1. Introduction
A number of studies in recent years investigated the potential impacts of high concentrations of hygroscopic aerosols acting as cloud condensation nuclei (CCN) on tropical cyclone (TC) intensity (e.g., Rosenfeld et al. 2011; Khain et al. 2010; Carrió and Cotton 2011; Rosenfeld et al. 2012; Cotton et al. 2012; Herbener et al. 2014; Zhang et al. 2007, 2009). In most of those studies enhanced CCN concentrations resulted in the weakening of the storm. Cotton et al. (2012), however, found that during the initial stages of their simulations the enhanced CCN concentrations led to an increase in storm intensity. They attributed the initial storm intensification to high concentrations of CCN penetrating the eyewall and inner rainband region. Later on the simulated storm weakened as the aerosols were washed out by the heavy rainfall in the storm core. Herbener et al. (2014) also found TC intensification following introduction of high concentrations of CCN. This also was a result of the penetration of high CCN concentrations into the storm core.
High CCN concentrations contribute to TC intensification by invigorating the updrafts of embedded convective cells. The hypothesized mechanisms for aerosol-induced convective invigoration remain controversial. Cotton and Walko (2021) provide an overview of the hypothesized mechanisms. The first school of thought is what is called “mixed-phase invigoration.” The hypothesized response to aerosols is as follows. High concentrations of particulate pollutants lead to high concentrations of cloud droplets, they suppress warm rain formation, transport greater quantities of cloud droplets to supercooled levels, which freezes thereby releasing greater amounts of latent heat, and the added buoyancy invigorates the updrafts of cumuli, and the invigorated updrafts process more water thereby enhancing rainfall. Enhanced rainfall can create colder and larger area cold pools, which can trigger more convection and lead to greater duration of rain events. The second school of thought is called “condensational invigoration.” The theory is that high concentrations of cloud droplets formed on numerous pollution-sized aerosols exhibit greater net surface area upon which condensation occurs, thereby enhancing net vapor deposition rates that leads to enhanced latent heat release by condensation in cumuli. At heights roughly 3 km above cloud base where droplet collection can be prevalent, the concentrations of cloud droplets is thereby reduced and supersaturations can exceed nominal near-cloud-base values, which can lead to appreciable enhancement of condensation in a polluted cloud relative to a clean cloud. Thus, latent heat by condensation of droplets can be enhanced enough to invigorate updrafts, lead to greater amounts of condensed water, produce broader, longer-lived cumuli and thereby enhance rainfall. As noted by Cotton and Walko (2021) for idealized tropical convection, these two mechanisms are intertwined such that condensational invigoration can lead to mixed-phase invigoration. However, Cotton and Walko (2021) found that condensational invigoration was dominant.
The impact of convective invigoration on TC intensity arises from the alteration of latent release in the rainbands, and from enhanced rainfall on low-level cold-pool intensity. Carrió and Cotton (2011) concluded that in the outer rainbands, enhanced cold pools impede the supply of warm moist air into the storm core resulting in a weakening of storm intensity. In the inner rainband and eyewall region, aerosol-induced enhanced latent heating contributes to intensification of the TC (Cotton et al. 2012; Herbener et al. 2014).
In this paper, we examine the potential impacts of aerosol pollution on TC Harvey. We chose TC Harvey because it stalled near Houston, Texas, for 4 days, dropping historic amounts of rainfall of more than 60 in. (~152 cm) over southeastern Texas (Blake and Zelinsky 2018). Houston is one of the most polluted cities in the United States where peak CCN concentrations measured during the Texas Air Quality Study/Gulf of Mexico Atmospheric Composition and Climate Study (TexAQS-GoMACCS; Lance et al. 2009) exceeded 25 000 cm−3. We hypothesize that, if very high CCN concentrations entered TC Harvey prior to landfall, the intensity of the storm would have been affected. Moreover, as TC Harvey hovered around the Houston basin, convective cells may have been invigorated, which then would contribute to the large rainfall amounts from the storm.
2. Model setup
a. OLAM
In this study, we use the global-to-storm-scale model called Ocean–Land–Atmosphere Model (OLAM). OLAM is a global nonhydrostatic weather and climate prediction model that has been developed since 2001 (Walko and Avissar 2008a,b; Walko and Avissar 2011; Ullrich et al. 2017). It is an outgrowth of the Regional Atmospheric Modeling System (RAMS; Pielke et al. 1992; Cotton et al. 2003), which was developed for investigating meso- and cloud-scale phenomena. OLAM incorporates the cloud physics scheme from RAMS (Walko et al. 1995, 2000; Cotton et al. 2003), and is thus well equipped to simulate moist convective systems at very high resolution (e.g., down to tens of meters scales). Recent upgrades in cloud physics (Saleeby and Cotton 2004, 2005, 2008; Saleeby et al. 2007; Ward et al. 2010) include explicit representation of aerosols and their impact on liquid and ice nucleation (including aerosol size, concentration, and chemistry via kappa), adding a new prognostic liquid water category that represents large cloud droplets or drizzle sizes, and use of numerous lookup tables that permit emulation of size-bin computations without compromising efficiency (see below for a more extensive discussion). Recently, the Rapid Radiative Transfer Model or RRTMG (Mlawer et al. 1997; Clough et al. 2005) was implemented in OLAM. Optical properties of OLAM’s ice categories were defined partly from Key et al. (2002). OLAM solves the fully compressible Navier–Stokes equations instead of the Boussinesq system used in RAMS, and it employs a finite volume discretization on a hexagonal mesh that enables spatially variable resolution.
