Annett Wania
Research Subjects
Annett Wania has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Akhil Singh Rana
Planet Labs Germany GmbH, 10719 Berlin, Germany
3 shared publications - Caglar Senaras
Ctr. for Biomedical Informatics, Wake Forest School of Medicine (United States)
3 shared publications - Maddie Grady
Planet Labs Germany GmbH, 10719 Berlin, Germany
2 shared publications - Piers Holden
Planet Labs Germany GmbH, 10719 Berlin, Germany
2 shared publications - Timothy Davis
Planet Labs Germany GmbH, 10719 Berlin, Germany
2 shared publications - Benjamin Bischke1 shared publication
Latest Publications
- Open Access
- 10 November 2026
- Published by Wiley in Ecology and Evolution
- Vol. 16 (11) , e74658
Building a Framework for Peatland Assessment in Europe by Combining Spectral and Molecular Data
Peatlands provide critical ecosystem services, including carbon storage, water regulation, and climate mitigation, yet spatial information on their ecological condition remains limited despite widespread degradation. Specific biogeochemical analyses (e.g., biomarkers, molecular compounds, compound specific stable isotopes) can characterize drainage‐related changes in peat properties but are time‐consuming, costly, linked to sophisticated analytical set‐up and thus spatially restricted. We therefore investigated whether high‐resolution PlanetScope imagery capture consistent spectral differences between pristine and drained areas and explored their multivariate correspondence with complementary site‐level biogeochemical properties. Spectral indices, including NDVI, EVI, gNDVI, Greenness Index, and albedo were derived from PlanetScope Analysis‐Ready Surface Reflectance (ARPS) data for nine European peatland sites spanning three climatic zones and contrasting open and sparsely treed vegetation structures. Peat biogeochemistry was characterized using elemental carbon and nitrogen contents and molecular composition determined by pyrolysis–GC/MS. NDVI and Greenness Index differed most consistently between the investigated pristine and drained areas, with higher values under drained conditions across both vegetation structures. In contrast, EVI and albedo, even though also suitable indicators, depended more strongly on vegetation structure and the patterns differed between open and sparsely treed systems. Spectral differences were most pronounced during May and June, identifying the early growing season as the period of strongest spectral separation within the investigated dataset. Drained areas were additionally characterized by higher nitrogen contents, lower C/N ratios, and a greater relative abundance of compounds indicative of metabolized organic matter. Exploratory principal component analyses indicated complementary site‐level patterns between spectral and biogeochemical variables associated with hydrological condition. Consequently, this study demonstrates the potential of high‐resolution PlanetScope spectral data to differentiate between pristine and drained peatland conditions across the investigated sites. However, predictive classification of previously unknown peatland areas will require dedicated modeling and independent validation especially in differentiating from surrounding transition peats and mineral grassland soils.
- 10 November 2026
- Published by Wiley in Ecology and Evolution
- Vol. 16 (11) , e74658
Open AccessBuilding a Framework for Peatland Assessment in Europe by Combining Spectral and Molecular Data
Peatlands provide critical ecosystem services, including carbon storage, water regulation, and climate mitigation, yet spatial information on their ecological condition remains limited despite widespread degradation. Specific biogeochemical analyses (e.g., biomarkers, molecular compounds, compound specific stable isotopes) can characterize drainage‐related changes in peat properties but are time‐consuming, costly, linked to sophisticated analytical set‐up and thus spatially restricted. We therefore investigated whether high‐resolution PlanetScope imagery capture consistent spectral differences between pristine and drained areas and explored their multivariate correspondence with complementary site‐level biogeochemical properties. Spectral indices, including NDVI, EVI, gNDVI, Greenness Index, and albedo were derived from PlanetScope Analysis‐Ready Surface Reflectance (ARPS) data for nine European peatland sites spanning three climatic zones and contrasting open and sparsely treed vegetation structures. Peat biogeochemistry was characterized using elemental carbon and nitrogen contents and molecular composition determined by pyrolysis–GC/MS. NDVI and Greenness Index differed most consistently between the investigated pristine and drained areas, with higher values under drained conditions across both vegetation structures. In contrast, EVI and albedo, even though also suitable indicators, depended more strongly on vegetation structure and the patterns differed between open and sparsely treed systems. Spectral differences were most pronounced during May and June, identifying the early growing season as the period of strongest spectral separation within the investigated dataset. Drained areas were additionally characterized by higher nitrogen contents, lower C/N ratios, and a greater relative abundance of compounds indicative of metabolized organic matter. Exploratory principal component analyses indicated complementary site‐level patterns between spectral and biogeochemical variables associated with hydrological condition. Consequently, this study demonstrates the potential of high‐resolution PlanetScope spectral data to differentiate between pristine and drained peatland conditions across the investigated sites. However, predictive classification of previously unknown peatland areas will require dedicated modeling and independent validation especially in differentiating from surrounding transition peats and mineral grassland soils.
