COMSIMM - Community Simulations of the last Millennium

Knowledge of past climate variability is crucial for understanding current and future climate trends. For the first time, sufficient computational resources are available to carry out millennial-scale simulations with a comprehensive Earth System Model. Within the Community Earth System Modelling (http://cosmos.enes.org) initiative, a community effort has been initiated to carry out such simulations using the COSMOS model.
Performing millennia-long integrations in ensemble mode is still a challenge, both in terms of required computing time, data transfer systems, and storage capacities. Innovative infrastructure and sophisticated environments, such as offered by DEISA, is necessary to perform and manage such a project.

More about Comsimm >>

CrossGrid

The CrossGrid project is oriented towards compute- and data-intensive applications that are characterized by the interaction with a person in a processing loop. Such applications require a response from the Grid to an action by a human agent in different time scales.

To see more information >>

DynlowPm - Dynamo effect and magnetic induction
at low magnetic Prandtl number

Primarily, we investigated numerical approaches to the dynamo problem at low magnetic Prandtl numbers. The difficulty of resolving a large range of scales is circumvented by combining direct numerical simulations and large-eddy simulations. The flow is generated by the Taylor-Green forcing; it combines a well defined structure at large scales and turbulent fluctuations at small scales.
Mainly, we want to study the statistical properties of the magnetic fluctuation of the induction response of an imposed magnetic field just below the dynamo threshold.

More about Dynolowpm >>

Regional Climate Simulations with a CAM+WRF Sequence
Sensitivity Studies to El Niño Phenomenone

The climate application of the EELA project deals with El Niño, which is a key climate factor for Latin-American countries. Understanding the local effects and sensitivities of this phenomenon requires a combination of global (~200 Km resolution) and regional (~30 Km) climate models run in sequence with precise dependencies.
For this purpose, we selected a global atmospheric model (Community Atmospheric Model, CAM, http://www.ccsm.ucar.edu/models/) and a regional model (Weather Research and Forecasting Model, WRF, http://www.wrf-model.org/), and deployed middleware solutions and portal facilities to run and monitor sequences of jobs with these models for an efficient exploration of El Niño effects in regional domains based on ensembles of hundreds of independent simulations. Global climate simulations are performed with the CAM model using perturbed Sea Surface Temperature (SST) conditions. In parallel, regional WRF simulations are automatically submitted to "zoom" over a zone as soon as partial CAM outputs are produced for prescribed periods.
These model simulations involve managing a complex workflow including long-term jobs and job dependencies in a user-transparent way. On one hand, a single climate simulation may last longer than the limit of the local workload manager system (usually 2 CPU-days). This requires the restart and reallocation of the jobs. On the other hand, the regional model uses as boundary conditions the output of the global model and, therefore, an event-oriented system should be used in order to wait for the global model outputs. In doing so, we identified the weaknesses of current middleware tools and developed a robust workflow by merging the optimal existing applications with an underlying self-developed workflow application. This application is based on the communication with metadata catalogs (currently AMGA) storing application status and dynamic model output generation.
As an illustrative scientific challenge, the application is applied to study the El Niño phenomenon, by simulating an El Niño year with different SST forcing conditions and analyzing the precipitation response over Latin-American countries subject to flooding risk.

Grid properties/requirements:

• Very long (days to weeks) for global simulations and DAGs for regional simulations
• Metadata catalogs (currently AMGA) for workflow management
• Large storage requirements

More about Regional Climate Simulations

GeoSys - Advanced Geosystem Analysis

Computer simulations have become a helpful tool in geosciences. They provide insight into areas which are difficult to access without these simulations. GeoSys is an application that is designed to simulate relevant processes in the area of geothermal processes and hydrogeology.
This project aims at simulating highly complex geological models that contain many geological processes.

