AIR POLLUTION MODEL

Nowadays VŠB – TU Ostrava is realizing uniquely extensive calculations within this task. The calculations are being carried out on the parallel clusters of MetaCentrum called CESNET (https://www.metacentrum.cz/en/about/) and the IT4Innovation computing centre at VSB (https://www.it4i.cz/?lang=en). MetaCentrum operates the distributed computing infrastructure which enables the use of computing and data sources to solve very demanding computational tasks that overcome the possibilities of respective workplaces in the Czech Republic. Under IT4Innovation, the project AIR TRITIA was granted by 10 000 core hours (https://www.it4i.cz/computing-resources-allocation/11th-open-access-grant-competition-results/?lang=en).

obr1
obr2

IT4 Innovation computing centre at VSB – TU Ostrava

The most demanding part of the modelling regarding the computing power is the calculation of pollution from domestic boilers from the Polish part of the TRITIA region.

obr 3

Domestic boilers

Here are interesting details of the computational task:

  • The calculation includes emissions from domestic boilers in the Opole, Silesian and Malopolska Voivodeships. This area comprises 8.9 million people and 1 652 thousands of local fireplaces. The calculation is performed in a rectangular area of 250 x 310 km.
  • For the purposes of modelling, the domestic boilers are represented by 317 800 100-meter surface sources.
  • The modelling is completed for 4 pollutants (PM10, PM2.5, NO2 and benzo[a]pyrene) and 3 model years (2006, 2010, 2015), thus 12 modelling tasks.
  • Each modelling task is divided into 6015 subtasks, which are computed in parallel. In this case it is necessary to solve 72,180 subtasks in total.
  • The used model calculates the contribution of individual sources in the network of receptor points covering the model area. Altogether 81.6 billion pollutions from the sources at the receptor points has to be evaluated, that is ~ 3.23 * 10 ^ 14 (323 trillion) calculations of the function describing the pollution distribution from sources.
  • The calculation is performed simultaneously at 175 computing nodes, where each computation node contains 8 processors. The total calculation time is more than 1 000 processor-days. On a top PC, a similar calculation would take approximately 3 years, at MetaCentrum 3 days.
  • The unprocessed modelling results take approximately 110 GB of HDD space. These results are subsequently processed in 72 180 subtasks. The outcome of each subtask is interpolated into a regular grid of 100 m covering the modelled area and summarised into the final bitmap.

These results together with results of pollution modelling from domestic boilers at the Czech and Slovak part of the region, from industrial sources and car traffic will be the basis for further analysis of the air quality, health risks and remedies.