BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.earthmonitor.org//global-workshop-2026//ZERAER
BEGIN:VTIMEZONE
TZID:Europe/Amsterdam
BEGIN:STANDARD
DTSTART:20001029T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T020000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-global-workshop-2026-TURRVV@pretalx.earthmonitor.org
DTSTART;TZID=Europe/Amsterdam:20261007T151500
DTEND;TZID=Europe/Amsterdam:20261007T153000
DESCRIPTION:Accurate characterization of tropical forest vertical structure
  is critical for carbon accounting and ecosystem monitoring\, yet most mac
 hine-learning pipelines reduce GEDI's rich waveform information to a singl
 e scalar\, typically canopy height or a high relative-height percentile. T
 his simplification discards the ordered height distribution that GEDI enco
 des across its full relative height (RH) profile\, and that its own biomas
 s algorithms depend on. We introduce Biomazon\, an open\, ML-ready multimo
 dal benchmark dataset at 20 m resolution over the Amazon Basin\, designed 
 to support joint prediction of the full GEDI RH profile (RH0 to RH100) tog
 ether with above-ground biomass density (AGBD). The dataset pairs GEDI-der
 ived targets with multi-sensor predictors including Sentinel-1\, Sentinel-
 2\, ALOS-2 PALSAR-2\, Copernicus DEM\, Dynamic World land cover\, and geos
 patial foundation model embeddings\, all co-registered on a common grid wi
 th standardized spatial splits and evaluation protocols to enable reproduc
 ible comparison of methods. We formulate RH prediction as structured outpu
 t learning with a monotonicity constraint that enforces physical consisten
 cy across percentiles\, and we provide baseline results from systematic ab
 lations over model scale\, sensor contributions\, and the role of AlphaEar
 th embeddings\, both as standalone predictors and in fusion with raw modal
 ities. Results are contextualized against existing gridded products to ass
 ess practical relevance. Biomazon addresses a gap in current benchmarking 
 by shifting the task formulation from scalar regression toward structure-a
 ware modeling\, and by providing the community with an open\, multi-sensor
  dataset and protocol for investigating when and how different data source
 s\, including learned representations\, contribute to forest structure and
  biomass retrieval in tropical forests.
DTSTAMP:20260624T071014Z
LOCATION:Rooms 12+14
SUMMARY:Biomazon: A Multimodal Benchmark for Full Vertical Structure and Bi
 omass Modeling in the Amazon Basin - Sayan Mandal\, Rocco Sedona
URL:https://pretalx.earthmonitor.org/global-workshop-2026/talk/TURRVV/
END:VEVENT
END:VCALENDAR