- Open Access
- 25 July 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (15) , 2730
Detection of Maize Crop Phenology Using Planet Fusion
Accurate identification of crop phenology timing is crucial for agriculture. While remote sensing tracks vegetation changes, linking these to ground-measured crop growth stages remains challenging. Existing methods offer broad overviews but fail to capture detailed phenological changes, which can be partially related to the temporal resolution of the remote sensing datasets used. The availability of higher-frequency observations, obtained by combining sensors and gap-filling, offers the possibility to capture more subtle changes in crop development, some of which can be relevant for management decisions. One such dataset is Planet Fusion, daily analysis-ready data obtained by integrating PlanetScope imagery with public satellite sensor sources such as Sentinel-2 and Landsat. This study introduces a novel method utilizing Dynamic Time Warping applied to Planet Fusion imagery for maize phenology detection, to evaluate its effectiveness across 70 micro-stages. Unlike singular template approaches, this method preserves critical data patterns, enhancing prediction accuracy and mitigating labeling issues. During the experiments, eight commonly employed spectral indices were investigated as inputs. The method achieves high prediction accuracy, with 90% of predictions falling within a 10-day error margin, evaluated based on over 3200 observations from 208 fields. To understand the potential advantage of Planet Fusion, a comparative analysis was performed using Harmonized Landsat Sentinel-2 data. Planet Fusion outperforms Harmonized Landsat Sentinel-2, with significant improvements observed in key phenological stages such as V4, R1, and late R5. Finally, this study showcases the method’s transferability across continents and years, although additional field data are required for further validation.
- 25 July 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (15) , 2730
Open AccessDetection of Maize Crop Phenology Using Planet Fusion
Accurate identification of crop phenology timing is crucial for agriculture. While remote sensing tracks vegetation changes, linking these to ground-measured crop growth stages remains challenging. Existing methods offer broad overviews but fail to capture detailed phenological changes, which can be partially related to the temporal resolution of the remote sensing datasets used. The availability of higher-frequency observations, obtained by combining sensors and gap-filling, offers the possibility to capture more subtle changes in crop development, some of which can be relevant for management decisions. One such dataset is Planet Fusion, daily analysis-ready data obtained by integrating PlanetScope imagery with public satellite sensor sources such as Sentinel-2 and Landsat. This study introduces a novel method utilizing Dynamic Time Warping applied to Planet Fusion imagery for maize phenology detection, to evaluate its effectiveness across 70 micro-stages. Unlike singular template approaches, this method preserves critical data patterns, enhancing prediction accuracy and mitigating labeling issues. During the experiments, eight commonly employed spectral indices were investigated as inputs. The method achieves high prediction accuracy, with 90% of predictions falling within a 10-day error margin, evaluated based on over 3200 observations from 208 fields. To understand the potential advantage of Planet Fusion, a comparative analysis was performed using Harmonized Landsat Sentinel-2 data. Planet Fusion outperforms Harmonized Landsat Sentinel-2, with significant improvements observed in key phenological stages such as V4, R1, and late R5. Finally, this study showcases the method’s transferability across continents and years, although additional field data are required for further validation.