b. GEOS-Chem
To estimate aerosol concentrations for clean conditions and for conditions where anthropogenic sources are prevalent, we use output from GEOS-Chem (www.geos-chem.org). GEOS-Chem is run globally with a resolution of 2° × 2.5° and 47 vertical layers extending from the surface to 0.01 hPa and is driven by assimilated meteorology from the GEOS5 reanalyses (http://gmao.gsfc.nasa.gov). The aerosol scheme includes sulfate, nitrate, ammonium, dust, sea salt, hydrophilic black carbon, hydrophilic organic carbon, hydrophobic black carbon, hydrophobic organic carbon and five groups of secondary organic aerosols (SOAs). Global anthropogenic emissions in GEOS-Chem are provided by the Emissions Database for Global Atmospheric Research (EDGAR) inventory, except where overwritten by the following regional inventories: The Environmental Protection Agency 2005 National Emissions Inventory (NEI05) (http://www.epa.gov/ttn/chief/net/2005inventory.html) over the United States, the criteria air contaminants (CAC) for anthropogenic emissions over Canada (http://www.ec.gc.ca/inrp-npri/), the Big Bend Regional Aerosol and Visibility Study (BRAVO) emissions inventory over Mexico and the southwestern United States, the Streets inventory for Asian emissions (Streets et al. 2003) over Asia, and the Cooperative Programme for Monitoring and Evaluation of the Long-Range Transmission of Air Pollutants in Europe (EMEP) over Europe (Auvray and Bey 2005). Biogenic emissions were from the Model of Emissions of Gases and Aerosols from Nature (MEGAN). Biomass burning emissions are from the Global Fire Emissions Database version 3 (GFEDv3; van der Werf et al. 2010).
c. Aerosol activation in OLAM
To understand how aerosols are activated in OLAM, we first must understand how supersaturation is computed in OLAM microphysics. Walko et al. (2000) gives a complete description. Water vapor specific density and air temperature, which together determine supersaturation, are governed by energy and water mass conservation laws that involve multiple constituents in the model, including dry air, water vapor, cloud droplets, larger liquid water droplets (drizzle and rain), ice crystals (small and large pure crystals plus aggregates), graupel, and hail. Diffusive fluxes of heat and water to or from each hydrometeor type and air are driven by gradients of temperature and vapor specific density, respectively, between the surface of each particle and the air. These flux equations, the mass and energy equations for each hydrometeor type and for the air, and the Clausius–Clapeyron equation for vapor specific density at the surface of hydrometeors are solved implicitly as a linear system of interactive equations for each grid cell and time step of an OLAM simulation, although when no hydrometeors are present, the system simplifies considerably. Latent heating or cooling is accounted for in all phase changes, as is the constraint that ice temperature does not exceed 0°C. Other source and sink terms, including gridscale vertical motion and other advective transport, gain or loss of hydrometeors in a parcel due to gravitational sedimentation, and radiative heating and cooling, are represented explicitly as forcing terms for the linear system. In its standard form, this system provides an accurate and computationally stable result that accounts for all governing processes in the model. Near the cloud base, this system of equations correctly predicts the rise of supersaturation and activation of aerosol particles to form cloud droplets whose concentration depends on the concentrations and chemistry of the aerosol particles. As cloud droplets grow by vapor deposition, they limit the growth of supersaturation resulting in peak supersaturations typically of less than 1% a few hundred meters above cloud base (Warner 1968; Paluch and Knight 1984; Korolev and Mazin 2003).
However, higher in the cloud when droplet collision and coalescence is prevalent and cloud droplet number is therefore greatly diminished, Clark (1973) first showed that supersaturations can rise well above near-cloud-base nominal values. The OLAM interactive system of equations explicitly represents the slow pace of vapor diffusion to the limited surface area of the existing hydrometeors and the consequent continued increase in supersaturation. Nucleated aerosols remain in the atmosphere if the cloud droplets subsequently evaporate, but aerosols are depleted in number when there is collisional growth of cloud droplets, scavenging of cloud droplets by rain and ice, and transport to the ground in falling precipitation. As aerosols and cloud droplets are removed by these processes, higher supersaturations occur in updrafts, both because the aerosols that remain tend to require higher supersaturation to nucleate (due to having smaller size and/or less hygroscopicity) and because without abundant cloud droplets, there is less combined hydrometeor surface area for vapor to diffuse to. The smallest and least hygroscopic aerosol particles represented in the GEOS-Chem aerosol fields are 5 nm black-carbon particles whose critical supersaturation is 4%. Consequently, once supersaturations reach this level in the simulations all aerosols are nucleated.
Specifically in OLAM, whenever supersaturation occurs in a grid cell, the nucleation bin model is activated for the purpose of determining how many aerosols will nucleate. A semi-Lagrangian environment is first reconstructed for that grid cell and time step in order to provide a Lagrangian rate of supersaturation production, which differs from the Eulerian rate in the grid cell, most often due primarily to convective vertical motion. Aerosols from each of the prognosed categories in OLAM, of which there are a total of nine due to some combining of GEOS-Chem aerosol species that have similar size and hygroscopic values, are sorted by size into 20 bins based on a median diameter and spectral variance specified for each category. Typically, aerosols in the bin representing the smallest size for a category are about 1/3 the median diameter for that category, while those in the bin with the largest size are about 3 times the median diameter. Counting all aerosol categories together, up to 180 bins are populated from the aerosols that are present in that grid cell. These bins are sorted in order of increasing critical saturation, which depends on both the aerosol size and hygroscopicity for each bin. If any cloud droplets are present in the grid cell, aerosols with the lowest critical saturation are assumed to have been already activated within those droplets, and these aerosols thus do not participate in the bin model nucleation process for the present time step. (OLAM does not deplete aerosols when they activate as cloud droplets because the droplets may subsequently evaporate and leave dry aerosols remaining. Aerosol depletion occurs when cloud droplets collide and/or are collected by precipitating hydrometeors). The remaining unactivated aerosols are processed in the bin model in which environmental production of supersaturation is consistent with local semi-Lagrangian processes for the given grid cell. Aerosol activation and vapor diffusional growth are represented explicitly as the bin model is integrated forward in time using (in the present case) time step increments 0.02 times the length of an OLAM time step. Integration is halted once supersaturation ceases to increase, at which time the number of newly activated aerosols has been determined, or when integration time equals that of an OLAM time step, whichever comes first. In the latter case, if conditions are such that nucleation continues beyond this time period, the procedure will be continued on subsequent OLAM time steps. New cloud droplets that result from nucleation in the bin model in one OLAM time step will become “preexisting” cloud droplets on the next time step where they take up some of the excess water vapor and leave less for new nucleation. OLAM prognoses both cloud droplet mass and number as a two-moment scheme, and accordingly, the nucleation bin model contributes both to OLAM’s representation of cloud water. We somewhat arbitrarily begin the nucleation bin model integration with exactly saturated conditions, whereas in reality conditions may be slightly below or slightly above saturation at the beginning of the OLAM time step. In tests of the nucleation procedure, we found sensitivity to the initial supersaturation to be reasonably small. The nucleation bin model is ultraefficient and has only a small impact on the overall speed of the model.