More about Geosys >>

ICAROS - Innovative Chemical Assimilation
of Remote Sensing Observations of the Stratosphere

To better characterize errors in chemical reanalyses of stratospheric ozone derived from remote sensing data a combination of four-dimensional variational (4Dvar) data assimilation with model ensemble calculations is proposed. In 4Dvar the initial conditions for a model forecast are adjusted to better comply with observations for the whole assimilation time window (e.g., 24 hours). Taking into account the known instrument and model uncertainties the most likely distribution of trace gases is derived.
The resulting analyses are a prerequisite to study chemistry and dynamics of the stratosphere. In close cooperation DLR and national partners have developed a novel 4Dvar chemical data assimilation system within the BMBF project SACADA (Synoptic analysis of Chemical Constituents by Advanced Data Assimilation). Meteorological and chemical differential equations are solved in the physical domain, pioneering a novel icosahedral spherical grid.

More about Icaros >>

Stratospheric ozone in polar regions

The first goal of this application is the prediction, in quasi-real time, of the ozone concentration in the stratospheric polar regions and is related to the time evolution of the polar ozone hole.
For each day during winter time and spring time (Arctic : 11/01 to 30/04 and Antarctic : 04/01 to 30/11) the computation of ozone concentration is achieved by running a simulation in both polar areas, Arctic and Antarctic. That simulation is based on a chemical-transport model called MIMOSA-CHIM using the daily meteorological outputs from the ECMWF (European Centre of Medium range Weather Forecasting), and the output of another simulation, called REPROBUS (Reactive Processes Ruling the Ozone Budget in the Stratosphere) for the initialization and boundary conditions. The results are the winter daily concentrations of around 10 constituents involved in the ozone photochemistry.
Then, the second aspect is to determine the trend of the ozone concentration in the last decades in both Arctic and Antarctic regions during winter and spring time, periods during which chemical destruction due to elevated halogen levels in the stratosphere takes place. In order to do that, one will run the simulation for every winter and both the poles during the 1985-2005 period, that is why the process for the daily prediction needs to be well adjusted first.

Grid properties/requirements:

• Currently being ported to EGEE
• High security and restricted access for data
• Licenced software

Validation of GOME ozone profiles

The GOME experiment aboard ERS2 provides total ozone contents.The GOME data were processed with three different algorithms, one based on inversion, OPERA (KNMI), and two versions of a Neural Network algorithm (ESA, Univ. Tor Vergata). The output results are the ozone profiles as a function of the altitude.
The ozone profiles retrieved from GOME spectra from January 1996 to June 2003 by means of neural network techniques have been validated with ozone lidar measurements performed in different lidar stations (7) belonging to the Network for Detection of Stratospheric Changes (NDSC). Nearly 2000 collocated profiles have been found, in tropical, mid-latitude and high-latitude regions; for each lidar station the relative difference between GOME and lidar profiles, as well as the seasonal variation of the relative difference, have been evaluated.

Grid properties/requirements:

• Large number of files (~40000 for 7 years of data)
• Metadatabase
• Geospatial search in the database

More about Gome >>

Korba aquifer

This application concerns the sustainable management of groundwater exploitation using Monte Carlo simulation of seawater intrusion in the Korba aquifer (Tunisia).
Coastal areas are the most densely-populated areas in our planet. In those zones, underground freshwater constitutes a vital resource, not only for populations and ecosystems but also and especially for agriculture and therefore for the economy. At the same time, groundwater resources in these areas are intensively exploited despite their extreme vulnerability to salinization by seawater intrusion.
In our application, we aim at establishing probabilistic modelling techniques to estimate uncertainty when forecasting the impact of potential management decisions on the long term evolution of the groundwater resources in coastal irrigated plains. The Stochastic modelling allows to estimate the uncertainty on all of the physical parameters used in a deterministic way to model the groundwater system and the effect of this uncertainty on the model results. This technique provides accurate tools to investigate optimal scenarios of the sustainable groundwater resources. Monte Carlo simulations are very often used.
The map of probability is calculated using the outputs 100 Monte Carlo simulations of flow and transport. Monte Carlo simulations are very often used to analyze the uncertainty of model parameters. In general we use 100 MC simulations, but the higher is the number of (equiprobables) those realizations, the better is the convergence of the statistical parameters. The map of probability allows to decision makers to investigate optimal management scenarios for sustainable groundwater resources.

