A coupled cloud physics–radiation parameterization of the bulk optical properties of cirrus and its impact on the Met Office unified model global atmosphere 5.0 configuration

A new coupled cloud physics–radiation parameterization of the bulk optical properties of ice clouds is presented. The parameterization is consistent with assumptions in the cloud physics scheme regarding particle size distributions (PSDs) and mass–dimensional relationships. The parameterization is based on a weighted ice crystal habit mixture model, and its bulk optical properties are parameterized as simple functions of wavelength and ice water content (IWC). This approach directly couples IWC to the bulk optical properties, negating the need for diagnosed variables, such as the ice crystal effective dimension. The parameterization is implemented into the Met Office Unified Model Global Atmosphere 5.0 (GA5) configuration. The GA5 configuration is used to simulate the annual 20-yr shortwave (SW) and longwave (LW) fluxes at the top of the atmosphere (TOA), as well as the temperature structure of the atmosphere, under various microphysical assumptions. The coupled parameterization is directly compared against the current operational radiation parameterization, while maintaining the same cloud physics assumptions. In this experiment, the impacts of the two parameterizations on the SW and LW radiative effects at TOA are also investigated and compared against observations. The 20-yr simulations are compared against the latest observations of the atmospheric temperature and radiative fluxes at TOA. The comparisons demonstrate that the choice of PSD and the assumed ice crystal shape distribution are as important as each other. Moreover, the consistent radiation parameterization removes a long-standing tropical troposphere cold temperature bias but slightly warms the southern midlatitudes by about 0.5 K.

Impact of biomass burning aerosols on precipitation in the Amazon: A modeling case study

A study of the potential role of aerosols in modifying clouds and precipitation is presented using a numerical atmospheric model. Measurements of cloud condensation nuclei (CCN) and cloud size distribution properties taken in the southwestern Amazon region during the transition from dry to wet seasons were used as guidelines to define the microphysical parameters for the simulations. Numerical simulations were carried out using the Brazilian Development on Regional Atmospheric Modeling System, and the results presented considerable sensitivity to changes in these parameters. High CCN concentrations, typical of polluted days, were found to result in increases or decreases in total precipitation, depending on the level of pollution used as a reference, showing a complexity that parallels the aerosol-precipitation interaction. Our results show that on the grids evaluated, higher CCN concentrations reduced low-to-moderate rainfall rates and increased high rainfall rates. The principal consequence of the increased pollution was a change from a warm to a cold rain process, which affected the maximum and overall mean accumulated precipitation. Under polluted conditions, cloud cover diminished, allowing greater amounts of solar radiation to reach the surface. Aerosol absorption of radiation in the lower layers of the atmosphere delayed convective evolution but produced higher maximum rainfall rates due to increased instability. In addition, the intensity of the surface sensible heat flux, as well as that of the latent heat flux, was reduced by the lower temperature difference between surface and air, producing greater energy stores at the surface.

Remote sensing the vertical profile of cloud droplet effective radius, thermodynamic phase, and temperature

Cloud-aerosol interaction is a key issue in the climate system, affecting the water cycle, the weather, and the total energy balance including the spatial and temporal distribution of latent heat release. Information on the vertical distribution of cloud droplet microphysics and thermodynamic phase as a function of temperature or height, can be correlated with details of the aerosol field to provide insight on how these particles are affecting cloud properties and their consequences to cloud lifetime, precipitation, water cycle, and general energy balance. Unfortunately, today's experimental methods still lack the observational tools that can characterize the true evolution of the cloud microphysical, spatial and temporal structure in the cloud droplet scale, and then link these characteristics to environmental factors and properties of the cloud condensation nuclei. Here we propose and demonstrate a new experimental approach (the cloud scanner instrument) that provides the microphysical information missed in current experiments and remote sensing options. Cloud scanner measurements can be performed from aircraft, ground, or satellite by scanning the side of the clouds from the base to the top, providing us with the unique opportunity of obtaining snapshots of the cloud droplet microphysical and thermodynamic states as a function of height and brightness temperature in clouds at several development stages. The brightness temperature profile of the cloud side can be directly associated with the thermodynamic phase of the droplets to provide information on the glaciation temperature as a function of different ambient conditions, aerosol concentration, and type. An aircraft prototype of the cloud scanner was built and flew in a field campaign in Brazil. The CLAIM-3D (3-Dimensional Cloud Aerosol Interaction Mission) satellite concept proposed here combines several techniques to simultaneously measure the vertical profile of cloud microphysics, thermodynamic phase, brightness temperature, and aerosol amount and type in the neighborhood of the clouds. The wide wavelength range, and the use of multi-angle polarization measurements proposed for this mission allow us to estimate the availability and characteristics of aerosol particles acting as cloud condensation nuclei, and their effects on the cloud microphysical structure. These results can provide unprecedented details on the response of cloud droplet microphysics to natural and anthropogenic aerosols in the size scale where the interaction really happens.

