; Niyogi, D.; Tewari, M.; et al. The community Noah land surface model with multiparameterization options (Noah–MP): 2. Evaluation over global river basins. Mon. Weather Rev. 2011, 116, D12110. [Google Scholar]
  42. Jiménez, P.A.; Dudhia, J.; Thompson, G.; Lee, J.A.; Brummet, T. Improving the cloud initialization in WRF-Solar with enhanced short-range forecasting functionality: The MAD-WRF model. Sol. Energy 2022, 239, 221–233. [Google Scholar] [CrossRef]
  43. van der Veen, S.H. Improving NWP model cloud forecasts using Meteosat Second-Generation imagery. Mon. Weather Rev. 2012, 141, 1545–1557. [Google Scholar] [CrossRef]
  44. Xu, J.; Shu, H. Assimilating MODIS-based albedo and snow coverfraction into the Common Land Model to improve snow depth simulation with direct insertion and deterministic ensemble Kalman filter methods. J. Geophys. Res. 2014, 119, 10684–10701. [Google Scholar] [CrossRef]
  45. Zidikheri, M.J.; Lucas, C. Improving ensemble volcanic ash forecasts by direct insertion of satellite data and ensemble filtering. Atmosphere 2021, 12, 1215. [Google Scholar] [CrossRef]
  46. Computational and Information Systems Laboratory. Cheyenne: HPE/SGI ICE XA System (NCAR Community Computing); National Center for Atmospheric Research: Boulder, CO, USA, 2021. [Google Scholar] [CrossRef]
Figure 1. Soil moisture modulations of the wind erosion threshold for GOCART (gray dashed line, Equation (2)) and GOCART−AFWA (black solid line, Equation (4)) for a soil porosity of 0.476, clay content of 0.25, and air density of 1.2 kg m 3 .
Figure 1. Soil moisture modulations of the wind erosion threshold for GOCART (gray dashed line, Equation (2)) and GOCART−AFWA (black solid line, Equation (4)) for a soil porosity of 0.476, clay content of 0.25, and air density of 1.2 kg m 3 .
Remotesensing 17 01345 g001
Figure 2. Number of SMAP soil moisture retrievals inserted over the period 1 April 2015–31 March 2016.
Figure 2. Number of SMAP soil moisture retrievals inserted over the period 1 April 2015–31 March 2016.
Remotesensing 17 01345 g002
Figure 3. Scatter plot of the mean retrieved or simulated soil moisture versus the mean observed soil moisture at the USCRN sites, for (a) SMAP, (b) GOCART, and (c) GOCART_SMAP experiments, over the period 1 April 2015−31 March 2016.
Figure 3. Scatter plot of the mean retrieved or simulated soil moisture versus the mean observed soil moisture at the USCRN sites, for (a) SMAP, (b) GOCART, and (c) GOCART_SMAP experiments, over the period 1 April 2015−31 March 2016.
Remotesensing 17 01345 g003
Figure 4. MBE (a) and MAE (c) as a result of comparing SMAP retrievals versus the USCRN observations over the period 1 April 2015–31 March 2016. The MBE (b) and MAE (d) calculated with the GOCART experiment are also shown.
Figure 4. MBE (a) and MAE (c) as a result of comparing SMAP retrievals versus the USCRN observations over the period 1 April 2015–31 March 2016. The MBE (b) and MAE (d) calculated with the GOCART experiment are also shown.
Remotesensing 17 01345 g004
Figure 5. Soil moisture skill score as a result of comparing the GOCART_SMAP experiment against the GOCART experiment over the period 1 April 2015–31 March 2016. The erodibility representing locations of dust sources is also shown (shaded in yellow to brown).
Figure 5. Soil moisture skill score as a result of comparing the GOCART_SMAP experiment against the GOCART experiment over the period 1 April 2015–31 March 2016. The erodibility representing locations of dust sources is also shown (shaded in yellow to brown).
Remotesensing 17 01345 g005
Figure 6. MBE calculated with the simulated AOD at 500 nm from (a) GOCART_SMAP, (b) GOCART-AFWA_SMAP, and (c) NODUST_SMAP experiments and the AERONET observations over the period 1 Apr 2015–31 Mar 2016.
Figure 6. MBE calculated with the simulated AOD at 500 nm from (a) GOCART_SMAP, (b) GOCART-AFWA_SMAP, and (c) NODUST_SMAP experiments and the AERONET observations over the period 1 Apr 2015–31 Mar 2016.
Remotesensing 17 01345 g006
Figure 7. MBE calculated with the simulated AOD at 550 nm from (a) GOCART_SMAP, (b) GOCART-AFWA_SMAP, and (c) NODUST_SMAP experiments and the MODIS-Terra retrievals over the period 1 April 2015–31 March 2016.
Figure 7. MBE calculated with the simulated AOD at 550 nm from (a) GOCART_SMAP, (b) GOCART-AFWA_SMAP, and (c) NODUST_SMAP experiments and the MODIS-Terra retrievals over the period 1 April 2015–31 March 2016.
