A5048243105- Remote Sensing and LiDAR Applications
- Remote Sensing in Agriculture
- Adaptive optics and wavefront sensing
- Meteorological Phenomena and Simulations
- Plant Water Relations and Carbon Dynamics
- Forest ecology and management
- Land Use and Ecosystem Services
- Gamma-ray bursts and supernovae
- Automated Road and Building Extraction
- Semantic Web and Ontologies
- Oil Palm Production and Sustainability
- Cocoa and Sweet Potato Agronomy
- Species Distribution and Climate Change
- Solar Radiation and Photovoltaics
- Smart Agriculture and AI
- Soil erosion and sediment transport
- Hydrology and Sediment Transport Processes
- Remote Sensing and Land Use
- Calibration and Measurement Techniques
- Horticultural and Viticultural Research
- Rough Sets and Fuzzy Logic
- Machine Learning and Algorithms
- Domain Adaptation and Few-Shot Learning
- Impact of Light on Environment and Health
- Remote-Sensing Image Classification
ETH Zurich
2017-2024
University of Copenhagen
2023-2024
University of Wuppertal
2017
Kentucky State University
1999
Eastern Illinois University
1999
Purdue University West Lafayette
1999
University of Guelph
1999
Fort Valley State University
1999
Texas A&M University System
1999
Abstract The worldwide variation in vegetation height is fundamental to the global carbon cycle and central functioning of ecosystems their biodiversity. Geospatially explicit and, ideally, highly resolved information required manage terrestrial ecosystems, mitigate climate change prevent biodiversity loss. Here we present a comprehensive canopy map at 10 m ground sampling distance for year 2020. We have developed probabilistic deep learning model that fuses sparse data from Global Ecosystem...
NASA's Global Ecosystem Dynamics Investigation (GEDI) is a key climate mission whose goal to advance our understanding of the role forests in global carbon cycle. While GEDI first space-based LIDAR explicitly optimized measure vertical forest structure predictive aboveground biomass, accurate interpretation this vast amount waveform data across broad range observational and environmental conditions challenging. Here, we present novel supervised machine learning approach interpret waveforms...
Trees are an integral part in European landscapes, but only forest resources systematically assessed by national inventories. The contribution of urban and agricultural trees to national-level carbon stocks remains largely unknown. Here we produced canopy cover, height above-ground biomass maps from 3-meter resolution nanosatellite imagery across Europe. Our estimates have a systematic bias 7.6% (overestimation;
The worldwide variation in vegetation height is fundamental to the global carbon cycle and central functioning of ecosystems their biodiversity. Geospatially explicit and, ideally, highly resolved information required manage terrestrial ecosystems, mitigate climate change, prevent biodiversity loss. Here, we present first global, wall-to-wall canopy map at 10 m ground sampling distance for year 2020. No single data source meets these requirements: dedicated space missions like GEDI deliver...
Côte d'Ivoire and Ghana, the world's largest producers of cocoa, account for two thirds global cocoa production. In both countries, is primary perennial crop, providing income to almost million farmers. Yet precise maps area planted with are missing, hindering accurate quantification expansion in protected areas, production yields limiting information available improved sustainability governance. Here we combine plantation data publicly satellite imagery a deep learning framework create...
Abstract. Grain size analysis is the key to understand sediment dynamics of river systems. We propose GRAINet, a data-driven approach analyze grain distributions entire gravel bars based on georeferenced UAV images. A convolutional neural network trained regress as well characteristic mean diameter from raw GRAINet allows for holistic bars, resulting in (i) high-resolution estimates and maps spatial distribution at large scale (ii) robust grading curves bars. To collect an extensive training...
Monitoring and managing Earth's forests in an informed manner is important requirement for addressing challenges like biodiversity loss climate change. While traditional situ or aerial campaigns forest assessments provide accurate data analysis at regional level, scaling them to entire countries beyond with high temporal resolution hardly possible. In this work, we propose a method based on deep ensembles that densely estimates structure variables country-scale 10-m resolution, using freely...
Abstract Trees are an integral part of almost all European landscapes, but only forest resources systematically assessed by national inventories, and the extent to which trees in urban agricultural areas contribute biomass carbon stocks at level remains largely unknown. Here we make use nanosatellite imagery generate canopy cover, height, above-ground maps for entire continent from 3-m resolution imagery. Our country-scale estimates have a systematic bias 7.6% (overestimation; R = 0.98) when...
Abstract. Grain size analysis is the key to understand sediment dynamics of river systems. We propose GRAINet, a data-driven approach analyze grain distributions entire gravel bars based on georeferenced UAV images. A convolutional neural network trained regress as well characteristic mean diameter from raw GRAINet allows holistic bars, resulting in (i) high-resolution maps spatial distribution at large scale, and (ii) robust grading curves for bars. To collect training dataset 1,491...
The increasing demand for commodities is leading to changes in land use worldwide. In the tropics, deforestation, which causes high carbon emissions and threatens biodiversity, often linked agricultural expansion. While need deforestation-free global supply chains widely recognized, making progress practice remains a challenge. Here, we propose an automated approach that aims support conservation sustainable planning decisions by mapping tropical landscapes at large scale spatial resolution...
C\^ote d'Ivoire and Ghana, the world's largest producers of cocoa, account for two thirds global cocoa production. In both countries, is primary perennial crop, providing income to almost million farmers. Yet precise maps planted area are missing, hindering accurate quantification expansion in protected areas, production yields, limiting information available improved sustainability governance. Here, we combine plantation data with publicly satellite imagery a deep learning framework create...
We present a new approach for matching tree instances across multiple street-view panorama images the ultimate goal of city-scale street-tree mapping with high positioning accuracy. What makes this task challenging is strong change in view-point, different lighting conditions, similarity neighboring trees, and variability scale. propose to turn (tree) instance into learning task, where image-appearance geometric relationships between views fruitfully interact. Our constructs Siamese...
The volume of unlabelled Earth observation (EO) data is huge, but many important applications lack labelled training data. However, EO offers the unique opportunity to pair from different modalities and sensors automatically based on geographic location time, at virtually no human labor cost. We seize this create a diverse multi-modal pretraining dataset global scale. Using new corpus 1.2 million locations, we propose Multi-Pretext Masked Autoencoder (MP-MAE) approach learn general-purpose...
Category discovery methods aim to find novel categories in unlabeled visual data. At training time, a set of labeled and images are provided, where the labels correspond present images. The data provides guidance during by indicating what types properties features relevant for performing As result, changing can have large impact on is ultimately discovered set. Despite its importance, selection has not been explored category literature date. We show that significantly performance. Motivated...