The impact of two coupled cirrus microphysics-radiation parameterizations on the temperature and specific humidity biases in the tropical tropopause layer in a climate model

The impact of two different coupled cirrus microphysics-radiation parameterizations on the zonally averaged temperature and humidity biases in the tropical tropopause layer (TTL) of a Met Office climate model configuration is assessed. One parameterization is based on a linear coupling between a model prognostic variable, the ice mass mixing ratio, qi, and the integral optical properties. The second is based on the integral optical properties being parameterized as functions of qi and temperature, Tc, where the mass coefficients (i.e. scattering and extinction) are parameterized as nonlinear functions of the ratio between qi and Tc. The cirrus microphysics parameterization is based on a moment estimation parameterization of the particle size distribution (PSD), which relates the mass moment (i.e. second moment if mass is proportional to size raised to the power of 2 ) of the PSD to all other PSD moments through the magnitude of the second moment and Tc. This same microphysics PSD parameterization is applied to calculate the integral optical properties used in both radiation parameterizations and, thus, ensures PSD and mass consistency between the cirrus microphysics and radiation schemes. In this paper, the temperature-non-dependent and temperature-dependent parameterizations are shown to increase and decrease the zonally averaged temperature biases in the TTL by about 1 K, respectively. The temperature-dependent radiation parameterization is further demonstrated to have a positive impact on the specific humidity biases in the TTL, as well as decreasing the shortwave and longwave biases in the cloudy radiative effect. The temperature-dependent radiation parameterization is shown to be more consistent with TTL and global radiation observations.