Remotesensing 17 01345 g007
Figure 8. MBE calculated with the simulated AOD at 550 nm from GOCART-AFWA_SMAP with emissions set to (a) 50%, (b) 25%, (c) 10%, and (d) 5% and the MODIS-Terra retrievals over the period 1 April 2015–31 March 2016.
Figure 8. MBE calculated with the simulated AOD at 550 nm from GOCART-AFWA_SMAP with emissions set to (a) 50%, (b) 25%, (c) 10%, and (d) 5% and the MODIS-Terra retrievals over the period 1 April 2015–31 March 2016.
Remotesensing 17 01345 g008
Figure 9. Skill score calculated with the simulated AOD at 500 nm from GOCART-AFWA_SMAP with (a) emissions set to 50% and reference 100%, (b) 25% and reference 50%, (c) 10% and reference 25%, and (d) 5% and reference 10% and the AERONET observations over the period 1 April 2015–31 March 2016.
Figure 9. Skill score calculated with the simulated AOD at 500 nm from GOCART-AFWA_SMAP with (a) emissions set to 50% and reference 100%, (b) 25% and reference 50%, (c) 10% and reference 25%, and (d) 5% and reference 10% and the AERONET observations over the period 1 April 2015–31 March 2016.
Remotesensing 17 01345 g009
Figure 10. Skill score calculated with the simulated AOD at 550 nm from GOCART-AFWA_SMAP with (a) emissions set to 50% and reference 100%, (b) 25% and reference 50%, (c) 10% and reference 25%, and (d) 5% and reference 10% and the MODIS retrievals over the period 1 Apr 2015–31 Mar 2016.
Figure 10. Skill score calculated with the simulated AOD at 550 nm from GOCART-AFWA_SMAP with (a) emissions set to 50% and reference 100%, (b) 25% and reference 50%, (c) 10% and reference 25%, and (d) 5% and reference 10% and the MODIS retrievals over the period 1 Apr 2015–31 Mar 2016.
Remotesensing 17 01345 g010
Figure 11. Histogram of the ratio of the mean dust emissions at every grid point calculated with GOCART-AFWA with SMAP insertion and GOCART-AFWA without SMAP insertion for the experiments reducing the emissions to 25%. Values larger (lower) than one indicate larger (lower) emissions when inserting SMAP retrievals.
Figure 11. Histogram of the ratio of the mean dust emissions at every grid point calculated with GOCART-AFWA with SMAP insertion and GOCART-AFWA without SMAP insertion for the experiments reducing the emissions to 25%. Values larger (lower) than one indicate larger (lower) emissions when inserting SMAP retrievals.
Remotesensing 17 01345 g011
Figure 12. Skill score calculated with the simulated AOD at 500 nm and AERONET observations for the GOCART-AFWA_SMAP simulations with (a) 25% emissions and (b) 10% emissions using as reference the simulation without SMAP retrievals. The skill score calculated with the simulated AOD at 550 nm and MODIS retrievals for the GOCART-AFWA_SMAP simulations with (c,e) 25% emissions and (d,f) 10% emissions, using as reference the simulation without SMAP retrievals, are also shown for the complete region and for the grid points with erodibility larger than zero (dust sources). The evaluation period is 1 April 2015–31 March 2016.
Figure 12. Skill score calculated with the simulated AOD at 500 nm and AERONET observations for the GOCART-AFWA_SMAP simulations with (a) 25% emissions and (b) 10% emissions using as reference the simulation without SMAP retrievals. The skill score calculated with the simulated AOD at 550 nm and MODIS retrievals for the GOCART-AFWA_SMAP simulations with (c,e) 25% emissions and (d,f) 10% emissions, using as reference the simulation without SMAP retrievals, are also shown for the complete region and for the grid points with erodibility larger than zero (dust sources). The evaluation period is 1 April 2015–31 March 2016.
Remotesensing 17 01345 g012
Figure 13. Histogram showing the relative frequency of the skill score calculated with the simulated AOD at 550 nm and MODIS retrievals from Terra (left column, (a,c)) and Aqua (right column, (b,d)) and the GOCART−AFWA_SMAP simulations with 25% (black solid boxes) and 10% (gray boxes) emissions (see legend in panel (b)) using as a reference the simulations with GOCART−AFWA without imposing SMAP retrievals over the period 1 April 2015−31 March 2016. The first row shows the histograms for the complete region over CONUS, whereas the second row shows the histograms over the dust sources (grid cells with erodibility higher than zero).
