Massive-Parallel Trajectory Calculations version 2.2 (MPTRAC-2.2): Lagrangian transport simulations on Graphics Processing Units (GPUs)

Lagrangian models are fundamental tools to study atmospheric transport processes and for practical applications such as dispersion modeling for anthropogenic and natural emission sources. However, conducting large-scale Lagrangian transport simulations with millions of air parcels or more can become numerically rather costly. In this study, we assessed the potential of exploiting graphics processing units (GPUs) to accelerate Lagrangian transport simulations. We ported the Massive-Parallel Trajectory Calculations (MPTRAC) model to GPUs using the open accelerator (OpenACC) programming model. The trajectory calculations conducted within the MPTRAC model were fully ported to GPUs, i.e., except for feeding in the meteorological input data and for extracting the particle output data, the code operates entirely on the GPU devices without frequent data transfers between CPU and GPU memory. Model verification, performance analyses, and scaling tests of the MPI/OpenMP/OpenACC hybrid parallelization of MPTRAC were conducted on the JUWELS Booster supercomputer operated by the Jülich Supercomputing Centre, Germany. The JUWELS Booster comprises 3744 NVIDIA A100 Tensor Core GPUs, providing a peak performance of 71.0 PFlop/s. As of June 2021, it is the most powerful supercomputer in Europe and listed among the most energy-efficient systems internationally. For large-scale simulations comprising 108 particles driven by the European Centre for Medium-Range Weather Forecasts' ERA5 reanalysis, the performance evaluation showed a maximum speedup of a factor of 16 due to the utilization of GPUs compared to CPU-only runs on the JUWELS Booster. In the large-scale GPU run, about 67 % of the runtime is spent on the physics calculations, conducted on the GPUs. Another 15 % of the runtime is required for file-I/O, mostly to read the large ERA5 data set from disk. Meteorological data preprocessing on the CPUs also requires about 15 % of the runtime. Although this study identified potential for further improvements of the GPU code, we consider the MPTRAC model ready for production runs on the JUWELS Booster in its present form. The GPU code provides a much faster time to solution than the CPU code, which is particularly relevant for near-real-time applications of a Lagrangian transport model.

Massive-Parallel Trajectory Calculations version 2.2 (MPTRAC-2.2): Lagrangian transport simulations on Graphics Processing Units (GPUs)

Lagrangian models are powerful tools to study atmospheric transport processes. However, conducting large-scaleLagrangian transport simulations with many air parcels can become numerically rather costly. In this study, we assessed the potential of exploiting graphics processing units (GPUs) to accelerate Lagrangian transport simulations. We ported the Massive-Parallel Trajectory Calculations (MPTRAC) model to GPUs using the open accelerator (OpenACC) programming model. The trajectory calculations conducted within the MPTRAC model have been fully ported to GPUs, i. e., except for feeding in the meteorological input data and for extracting the particle output data, the code operates entirely on the GPU devices without frequent data transfers between CPU and GPU memory. Model verification, performance analyses, and scaling tests of the MPI/OpenMP/OpenACC hybrid parallelization of MPTRAC have been conducted on the JUWELS Booster supercomputer operated by the Jülich Supercomputing Centre, Germany. The JUWELS Booster comprises 3744 NVIDIA A100 Tensor CoreGPUs, providing a peak performance of 71.0 PFlop/s. As of June 2021, it is the most powerful supercomputer in Europe and listed among the most energy-efficient systems internationally. For large-scale simulations comprising 100 million particles driven by the European Centre for Medium-Range Weather Forecasts’ ERA5 reanalysis, the performance evaluation showed a maximum speedup of a factor of 16 due to the utilization of GPUs compared to CPU-only runs on the JUWELS Booster. In the large-scale GPU run, about 67 % of the runtime is spent on the physics calculations, being conducted on the GPUs. Another 15 % of the runtime is required for file-I/O, mostly to read the ERA5 data from disk. Meteorological data preprocessing on the CPUs also requires about 15 % of the runtime. Although this study identified potential for further improvements of the GPU code, we consider the MPTRAC model to be ready for production runs on the JUWELS Booster in its present form. The GPU code provides a much faster time to solution than the CPU code, which is particularly relevant for near-real-time applications of a Lagrangian transport model