Numerical simulations of microphysical processes in pyro-convective clouds

Hochreichende Konvektion über Waldbränden ist eine der intensivsten Formen von atmosphärischer Konvektion. Die extreme Wolkendynamik mit hohen vertikalen Windgeschwindigkeiten (bis 20 m/s) bereits an der Wolkenbasis, hohen Wasserdampfübersättigungen (bis 1%) und die durch das Feuer hohen Anzahlkonzentration von Aerosolpartikeln (bis 100000 cm^-3) bilden einen besonderen Rahmen für Aerosol-Wolken Wechselwirkungen.Ein entscheidender Schritt in der mikrophysikalischen Entwicklung einer konvektiven Wolke ist die Aktivierung von Aerosolpartikeln zu Wolkentropfen. Dieser Aktivierungsprozess bestimmt die anfängliche Anzahl und Größe der Wolkentropfen und kann daher die Entwicklung einer konvektiven Wolke und deren Niederschlagsbildung beeinflussen. Die wichtigsten Faktoren, welche die anfängliche Anzahl und Größe der Wolkentropfen bestimmen, sind die Größe und Hygroskopizität der an der Wolkenbasis verfügbaren Aerosolpartikel sowie die vertikale Windgeschwindigkeit. Um den Einfluss dieser Faktoren unter pyro-konvektiven Bedingungen zu untersuchen, wurden numerische Simulationen mit Hilfe eines Wolkenpaketmodells mit detaillierter spektraler Beschreibung der Wolkenmikrophysik durchgeführt. Diese Ergebnisse können in drei unterschiedliche Bereiche abhängig vom Verhältnis zwischen vertikaler Windgeschwindigkeit und Aerosolanzahlkonzentration (w/NCN) eingeteilt werden: (1) ein durch die Aerosolkonzentration limitierter Bereich (hohes w/NCN), (2) ein durch die vertikale Windgeschwindigkeit limitierter Bereich (niedriges w/NCN) und (3) ein Übergangsbereich (mittleres w/NCN). Die Ergebnisse zeigen, dass die Variabilität der anfänglichen Anzahlkonzentration der Wolkentropfen in (pyro-) konvektiven Wolken hauptsächlich durch die Variabilität der vertikalen Windgeschwindigkeit und der Aerosolkonzentration bestimmt wird. rnUm die mikrophysikalischen Prozesse innerhalb der rauchigen Aufwindregion einer pyrokonvektiven Wolke mit einer detaillierten spektralen Mikrophysik zu untersuchen, wurde das Paketmodel entlang einer Trajektorie innerhalb der Aufwindregion initialisiert. Diese Trajektore wurde durch dreidimensionale Simulationen eines pyro-konvektiven Ereignisses durch das Model ATHAM berechnet. Es zeigt sich, dass die Anzahlkonzentration der Wolkentropfen mit steigender Aerosolkonzentration ansteigt. Auf der anderen Seite verringert sich die Größe der Wolkentropfen mit steigender Aerosolkonzentration. Die Reduzierung der Verbreiterung des Tropfenspektrums stimmt mit den Ergebnissen aus Messungen überein und unterstützt das Konzept der Unterdrückung von Niederschlag in stark verschmutzen Wolken.Mit Hilfe des Models ATHAM wurden die dynamischen und mikrophysikalischen Prozesse von pyro-konvektiven Wolken, aufbauend auf einer realistischen Parametrisierung der Aktivierung von Aerosolpartikeln durch die Ergebnisse der Aktivierungsstudie, mit zwei- und dreidimensionalen Simulationen untersucht. Ein modernes zweimomenten mikrophysikalisches Schema wurde in ATHAM implementiert, um den Einfluss der Anzahlkonzentration von Aerosolpartikeln auf die Entwicklung von idealisierten pyro-konvektiven Wolken in US Standardamtosphären für die mittleren Breiten und den Tropen zu untersuchen. Die Ergebnisse zeigen, dass die Anzahlkonzentration der Aerosolpartikel die Bildung von Regen beeinflusst. Für geringe Aerosolkonzentrationen findet die rasche Regenbildung hauptsächlich durch warme mikrophysikalische Prozesse statt. Für höhere Aerosolkonzentrationen ist die Eisphase wichtiger für die Bildung von Regen. Dies führt zu einem verspäteten Einsetzen von Niederschlag für verunreinigtere Atmosphären. Außerdem wird gezeigt, dass die Zusammensetzung der Eisnukleationspartikel (IN) einen starken Einfluss auf die dynamische und mikrophysikalische Struktur solcher Wolken hat. Bei sehr effizienten IN bildet sich Regen früher. Die Untersuchung zum Einfluss des atmosphärischen Hintergrundprofils zeigt eine geringe Auswirkung der Meteorologie auf die Sensitivität der pyro-konvektiven Wolken auf diernAerosolkonzentration. Zum Abschluss wird gezeigt, dass die durch das Feuer emittierte Hitze einen deutlichen Einfluss auf die Entwicklung und die Wolkenobergrenze von pyro-konvektive Wolken hat. Zusammenfassend kann gesagt werden, dass in dieser Dissertation die Mikrophysik von pyrokonvektiven Wolken mit Hilfe von idealisierten Simulation eines Wolkenpaketmodell mit detaillierte spektraler Mikrophysik und eines 3D Modells mit einem zweimomenten Schema im Detail untersucht wurde. Es wird gezeigt, dass die extremen Bedingungen im Bezug auf die vertikale Windgeschwindigkeiten und Aerosolkonzentrationen einen deutlichen Einfluss auf die Entwicklung von pyro-konvektiven Wolken haben.

Investigations of Cloud Microphysical Response to Mixing Using Digital Holography