d. OLAM numerical experiments
Our original plan was to run continuous simulations of Harvey beginning as a tropical cyclone just prior to its hurricane stage (24 August) and ending after most of the extreme rainfall accumulation in southeast Texas (31 August). This would enable numerical investigation of the impact of aerosol concentration on both the intensification phase of the cyclone prior to landfall and the extreme precipitation it produced in southeast Texas. However, we found that the simulated cyclone tracked a few degrees to the left of the observed best track as it approached the Texas coast. As a result, the simulated cyclone made landfall farther southwest and a few hours earlier than observed, and then the day after landfall, the simulated storm did not turn eastward as observed. Consequently, we instead performed shorter simulations over two separate periods, the first (P1) covering the rapid intensification of Harvey during 24–26 August, and the second (P2) covering the majority of precipitation in southeast Texas during 28–31 August.
OLAM was configured with a variable-resolution grid in which hexagonal cells of width 1.6 km covered nearly the entire region occupied by Harvey throughout the simulation period, and resolution gradually coarsened with increasing distance from that region. For P1, the high-resolution part of the grid covered all locations within 300 km of at least one point along a great circle segment with endpoints at latitude–longitude coordinates (25.0°N, 94.4°W) and (29.0°N, 97.5°W). For P2, the high-resolution part of the grid was all locations within 400 km of the segment with endpoints at (29.0°N, 96.0°W) and (29.0°N, 94.0°W). Grid cells of double that size (i.e., 3.2 km) covered the next 200 km outward from that region (in both P1 and P2), although the transition in resolution is gradual over a distance of four grid cells. No convective parameterization was applied over either of those resolutions, but it was applied farther out starting with grid cells 6.4 km in width. The maximum cell width was 100 km and most of the globe was covered at that resolution. Vertical grid spacing in P1 (P2) was 100 m at sea level, stretching gradually to 2000 m (1000 m) at 20 km, and a constant 2000 m (1000 m) above 20 km. The model top was near 45 km. Initial conditions were obtained by interpolating the NCEP Climate Forecast System Reanalysis (CFSR) 0.5° global dataset to the OLAM grid and performing an initial hydrostatic balance. For initialization at times where Harvey contained a core structure that was underresolved by the CFSR data, an axisymmetric perturbation vortex was added to the interpolated CFSR fields in order to recover the intensity and core diameter of the observed vortex at that time. Hydrostatic and gradient-wind balance procedures were applied to the resulting fields in order to obtain a vortex in approximate equilibrium.
To test the sensitivity of the system to aerosol concentration, we ran a set of three simulations for each period that differed only in the selection and concentration of aerosols that were input. OLAM simulations, which we designate with “N,” used aerosols generated by a GEOS-Chem simulation with natural aerosols only, while OLAM simulations designated with “NA” used aerosols from a separate GEOS-Chem simulation with both natural and anthropogenic aerosols included. Owing to the small differences that we found in hurricane intensity and trajectory between N and NA simulations (as will be shown in Figs. 2 and 3), we performed “NA10X” in which the aerosol concentrations for NA were increased 10 times (see Table 1 for a complete list of simulations performed). Although aerosol concentrations for NA10X were assigned somewhat arbitrarily, we note that they rarely exceed 20 000 mg−1 and then only in localized regions at low levels in the storm periphery. Given that measured CCN concentrations over Houston exceeded 25 000 cm−3 during the Texas Air Quality Study/Gulf of Mexico Atmospheric Composition and Climate Study (TexAQS-GoMACCS; Lance et al. 2009), the NA10X CCN concentrations are not excessive.
Designation simulations for two time periods and three aerosol specifications.
GEOS-Chem aerosol fields were used in OLAM both as an initial condition and additionally as an assimilated time-dependent value by adding a Newtonian relaxation term to the aerosol prognostic equations. The relaxation time scale was chosen as 24 h in order to balance the requirement that aerosol scavenging on convective time scales be capable of depleting nearly all aerosols against the requirement that continuous sources of aerosol be represented over the multiday period of the simulations. GEOS-Chem fields are available at 6-h increments, and the values to which OLAM is nudged are a time interpolation between the nearest past and future GEOS-Chem datasets.
