12 research outputs found
DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications
No full textSignificant efforts have been directed towards adapting self-supervised multimodal learning for Earth observation applications. However, most current methods produce coarse patch-sized embeddings, limiting their effectiveness and integration with other modalities like LiDAR. To close this gap, we present DUNIA, an approach to learn pixel-sized embeddings through cross-modal alignment between images and full-waveform LiDAR data. As the model is trained in a contrastive manner, the embeddings can be directly leveraged in the context of a variety of environmental monitoring tasks in a zero-shot setting. In our experiments, we demonstrate the effectiveness of the embeddings for seven such tasks: canopy height mapping, fractional canopy cover, land cover mapping, tree species identification, plant area index, crop type classification, and per-pixel waveform-based vertical structure mapping. The results show that the embeddings, along with zero-shot classifiers, often outperform specialized supervised models, even in low-data regimes. In the fine-tuning setting, we show strong performances near or better than the state-of-the-art on five out of six tasks
DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications
No full textSignificant efforts have been directed towards adapting self-supervised multimodal learning for Earth observation applications. However, most current methods produce coarse patch-sized embeddings, limiting their effectiveness and integration with other modalities like LiDAR. To close this gap, we present DUNIA, an approach to learn pixel-sized embeddings through cross-modal alignment between images and full-waveform LiDAR data. As the model is trained in a contrastive manner, the embeddings can be directly leveraged in the context of a variety of environmental monitoring tasks in a zero-shot setting. In our experiments, we demonstrate the effectiveness of the embeddings for seven such tasks: canopy height mapping, fractional canopy cover, land cover mapping, tree species identification, plant area index, crop type classification, and per-pixel waveform-based vertical structure mapping. The results show that the embeddings, along with zero-shot classifiers, often outperform specialized supervised models, even in low-data regimes. In the fine-tuning setting, we show strong performances near or better than the state-of-the-art on five out of six tasks
DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications
No full textSignificant efforts have been directed towards adapting self-supervised multimodal learning for Earth observation applications. However, most current methods produce coarse patch-sized embeddings, limiting their effectiveness and integration with other modalities like LiDAR. To close this gap, we present DUNIA, an approach to learn pixel-sized embeddings through cross-modal alignment between images and full-waveform LiDAR data. As the model is trained in a contrastive manner, the embeddings can be directly leveraged in the context of a variety of environmental monitoring tasks in a zero-shot setting. In our experiments, we demonstrate the effectiveness of the embeddings for seven such tasks: canopy height mapping, fractional canopy cover, land cover mapping, tree species identification, plant area index, crop type classification, and per-pixel waveform-based vertical structure mapping. The results show that the embeddings, along with zero-shot classifiers, often outperform specialized supervised models, even in low-data regimes. In the fine-tuning setting, we show strong performances near or better than the state-of-the-art on five out of six tasks
DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications
No full textSignificant efforts have been directed towards adapting self-supervised multimodal learning for Earth observation applications. However, most current methods produce coarse patch-sized embeddings, limiting their effectiveness and integration with other modalities like LiDAR. To close this gap, we present DUNIA, an approach to learn pixel-sized embeddings through cross-modal alignment between images and full-waveform LiDAR data. As the model is trained in a contrastive manner, the embeddings can be directly leveraged in the context of a variety of environmental monitoring tasks in a zero-shot setting. In our experiments, we demonstrate the effectiveness of the embeddings for seven such tasks: canopy height mapping, fractional canopy cover, land cover mapping, tree species identification, plant area index, crop type classification, and per-pixel waveform-based vertical structure mapping. The results show that the embeddings, along with zero-shot classifiers, often outperform specialized supervised models, even in low-data regimes. In the fine-tuning setting, we show strong performances near or better than the state-of-the-art on five out of six tasks
High-resolution canopy height map in the Landes forest (France) based on GEDI, Sentinel-1, and Sentinel-2 data with a deep learning approach
No full textInternational audienceIn intensively managed forests in Europe, where forests are divided into stands of small size and may show heterogeneity within stands, a high spatial resolution (10---20 m) is needed to capture the differences in canopy height. In this work, we developed a deep learning model based on multi -sensor remote sensing measurements to create a high -resolution canopy height map over the "Landes de Gascogne" forest in France, a large maritime pine plantation of 13,000 km2 with flat terrain and intensive management. This area is characterized by even -aged and mono -specific stands, of a typical length of a few hundred meters, harvested every 35 to 50 years. Our deep learning U -Net model uses multi -band images from Sentinel -1 and Sentinel -2 with composite time averages as input to predict tree height derived from GEDI waveforms. The evaluation is performed with external validation data from forest inventory plots and a stereo 3D reconstruction model based on Skysat imagery available at specific locations. We trained seven different U -Net models based on combinations of Sentinel -1 and Sentinel -2 bands to evaluate the importance of each sensor in the dominant height retrieval. The model outputs allow us to generate a 10 m resolution canopy height map of the whole "Landes de Gascogne" forest area for 2020 with a mean absolute error of 2.02 m on the test dataset. The best predictions were obtained using all available bands from Sentinel -1 and Sentinel -2 but using only one satellite source also provided good predictions. For all validation datasets in coniferous forests, our model showed better metrics than previous canopy height models available in the same region