- Open Access
- 15 May 2023
- Published by Copernicus GmbH
Geography-Aware Masked Autoencoders for Change Detection in Remote Sensing
Increasing coverage of commercial and public satellites allows us to monitor the pulse of the Earth in ever-shorter frequency (Zhu et al., 2017). Together with the rise of deep learning in artificial intelligence (AI) (LeCun et al., 2015), the field of AI for Earth Observation (AI4EO) is growing rapidly. However, many supervised deep learning techniques are data-hungry, which means that annotated data in large quantities are necessary to help these algorithms reach their full potential. In many Earth Observation applications such as change detection, this is often infeasible because high-quality annotations require manual labeling which is time-consuming and costly. Self-supervised learning (SSL) can help tackle the issue of limited label availability in AI4EO. In SSL, an algorithm is pretrained with tasks that only require the input data without annotation. Notably, Masked Autoencoders (MAE) have shown promising performances recently where a Vision Transformer learns to reconstruct a full image with only 25% of it as input. We hypothesize that the success of MAEs also extends to satellite imagery and evaluate this with a change detection downstream task. In addition, we provide a multitemporal DINO baseline which is another widely successful SSL method. Further, we test a second version of MAEs, which we call GeoMAE. GeoMAE incorporates the location and date of the satellite image as auxiliary information in self-supervised pretraining. The coordinates and date information are passed as additional tokens to the MAE model similar to the positional encoding. The pretraining dataset used is the RapidAI4EO corpus which contains multi-temporal Planet Fusion imagery for a variety of locations across Europe. The dataset for the downstream task also uses Planet Fusion in pairs as input data. These are provided on a 600m * 600m patch level three months apart together with a classification if the respective patch has changed in this period. Self-supervised pretraining is done for up to 150 epochs where we take the model with the best validation performance on the downstream task as a starting point for the test set. We find that the regular MAE model scores the best on the test set with an accuracy of 81.54% followed by DINO with 80.63% and GeoMAE with 80.02%. Pretraining MAE with ImageNet data instead of satellite images results in a notable performance loss down to 71.36%. Overall, our current pretraining experiments can not yet confirm our hypothesis that GeoMAE is advantageous compared to regular MAE. However, in similar spirit, Cong et al. (2022) recently introduced SatMAE which outlines that for other remote sensing applications, the combination of auxiliary information and novel masking strategies is a key factor. Therefore, it seems that a combination of location and time inputs together with adapted masking may also hold the most potential for change detection. There is ample potential for future research in geo-specific applications of MAEs and we provide a starting point for this with our experimental results for change detection.
- 15 May 2023
- Published by Copernicus GmbH
Open AccessGeography-Aware Masked Autoencoders for Change Detection in Remote Sensing
Increasing coverage of commercial and public satellites allows us to monitor the pulse of the Earth in ever-shorter frequency (Zhu et al., 2017). Together with the rise of deep learning in artificial intelligence (AI) (LeCun et al., 2015), the field of AI for Earth Observation (AI4EO) is growing rapidly. However, many supervised deep learning techniques are data-hungry, which means that annotated data in large quantities are necessary to help these algorithms reach their full potential. In many Earth Observation applications such as change detection, this is often infeasible because high-quality annotations require manual labeling which is time-consuming and costly. Self-supervised learning (SSL) can help tackle the issue of limited label availability in AI4EO. In SSL, an algorithm is pretrained with tasks that only require the input data without annotation. Notably, Masked Autoencoders (MAE) have shown promising performances recently where a Vision Transformer learns to reconstruct a full image with only 25% of it as input. We hypothesize that the success of MAEs also extends to satellite imagery and evaluate this with a change detection downstream task. In addition, we provide a multitemporal DINO baseline which is another widely successful SSL method. Further, we test a second version of MAEs, which we call GeoMAE. GeoMAE incorporates the location and date of the satellite image as auxiliary information in self-supervised pretraining. The coordinates and date information are passed as additional tokens to the MAE model similar to the positional encoding. The pretraining dataset used is the RapidAI4EO corpus which contains multi-temporal Planet Fusion imagery for a variety of locations across Europe. The dataset for the downstream task also uses Planet Fusion in pairs as input data. These are provided on a 600m * 600m patch level three months apart together with a classification if the respective patch has changed in this period. Self-supervised pretraining is done for up to 150 epochs where we take the model with the best validation performance on the downstream task as a starting point for the test set. We find that the regular MAE model scores the best on the test set with an accuracy of 81.54% followed by DINO with 80.63% and GeoMAE with 80.02%. Pretraining MAE with ImageNet data instead of satellite images results in a notable performance loss down to 71.36%. Overall, our current pretraining experiments can not yet confirm our hypothesis that GeoMAE is advantageous compared to regular MAE. However, in similar spirit, Cong et al. (2022) recently introduced SatMAE which outlines that for other remote sensing applications, the combination of auxiliary information and novel masking strategies is a key factor. Therefore, it seems that a combination of location and time inputs together with adapted masking may also hold the most potential for change detection. There is ample potential for future research in geo-specific applications of MAEs and we provide a starting point for this with our experimental results for change detection.
