Register now After registration you will be able to apply for this opportunity online.
Quantifying forest diversity shifts after a storm using aerial images
Forest ecosystems are facing rapid and severe changes due to climate change Increasingly frequent disturbances in Central Europe (Seidl et al., 2014; Senf and Seidl, 2021) affect ecosystem structure with cascading effects on ecosystem functioning and the provision of ecosystem services (Paul et al., 2020).
Diversification strategies are discussed as a predictor for forest resilience (Lloret et al., 2024). Such strategies may be implemented in forest adaptation by diversifying the local tree-species composition (e.g., species mixtures vs. monocultures), the local forest structure (e.g., height distribution), and the management types across a landscape; all of which are discussed as strategies to increase resistance and resilience to disturbances (Elmqvist et al., 2003; Hof et al., 2017; Morin et al., 2011), and to reduce trade-offs between ecosystem functioning and multiple ecosystem services (Knoke et al., 2017; Topanotti, 2024).
Transitions towards new forest management regimes, and in particular changes in tree-species composition, are slow processes due to the long production and planning periods in forest management. This contrasts with recent observations of increasingly severe events, such as the extreme mortality events following the drought in the year 2018 in Central Europe (Buras et al., 2020; Schuldt et al., 2020). Consequently, a new type of thinking developed that interprets disturbances as drivers of forest change rather than disruptions to a status-quo system that needs to be preserved (Buma and Schultz, 2020; Thom et al., 2017). Thom et al. (2017) showed that disturbance characteristics influence this adaptation-fostering effect of disturbances: An increasing frequency and intensity accelerated the adaptation process while an increasing size slowed it down. The study is based on simulation modeling in an unmanaged forest. In managed forests, however, post-disturbance management, such as salvage logging and planting, have a strong potential to alter these dynamics.
Keywords: Forest dynamics; Deep Learning; CNN; species identification
Research questions:
• Are time series of aerial images suitable to assess the influence of historic storm events on tree-species compositions?
• How did storm Lothar and possible post-disturbance management influence local tree-species diversity?
• How did storm Lothar and the possible post-disturbance management influence landscape diversity?
• Do these influences depend on the patch size of the disturbance?
• Do these influences differ between regions and their predominant forest management goals?
**Methods**:
Identify disturbed areas: use the BAFU Storm Lothar map to delineate affected areas as polygons and select undisturbed control sites for comparison.
Data collection: Gather aerial image time series, LiDAR data for canopy structure, and other relevant datasets.
Tree-species identification: Apply current machine learning methods to classify tree species using aerial images, validating with available ground-truth data available from cantonal offices.
Analyze structural and landscape changes: Derive metrics for canopy height, cover, and landscape diversity to assess spatial and structural changes.
Statistical analysis of tree-species diversity (Shannon, evenness, or Bray Curtis for changes) (1) before and after the event and (2) across the forest landscape (gamma-diversity).
Identification of relevant covariates (incl. management goals in the forest area) to predict post-disturbance forest diversity.
Research questions:
• Are time series of aerial images suitable to assess the influence of historic storm events on tree-species compositions?
• How did storm Lothar and possible post-disturbance management influence local tree-species diversity?
• How did storm Lothar and the possible post-disturbance management influence landscape diversity?
• Do these influences depend on the patch size of the disturbance?
• Do these influences differ between regions and their predominant forest management goals?
**Methods**:
Identify disturbed areas: use the BAFU Storm Lothar map to delineate affected areas as polygons and select undisturbed control sites for comparison.
Data collection: Gather aerial image time series, LiDAR data for canopy structure, and other relevant datasets.
Tree-species identification: Apply current machine learning methods to classify tree species using aerial images, validating with available ground-truth data available from cantonal offices.
Analyze structural and landscape changes: Derive metrics for canopy height, cover, and landscape diversity to assess spatial and structural changes.
Statistical analysis of tree-species diversity (Shannon, evenness, or Bray Curtis for changes) (1) before and after the event and (2) across the forest landscape (gamma-diversity). Identification of relevant covariates (incl. management goals in the forest area) to predict post-disturbance forest diversity.
With this thesis, we aim to collect empirical datasets from managed and unmanaged forests and to assess how a large storm event (Lothar, 1999), and possibly triggered post-disturbance management, altered the local and regional tree-species diversity.
**Wanted**:
We look for a highly motivated student with an interest in remote sensing, deep learning, forest ecosystems, and statistical analysis. You should be willing to learn new techniques and methods. A solid knowledge of GIS is a plus.
The project has a flexible starting date.
**You will get to**:
• Learn new and highly required theoretical and practical knowledge about remote sensing and deep learning, and how to apply these skills to up-to-date questions in forest ecology and management.
• Gain important statistical knowledge for future work.
• Expand your network by discussing your work with experts from the intersecting fields of forest sciences and computer sciences.
• Be a co-author on a publication resulting from this work.
• Be part of a motivated team of scientists.
With this thesis, we aim to collect empirical datasets from managed and unmanaged forests and to assess how a large storm event (Lothar, 1999), and possibly triggered post-disturbance management, altered the local and regional tree-species diversity.
**Wanted**:
We look for a highly motivated student with an interest in remote sensing, deep learning, forest ecosystems, and statistical analysis. You should be willing to learn new techniques and methods. A solid knowledge of GIS is a plus. The project has a flexible starting date.
**You will get to**:
• Learn new and highly required theoretical and practical knowledge about remote sensing and deep learning, and how to apply these skills to up-to-date questions in forest ecology and management.
• Gain important statistical knowledge for future work.
• Expand your network by discussing your work with experts from the intersecting fields of forest sciences and computer sciences.
• Be a co-author on a publication resulting from this work.
• Be part of a motivated team of scientists.
Supervisor: Dr. Mirela Beloiu Schwenke, Dr. Jasper Fuchs
If the idea of participating in cutting-edge interdisciplinary research excites you, please contact mirela.beloiu(at)usys.ethz.ch. The FORM team is looking forward to hearing from you!
Paper: Sparse subalpine forest recovery pathways, plant communities, and carbon stocks 34 years after stand-replacing fire (link).
Supervisor: Dr. Mirela Beloiu Schwenke, Dr. Jasper Fuchs
If the idea of participating in cutting-edge interdisciplinary research excites you, please contact mirela.beloiu(at)usys.ethz.ch. The FORM team is looking forward to hearing from you!
Paper: Sparse subalpine forest recovery pathways, plant communities, and carbon stocks 34 years after stand-replacing fire (link).