Cloud edge mixing plays an important role in the life cycle and development of clouds. Entrainment of subsaturated air affects the cloud at the microscale, altering the number density and size distribution of its droplets. The resulting effect is determined by two timescales: the time required for the mixing event to complete, and the time required for the droplets to adjust to their new environment. If mixing is rapid, evaporation of droplets is uniform and said to be homogeneous in nature. In contrast, slow mixing (compared to the adjustment timescale) results in the droplets adjusting to the transient state of the mixture, producing an inhomogeneous result. Studying this process in real clouds involves the use of airborne optical instruments capable of measuring clouds at the `single particle' level. Single particle resolution allows for direct measurement of the droplet size distribution. This is in contrast to other `bulk' methods (i.e. hot-wire probes, lidar, radar) which measure a higher order moment of the distribution and require assumptions about the distribution shape to compute a size distribution. The sampling strategy of current optical instruments requires them to integrate over a path tens to hundreds of meters to form a single size distribution. This is much larger than typical mixing scales (which can extend down to the order of centimeters), resulting in difficulties resolving mixing signatures. The Holodec is an optical particle instrument that uses digital holography to record discrete, local volumes of droplets. This method allows for statistically significant size distributions to be calculated for centimeter scale volumes, allowing for full resolution at the scales important to the mixing process. The hologram also records the three dimensional position of all particles within the volume, allowing for the spatial structure of the cloud volume to be studied. Both of these features represent a new and unique view into the mixing problem. In this dissertation, holographic data recorded during two different field projects is analyzed to study the mixing structure of cumulus clouds. Using Holodec data, it is shown that mixing at cloud top can produce regions of clear but humid air that can subside down along the edge of the cloud as a narrow shell, or advect down shear as a `humid halo'. This air is then entrained into the cloud at lower levels, producing mixing that appears to be very inhomogeneous. This inhomogeneous-like mixing is shown to be well correlated with regions containing elevated concentrations of large droplets. This is used to argue in favor of the hypothesis that dilution can lead to enhanced droplet growth rates. I also make observations on the microscale spatial structure of observed cloud volumes recorded by the Holodec.

Hydrometeor and latent heat profiles of tropical cyclones Conson, Ivan and Catarina using PR/TRMM data

ABSTRACT Microphysical and thermodynamical features of two tropical systems, namely Hurricane Ivan and Typhoon Conson, and one sub-tropical, Catarina, have been analyzed based on space-born radar PR measurements available on the TRMM satellite. The procedure to classify the reflectivity profiles followed the Heymsfield et al (2000) and Steiner et al (1995) methodologies. The water and ice content have been calculated using a relationship obtained with data of the surface SPOL radar and PR in Rondonia State in Brazil. The diabatic heating rate due to latent heat release has been estimated using the methodology developed by Tao et al (1990). A more detailed analysis has been performed for Hurricane Catarina, the first of its kind in South Atlantic. High water content mean value has been found in Conson and Ivan at low levels and close to their centers. Results indicate that hurricane Catarina was shallower than the other two systems, with less water and the water was concentrated closer to its center. The mean ice content in Catarina was about 0.05 g kg-1 while in Conson it was 0.06 g kg-1 and in Ivan 0.08 g kg-1. Conson and Ivan had water content up to 0.3 g kg-1 above the 0ºC layer, while Catarina had less than 0.15 g kg-1. The latent heat released by Catarina showed to be very similar to the other two systems, except in the regions closer to the center.

RACORO extended-term, aircraft observations of boundary-layer clouds

A first-of-a-kind, extended-term cloud aircraft campaign was conducted to obtain an in-situ statistical characterization of continental boundary-layer clouds needed to investigate cloud processes and refine retrieval algorithms. Coordinated by the Atmospheric Radiation Measurement (ARM) Aerial Facility (AAF), the Routine AAF Clouds with Low Optical Water Depths (CLOWD) Optical Radiative Observations (RACORO) field campaign operated over the ARM Southern Great Plains (SGP) site from 22 January to 30 June 2009, collecting 260 h of data during 59 research flights. A comprehensive payload aboard the Center for Interdisciplinary Remotely-Piloted Aircraft Studies (CIRPAS) Twin Otter aircraft measured cloud microphysics, solar and thermal radiation, physical aerosol properties, and atmospheric state parameters. Proximity to the SGP's extensive complement of surface measurements provides ancillary data that supports modeling studies and facilitates evaluation of a variety of surface retrieval algorithms. The five-month duration enabled sampling a range of conditions associated with the seasonal transition from winter to summer. Although about two-thirds of the cloud flights occurred in May and June, boundary-layer cloud fields were sampled under a variety of environmental and aerosol conditions, with about 77% of the flights occurring in cumulus and stratocumulus. Preliminary analyses illustrate use of these data to analyze cloud-aerosol relationships, characterize the horizontal variability of cloud radiative impacts, and evaluate surface-based retrievals. We discuss how an extended-term campaign requires a simplified operating paradigm that is different from that used for typical, short-term, intensive aircraft field programs.