- Open Access
- 21 April 2023
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-M-1, 309-315
SELF-SUPERVISED LEARNING FOR CROP CLASSIFICATION USING PLANET FUSION
Benefiting from the high cadence and spatial resolution of the new generation of Earth observation satellites, remote sensing technology is allowing us to derive more valuable information for the agricultural sector. Crop classification is one of the fundamental information derivatives from Earth observation data researchers used for food security, crop monitoring, and economic assessment. The robustness of a crop classification model to variations in environmental and management conditions due to time and location is one of the crucial requirements. To achieve this, we developed a novel self-supervised method using the advantage of unlabeled samples and transformer architectures. We used six different areas in Germany and four years to evaluate the robustness of the model. Our experiments showed that self-supervised deep learning methods could provide a significant advantage in handling these variations. In some cases, we observed around 30 percentage points improvements in F1-score performance compared to a Random Forest based model.
- 21 April 2023
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-M-1, 309-315
Open AccessSELF-SUPERVISED LEARNING FOR CROP CLASSIFICATION USING PLANET FUSION
Benefiting from the high cadence and spatial resolution of the new generation of Earth observation satellites, remote sensing technology is allowing us to derive more valuable information for the agricultural sector. Crop classification is one of the fundamental information derivatives from Earth observation data researchers used for food security, crop monitoring, and economic assessment. The robustness of a crop classification model to variations in environmental and management conditions due to time and location is one of the crucial requirements. To achieve this, we developed a novel self-supervised method using the advantage of unlabeled samples and transformer architectures. We used six different areas in Germany and four years to evaluate the robustness of the model. Our experiments showed that self-supervised deep learning methods could provide a significant advantage in handling these variations. In some cases, we observed around 30 percentage points improvements in F1-score performance compared to a Random Forest based model.
- Open Access
- 27 May 2021
- Published by MDPI AG in Remote Sensing
- Vol. 13 (11) , 2114
Increasing Timeliness of Satellite-Based Flood Mapping Using Early Warning Systems in the Copernicus Emergency Management Service
Remotely sensed images have become an important source of information for actors involved in disaster management and satellite-based emergency mapping (SEM) is increasingly used to support the response phase in the first hours and days after a disaster occurs. The delivery timeliness of the crisis information is key to the success of SEM. In the Copernicus Emergency Management Service (CEMS), a procedure was tested during the past 5 years which links the European Flood Awareness System (EFAS) with the on-demand Rapid Mapping module in order to anticipate satellite tasking in view of an upcoming activation of the service for riverine floods. This study aims at assessing if the procedure has helped to improve the delivery timeliness of the first flood impact information. For the assessment, we used the Rapid Mapping performance statistics recorded from June 2016 to December 2020. Standard Rapid Mapping activations for floods were compared with those preceded by an EFAS based pre-tasking request. The focus was on essential time stamps such as activation start, provision of imagery and the availability of derived information products. For the pre-tasking-related activations, we further compared the EFAS predictions with the Rapid Mapping user request and compared flood predictions with actual observations. Our results show that the EFAS based pre-tasking improves the timeliness of the first product delivery due to the fact that satellite images could be acquired earlier compared to activations without pre-tasking.
- 27 May 2021
- Published by MDPI AG in Remote Sensing
- Vol. 13 (11) , 2114
Open AccessIncreasing Timeliness of Satellite-Based Flood Mapping Using Early Warning Systems in the Copernicus Emergency Management Service
Remotely sensed images have become an important source of information for actors involved in disaster management and satellite-based emergency mapping (SEM) is increasingly used to support the response phase in the first hours and days after a disaster occurs. The delivery timeliness of the crisis information is key to the success of SEM. In the Copernicus Emergency Management Service (CEMS), a procedure was tested during the past 5 years which links the European Flood Awareness System (EFAS) with the on-demand Rapid Mapping module in order to anticipate satellite tasking in view of an upcoming activation of the service for riverine floods. This study aims at assessing if the procedure has helped to improve the delivery timeliness of the first flood impact information. For the assessment, we used the Rapid Mapping performance statistics recorded from June 2016 to December 2020. Standard Rapid Mapping activations for floods were compared with those preceded by an EFAS based pre-tasking request. The focus was on essential time stamps such as activation start, provision of imagery and the availability of derived information products. For the pre-tasking-related activations, we further compared the EFAS predictions with the Rapid Mapping user request and compared flood predictions with actual observations. Our results show that the EFAS based pre-tasking improves the timeliness of the first product delivery due to the fact that satellite images could be acquired earlier compared to activations without pre-tasking.