Initial GEOS-Chem aerosol concentrations, summed over all prognosed categories, for P1N and P1NA are plotted in Fig. 1 over a region including southeast Texas and the northwestern Gulf of Mexico. The higher concentrations in P1NA are solely due to the addition of anthropogenic aerosols. High concentrations are generally confined to the lowest 2 km. The impact of Tropical Cyclone Harvey can be seen in both of the horizontal cross section plots; this is due to the realism of the wind fields used in the GEOS-Chem simulations. Initial aerosol fields for P1NA10X are not shown, but of course they are like P1NA except with exactly 10 times the value at each location.
Initial concentrations (number per mg of air) of aerosols for (left) P1N and (right) P1NA. (top) Horizontal cross sections at 1 km above sea level over a region centered at 28°N, 96°W. (bottom) North–south vertical cross sections as seen looking westward centered at the same location.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Initial concentrations (number per mg of air) of aerosols for (left) P1N and (right) P1NA. (top) Horizontal cross sections at 1 km above sea level over a region centered at 28°N, 96°W. (bottom) North–south vertical cross sections as seen looking westward centered at the same location.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Initial concentrations (number per mg of air) of aerosols for (left) P1N and (right) P1NA. (top) Horizontal cross sections at 1 km above sea level over a region centered at 28°N, 96°W. (bottom) North–south vertical cross sections as seen looking westward centered at the same location.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
3. Results
a. Rapid intensification phase
Simulated cyclone tracks for P1N, P1NA, and P1NA10X are shown in Fig. 2. For comparison, simulated tracks from selected operational forecast models are also shown (based on data contained in
Harvey trajectories over 48 h for P1N (OL01), P1NA (OL02), P1NA10X (OL03), and selected operational forecast models. Observed best track locations at 6-h intervals are numbered consecutively on the plot. Plot is centered at 27.2°N, 96.8°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Harvey trajectories over 48 h for P1N (OL01), P1NA (OL02), P1NA10X (OL03), and selected operational forecast models. Observed best track locations at 6-h intervals are numbered consecutively on the plot. Plot is centered at 27.2°N, 96.8°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Harvey trajectories over 48 h for P1N (OL01), P1NA (OL02), P1NA10X (OL03), and selected operational forecast models. Observed best track locations at 6-h intervals are numbered consecutively on the plot. Plot is centered at 27.2°N, 96.8°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Figure 3 shows the time history of minimum sea level pressure for P1N, P1NA, and P1NA10X in comparison with observed values. While P1N and P1NA have similar rates of intensification, P1NA10X strengthens more rapidly after 24 h and reaches greater intensity. All three cases achieve landfall within 1 h of each other, so the timing of landfall is not a major factor influencing the maximum intensity. However, landfall occurred approximately 5 h earlier than observed because of the leftward trajectory deviation and the orientation of the coastline, and this contributed to all three cases deepening less than the observed lowest pressure of 937 hPa.
Time history of minimum sea level pressure in P1N (black), P1NA (red), P1NA10X (blue), and Harvey observations (green).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time history of minimum sea level pressure in P1N (black), P1NA (red), P1NA10X (blue), and Harvey observations (green).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time history of minimum sea level pressure in P1N (black), P1NA (red), P1NA10X (blue), and Harvey observations (green).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Figures 4–9 show aerosol and hydrometeor abundances in P1N and P1NA10X at 2200 UTC 25 August, or 34 h after initialization, which is very close to the time of maximum intensity and landfall in both simulations. Horizontal cross sections in Fig. 4 (P1N) and 7 (P1NA10X) are taken at a height of 1 km and show aerosol and cloud droplet concentrations, both in number per milligram of air or approximately per cubic centimeter, and cloud water and precipitation mixing ratios in grams per kilogram of air. Precipitation here is the sum of all liquid and ice condensate other than cloud water. The location of the hurricane eye is evident in all plots. The heaviest precipitation occurs in a ring containing the eyewall and the region immediately outside it. Both aerosol and cloud droplet concentrations are extremely low in this region, rarely exceeding 1 or 2 per milligram of air. This is because of rapid scavenging by the precipitation. Farther outward, areas of heavy precipitation organized in spiral bands clearly coincide with local minima of aerosol concentration, again showing near total aerosol scavenging. Outside the eyewall, cloud droplet concentrations and mixing ratio are generally higher in P1NA10X than in P1N, and this is supported by the higher concentrations of aerosols. The eye itself has a secondary maximum of aerosol concentration, especially in P1NA10X; this is due to the aerosol nudging source term and the absence of any scavenging precipitation inside the eye. Note that with landfall occurring at this time, the eye is located where pollution is elevated in GEOS-Chem. That aerosol concentrations are far lower in the nearby eyewall attests to the rapidity at which precipitation scavenges aerosols relative to the 24-h aerosol nudging time scale. Nevertheless, it must be pointed out that the nudging approach chosen for this study introduces aerosols directly to the inner region of the hurricane without subjecting it to potential scavenging that it would normally encounter in an inward trajectory through the outer precipitation bands of the storm. This is particularly true for P1NA10X, which has the highest rate of aerosol replenishment through nudging.
