We present the dust module in the Multiscale Online Non-hydrostatic AtmospheRe CHemistry model (MONARCH) version 2.0, a chemical weather prediction system that can be used for regional and global modeling at a range of resolutions. The representations of dust processes in MONARCH were upgraded with a focus on dust emission (emission parameterizations, entrainment thresholds, considerations of soil moisture and surface cover), lower boundary conditions (roughness, potential dust sources), and dust–radiation interactions. MONARCH now allows modeling of global and regional mineral dust cycles using fundamentally different paradigms, ranging from strongly simplified to physics-based parameterizations. We present a detailed description of these updates along with four global benchmark simulations, which use conceptually different dust emission parameterizations, and we evaluate the simulations against observations of dust optical depth. We determine key dust parameters, such as global annual emission/deposition flux, dust loading, dust optical depth, mass-extinction efficiency, single-scattering albedo, and direct radiative effects. For dust-particle diameters up to 20
The Multiscale Online Non-hydrostatic AtmospheRe CHemistry model (MONARCH) is a chemical weather modeling system that can be used at multiple spatial scales, ranging from regional scales at single-digit kilometer resolutions with explicit convection to coarse-resolution global scales with parameterized convection
Mineral soil dust is the most abundant aerosol type in terms of global mass, competing only with sea salt
Mineral dust is emitted as soon as the forces that act to retain the soil particles at the surface (gravity and interparticle cohesion) are overcome either by atmospheric lifting forces generated by wind and turbulence (aerodynamic entrainment) or by the force generated by other impacting particles, i.e., sand grains or particle aggregates (saltation bombardment/aggregate disintegration)
Once airborne, mineral dust particles interact with short- and longwave radiation through scattering and absorption
Available physics schemes in MONARCH.
Existing dust emission parameterizations range from formulations that are strongly simplified
In this work, we introduce recent advancements in the treatment of mineral dust in MONARCH. The model now has diverse available model configurations, in particular to estimate dust emission, which makes MONARCH unique among state-of-the-art models, and which makes it suitable for a variety of applications that range from process studies to operational forecasting and climate research. In the following sections, we briefly present the MONARCH modeling system and subsequently focus on the mineral dust cycle. We then demonstrate and evaluate MONARCH's dust modeling capabilities based on four annual global model runs.
Summary of the six available dust emission schemes in MONARCH.
MONARCH (previously known as NMMB/BSC-CTM) consists of advanced chemistry and aerosol packages coupled online with the Non-hydrostatic Multiscale Model on the B-grid (NMMB)
The gas-phase chemistry in MONARCH solves the Carbon Bond 2005 chemical mechanism
In addition to single model runs, MONARCH can be run in an ensemble mode for data assimilation applications, where the ensemble of model states is used to derive a flow-dependent background error covariance at the assimilation time, which evolves during the model forecast. The background error covariance is used to express model uncertainty within the data assimilation framework. Model uncertainty, together with observational uncertainty, is a key ingredient in the optimal integration of model simulations and observations for the production of an analysis that best represents the atmospheric state. The MONARCH ensemble is coupled with the local ensemble transform Kalman filter (LETKF) scheme
The dust module in MONARCH (previously known as NMMB/BSC-Dust), initially described by dust generation and uplift by surface wind and turbulence, horizontal and vertical advection, horizontal diffusion and vertical transport by turbulence and convection, dry deposition and gravitational settling, and wet removal by convective and stratiform clouds.
The dust size distribution is represented with eight bins ranging up to 20
Our new developments presented below have mostly focused on aspect (1): dust generation and uplift by surface wind and turbulence. In particular, we have implemented and tested a variety of dust emission and drag partition parameterizations, along with new data sets for dust source areas, source type (i.e., natural and anthropogenic), surface roughness, and vegetation. Additional upgrades include the option to calculate dust extinction assuming non-spherical particle shape, as well as new diagnostic capabilities (output of three-dimensional single-scattering albedo and extinction, clear-sky aerosol optical depth (clear-sky AOD), and AOD at satellite overpass times). In the following, we present the MONARCH dust module. We first describe the treatment of dust emission, summarizing previous and detailing new developments. Then, we recapitulate the implementation of dust transport and deposition, and interactions with radiation.
Several different parameterizations of dust emission are available in MONARCH, which cover different paradigms and range from more simplified to more physics-based descriptions. To describe dust emission generated by saltation, MONARCH includes the parameterizations from
Coefficients for minimally dispersed particle-size distributions as assigned to the 12 USGS soil texture classes. Each particle-size distribution (PSD) is composed of four lognormal distributions (
Summary of available options in MONARCH to account for soil moisture in the particle entrainment threshold.