- Open Access
- 4 March 2021
- Published by Copernicus GmbH
RapidAI4EO: Advancing the State-of-the-Art in Continuous Land Monitoring
New catalogues of nearly daily or even intraday temporal data will soon dominate the global archives. However, there has been little exploration of artificial intelligence (AI) techniques to leverage the high cadence that is already possible to achieve through the fusion of multiscale, multimodal sensors. Under the sponsorship of the European Union’s Horizon 2020 programme, RapidAI4EO will establish the foundations for the next generation of Copernicus Land Monitoring Service (CLMS) products. Focus is on the CORINE Land Cover programme, which is the flagship of CLMS. Specific objectives of the project are to: 1) explore and stimulate the development of new spatiotemporal monitoring applications based on the latest advances in AI and Deep Learning (DL); 2) demonstrate the fusion of Copernicus high resolution satellite imagery and third party very high resolution imagery; 3) provide intensified monitoring of Land Use and Land Cover, and Land Use change at a much higher level of detail and temporal cadence than it is possible today. Our strategy is two-fold. The first aspect involves developing vastly improved DL architectures to model the phenomenology inherent in high cadence observations with focus on disentangling phenology from structural change. The second involves providing critical training data to drive advancement in the Copernicus community and ecosystem well beyond the lifetime of this project. To this end we will create the most complete and dense spatiotemporal training sets ever, combining Sentinel-2 with daily, harmonized, cloud-free, gap filled, multispectral 3m time series resulting from fusion of open satellite data with Planet imagery at as many as 500,000 patch locations over Europe. The daily time series will span the entire year 2018, to coincide with the latest release of CORINE. We plan to open source these datasets for the benefit of the entire remote sensing community.This talk focuses on the description of the datasets whose inspirations comes from the recently released EuroSAT (Helbert et al, 2019) and BigEarthNet corpora (Sumbul et al, 2019). The new corpora will look at the intersection of CORINE 2018 with all the countries in the EU, balancing relative country surface with relative LULC distribution and most notably adding the daily high resolution time series at all locations for the year 2018. Annotations will be based on the CORINE ontology. The higher spatial resolution will support modeling of more LC classes, while the added temporal dimension should enable disambiguation of land covers across diverse climate zones, as well as an improved understanding of land use.This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101004356.
- 4 March 2021
- Published by Copernicus GmbH
Open AccessRapidAI4EO: Advancing the State-of-the-Art in Continuous Land Monitoring
New catalogues of nearly daily or even intraday temporal data will soon dominate the global archives. However, there has been little exploration of artificial intelligence (AI) techniques to leverage the high cadence that is already possible to achieve through the fusion of multiscale, multimodal sensors. Under the sponsorship of the European Union’s Horizon 2020 programme, RapidAI4EO will establish the foundations for the next generation of Copernicus Land Monitoring Service (CLMS) products. Focus is on the CORINE Land Cover programme, which is the flagship of CLMS. Specific objectives of the project are to: 1) explore and stimulate the development of new spatiotemporal monitoring applications based on the latest advances in AI and Deep Learning (DL); 2) demonstrate the fusion of Copernicus high resolution satellite imagery and third party very high resolution imagery; 3) provide intensified monitoring of Land Use and Land Cover, and Land Use change at a much higher level of detail and temporal cadence than it is possible today. Our strategy is two-fold. The first aspect involves developing vastly improved DL architectures to model the phenomenology inherent in high cadence observations with focus on disentangling phenology from structural change. The second involves providing critical training data to drive advancement in the Copernicus community and ecosystem well beyond the lifetime of this project. To this end we will create the most complete and dense spatiotemporal training sets ever, combining Sentinel-2 with daily, harmonized, cloud-free, gap filled, multispectral 3m time series resulting from fusion of open satellite data with Planet imagery at as many as 500,000 patch locations over Europe. The daily time series will span the entire year 2018, to coincide with the latest release of CORINE. We plan to open source these datasets for the benefit of the entire remote sensing community.This talk focuses on the description of the datasets whose inspirations comes from the recently released EuroSAT (Helbert et al, 2019) and BigEarthNet corpora (Sumbul et al, 2019). The new corpora will look at the intersection of CORINE 2018 with all the countries in the EU, balancing relative country surface with relative LULC distribution and most notably adding the daily high resolution time series at all locations for the year 2018. Annotations will be based on the CORINE ontology. The higher spatial resolution will support modeling of more LC classes, while the added temporal dimension should enable disambiguation of land covers across diverse climate zones, as well as an improved understanding of land use.This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101004356.