BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//TUC//Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Europe/Athens
TZNAME:EEST
DTSTART:19700329T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0300
TZNAME:EET
DTSTART:19701025T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CREATED:20230626T112756Z
LAST-MODIFIED:20230626T112756Z
DTSTAMP:20260714T213326Z
UID:1784054006@tuc.gr
SUMMARY:Παρουσίαση Διδακτορικής Διατριβής κα
 ς Βασιλικής Άγου-Σχολή ΜΗΧΟΠ
LOCATION:Εξ αποστάσεως με τηλεδιάσκεψη
DESCRIPTION:https://www.tuc.gr/el/to-polytechnei
 o/ilektronikes-ypiresies/imerologio/
 imerologio-ekdiloseon-1?tx_tucevents
 2_tuceventsdisplay%5Baction%5D=show&
 tx_tucevents2_tuceventsdisplay%5Bcon
 troller%5D=Event&tx_tucevents2_tucev
 entsdisplay%5Bevent%5D=6266&cHash=4a
 01fd010fe4a18ba4ec04a919245c84\nΤίτλ
 ος: Χωροχρονική Ανάλυση Μετεωρολογικ
 ών Παραμέτρων με Κλασικές Γεωστατιστ
 ικές Μεθόδους και Μεθόδους Μηχανικής
  Μάθησης.\nTitle: Space-time analysi
 s of meteorological parameters with 
 geostatistical and machine learning 
 methods.\n Επταμελής Εξεταστική Επιτ
 ροπή:\n 1. Καθηγητής Διονύσιος Χριστ
 όπουλος (επιβλέπων), Σχολή ΗΜΜΥ\n 2.
  Καθηγητής Γεώργιος Καρατζάς (Σχολή 
 ΧΗΜΗΠΕΡ Π.Κ)\n 3. Καθηγητής Παναγιώτ
 ης Παρτσινέβελος (Σχολή ΜΗΧΟΠ Π.Κ.)\
 n 4. Καθηγητής Νικόλαος Νικολαΐδης (
 Σχολή ΧΗΜΗΠΕΡ Π.Κ.)\n 5. Αναπληρώτρι
 α Καθηγήτρια Αναστασία Μπαξεβάνη (Τμ
 ήμα Μαθηματικών και Στατιστικής, Παν
 επιστήμιο Κύπρου)\n 6. Αναπληρωτής Κ
 αθηγητής, Τρύφωνας Δάρας (Σχολή ΧΗΜΗ
 ΠΕΡ Π.Κ.)\n 7. Επ. Καθηγητής Εμμανου
 ήλ Βαρουχάκης, (Σχολή ΜΗΧΟΠ Π.Κ.)\n 
 Abstract:\n Technological advancemen
 ts have increased the availability o
 f spatiotemporal data. However, mete
 orological data are usually non-Gaus
 sian and correlated in space and tim
 e. In this dissertation, state-of-th
 e-art geostatistical and machine-lea
 rning methodologies were utilized to
  analyze large-scale non-Gaussian me
 teorological space-time data. We car
 ried out a series of numerical inves
 tigations utilizing 26 surface varia
 bles from the ERA5 reanalysis data s
 ets collected for 65 grid locations 
 on the island of Crete, Greece. The 
 data sets correspond to multiple tem
 poral scales (hourly to annually) an
 d span the period from 1979 until 20
 19.\n Four distinct approaches were 
 implemented for the analysis of the 
 meteorological parameters:\n \nThe E
 RA5 data set was used for the estima
 tion of the standardized precipitati
 on index (SPI) and the standardized 
 precipitation evapotranspiration ind
 ex (SPEI) to reveal the spatiotempor
 al patterns of drought in Crete.\n \
 nGaussian Anamorphosis with Hermite 
 polynomials (GAH) was employed to tr
 ansform non-Gaussian precipitation d
 ata into normally distributed variab
 les. Ten processing scenarios were i
 nvestigated and their performance wi
 th respect to spatial interpolation 
 (based on Ordinary kriging) was eval
 uated. The scenarios include the app
 lication or exclusion of GAH with va
 rying polynomial degrees, the utiliz
 ation of either the exponential or S
 partan variogram models, and the inc
 orporation or omission of Monte Carl
 o simulations.\n \nTwelve machine le
 arning (ML) techniques were compared
  for the classification of precipita
 tion data into eight classes. Twenty
 -six (26) numerical and categorical 
 variables were used in a spatiotempo
 ral predictive framework for precipi
 tation. Due to pronounce class imbal
 ance (dominance of ``no rain'' event
 s), we first divided the data into t
 wo classes that represent the absenc
 e or occurrence of precipitation eve
 nts. Then, the occurrence data set w
 as split in five different classes t
 o characterize the intensity of prec
 ipitation events.\n \nFinally, we ap
 plied the Stochastic Local Interacti
 on (SLI) model to perform temporal (
 precipitation, temperature and solar
  radiation) and spatiotemporal (prec
 ipitation and temperature) estimatio
 n of missing values (data gaps). \n 
 The most important conclusions deriv
 ed in this dissertation are as follo
 ws:\n \nThe dry climate of Crete was
  confirmed by the estimation of the 
 SPI and SPEI drought indices. It was
  found that the eastern part of the 
 island is more prone to desertificat
 ion than the north-western part. Mor
 eover, a temperature-inclusive droug
 ht index was shown to be more approp
 riate than a purely precipitation-ba
 sed index for the study area.\n \nUs
 ing higher-order (35 versus 20) poly
 nomials in GAH has little effect on 
 the cross-validation results for the
  monthly total precipitation data. I
 n addition, the incorporation of Mon
 te Carlo simulations does not univer
 sally improve the statistical measur
 es.\n \nWith respect to the classifi
 cation of hourly precipitation data,
  the method of Random Forests (Bagge
 d Ensemble Trees) performs best for 
 both the ``Binary'' (``rain'' versus
  ``no rain'') and the ``Only Rain'' 
 classification cases.\n \nSLI is a c
 ompetitive method for interpolating 
 large temporal and spatiotemporal da
 ta since it is fast and it performed
  very well (compared to nearest-neig
 hbor interpolation) across all the d
 ifferent hourly data sets (temperatu
 re, precipitation, and solar radiati
 on).\n The present study investigate
 s a variety of methodological approa
 ches for the analysis of non-Gaussia
 n, large-scale meteorological variab
 les. It provides an extensive analys
 is of precipitation, temperature, an
 d solar radiation for the island of 
 Crete using the ERA5 reanalysis data
  set. The meteorological data used i
 nvolve multiple timescales. Two drou
 ght indices are evaluated and compar
 ed in order to assess the effect of 
 warming trends on drought events. Va
 rious data processing scenarios that
  combine GAH, kriging interpolation 
 and bootstrapping are studied and as
 sessed. In addition, a comparison of
  twelve machine learning methods for
  the classification of precipitation
  data supported by 26 meteorological
  variables is conducted. Lastly, the
  computationally efficient SLI model
 s are herein applied for the first t
 ime to spatiotemporal precipitation 
 and solar radiation data.\n
STATUS:CONFIRMED
ORGANIZER;RSVP=FALSE;CN=TUC;CUTYPE=TUC:mailto:webmaster@tuc.gr
DTSTART:20230704T120000
DTEND:20230704T130000
TRANSP:OPAQUE
CLASS:DEFAULT
END:VEVENT
END:VCALENDAR