Figure 13. Histogram showing the relative frequency of the skill score calculated with the simulated AOD at 550 nm and MODIS retrievals from Terra (left column, (a,c)) and Aqua (right column, (b,d)) and the GOCART−AFWA_SMAP simulations with 25% (black solid boxes) and 10% (gray boxes) emissions (see legend in panel (b)) using as a reference the simulations with GOCART−AFWA without imposing SMAP retrievals over the period 1 April 2015−31 March 2016. The first row shows the histograms for the complete region over CONUS, whereas the second row shows the histograms over the dust sources (grid cells with erodibility higher than zero).
Remotesensing 17 01345 g013
Table 1. WRF-Chem experiments performed. The experiments frequently cited in the text are labeled (see column 1).
Table 1. WRF-Chem experiments performed. The experiments frequently cited in the text are labeled (see column 1).
Experiment NameEmissionsSMAPEmissions Factor
Exp1 (GOCART)GOCARTNo-
Exp2 (GOCART_SMAP)GOCARTYes-
Exp3 (GOCART-AFWA)GOCART-AFWANo1.0
Exp4 (GOCART-AFWA_SMAP)GOCART-AFWAYes1.0
Exp5GOCART-AFWANo0.5
Exp6GOCART-AFWAYes0.5
Exp7GOCART-AFWANo0.25
Exp8GOCART-AFWAYes0.25
Exp9GOCART-AFWANo0.1
Exp10GOCART-AFWAYes0.1
Exp11GOCART-AFWANo0.05
Exp12GOCART-AFWAYes0.05
Exp13 (NODUST)-Yes-
Table 2. Statistics comparing SMAP soil moisture and the simulated soil moisture using GOCART experiments versus USCRN observations. The statistics are calculated using only the times when SMAP, WRF-Chem, and USCRN data are all available. The MBE, MAE, and RMSE are expressed in volumetric percent of water in the soil [%]. The evaluation period is 1 April 2015–31 March 2016.
Table 2. Statistics comparing SMAP soil moisture and the simulated soil moisture using GOCART experiments versus USCRN observations. The statistics are calculated using only the times when SMAP, WRF-Chem, and USCRN data are all available. The MBE, MAE, and RMSE are expressed in volumetric percent of water in the soil [%]. The evaluation period is 1 April 2015–31 March 2016.
ExperimentMBEMAERMSECorr
SMAP−0.57.29.20.65
GOCART2.07.18.90.68
GOCART_SMAP−0.56.58.30.71
Table 3. Statistics summarizing the performance of WRF-Chem in reproducing the wind speed/wind direction from METAR reports over the 10 EPA regions. The evaluation period is 1 April 2015–31 March 2016. The unit for the wind speed MBE, MAE, and RMSE is m s 1 , whereas for the wind direction is degrees.
Table 3. Statistics summarizing the performance of WRF-Chem in reproducing the wind speed/wind direction from METAR reports over the 10 EPA regions. The evaluation period is 1 April 2015–31 March 2016. The unit for the wind speed MBE, MAE, and RMSE is m s 1 , whereas for the wind direction is degrees.
RegionMBEMAERMSECorr
R1: New England0.05/80.60/260.85/380.82/0.80
R2: New York, New Jersey0.02/20.68/260.85/370.85/0.83
R3: Mid-Atlantic0.78/30.85/201.01/280.89/0.90
R4: Southeast0.74/20.78/160.90/220.88/0.91
R5: Upper Midwest/Great Lakes0.53/30.64/120.76/160.93/0.96
R6: South Central0.51/60.66/170.80/230.89/0.86
R7: Midwest−0.02/50.60/160.76/230.90/0.92
R8: Mountains and Plains0.01/60.44/150.56/200.90/0.84
R9: Pacific Southwest0.28/100.62/190.75/230.84/0.83
R10: Pacific Northwest0.24/90.51/240.63/290.84/0.72
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Jiménez y Muñoz, P.A.; Kumar, R.; He, C.; Lee, J.A. Revisiting the Role of SMAP Soil Moisture Retrievals in WRF-Chem Dust Emission Simulations over the Western U.S. Remote Sens. 2025, 17, 1345. https://doi.org/10.3390/rs17081345

AMA Style

Jiménez y Muñoz PA, Kumar R, He C, Lee JA. Revisiting the Role of SMAP Soil Moisture Retrievals in WRF-Chem Dust Emission Simulations over the Western U.S. Remote Sensing. 2025; 17(8):1345. https://doi.org/10.3390/rs17081345

Chicago/Turabian Style

Jiménez y Muñoz, Pedro A., Rajesh Kumar, Cenlin He, and Jared A. Lee. 2025. "Revisiting the Role of SMAP Soil Moisture Retrievals in WRF-Chem Dust Emission Simulations over the Western U.S." Remote Sensing 17, no. 8: 1345. https://doi.org/10.3390/rs17081345

APA Style

Jiménez y Muñoz, P. A., Kumar, R., He, C., & Lee, J. A. (2025). Revisiting the Role of SMAP Soil Moisture Retrievals in WRF-Chem Dust Emission Simulations over the Western U.S. Remote Sensing, 17(8), 1345. https://doi.org/10.3390/rs17081345

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop