BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.earthmonitor.org//soil-health-now-2026//NX83KF
BEGIN:VTIMEZONE
TZID:Europe/Athens
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:EET
TZOFFSETFROM:+0300
TZOFFSETTO:+0200
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:EEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0300
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-soil-health-now-2026-ECXQB7@pretalx.earthmonitor.org
DTSTART;TZID=Europe/Athens:20261207T170000
DTEND;TZID=Europe/Athens:20261207T180000
DESCRIPTION:Materials and Methods \n\nLand use classification is a key tool
  for understanding and monitoring both natural and human-modified landscap
 es. It supports environmental protection\, sustainable agricultural manage
 ment\, and urban planning. Its importance is also reflected in the Land Us
 e\, Land-Use Change and Forestry (LULUCF) framework\, which is closely lin
 ked to European Union climate policies\, carbon stock changes\, greenhouse
  gas emissions\, and ecosystem resilience. Satellite remote sensing provid
 es an efficient and cost-effective way to monitor land use over large area
 s. However\, accurately separating similar land cover types\, especially i
 n agricultural areas with small parcels and strong seasonal changes\, rema
 ins a difficult task. This study investigates the use of deep learning met
 hods to improve land use classification using Sentinel-2 imagery\, support
 ed by suitable preprocessing steps and additional ancillary data.\nWe comp
 iled a time series of Sentinel-2 imagery\, computing monthly median reflec
 tance values for each spectral band\, resulting in 12 composite images. To
  ensure data quality\, individual images with more than 20% cloud cover we
 re excluded before median calculation. Further filtering was applied at th
 e pixel level using the Normalized Difference Snow Index (NDSI) and the No
 rmalized Difference Water Index (NDWI)\, setting a <30% threshold to remov
 e pixels associated with snow cover and water bodies. The study was conduc
 ted in the Eastern Macedonia and Thrace Region\, utilizing 2022 Land Parce
 l Information System (LPIS) data provided by the National Paying Agency of
  Greece as ground truth\, including the field boundaries and crop type lab
 els.\nWe evaluated multiple deep learning models for land use classificati
 on\, including and Temporal CNN Bidirectional Long Short-Term Memory (BiLS
 TM) Bidirectional GRU (BiGRU) Multi-Head Attention (MHA) and Transformer E
 ncoder. Performance was assessed using the Classification Report from the 
 scikit-learn library\, which provided key metrics such as precision\, reca
 ll\, F1-score\, and support. By analyzing results at the crop type level\,
  we identified underperforming classes and refined the classification stra
 tegy by merging related categories\, ultimately enhancing overall model ac
 curacy.\n \nResults and Evaluation\n\nAs the dataset is highly imbalanced 
 among crop-type classes\, we utilized the F1-score to select the best perf
 orming model which was temporal CNN achieving the value of 0.85\, which is
  comparable to the reported State-of-the-Art at various works.\n\nConclusi
 on\n\nOverall\, the study shows  that multitemporal deep learning applied 
 to Sentinel-2 parcel sequences provides a robust and scalable framework fo
 r operational crop type classification.
DTSTAMP:20260825T180212Z
LOCATION:Basement (Foyer)
SUMMARY:Deep Learning-Based Crop Type Classification Using Multitemporal Se
 ntinel-2 Data - Giorgos Varras
URL:https://pretalx.earthmonitor.org/soil-health-now-2026/talk/ECXQB7/
END:VEVENT
END:VCALENDAR