Summary of available options in MONARCH to account for surface roughness in particle entrainment.
In saltation-based dust emission schemes, the vertical dust emission flux
The G01 dust emission scheme does not include an explicit formulation of
The S01 scheme is a physics-based dust emission scheme, which calculates size-resolved dust emission based on the soil volume removed by impacting saltation particles and explicitly considers aggregate disintegration as a dust emission process in addition to saltation bombardment. The emission of dust particles of size
The S04 scheme is a simplification of the S01 scheme in which the saltation bombardment efficiency,
The K14 dust emission scheme uses the concept of the fragmentation of brittle material
For the schemes that contain explicit representations of the saltation flux (MB95, S01, S04, S11), the saltation flux of particles with diameter
The implementation of the threshold friction velocity for ideal (dry) conditions,
Models are known to underestimate the strong-wind tail of the wind speed distribution by different degrees depending on their resolution. This is particularly relevant for dust emission
When the soil is moist, the threshold friction velocity above which particles are lifted is higher than under dry conditions, because soil-water capillary forces increase the cohesion between the soil particles
The soil moisture correction from
Shao and Jung (2000, unpublished manuscript) and
Surface roughness through, e.g., vegetation, pebbles, or rocks, absorbs momentum from the air flow and reduces the atmospheric momentum available for particle entrainment. We account for this drag partitioning using either the scheme of
In the parameterization of
Figure
Comparison of roughness input and drag partition approaches: panels
Normalized particle-size distributions (PSDs) based on
In the formulation from
The correction
Apart from the effect of vegetation or other roughness elements to absorb atmospheric momentum, they also directly prohibit particle entrainment from the area they cover. Similarly, areas covered by snow/ice (
The particle-size distribution of emitted dust is key to quantifying the emitted dust mass, dust loading in the atmosphere, dust interactions with the energy and water cycles, along with more general impacts of dust upon climate. Whether or not the emitted dust PSD changes with the magnitude of atmospheric forces is still debated
Physical and optical particle properties available in MONARCH for eight particle-size bins: equivalent volume radius (
Total annual dust emission (left), dust deposition (center), and annual average column dust load for 2012 using the configurations described in Sect.
Statistical dust parameters of four global model simulations using the dust emission schemes MB95, G01-UST, S04, and K14 with the configurations described in Sect.
In MONARCH, areas from which dust emission is possible are described using
a map obtained from the climatological (for the years 2003–2015) frequency of occurrence (FoO) of the MODIS Deep Blue dust optical depth (DOD) greater than 0.2
In addition to the definition of areas from which dust emission is possible, a scaling of the calculated dust emission fluxes with the above-mentioned dust source functions is deployed in the MB95 and G01 schemes. The preferential source map from
An additional special feature of MONARCH is its ability to tag dust originating from natural and anthropogenic (agricultural) sources. For this purpose, the MODIS FoO-based map is linked with fractions of anthropogenic land use, following the approach described in
Vegetation in MONARCH is prescribed based on satellite data, using either an AVHRR monthly climatology of green vegetation cover fraction
Soil texture class information in MONARCH is obtained from the hybrid STATSGO-FAO data set at a resolution of 30 arc seconds (0.0083
Dust transport and deposition in MONARCH have been thoroughly described in
The model's radiation scheme is RRTM-G
A range of global
model simulations was performed with MONARCH for 1 year (2012) to demonstrate MONARCH's dust modeling capabilities. We used different configurations in the runs covering different dust emission schemes. We evaluate the presented simulations against MODIS
The global model runs performed with MONARCH were conducted at a horizontal resolution of 1
Here we present results of global MONARCH simulations using the MB95, G01-UST,
S04, and K14 dust emission schemes, a set of well-known and frequently used
parameterizations. In all runs, we scaled soil moisture using
The dust fields of all model runs were calibrated using experiment-specific global calibration factors, which were obtained by comparing monthly averages of modeled coarse DOD (size range 1.2–20
The total mass of dust emitted globally during 2012 was 3489, 3627, 5994, and 3739
The similarity in global dust emission between the MB95 and G01-UST schemes is a result of the scaling with the topographic source mask. Nevertheless, differences in the magnitude of dust emission are evident, in particular in the Middle East, central Asia, and Australia. Neither the S04 nor K14 scheme uses a preferential source function besides the binary treatment explained in Sect.
