BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//talks.osgeo.org//foss4g-europe-2026//talk//BGWUCE BEGIN:VTIMEZONE TZID:EET BEGIN:STANDARD DTSTART:20001029T050000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:EET TZOFFSETFROM:+0300 TZOFFSETTO:+0200 END:STANDARD BEGIN:DAYLIGHT DTSTART:20000326T040000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:EEST TZOFFSETFROM:+0200 TZOFFSETTO:+0300 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT UID:pretalx-foss4g-europe-2026-BGWUCE@talks.osgeo.org DTSTART;TZID=EET:20260629T154000 DTEND;TZID=EET:20260629T154500 DESCRIPTION:*Introduction*\nEuropean landscapes have experienced drastic an d accelerating changes in recent decades\, particularly since World War II . Understanding landscape change from a historical perspective is essentia l in assessing the impact of previous land management choices on current-d ay landscapes\, habitats and biodiversity. While conservation policies typ ically consider current situations\, the lack of historical context might lead to shifting baselines to fit a deteriorating trend. \nThe Habitat Map of Switzerland is a high-resolution thematic product mapping the differen t Swiss habitats [1]. The classification is based on TypoCH\, a Swiss-spec ific hierarchical habitat typology\, which can be translated into the pan- European EUNIS classification system. TypoCH contains land cover classes o n the first level and becomes increasingly more detailed\, with plant comm unities and species on the lower levels [2]. Recently declassified Swiss-w ide imagery from 1946 with 1m spatial resolution\, in conjunction with reg ularly updated aerial imagery since the 1980s\, offers an unprecedented op portunity to map landscape and habitat change in the past century. In the perspective of multi-temporal classification\, a proof-of-concept study wa s performed on selected study areas in Switzerland\, using grayscale 1946 aerial imagery\, and object-oriented analysis and classification [3]. \nTh e gap bridged by this project is creating consistent multi-temporal mappin g for Switzerland\, with potential to apply the methodology in other count ries with similar historical data. The inherent inconsistencies in histori cal aerial image quality\, the limited spectral information (grayscale) an d habitat heterogeneity and complexity are the most challenging. Consisten t mapping is important for robust change detection and comparison between time steps. Therefore\, a flexible habitat typology and scalable method sh ould be determined.\nThis study aims to map habitat status and explore hab itat change in Switzerland over 5 times steps (1946 – present) using dee p learning image segmentation methods. Given the varying quality and spect ral resolution of imagery over time\, the first aim of this project is to determine which habitats can be consistently mapped from 1946 to the prese nt. This will be done in a data-driven approach\, developing a hierarchica l\, modular and flexible open-source deep learning methodology to check ha bitat mapping feasibility. The further aims are to develop a method to con sistently classify habitats across this time-series\, detect landscape cha nge\, and relate results to the current biodiversity status. \n\n*Methodol ogy*\nTo determine which habitats can be mapped from grayscale imagery\, a data-driven approach will be used\, starting from current-day aerial imag ery and Habitat Map of Switzerland\, integrated into a preliminary archite cture based on a hierarchical U-Net structure. The current-day aerial imag ery will be degraded to simulate historical aerial imagery\, using state-o f-the-art algorithms to add noise\, scratches\, distortions and blurring [ 4]. The Habitat Map of Switzerland will be used as training\, validation a nd test data. \nThe first level of the U-Net architecture will segment the first level of the habitat typology\, which mainly corresponds to land-co ver classes. The current Habitat Map of Switzerland includes habitats up t o the third level of the TypoCH typology. The U-Net will be trained to fir st separate the TypoCH classes\, which correspond to the first level of th e TypoCH typology. For each class\, a new U-Net will be trained to separat e the groups within the class\, which correspond to the second level of th e TypoCH typology. Then\, for each group\, new U-Nets will be trained to s eparate types within each group\, which correspond to the third level of t he TypoCH typology.\nThis approach will inform about habitats which are mo re difficult to map\, as well as habitats which might be easily confused w ith other habitats. This in turn will inform on class-specific or group-sp ecific variables which would need to be added to address uncertainty and t he limited spectral information. In conjunction with a previous ecological priority review\, adjacent data will be researched for habitats which are more difficult to map but are of high ecological importance in terms of l ong-term landscape change. Potential auxiliary sources include digitized d ata [5] or even automated feature extraction from topographical maps (Sieg fried maps).\nThe data for the model was chosen using a stratified random sampling technique. The extent of Switzerland was gridded into 512x512px t iles. The choice of the tiles was stratified in a first phase using the 12 biogeographical regions of Switzerland. Upon preliminary results\, additi onal stratification on habitat composition\, certain habitat coverage perc entage or elevation might be considered. 5% of tiles of each biogeographic al region were randomly selected\, resulting in 7356 tile-mask pairs. Per region\, the tile-mask pairs were separated into 70% for training\, 20% fo r validation and 10% for testing. \nThe lack of validation data for histor ical imagery is one of the biggest challenges of the project. Therefore\, in the first step\, the model will be trained and tested on current aerial imagery converted to grayscale and degraded with artifacts common to hist orical imagery. This way\, the feasibility of mapping the wide range of Ty poCH habitats will be tested with robust validation based on the current H abitat Map of Switzerland. In further steps\, the model will be strategica lly applied on historical imagery and potentially active learning and/or z ero-shot segmentation algorithms will be used to generate historical train ing data\, aided by time-series comparisons. \n\n*Expected results\, impl ications and conclusions*\nThe first part of the project will show which h abitats can be mapped from historical imagery\, providing a methodology wh ich can be transferred in other areas with historical data availability. T hen the method will be scaled Swiss-wide on multiple time steps\, showing types and rates of habitat change and link the results to management pract ice and the current biodiversity status. The results will have broad impli cations on future conservation measures\, land management policies\, and r estoration actions. Given the amount of data available for Switzerland\, t raining a geofoundation model specialized on historical grayscale imagery and object change detection would be an idea to be explored as a future pa rt of this project using the knowledge and data obtained from preliminary model testing. DTSTAMP:20260605T022350Z LOCATION:A01 SUMMARY:Habitat Change Mapping Using Historical Aerial Imagery and Deep Lea rning - Francesca Drăguț URL:https://talks.osgeo.org/foss4g-europe-2026/talk/BGWUCE/ END:VEVENT END:VCALENDAR