Improved estimation of volcanic SO₂ injections from satellite observations and Lagrangian transport simulations: the 2019 Raikoke eruption

Monitoring and modeling of volcanic plumes is important for understanding the impact of volcanic activity on climate and for practical concerns, such as aviation safety or public health. Here, we applied the Lagrangian transport model Massive-Parallel Trajectory Calculations (MPTRAC) to estimate the SO2 injections into the upper troposphere and lower stratosphere by the eruption of the Raikoke volcano (48.29° N, 153.25° E) in June 2019 and its subsequent long-range transport and dispersion. First, we used SO2 observations from the AIRS (Atmospheric Infrared Sounder) and TROPOMI (TROPOspheric Monitoring Instrument) satellite instruments together with a backward trajectory approach to estimate the altitude-resolved SO2 injection time series. Second, we applied a scaling factor to the initial estimate of the SO2 mass and added an exponential decay to simulate the time evolution of the total SO2 mass. By comparing the estimated SO2 mass and the observed mass from TROPOMI, we show that the volcano injected 2.1 ± 0.2 Tg SO2 and the e-folding lifetime of the SO2 was about 13 to 17 days. The reconstructed injection time series are consistent between the AIRS nighttime and the TROPOMI daytime measurements. Further, we compared forward transport simulations that were initialized by AIRS and TROPOMI satellite observations with a constant SO2 injection rate. The results show that the modeled SO2 change, driven by chemical reactions, captures the SO2 mass variations observed by TROPOMI. In addition, the forward simulations reproduce the SO2 distributions in the first ~10 days after the eruption. However, diffusion in the forward simulations is too strong to capture the internal structure of the SO2 clouds, which is further quantified in the simulation of the compact SO2 cloud from late July to early August. Our study demonstrates the potential of using combined nadir satellite observations and Lagrangian transport simulations to further improve SO2 time- and height-resolved injection estimates of volcanic eruptions.

Experimental Investigation of Lagrangian Coherent Structures and Lobe Dynamics in Perturbed Rayleigh-Benard Convection

It is well known that Rayleigh-Benard convection with perturbations yields Lagrangian chaotic transport, and the mechanism of inducing chaotic transport has been numerically clarified by lobe dynamics [2]. On the other hand, the mechanism of such Lagrangian transport has not been enough studied by experiments. In our previous work [16], we made an experimental study to investigate the Lagrangian transport appeared in the two-dimensional Rayleigh-Benard convection by giving oscillation on the velocity fields and showed that there exist Lagrangian Coherent Structures (LCSs) which correspond to invariant manifolds of non-autonomous systems. We also showed that the LCSs entangle with each other around cell boundaries. In this paper, we further explore the global invariant structures of the perturbed Rayleigh-Benard convection by clarifying the details on the LCSs and explain how the fluid transport obeys lobe dynamics. Finally, we propose a novel Hamiltonian model for the two-dimensional perturbed Rayleigh-Benard convection that enables to elucidate the global structures detected by experiments.