Grid properties/requirements:

• Large dataset volume
• MPI
• Long running jobs
• Virtual data model
• Monitoring needs
• Collaborative environment
• Use of licensed software (not yet on the grid)

Contact the author for more information.

Flood Forecasting

Floods are among the most frequent and costly natural disasters in terms of human hardship and economic loss. As much as 90 percent of the damage related to natural disasters (excluding droughts) is caused by floods and associated mud and debris flows. Timely warnings and forecasts save lives and aid disaster preparedness, which decreases damage.
Whereas many weather phenomena have specific geographical locations where they occur, rainfall is an event that occurs virtually everywhere. If rainfall is possible, there are going to be occasions when it becomes intense and that intensity is maintained long enough to create the potential for floods and even flash floods. Hydrology plays a large role in the flood problem; a given amount of rainfall in a given time may or may not result in a flood, owing to such factors as antecedent precipitation, soil permeability, terrain gradients, and so on.
The Flood Forecasting Simulation Cascade (FFSC) is a hydro-meteorological application, which tries to predict oncoming floods based on series of simulations of meteorological, hydrological and hydraulic conditions in the target area. The application consists of several simulation steps (with more choices of used models for each step), attached pre-processing and post-processing and visualization tools.
The models used are the following:

• Meteorological models ALADIN (MPI-parallel), MM5 (MPI-parallel), hydrological models HSPF (sequential-parametric), NLC (sequential-parametric), hydraulic models DaveF (MPI-parallel), FESWMS (MPI-parallel)
• Input data: boundary condition of meteorology, river network, digital terrain maps
• Output: weather forecasting, precipitation forecasting, hydrograph, water level and velocity of flooding area

Grid properties/requirements:

• Complex workflow
• MPI
• Knowledge management

Non- thermal sources of hot oxygen
in the Martian upper atmosphere

Due to absence of a planetary magnetic field at mars, coronal ionized particles accelerated by the solar wind electric field can impact the atmospheric background with energies around 1 keV. The energy deposited during the binary collisions is transferred to neutral atoms and molecules that in some cases can escape the planetary gravity field. This phenomena is known as sputtering. Another process that produces energetic neutral atoms, is the dissociative recombination.
The different processes are simulated with a test particle type code (each simulated particle represent roughly 10 26 real particles). Trajectories of hot atoms and molecules under the action of gravity are calculated in the upper atmospheric background (bottom level at 100 km) up to 2 Martian radius. Typically, the movement of around 200 000 particles are followed in the gravitation field at each time step. We use a Direct Simulation Monte Carlo scheme to evaluate the maximum number of binary collisions between each type of hot particle and each component of the background gas (O, CO2, CO), in each of the 30 atmospheric cells. We use this Monte Carlo procedure to calculate the density versus altitude profiles of O, CO, CO2 and C, as well as their velocity and energy distributions.
4 atmospheric models and two solar configurations were tested that may not give the same results for each process.

Grid properties/requirements:

• Very long job (1 day to 3 days)

Meteorological – Air quality applications for GRID

The Atmospheric Modeling and Weather Forecasting Group (AM&WFG) has various activities ranging from atmospheric, air quality, sea waves and climate model development and operations. The operations are in hindcasting and forecasting mode, all related to parallel processing. The major modeling systems used in development and operations are: the weather and dust forecasting system SKIRON, the limited area model RAMS, the atmospheric chemistry and transport model CAMx and the wave analysis and prediction model WAM. The SKIRON system has been already ported in GRID platforms. Recently, the RAMS modeling system was ported for the requirements of the CIRCE climate study project (www.circeproject.eu). In CIRCE, a new generation limited area model is been developed based on RAMS, with unique capabilities such as the direct coupling of gas, aqueous, aerosol chemical processes that are resolved explicitly. The development of the new model requires computer power that might be provided by the GRID platforms. The future plans of AM&WFG related to GRID include the porting of the other modeling systems (CAMx, WAM) and the use of GRID storage facilities.

Grid properties/requirements:

• Duration, up to 6h/job

• Small storage requirements (aprox. 0.5G/job)

• MPI

AM&WFG home http://forecast.uoa.gr