The COnvective Precipitation Experiment (COPE): investigating the origins of heavy precipitation in the southwestern UK

A recent field campaign in southwest England used numerical modeling integrated with aircraft and radar observations to investigate the dynamic and microphysical interactions that can result in heavy convective precipitation. The COnvective Precipitation Experiment (COPE) was a joint UK-US field campaign held during the summer of 2013 in the southwest peninsula of England, designed to study convective clouds that produce heavy rain leading to flash floods. The clouds form along convergence lines that develop regularly due to the topography. Major flash floods have occurred in the past, most famously at Boscastle in 2004. It has been suggested that much of the rain was produced by warm rain processes, similar to some flash floods that have occurred in the US. The overarching goal of COPE is to improve quantitative convective precipitation forecasting by understanding the interactions of the cloud microphysics and dynamics and thereby to improve NWP model skill for forecasts of flash floods. Two research aircraft, the University of Wyoming King Air and the UK BAe 146, obtained detailed in situ and remote sensing measurements in, around, and below storms on several days. A new fast-scanning X-band dual-polarization Doppler radar made 360-deg volume scans over 10 elevation angles approximately every 5 minutes, and was augmented by two UK Met Office C-band radars and the Chilbolton S-band radar. Detailed aerosol measurements were made on the aircraft and on the ground. This paper: (i) provides an overview of the COPE field campaign and the resulting dataset; (ii) presents examples of heavy convective rainfall in clouds containing ice and also in relatively shallow clouds through the warm rain process alone; and (iii) explains how COPE data will be used to improve high-resolution NWP models for operational use.

Relationship between Amazon biomass burning aerosols and rainfall over La Plata Basin

High aerosol loads are discharged into the atmosphere by biomass burning in Amazon and Central Brazil during the dry season. These particles can interact with clouds as cloud condensation nuclei (CCN) changing cloud microphysics and radiative properties and, thereby, affecting the radiative budget of the region. Furthermore, the biomass burning aerosols can be transported by the low level jet (LLJ) to La Plata Basin where many mesoscale convective systems (MCS) are observed during spring and summer. This work proposes to investigate whether the aerosols from biomass burning may affect the MCS in terms of rainfall over La Plata Basin during spring. Since the aerosol effect is very difficult to isolate because convective clouds are very sensitive to small environment disturbances, detailed analyses using different techniques are used. The binplot, 2D histograms and combined empirical orthogonal function (EOF) methods are used to separate certain environment conditions with the possible effects of aerosol loading. Reanalysis 2, TRMM-3B42 and AERONET data are used from 1999 up to 2012 during September-December. The results show that there are two patterns associated to rainfall-aerosol interaction in La Plata Basin: one in which the dynamic conditions are more important than aerosols to generate rain; and a second one where the aerosol particles have a role in rain formation, acting mainly to suppress rainfall over La Plata Basin.

Airborne in situ measurements of ice particles in the tropical tropopause layer

Ice clouds have a strong effect on the Earth-atmosphere radiative energy balance, on the distribution of condensable gases in the atmosphere, as well as on the chemical composition of the air. The ice particles in these clouds can take on a variety of shapes which makes the description of the cloud microphysical properties more difficult. In the tropical upper troposphere/lower stratosphere (UTLS), a region where ice cloud abundance is relatively high, different types of ice clouds can be observed. However, in situ measurements are rare due to the high altitude of these clouds and the few available research aircraft, only three worldwide, that can fly at such altitudes.rnThis work focuses on in situ measurements of the tropical UTLS clouds performedrnwith a Cloud Imaging Probe (CIP) and a Forward Scattering Spectrometer Probern(FSSP-100), whereof the CIP is the key instrument of this thesis. The CIP is anrnairborne in situ instrument that obtains two-dimensional shadow images of cloud particles. Several cloud microphysical parameters can be derived from these measurements, e.g. number concentrations and size distributions. In order to obtain a high quality data set, a careful image analysis and several corrections need to be applied to the CIP observations. These methods are described in detail.rnMeasurements within the tropical UTLS have been performed during two campaigns:rnSCOUT-O3, 2005 in Northern Australia and SCOUT-AMMA, 2006 inWest Africa. Thernobtained data set includes first observations of subvisible cirrus clouds over a continental area and observations of the anvils of deep convective clouds. The latter can be further divided into clouds in mesoscale convective system outflows of different ages and clouds in overshooting cloud turrets that even penetrated the stratosphere. The microphysical properties of these three cloud types are discussed in detail. Furthermore, the vertical structure of the ice clouds in the UTLS is investigated. The values of the microphysical parameters were found to decrease with increasing altitude in the upper troposphere. Particle numbers and maximum sizes were also decreasing with increasing age of the outflow clouds. Further differences between the deep convective clouds and subvisible cirrus were found in the particle morphology as well as in the ratio of the observed aerosol particles to cloud particles which indicates that the different freezing processes (deposition, contact, immersion freezing) play different roles in the formation of the respective clouds. For the achievementrnof a better microphysical characterisation and description numerical fits have been adjusted onto the cloud particle size distributions of the subvisible cirrus as well as on the size distributions of the clouds at different altitudes in the UTLS.