High-resolution data reveal a surge of biomass loss from temperate and Atlantic pine forests, contextualizing the 2022 fire season distinctiveness in France
No full textThe frequency and intensity of summer droughts and heat
waves in Western Europe have been increasing, raising concerns about the
emergence of fire hazard in less fire-prone areas. This exposure of
old-growth forests hosting unadapted tree species may cause
disproportionately large biomass losses compared to those observed in
frequently burned Mediterranean ecosystems. Therefore, analyzing fire
seasons from the perspective of exposed burned areas alone is insufficient;
we must also consider impacts on biomass loss. In this study, we focus on
the exceptional 2022 summer fire season in France and use very
high-resolution (10 m) satellite data to calculate the burned area, tree
height at the national level, and subsequent ecological impact based on
biomass loss during fires. Our high-resolution semi-automated detection
estimated 42 520 ha of burned area, compared to the 66 393 ha estimated by
the European automated remote sensing detection system (EFFIS), including
48 330 ha actually occurring in forests. We show that Mediterranean forests
had a lower biomass loss than in previous years, whereas there was a drastic
increase in burned area and biomass loss over the Atlantic pine forests and
temperate forests. High biomass losses in the Atlantic pine forests were
driven by the large burned area (28 600 ha in 2022
vs. 494 ha yr−1 in
2006–2021 period) but mitigated by a low exposed tree biomass mostly located
on intensive management areas. Conversely, biomass loss in temperate forests
was abnormally high due to both a 15-fold increase in burned area compared
to previous years (3300 ha in 2022 vs. 216 ha in the 2006–2021 period) and
a high tree biomass of the forests which burned. Overall, the biomass loss
(i.e., wood biomass dry weight) was 0.25 Mt in Mediterranean forests and
shrublands, 1.74 Mt in the Atlantic pine forest, and 0.57 Mt in temperate
forests, amounting to a total loss of 2.553 Mt, equivalent to a 17 %
increase of the average natural mortality of all French forests, as reported
by the national inventory. A comparison of biomass loss between our
estimates and global biomass/burned areas data indicates that higher
resolution improves the identification of small fire patches, reduces the
commission errors with a more accurate delineation of the perimeter of each
fire, and increases the biomass affected. This study paves the way for the
development of low-latency, high-accuracy assessment of biomass losses and
fire patch contours to deliver a more informative impact-based
characterization of each fire year.</p
DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications
No full textIB-AGC: Annual 25 km global live biomass carbon product from SMOS L-band passive microwave vegetation optical depth
No full textInternational audienceMonitoring aboveground biomass carbon (AGC) stocks and their changes is crucial for understanding the global carbon cycle and the impact of climate change. Among remotely sensed methods, the use of the L-band (1.4 GHz) vegetation optical depth (L-VOD) derived from passive microwave satellite observations, offers rapid updates for timely monitoring of interannual AGC changes. L-VOD is sensitive to changes in the total water content of vegetation, which is determined by both AGC (the biomass of vegetation) and vegetation moisture content. While several methods have been used to understand and correct for the influence of the latter parameter when inferring AGC from L-VOD, there is still a lack of quantified corrections for its impact. Moreover, varying benchmark biomass datasets and fitting functions are currently used for converting L-VOD to AGC, making it difficult to harmonize or compare AGC estimates at regional and global scales. To address these issues, we first corrected the L-VOD time series for changes in the vegetation moisture content and then implemented a systematic global-scale calibration, resulting in annual AGC data set (called INRAE-BORDEAUX AGC, hereafter IB AGC) from 2010 to 2020 at a 25 km resolution. The accuracy assessments showed that IB AGC had a reasonably good spatial agreement with LiDAR referenced AGC data (R 2 = 0.60). Moreover, when aggregated at the national level, IB AGC exhibited stronger consistency with long-term net changes from country-level forest inventory data (R 2 = 0.62) than other mainstream satellite products. It is expected that IB AGC will provide an independent means for better monitoring the global vegetation carbon stocks and their variability in response to climate change. Background & SummaryAbove-ground biomass carbon (AGC) in live vegetation is a main land carbon stock and can increase or decrease from environmental changes and direct human activities such as deforestation or forest management 1,2 . The carrier of AGC, above-ground biomass (AGB), has been recognized as an Essential Climate Variable (ECV) by the Global Climate Observing System (GCOS) 3 , serving as a critical input for constraining Earth system models 4,5 and a requisite for countries in their mandated carbon accounting 6-8 . Most of the Earth biomass is in forests, with large regional differences of carbon stocks and stocks density per unit area. Accurately monitoring changes</div
Impact of Fecal Microbiota Transplantation on Digestive Tract Colonization due to Carbapenem-resistant Enterobacteriacae and Vancomycin-resistant Enterococci
No full textAbstract
Background
Fecal Microbiota Transplantation (FMT) has proved to be an efficient therapy for recurrent C. difficile infection. Its indication is currently discussed for the decolonization of Multidrug-resistant organisms (MDRO) on the basis of mice experiments. Two recent publications suggest that it could be an efficient strategy for patients colonized with digestive MDRO colonization but few data are available for Carbapenem-Resistant Enterobacteria (CRE) and Vancomycin-Resistant Enterococcus (VRE) colonization.
