BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20220812T074334Z
LOCATION:Sydney Room
DTSTART;TZID=Europe/Stockholm:20220627T111500
DTEND;TZID=Europe/Stockholm:20220627T114500
UID:submissions.pasc-conference.org_PASC22_sess171_pap124@linklings.com
SUMMARY:Parallel Space-Time Likelihood Optimization for Air Pollution Pred
 iction on Large-Scale Systems
DESCRIPTION:Paper\n\nParallel Space-Time Likelihood Optimization for Air P
 ollution Prediction on Large-Scale Systems\n\nSalvaña, Abdulah, Ltaief, Su
 n, Genton...\n\nGaussian geostatistical space-time modeling is an effectiv
 e tool for performing statistical inference of data evolving in space and 
 time, generalizing spatial modeling alone at the cost of the greater compl
 exity of operations and storage, and pushing geostatistical modeling even 
 further into the arms of high-performance computing. It makes inferences f
 or missing data by leveraging space-time measurements of one or more field
 s. We propose a high-performance implementation of a space-time model for 
 large-scale systems using a two-level parallelization technique. At the in
 ner level, we rely on parallel linear algebra libraries and runtime system
 s to perform complex matrix operations required to evaluate the maximum li
 kelihood estimation (MLE). At the outer level, we parallelize the optimiza
 tion process using a distributed implementation of the particle swarm opti
 mization (PSO) algorithm. At this level, parallelization is accomplished u
 sing MPI sub-communicators, such that the nodes in each sub-communicator p
 erform a single MLE iteration at a time. We assess the accuracy of the imp
 lemented space-time model on large-scale synthetic space-time datasets. Mo
 reover, we use the proposed implementation to model two air pollution data
 sets from the Middle East and US regions with 550 spatial locations X 730 
 time slots and 945 spatial locations X 500 time slots, respectively. The e
 valuation shows that the proposed implemntation satisfies high prediction 
 accuracy on both synthetic and real particulate matter (PM) datasets in th
 e context of the air pollution problem. We achieve up to 757.16 TFLOPS/s u
 sing 1024 nodes (75% of the peak performance) using 490K geospatial locati
 ons on a Cray XC40 system.\n\nDomain: Climate, Weather and Earth Sciences
END:VEVENT
END:VCALENDAR