Aerosol properties associated with air masses arriving into the North East Atlantic during the 2008 Mace Head EUCAARI intensive observing period: an overview

As part of the EUCAARI Intensive Observing Period, a 4-week campaign to measure aerosol physical, chemical and optical properties, atmospheric structure, and cloud microphysics was conducted from mid-May to mid-June, 2008 at the Mace Head Atmospheric Research Station, located at the interface of Western Europe and the N. E. Atlantic and centered on the west Irish coastline. During the campaign, continental air masses comprising both young and aged continental plumes were encountered, along with polar, Arctic and tropical air masses. Polluted-continental aerosol concentrations were of the order of 3000 cm(-3), while background marine air aerosol concentrations were between 400-600 cm(-3). The highest marine air concentrations occurred in polar air masses in which a 15 nm nucleation mode, with concentration of 1100 cm(-3), was observed and attributed to open ocean particle formation. Continental air submicron chemical composition (excluding refractory sea salt) was dominated by organic matter, closely followed by sulphate mass. Although the concentrations and size distribution spectral shape were almost identical for the young and aged continental cases, hygroscopic growth factors (GF) and cloud condensation nuclei (CCN) to total condensation nuclei (CN) concentration ratios were significantly less in the younger pollution plume, indicating a more oxidized organic component to the aged continental plume. The difference in chemical composition and hygroscopic growth factor appear to result in a 40-50% impact on aerosol scattering coefficients and Aerosol Optical Depth, despite almost identical aerosol microphysical properties in both cases, with the higher values been recorded for the more aged case. For the CCN/CN ratio, the highest ratios were seen in the more age plume. In marine air, sulphate mass dominated the sub-micron component, followed by water soluble organic carbon, which, in turn, was dominated by methanesulphonic acid (MSA). Sulphate concentrations were highest in marine tropical air - even higher than in continental air. MSA was present at twice the concentrations of previously-reported concentrations at the same location and the same season. Both continental and marine air exhibited aerosol GFs significantly less than ammonium sulphate aerosol pointing to a significant organic contribution to all air mass aerosol properties.

Computação de alto desempenho aplicada a modelagem numérica de fenômenos atmosféricos

A constante evolução da tecnologia disponibilizou, atualmente, ferramentas computacionais que eram apenas expectativas há 10 anos atrás. O aumento do potencial computacional aplicado a modelos numéricos que simulam a atmosfera permitiu ampliar o estudo de fenômenos atmosféricos, através do uso de ferramentas de computação de alto desempenho. O trabalho propôs o desenvolvimento de algoritmos com base em arquiteturas SIMT e aplicação de técnicas de paralelismo com uso da ferramenta OpenACC para processamento de dados de previsão numérica do modelo Weather Research and Forecast. Esta proposta tem forte conotação interdisciplinar, buscando a interação entre as áreas de modelagem atmosférica e computação científica. Foram testadas a influência da computação do cálculo de microfísica de nuvens na degradação temporal do modelo. Como a entrada de dados para execução na GPU não era suficientemente grande, o tempo necessário para transferir dados da CPU para a GPU foi maior do que a execução da computação na CPU. Outro fator determinante foi a adição de código CUDA dentro de um contexto MPI, causando assim condições de disputa de recursos entre os processadores, mais uma vez degradando o tempo de execução. A proposta do uso de diretivas para aplicar computação de alto desempenho em uma estrutura CUDA parece muito promissora, mas ainda precisa ser utilizada com muita cautela a fim de produzir bons resultados. A construção de um híbrido MPI + CUDA foi testada, mas os resultados não foram conclusivos.