Connectivity networks and delineation of distinct coastal provinces along the Indian coastline using large-scale Lagrangian transport simulations

Ocean circulation defines the scale of population connectivity in marine ecosystems, and is essential for conservation planning. We performed Lagrangian transport simulations and built connectivity networks to understand the patterns of oceanographic connectivity along the Indian coastline. In these networks, nodes are coastal polygons and the edges connecting them represent the magnitude of larval transfer between them. We assessed the variation in connectivity networks within and between two monsoonal seasons, across El Nino-Southern Oscillation (ENSO) years and for pelagic larval durations (PLD) up to 50 days. We detected well-connected communities, mapped frequent connectivity breaks and ranked coastal areas by their functional role using network centrality measures. Network characteristics did not differ based on the ENSO year, but varied based on season and PLD. Large scale connectance (entire Indian coastline) was small, ranging from 0.5% to 3.4%, and the number of cohesive coastal communities decreased from 60 (PLD <4 days) to 30 (PLD >20 days) with increasing PLD. Despite intra-seasonal variation in connectivity breaks, four disconnected provinces were consistently identified across the entire PLD range, which partially overlapped with observed genetic and biogeographic breaks along the Indian coastline. Our results support the adoption of an adaptive regional management framework guided by fine-scale analysis of connectivity within the four provinces delineated in the present study. A few sites within each province displayed notably higher centrality values than other nodes of the network, but showed variation with season and PLD, and could be targeted for national and transnational conservation and management plans.

Incorporating features from the Stochastic Time-Inverted Lagrangian Transport (STILT) model into the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model: a unified dispersion model for time-forward and time-reversed applications

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model is a state-of-the-science atmospheric dispersion model that is developed and maintained at the National Oceanic Atmospheric Administration’s (NOAA) Air Resources Laboratory (ARL). In the early 2000s, HYSPLIT served as the starting point for development of the Stochastic Time-Inverted Lagrangian Transport (STILT) model that emphasizes backward-in-time dispersion simulations to determine source regions of receptors. STILT continued its separate development and gained a wide user base. Since STILT was built on a now outdated version of HYSPLIT and lacks long-term institutional support to maintain the model, incorporating STILT features into HYSPLIT allows these features to stay up to date. This paper describes the STILT features incorporated into HYSPLIT, which include: a new vertical interpolation algorithm for WRF derived meteorological input files, a detailed algorithm for estimating boundary layer height, a new turbulence parameterization, a vertical Lagrangian timescale that varies in time and space, a complex dispersion algorithm, and two new convection schemes. An evaluation of these new features was performed using tracer release data from the Cross Appalachian Tracer Experiment and the Across North America Tracer Experiment. Results show the dispersion module from STILT, which takes up to double the amount of time to run, is less dispersive in the vertical and in better agreement with observations than the existing HYSPLIT option. The other new modeling features from STILT were not consistently statistically different than existing HYSPLIT options. Forward-time simulations from the new model were also compared against backward-time equivalents and found to be statistically comparable to one another.

Improved estimation of volcanic SO2 emissions from satellite observations and Lagrangian transport simulations: The 2019 Raikoke eruption case study

&lt;p&gt;Monitoring and modeling of volcanic aerosols is important for understanding the influence of volcanic activity on climate. Here, we applied the Lagrangian transport model Massive-Parallel Trajectory Calculations (MPTRAC) to estimate the total injected SO2 by the stratosphere reaching eruption of the Raikoke volcano (48N, 153E) in June 2019 and its subsequent transport. We used SO2 observations from the AIRS and TROPOMI satellite instruments together with a backward trajectory approach to estimate the altitude-resolved SO2 emission timeseries. Then we applied a scaling factor to the initial estimate of the SO2 mass and added an exponential decay to simulate the time evolution of the total SO2 mass. By comparing the estimated SO2 mass and the observed mass from TROPOMI, we show that the volcano injected 2.1&amp;#177;0.2 Tg SO2 and the e-folding lifetime of the SO2 was about 13~17 days. Further, we compared simulations that were initialized by AIRS and TROPOMI satellite observations with a constant SO2 emission rate. The results show that the model captures the SO2 distributions in the first ~10 days after the eruption. The simulations using AIRS nighttime and TROPOMI measurements show comparable results and model skills which outperform the simulation using a constant emission rate. Our study demonstrates the potential of using combined satellite observations and transport simulations to further improve SO2 time- and height-resolved emission estimates of volcanic eruptions.&lt;/p&gt;