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:20190719T085743Z
LOCATION:HG E 1.1
DTSTART;TZID=Europe/Stockholm:20190612T114500
DTEND;TZID=Europe/Stockholm:20190612T121500
UID:submissions.pasc-conference.org_PASC19_sess109_pap_jan129@linklings.co
 m
SUMMARY:A Multivariate Approach to Ensure Statistical Reproducibility of C
 limate Model Simulations
DESCRIPTION:Paper\nChemistry and Materials, Climate and Weather\n\nA Multi
 variate Approach to Ensure Statistical Reproducibility of Climate Model Si
 mulations\n\nMahajan, Evans, Kennedy, Norman, Xu\n\nEffective utilization 
 of novel hybrid architectures of pre-exascale and exascale machines requir
 es transformations to global climate modeling systems that may not reprodu
 ce the original model solution bit-for-bit. Round-off level differences gr
 ow rapidly in these non-linear and chaotic systems. This makes it difficul
 t to isolate bugs/errors from innocuous growth expected from round-off lev
 el differences. Here, we apply two modern multivariate two sample equality
  of distribution tests to evaluate statistical reproducibility of global c
 limate model simulations using standard monthly output of short (~ 1-year)
  simulation ensembles of a control model and a modified model of US Depart
 ment of Energy's Energy Exascale Earth System Model (E3SM). Both the tests
  are able to identify changes induced by modifications to some model tunin
 g parameters. We also conduct formal power analyses of the tests by applyi
 ng them on designed suites of short simulation ensembles each with an incr
 easingly different climate from the control ensemble. The results are comp
 ared against those from another such test. These power analyses provide a 
 framework to quantify the degree of differences that can be detected confi
 dently by the tests for a given ensemble size (sample size). This will all
 ow model developers using the tests to make an informed decision when acce
 pting/rejecting an unintentional non-bit-for-bit change to the model solut
 ion.<br /><br />Full paper: https://doi.org/10.1145/3324989.3325724
END:VEVENT
END:VCALENDAR