Consistent with the differences in dust emission between the four runs, the annual total dust deposition and annual average dust load are similar in the MB95 and G01-UST runs, with pronounced individual source regions such as the Bodélé Depression. In comparison, deposition and dust loading are more intense in northwestern Africa and the Middle East in the S04 scheme, and more homogeneous in the K14 scheme.
Figure
The global annual average of DOD is 0.034, 0.032, 0.041, and 0.035 in the MB95, G01-UST, S04, and K14 runs. This results in an average mass-extinction efficiency
Seasonally averaged MODIS Deep Blue DOD (left), MONARCH all-sky DOD
Seasonally averaged FoO of DOD
Globally averaged monthly global DOD
Comparison of 3-hourly DOD between MONARCH (average (turquoise line) and standard deviation (shading)) and AERONET direct-Sun V3 level 2.0 for selected stations covering Cabo Verde and the Canary Islands (Capo Verde, Santa Cruz de Tenerife), the Sahara and Sahel (Ouarzazate, Tamanrasset, Cinzana, Banizoumbou), the Middle East (Eilat, Solar Village, Masdar Institute), Asia (Karachi, Issyk-Kul, Dalanzadgad), Europe (Granada), southern Africa (Henties Bay), Australia (Birdsville), and North and South America including the Caribbean (Railroad Valley, CASLEO, Ragged Point). The direct-Sun DOD is filtered for dust aerosol using Ångström exponent (AE)
We estimate DOD from MODIS using daily AOD and SSA at 550
To enable a direct comparison between MODIS satellite observations and MONARCH
results independent of the model output frequency, MONARCH internally
diagnoses the all-sky DOD for a given satellite overpass time for each
day. The sampling of the satellite overpass time follows Quaas (2011,
Figure
Differences in the modeled all-sky co-located and clear-sky DOD
Figure
The FoO obtained from the modeled clear-sky DOD
Figure
Annual average longwave, shortwave, and total direct radiative effect [
Global averages of MONARCH-derived DRE [
AERONET is a global network of ground-based solar photometer stations
Through AOD, AERONET gives information about the aerosol content and the mode-dominant type (i.e., fine or coarse modes) in the atmospheric column, but not the atmospheric dust burden. Almost pure mineral dust is difficult to find, except in specific areas close to desert dust sources. Instead, dust is often mixed in variable percentages with other aerosols. To isolate the atmospheric dust burden and estimate the DOD, two approaches are typically used.
The first approach aims to identify records in which the measured aerosol is dominated by mineral dust based on AE. AE is in general inversely related to the average size of the airborne particles and can be used to distinguish species with large particles like dust and sea salt. As a rule of thumb, a larger AE indicates smaller particle size. AE is typically in the range 0–4, where the upper limit corresponds to molecular extinction, and the lower limit corresponds to coarse-mode aerosols (sea salt and mineral dust), indicating no wavelength dependence of AOD
The second widely used methodology to estimate AERONET DOD is based on the spectral deconvolution algorithm (SDA) retrievals
For comparison with AERONET, we use bilinear interpolation to extract time series from the 3-hourly global model DOD and DOD
Taking into account the entire station list (Appendix
Figure
We note that the mineralogy-based set of refractive indices used in this work describes a more scattering dust in the shortwave with respect to other widely used prescriptions
We presented the description of mineral dust in the Multiscale Online Non-hydrostatic AtmospheRe CHemistry model (MONARCH) version 2.0. MONARCH contains multiple state-of-the-art options to represent dust emissions on global and regional scales, ranging from more simplified to more complex parameterizations based on physical processes. We tested and evaluated a set of four global model configurations for the year 2012. Comparison with observations of dust optical depth from MODIS and AERONET showed a good model reproduction of key features of the observed dust cycle. Global annual dust emissions ranged between around 3500 and 6000
The multifaceted options of MONARCH and its dust component, combined with an advanced workflow management for use in high-performance computing environments, makes it a powerful and versatile tool applicable for process studies, operational forecasting, and climate research. In the following, we outline a few ongoing activities related to the MONARCH dust component to demonstrate its capabilities.