Horizontal cross sections at 1 km height, centered at 28°N, 96°W, of simulated fields from P1N at 2200 UTC 25 Aug. (top left) CCN concentration (number mg of air), (top right) cloud droplet concentration (number per mg of air), (bottom right) cloud water mixing ratio (g per kg of air), and (bottom left) precipitation mixing ratio (g per kg of air).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Horizontal cross sections at 1 km height, centered at 28°N, 96°W, of simulated fields from P1N at 2200 UTC 25 Aug. (top left) CCN concentration (number mg of air), (top right) cloud droplet concentration (number per mg of air), (bottom right) cloud water mixing ratio (g per kg of air), and (bottom left) precipitation mixing ratio (g per kg of air).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Horizontal cross sections at 1 km height, centered at 28°N, 96°W, of simulated fields from P1N at 2200 UTC 25 Aug. (top left) CCN concentration (number mg of air), (top right) cloud droplet concentration (number per mg of air), (bottom right) cloud water mixing ratio (g per kg of air), and (bottom left) precipitation mixing ratio (g per kg of air).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
P1N results at 2200 UTC 25 Aug corresponding to Fig. 4, except north–south vertical cross sections (as viewed looking westward) passing through center of hurricane at coordinate x = −162 km. (top left) CCN concentration (number per mg of air), (top right) cloud droplet concentration (number per mg of air), (bottom left) low-density ice concentration (number per g of air), and (bottom right) high-density ice concentration ratio (number per kg of air).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
P1N results at 2200 UTC 25 Aug corresponding to Fig. 4, except north–south vertical cross sections (as viewed looking westward) passing through center of hurricane at coordinate x = −162 km. (top left) CCN concentration (number per mg of air), (top right) cloud droplet concentration (number per mg of air), (bottom left) low-density ice concentration (number per g of air), and (bottom right) high-density ice concentration ratio (number per kg of air).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
P1N results at 2200 UTC 25 Aug corresponding to Fig. 4, except north–south vertical cross sections (as viewed looking westward) passing through center of hurricane at coordinate x = −162 km. (top left) CCN concentration (number per mg of air), (top right) cloud droplet concentration (number per mg of air), (bottom left) low-density ice concentration (number per g of air), and (bottom right) high-density ice concentration ratio (number per kg of air).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
P1N condensate mixing ratios (g per kg of air) at 2200 UTC 25 Aug in vertical cross section identical to Fig. 5. (top left) Cloud water, (top right) low-density ice, (bottom left) high-density ice, and (bottom right) rain.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
P1N condensate mixing ratios (g per kg of air) at 2200 UTC 25 Aug in vertical cross section identical to Fig. 5. (top left) Cloud water, (top right) low-density ice, (bottom left) high-density ice, and (bottom right) rain.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
P1N condensate mixing ratios (g per kg of air) at 2200 UTC 25 Aug in vertical cross section identical to Fig. 5. (top left) Cloud water, (top right) low-density ice, (bottom left) high-density ice, and (bottom right) rain.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 4, but from P1NA10 X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 4, but from P1NA10 X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 4, but from P1NA10 X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 5, but from P1NA10X and slab is located at x = −137 km in the depiction of Fig. 7.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 5, but from P1NA10X and slab is located at x = −137 km in the depiction of Fig. 7.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 5, but from P1NA10X and slab is located at x = −137 km in the depiction of Fig. 7.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 6, but from P1NA10X and slab is located at x = −137 km in the depiction of Fig. 7.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 6, but from P1NA10X and slab is located at x = −137 km in the depiction of Fig. 7.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 6, but from P1NA10X and slab is located at x = −137 km in the depiction of Fig. 7.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Figures 5 and 6 show number concentration and mixing ratio for various hydrometeors in a north–south vertical cross section (as seen looking westward) that passes through the eye center location of P1N, which is at coordinate (x = −162 km, y = −90 km) as referenced in Fig. 4. Figures 8 and 9 show the corresponding plots for P1NA10X, where the eye center is at (x = −137 km, y = −75 km), corresponding to Fig. 7. Very low aerosol and cloud droplet concentrations are seen to extend through all heights in the inner region of the hurricane, although they are higher in P1NA10X. Low-density ice is more abundant in both number and mixing ratio in P1NA10X than in P1N, and this is supported by the higher aerosol concentration (note that the units for number concentration are 1000 times lower for low-density ice compared to aerosols and cloud droplets). High-density ice and rain are also more abundant in P1NA10X. It would be plausible to argue that increased latent heat release from the more abundant condensate caused P1NA10X to intensify relative to P1N. However, these higher amounts of condensate are also supported by the greater intensity of P1NA10X. Wind, thermodynamic, and condensate fields intensified simultaneously as P1NA10X strengthened relative to P1N, which complicated efforts to determine clear cause-and-effect mechanisms to explain how more aerosols led to greater intensity.
To examine whether the greater intensification in P1NA10X might result (at least in part) from increased latent heat release associated with the prevention of high supersaturations by higher aerosol concentrations, we plot time–height cross sections of maximum supersaturation in Fig. 10 and the fractional area inside a 100 km radius from the cyclone center having a supersaturation greater than 10% in Fig. 11. We see that high supersaturations do occur in all simulations but they are progressively lower in magnitude and less frequent as aerosol concentration increases. We note, however, that very high supersaturations (e.g., >10%) do not occur in the lower troposphere where saturation vapor density is highest and the impact on latent heat release would be greatest. To better quantify the combined effects of supersaturation and saturation vapor density, the potential temperature deficit associated with the peak supersaturation in Fig. 10 is plotted in Fig. 12, with the excess of potential deficit of P1N over P1NA10X shown in the top-right panel. Some deficit of potential temperature due to high supersaturation occurs in all three simulations, but it is least prevalent in P1NA10X. These analyses results are consistent with the condensational invigoration mechanism contributing to strengthening the hurricane in P1NA10X relative to the other two cases.
