BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.earthmonitor.org//YDNNYJ
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-YDNNYJ@pretalx.earthmonitor.org
DTSTART;TZID=Europe/Athens:20261209T160000
DTEND;TZID=Europe/Athens:20261209T161500
DESCRIPTION:Accurate\, scalable estimation of soil organic carbon (SOC) and
  clay content from remote sensing data is crucial for monitoring soil heal
 th at large continental scales. Standard baseline methods such as PLSR\, S
 VR\, Random Forest\, Gradient Boosting etc. struggle to exploit the inhere
 nt relationship between spectral signatures and spatial neighborhoods of t
 he corresponding samples. In order to overcome this limitation\, we propos
 e MVHGCN\, a Multi-View Hypergraph Convolutional Network developed for joi
 nt SOC and clay content prediction from Sentinel-2 imagery.\n\nThe propose
 d architecture addresses three main drawbacks of standard approaches via t
 he following contributions. First\, hyperedges can capture higher-order re
 lationships between soil samples\, connecting pixels through both spectral
  or geographic similarity within a single cluster\, instead of decomposing
  those relationships into pairwise links. Second\, a multi-view formulatio
 n processes six spectral transforms of each sample (continuum-removal\, ab
 sorbance\, Savitzky-Golay derivatives\, SNV-normalized variants)\, encoded
  via a 1D-CNN\, as parallel views fused through a learned View-Based Atten
 tion Fusion module that adaptively weights each transform via an attention
  mechanism. Third\, the model operates on two parallel branches\, a global
  cosine-similarity branch and local multi-scale spatial branch (within-pix
 el and within-tile)\, which are fused at every block via a Cross-Talk blen
 ding unit\, so local geographic proximity and global spectral similarity c
 an be simultaneously exploited. Finally\, we introduce Adaptive Hypergraph
  Learning (AHL)\, an additional attention mechanism that refines incidence
  matrices during training\, and a multi-task head that predicts SOC and cl
 ay jointly under an uncertainty-weighted loss with learnable per-task vari
 ance. \n\nWe evaluate our approach on data from the LUCAS 2015 Topsoil sur
 vey matched to satellite-derived reflectance composites\, generated from m
 ulti-temporal Sentinel-2 imagery. The proposed model is compared against a
  series of 1) Standard machine learning methods (PLSR\, SVR\, Random Fores
 t\, Gradient Boosted Trees)\, a state-of-the-art 1D-CNN developed specific
 ally for estimation of soil-properties from remote sensing data\, and seve
 ral baseline graph and hypergraph networks. The results are reported on th
 e original target scale\, after inverting the log1p and z-score transforms
 . The proposed MVHGCN is consistently found to outperform all standard bas
 elines\, the respective graph and hypergraph base models\, as well as the 
 state-of-the-art 1D-CNN model.
DTSTAMP:20260825T194154Z
LOCATION:Amphitheater I
SUMMARY:A Multi-View Hypergraph Convolutional Network for Joint SOC and Cla
 y Prediction from Sentinel-2 Imagery - Konstantinos Karyotis\, Christos Ch
 adoulos
URL:https://pretalx.earthmonitor.org/soil-health-now-2026/talk/YDNNYJ/
END:VEVENT
END:VCALENDAR