Dust ensemble runs can be generated with MONARCH by utilizing the diverse model configurations and by perturbing model parameters related to, for example, surface winds, soil humidity, and the spatial distribution of dust emission, which are deemed to be uncertain. In
Airborne dust is not a homogeneous entity but a mixture of minerals, the relative amounts of which depend on the source region. Mineralogy affects a variety of dust-related impacts, e.g., interaction with radiation, atmospheric chemistry, or nutrient supply to certain ecosystems. The capability to explicitly represent dust composition was recently added to MONARCH allowing the tagging of up to 12 different minerals. This new feature is currently used to assess the relevance of dust mineralogy for dust impacts and to provide insights for the near-term atmospheric and climate modeling communities
The combination of different vegetation input data sets, drag partition approaches, and the source tagging capability allows us to represent the seasonal vegetation dynamics and provides an ideal basis to investigate the importance of dust from anthropogenic (agricultural) sources, for which a key driver is the seasonal vegetation growth and decay. The benefit of online estimates within a modeling framework is that not only the emission but also the transport, deposition, and effect of anthropogenic dust can be investigated
Figure
Table
Percentage contributions of different dust source regions (definitions as in
Percent contribution of dust emission (left column) and deposition (right column) to their respective global and annual totals for the four experiments described in Sect.
Figures
Seasonally averaged MODIS Deep Blue DOD (left) and MONARCH all-sky DOD
Seasonally averaged FoO of DOD
The AERONET stations used for comparison with MONARCH and to obtain the global calibration factor are listed in Table
Figure
List of AERONET stations used for comparison with MONARCH results.
AERONET stations (direct-Sun V3 level 2.0) available in 2012. The station subset used for comparison with MONARCH are shown in turquoise, whereas all other stations are marked in red.
Scatter plots (top) of 3-hourly DOD estimated from AERONET direct-Sun V3 level 2.0 (AE
Table
Global averages of MONARCH-derived DRE [
Continued.
MONARCH source code used in this work (v2.0.0) is available at
MONARCH output presented in this paper is available via EUDAT B2SHARE (
MK implemented most of the dust-module upgrades, ran, analyzed, and evaluated the simulations, and wrote the manuscript. OJ contributed to the model upgrade and leads the MONARCH development together with CPGP. MGA, JE, MLD, VO, and EDT contributed to the model upgrades. SB contributed to the model evaluation. GMP and FM supported the development of MONARCH and MONARCH workflow. PG, JG, and CP contributed data used in the model/for model evaluation. YH and JFK contributed dust optical properties used in MONARCH. CPGP and RLM co-designed the model changes together with MK. In addition, CPGP implemented part of the model upgrades and contributed to manuscript writing. All authors commented on the manuscript.
The authors declare that they have no competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
BSC co-authors acknowledge PRACE (eFRAGMENT2) and RES (AECT-2019-3-0001 and AECT-2020-1-0018) for awarding access to MareNostrum at the Barcelona Supercomputing Center and for providing technical support. Martina Klose, Carlos Pérez García-Pando, Sara Basart, Paul Ginoux, Catherine Prigent, and Ron L. Miller appreciated the opportunity for exchange and discussion at the Institut Pascal at Université Paris-Saclay with support of the program “Investissements d'avenir” (ANR-11-IDEX-0003-01). The authors thank all the principal investigators and their staff for establishing and maintaining the AERONET sites, as well as the MODIS mission scientists and associated NASA personnel for the production of the AOD, SSA, and AE data used in this study.
This research has been supported by the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 789630. Martina Klose has received funding through the Helmholtz Association's Initiative and Networking Fund (grant agreement no. VH-NG-1533). Jeronimo Escribano and Matthew L. Dawson have received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreements no. 754433 (Jeronimo Escribano) and no. 747048 (Matthew Dawson). BSC co-authors acknowledge funding from the following: the European Research Council (FRAGMENT (grant no. 773051)); the AXA Research Fund; the Spanish Ministry of Science, Innovation and Universities (grant no. CGL2017-88911-R); the EU H2020 project FORCES (grant no. 821205); the CMUG-CCI3-TECHPROP contract, an activity carried out under a program of and funded by the European Space Agency (ESA); and the DustClim project, which is part of ERA4CS, an ERA-NET initiated by JPI Climate and funded by FORMAS (SE), DLR (DE), BMWFW (AT), IFD (DK), MINECO (ES), and ANR (FR), with co-funding by the European Union (grant no. 690462). Ron L. Miller is funded by the NASA Modeling, Analysis and Prediction Program (NNG14HH42I). Jasper F. Kok acknowledges support from the National Science Foundation (NSF) under grant nos. 1552519 and 1856389 and the Army Research Office (grant no. W911NF-20-2-0150). Yue Huang has received funding from the Columbia University Earth Institute Postdoctoral Research Fellowship and from NASA (grant no. 80NSSC19K1346) awarded under the Future Investigators in NASA Earth and Space Science and Technology (FINESST) program.
This paper was edited by Axel Lauer and reviewed by Yves Balkanski and one anonymous referee.