Time–height cross section of maximum supersaturation (%) if greater than zero for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time–height cross section of maximum supersaturation (%) if greater than zero for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time–height cross section of maximum supersaturation (%) if greater than zero for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time–height cross section of fractional horizontal area (%) within 100 km of cyclone center in which supersaturation is greater than 10% for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time–height cross section of fractional horizontal area (%) within 100 km of cyclone center in which supersaturation is greater than 10% for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Time–height cross section of fractional horizontal area (%) within 100 km of cyclone center in which supersaturation is greater than 10% for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Deficit of potential temperature (K) associated with peak supersaturation in Fig. 7. (top right) The difference with P1NA10X deficits subtracted from those of P1N.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Deficit of potential temperature (K) associated with peak supersaturation in Fig. 7. (top right) The difference with P1NA10X deficits subtracted from those of P1N.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Deficit of potential temperature (K) associated with peak supersaturation in Fig. 7. (top right) The difference with P1NA10X deficits subtracted from those of P1N.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Figures 13 and 14 show latent heating rates horizontally averaged over the innermost 100 km of the cyclone center for liquid and ice processes, respectively, as a function of height and time. After 25 h, which is when P1NA10X begins to deepen much more rapidly than P1N (and P1NA), latent heating rates due to both liquid and ice processes are greater in P1NA10X. While it is tempting to argue that this is evidence of both condensational and mixed-phase invigoration mechanisms contributing to invigoration of the cyclone, it is difficult to separate the cause and effect: The increase in cyclone intensity feeds back by dynamically strengthening the convection and hence the latent heating rates. As noted by Cotton and Walko (2021), this illustrates how the two hypothesized invigoration mechanisms are intertwined. This is particularly true in the well-organized system of a tropical cyclone.
Latent heating rate associated with liquid processes (K day−1) horizontally averaged over the innermost 100 km of the cyclone for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X. (top right) The difference P1NA10X minus P1N.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Latent heating rate associated with liquid processes (K day−1) horizontally averaged over the innermost 100 km of the cyclone for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X. (top right) The difference P1NA10X minus P1N.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Latent heating rate associated with liquid processes (K day−1) horizontally averaged over the innermost 100 km of the cyclone for (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X. (top right) The difference P1NA10X minus P1N.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 10, but for ice processes.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 10, but for ice processes.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
As in Fig. 10, but for ice processes.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
b. Texas precipitation phase
Simulated cyclone tracks for P2N, P2NA, and P2NA10X are shown in Fig. 15. For comparison, the observed best track and simulated tracks from selected operational forecast models are also shown. During this period, Harvey was observed to track to the ESE for about 24 h, moving offshore, and then to gradually turn to the east and northeast, making landfall for a second time. The three OLAM trajectories approximately followed this motion only for the first 24 h, and then motion slowed and the cyclone remained nearly stationary with some erratic movement. Other models varied widely for this forecast period, indicating relatively low predictability, although a few models matched the observed track fairly well. Because OLAM failed to move Harvey eastward on 29 and 30 August, it produced surface precipitation maxima (in all three simulations) very close to Houston, which is approximately 100 km to the west of the observed maximum, as shown in Fig. 16. However, the values of the maxima, which range from 1.0 to 1.2 m, are roughly comparable to the observed maximum of approximately 1.5 m, which accumulated over a somewhat longer time span than the 3-day-long simulations. Consequently, we deemed the OLAM simulations suitable for examining the sensitivity to aerosol concentrations, the main focus here being on precipitation. However, the main result was that very little sensitivity was found in precipitation or any other TC characteristic. The area-averaged accumulated precipitation within a radius of 600 km from the center location of Fig. 16, which encompasses the entire tropical cyclone through the 3-day simulation period, is shown in Fig. 17. Not only are these totals nearly identical among the three simulations, but the spatial patterns of accumulation at the end of the simulation period (Fig. 16) are also very similar. The differences between the three panels in Fig. 16 can largely be attributed to small differences in the simulated trajectories of the simulated storms. In all three OLAM simulations and in the observed cyclone, maximum winds remained at TS intensity through the period.
Harvey trajectories over 48 h for P1N (OL01), P1NA (OL02), P1NAX (OL03), and selected operational forecast models. Observed best track locations at 6-h intervals are numbered consecutively on the plot. Plot is centered at 29.0°N, 95.0°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Harvey trajectories over 48 h for P1N (OL01), P1NA (OL02), P1NAX (OL03), and selected operational forecast models. Observed best track locations at 6-h intervals are numbered consecutively on the plot. Plot is centered at 29.0°N, 95.0°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Harvey trajectories over 48 h for P1N (OL01), P1NA (OL02), P1NAX (OL03), and selected operational forecast models. Observed best track locations at 6-h intervals are numbered consecutively on the plot. Plot is centered at 29.0°N, 95.0°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Accumulated precipitation (mm) over a 3-day period of (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X. Plot is centered at 29.0°N, 94.0°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Accumulated precipitation (mm) over a 3-day period of (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X. Plot is centered at 29.0°N, 94.0°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Accumulated precipitation (mm) over a 3-day period of (top left) P1N, (bottom left) P1NA, and (bottom right) P1NA10X. Plot is centered at 29.0°N, 94.0°W.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Area-averaged accumulated precipitation (mm) vs time within a 600 km radius of the center location of Fig. 13 for P2N (black), P2NA (red), and P2NA10X (blue).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Area-averaged accumulated precipitation (mm) vs time within a 600 km radius of the center location of Fig. 13 for P2N (black), P2NA (red), and P2NA10X (blue).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
Area-averaged accumulated precipitation (mm) vs time within a 600 km radius of the center location of Fig. 13 for P2N (black), P2NA (red), and P2NA10X (blue).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0076.1
4. Summary and conclusions
Using the global model, OLAM, with aerosol estimates from the global atmospheric chemistry model GEOS-Chem, two sets of simulations of Hurricane Harvey were performed. The first set of simulations were performed during the rapid intensification phase of Harvey, and the second set were performed after landfall over Texas. During each period simulations were performed with no anthropogenic sources of aerosols (P1N and P2N), with both natural and anthropogenic aerosol sources (P1NA and P2NA), and with both sources enhanced 10 times (P1NA10X and P2NA10X).