Methods
We performed a FMT among patients colonized by CRE or VRE documented by at least 3 nonconsecutive positive swabs (including one in the week prior to the FMT).
Procedure: 2 days prior to the FMT, patients received a proton pump inhibitor and a naso-duodenal tube was inserted to perform a bowel lavage with X-prep. FMT was performed with frozen feces from 4 donors previously screened for potential diseases using 5 syringes of 50 cc of feces diluted with saline. Patients were discharged after 24h and benefited of outpatient control swabs (PCR + culture) on day 7, 14, 21, 28 and each month during 3 months in order to assess the decolonization. The study is registered at ClinicalTrials.gov (NCT03029078).
Results
Seventeen individuals were included. Mean age was 69 ± 12.7 (SD) years.
Eight patients were positive for CRE (KPC, OXA48 or NDM-1) and 9 for VRE.
All suffered from severe underlying condition (hemodialysis, dementia, cirrhosis) or chronic wounds. Median functional autonomy scale was evaluated using the French Iso-Resources Groups (GIR)=4/6 IQR[3–6] supporting they were dependent persons.
At 1-month follow-up, 3/8 patients were free from CRE and 5/9 from VRE. At 3-month follow-up, 3/8 patients were still free from CRE whereas 7/8 were free from VRE, considering one death from cirrhosis.
Moreover, one of them received antibiotics during a week for a hospital-acquired infection a long time after FMT. No adverse events were reported.
Conclusion
FMT seems to be an attractive option to eradicate colonization of MDRO, especially for VRE. Limited data are available in the literature to determine response factors. Meanwhile its efficacy is moderate; it provides an alternative solution to quarantine for fragile and frequently hospitalized patients. More data and a controlled trial are required.
Disclosures
All authors: No reported disclosures.
</jats:sec
High-resolution canopy height map in the Landes forest (France) based on GEDI, Sentinel-1, and Sentinel-2 data with a deep learning approach
No full textIn intensively managed forests in Europe, where forests are divided into
stands of small size and may show heterogeneity within stands, a high spatial
resolution (10 - 20 meters) is arguably needed to capture the differences in
canopy height. In this work, we developed a deep learning model based on
multi-stream remote sensing measurements to create a high-resolution canopy
height map over the "Landes de Gascogne" forest in France, a large maritime
pine plantation of 13,000 km with flat terrain and intensive management.
This area is characterized by even-aged and mono-specific stands, of a typical
length of a few hundred meters, harvested every 35 to 50 years. Our deep
learning U-Net model uses multi-band images from Sentinel-1 and Sentinel-2 with
composite time averages as input to predict tree height derived from GEDI
waveforms. The evaluation is performed with external validation data from
forest inventory plots and a stereo 3D reconstruction model based on Skysat
imagery available at specific locations. We trained seven different U-net
models based on a combination of Sentinel-1 and Sentinel-2 bands to evaluate
the importance of each instrument in the dominant height retrieval. The model
outputs allow us to generate a 10 m resolution canopy height map of the whole
"Landes de Gascogne" forest area for 2020 with a mean absolute error of 2.02 m
on the Test dataset. The best predictions were obtained using all available
satellite layers from Sentinel-1 and Sentinel-2 but using only one satellite
source also provided good predictions. For all validation datasets in
coniferous forests, our model showed better metrics than previous canopy height
models available in the same region.Comment: 39 pages, 16 figures + supplementary content