During the rapid intensification phase the results indicate that aerosol amounts had very little impact on storm motion. Moreover, very little difference was found in the intensity of the simulated storm to aerosol amounts for the no-anthropogenic versus the GEOS-Chem estimated amounts with anthropogenic sources. Owing to the lack of sensitivity of the simulated storm to aerosol amounts, GEOS-Chem estimated aerosol amounts were enhanced 10 times. Some justification for such high aerosol concentrations is that observed concentrations in Houston were this high on at least one occasion. The resultant simulated storm intensity was enhanced appreciably in terms of minimum sea level pressure. Examination of supersaturations and potential temperature for the P1NA10X simulation suggests that supersaturations are of lesser magnitude and potential temperatures are warmer, which supports the hypothesis that the condensational invigoration mechanism contributes to strengthening the hurricane relative to the other two cases.
During the second period of the simulation, during which Harvey remained a tropical storm, the main result was that very little sensitivity was found in precipitation or any other TC characteristic to aerosol concentrations. Thus, the simulations do not support the hypothesis that as TC Harvey hovered around the Houston basin; convective cells may have been invigorated by aerosol pollution, which then would contribute to the large rainfall amounts from the storm. We cannot definitively state why the individual convective cells did not respond to high aerosol concentrations during this phase of the storm. However, the abundant precipitation in all three simulations scavenged the vast majority of aerosols as it flowed radially inward, and we speculate that this modulated the potential impact of aerosols on the then tropical storm.
During the rapid intensification phase of the simulation, the simulated storm track resulted in storm landfall occurring farther southwest and somewhat earlier than observed landfall. Nonetheless, the simulated storm track was within the statistical spread of operational model forecasts. The overall spread of model trajectories is fairly wide, indicating a moderate amount of uncertainty for this situation. OLAM and most other models accurately replicated the observed propagation speed of the hurricane. After landfall, OLAM failed to move Harvey eastward on 29 and 30 August, and this resulted in a simulated precipitation maximum that was located about 100 km west of the observed maximum. Nevertheless, OLAM simulations of Harvey all produced copious amounts of precipitation that were comparable in amount to the historic accumulations observed.
Overall the simulated response of Hurricane Harvey to aerosols was far less spectacular than what has been simulated in the past. We conclude that this is because Hurricane Harvey was a strongly dynamically driven storm system that as a result was relatively impervious to the effects of aerosols.
Acknowledgments
This research was supported by NSF Grants AGS 1547752 and AGS 1547903. We thank Dr. Jeff Pierce and Dr. Jack Kodros for running GEOS-Chem and providing model output data.
REFERENCES
Auvray, M., and L. Bey, 2005: Long-range transport to Europe: Seasonal variations and implications for the European ozone budget. J. Geophys. Res., 110, D11303, https://doi.org/10.1029/2004JD005503.
Blake, E. S., and D.A. Zelinsky, 2018: Hurricane Harvey (AL092017). National Hurricane Center Rep., 77 pp.
Carrió, G. G., and W. R. Cotton, 2011: Investigations of aerosol impacts on hurricanes: Virtual seeding flights. Atmos. Chem. Phys., 11, 2557–2567, https://doi.org/10.5194/acp-11-2557-2011.
Clark, T. L., 1973: Numerical modeling of the dynamics and microphysics of warm cumulus convection. J. Atmos. Sci., 30, 857–878, https://doi.org/10.1175/1520-0469(1973)030<0857:NMOTDA>2.0.CO;2.
Clough, S. A., M. W. Shephard, E. J. Mlawer, J. S. Delamere, M. J. Iacono, K. Cady-Pereira, S. Boukabara, and P. D. Brown, 2005: Atmospheric radiative transfer modeling: A summary of the AER codes. J. Quant. Spectrosc. Radiat. Transfer, 91, 233–244, https://doi.org/10.1016/j.jqsrt.2004.05.058.
Cotton, W. R., and R. Walko, 2021: Examination of aerosol-induced convective invigoration using idealized simulations. J. Atmos. Sci., 78, 287–298, https://doi.org/10.1175/JAS-D-20-0023.1.
Cotton, W. R., and Coauthors, 2003: RAMS 2001: Current status and future directions. Meteor. Atmos. Phys., 82, 5–29, https://doi.org/10.1007/s00703-001-0584-9.
Cotton, W. R., G. M. Krall, and G. G. Carrió, 2012: Potential indirect effects of aerosol on tropical cyclone intensity: Convective fluxes and cold-pool activity. Trop. Cyclone Res. Rev., 1, 293–306, https://doi.org/10.6057/2012TCRR03.05.
Herbener, S. R., S. C. van den Heever, G. G. Carrió, S. M. Saleeby, and W. R. Cotton, 2014: Aerosol indirect effects on idealized tropical cyclone dynamics. J. Atmos. Sci., 71, 2040–2055, https://doi.org/10.1175/JAS-D-13-0202.1.
Key, J. R., P. Yang, B. A. Baum, and S. L. Nasiri, 2002: Parameterization of shortwave ice cloud optical properties for various particle habits. J. Geophys. Res., 107, 4181, https://doi.org/10.1029/2001JD000742.
Khain, A., B. Lynn, and J. Dudhia, 2010: Aerosol effects on intensity of landfalling hurricanes as seen with the WRF Model with spectral bin microphysics. J. Atmos. Sci., 67, 365–384, https://doi.org/10.1175/2009JAS3210.1.
Korolev, A. V., and I. P. Mazin, 2003: Supersaturation of water vapor in clouds. J. Atmos. Sci., 60, 2957–2974, https://doi.org/10.1175/1520-0469(2003)060<2957:SOWVIC>2.0.CO;2.
Lance, S., and Coauthors, 2009: CCN activity, closure and droplet growth kinetics of Houston aerosol during the Gulf of Mexico Atmospheric Composition and Climate Study (GoMACCS). J. Geophys. Res., 114, D00F15, https://doi.org/10.1029/2008JD011699.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, https://doi.org/10.1029/97JD00237.
Paluch, I. R., and C. A. Knight, 1984: Mixing and evolution of droplet spectra in a vigorous continental cumulus. J. Atmos. Sci., 41, 1801–1815, https://doi.org/10.1175/1520-0469(1984)041<1801:MATEOC>2.0.CO;2.
Pielke, R. A., and Coauthors, 1992: A comprehensive meteorological modeling system—RAMS. Meteor. Atmos. Phys., 49, 69–91, https://doi.org/10.1007/BF01025401.
Rosenfeld, D., M. Clavner, and R. Nirel, 2011: Pollution and dust aerosols modulating tropical cyclone intensities. Atmos. Res., 102, 66–76, https://doi.org/10.1016/j.atmosres.2011.06.006.
Rosenfeld, D., W. L. Woodley, A. Khain, W. R. Cotton, G. Carrió, I. Ginis and J. H. Golden , 2012: Aerosol effects on microstructure and intensity of tropical cyclones. Bull. Amer. Meteor. Soc., 93, 987–1001, https://doi.org/10.1175/BAMS-D-11-00147.1.
Saleeby, S. M., and W. R. Cotton, 2004: A large droplet mode and prognostic number concentration of cloud droplets in the Colorado State University Regional Atmospheric Modeling System (RAMS). Part I: Module descriptions and supercell test simulations. J. Appl. Meteor., 43, 182–195, https://doi.org/10.1175/1520-0450(2004)043<0182:ALMAPN>2.0.CO;2.
Saleeby, S. M., and W. R. Cotton, 2005: A large droplet mode and prognostic number concentration of cloud droplets in the Colorado State University Regional Atmospheric Modeling System (RAMS). Part II: Sensitivity to a Colorado winter snowfall event. J. Appl. Meteor., 44, 1912–1929, https://doi.org/10.1175/JAM2312.1.
Saleeby, S. M., and W. R. Cotton, 2008: A binned approach to cloud droplet riming implemented in a bulk microphysics model. J. Appl. Meteor. Climatol., 47, 694–703, https://doi.org/10.1175/2007JAMC1664.1.
Saleeby, S. M., W. Y. Y. Cheng, and W. R. Cotton, 2007: New developments in the regional atmospheric modeling system suitable for simulations of snowpack augmentation over complex terrain. J. Wea. Mod, 39, 37–49, https://journalofweathermodification.org/index.php/JWM/article/view/196.
Streets, D. G., K. F. Yarber, J.-H. Woo, and G. R. Carmichael, 2003: Biomass burning in Asia: Annual and seasonal estimates and atmospheric emissions. Global Biogeochem. Cycles, 17, 1099, https://doi.org/10.1029/2003GB002040.
Ullrich, P. A., and Coauthors, 2017: DCMIP2016: A review of non-hydrostatic dynamical core design and intercomparison of participating models. Geosci. Model Dev., 10, 4477–4509, https://doi.org/10.5194/gmd-10-4477-2017.
van der Werf, G. R., and Coauthors, 2010: Global fire emissions and the contribution of deforestation, savanna, forest, agricultural, and peat fires (1997–2009). Atmos. Chem. Phys., 10, 11 707–11 735, https://doi.org/10.5194/acp-10-11707-2010.
Walko, R. L., and R. Avissar, 2008a: The Ocean–Land–Atmosphere Model (OLAM). Part I: Shallow water tests. Mon. Wea. Rev., 136, 4033–4044, https://doi.org/10.1175/2008MWR2522.1.
Walko, R. L., and R. Avissar, 2008b: The Ocean–Land–Atmosphere Model (OLAM). Part II: Formulation and tests of the nonhydrostatic dynamic core. Mon. Wea. Rev., 136, 4045–4062, https://doi.org/10.1175/2008MWR2523.1.
Walko, R. L., and R. Avissar, 2011: A direct method for constructing refined regions in unstructured conforming triangular-hexagonal computational grids: Application to OLAM. Mon. Wea. Rev., 139, 3923–3937, https://doi.org/10.1175/MWR-D-11-00021.1.
Walko, R. L., W. R. Cotton, J. L. Harrington, and M. P. Meyers, 1995: New RAMS cloud microphysics parameterization. Part I: The single-moment scheme. Atmos. Res., 38, 29–62, https://doi.org/10.1016/0169-8095(94)00087-T.
Walko, R. L., W. R. Cotton, G. Feingold, and B. Stevens, 2000: Efficient computation of vapor and heat diffusion between hydrometeors in a numerical model. Atmos. Res., 53, 171–183, https://doi.org/10.1016/S0169-8095(99)00044-7.
Ward, D. S., T. Eidhammer, W. R. Cotton, and S. M. Kreidenweis, 2010: The role of the particle size distribution in assessing aerosol composition effects on simulated droplet activation. Atmos. Chem. Phys., 10, 5435–5447, https://doi.org/10.5194/acp-10-5435-2010.
Warner, J., 1968: The supersaturation in natural clouds. J. Rech. Atmos., 3, 233–238.
Zhang, H., G. M. McFarquhar, S. M. Saleeby, and W. R. Cotton, 2007: Impacts of Saharan dust as CCN on the evolution of an idealized tropical cyclone. Geophys. Res. Lett., 34, L14812, https://doi.org/10.1029/2007GL029876.
Zhang, H., G. M. McFarquhar, W. R. Cotton, and Y. Deng, 2009: Direct and indirect impacts of Saharan dust acting as cloud condensation nuclei on tropical cyclone eyewall development. Geophys. Res. Lett., 36, L06802, https://doi.org/10.1029/2009GL037276.
