Deepak R. Mishra

Antibiotic resistance is a significant threat to global public health and can disproportionately affect low‐ and middle‐ income countries (LMICs). There is a lack of studies focusing on antibiotic resistance in coral reef regions and environmental reservoirs in Central America. This study followed modified World Health Organization (WHO)'s Global Tricycle Surveillance protocols for the environmental sector to address these gaps. Water samples were collected from key areas in the lower Belize River, including above and below the Belize City, an open fish market, and sewage lagoon outfall, and coral reefs. Water samples underwent qPCR analysis for a suite of antibiotic resistance gene classes (sul1, sul2, ermF, tetA, and blaSHV), intI1, and 16S rRNA. Additionally, a subset of samples were tested for extended‐spectrum β‐lactamase (ESBL) E. coli and underwent shotgun sequencing. Results show that antibiotic resistance genes (ARGs) were highest and most diverse near Belize City, particularly near the treatment lagoons and open fish market. The coral reef regions had lower levels of antibiotic resistance though not void of their presence. This study is an application of a modified Global Tricycle Surveillance protocol integrated with qPCR‐ and metagenomics‐based characterization of environmental antibiotic resistance in understudied areas. Notably, data from this study indicated that ESBL‐ E. coli could potentially be used as a screening tool for environmental antibiotic resistance, as it was only present at sites that had the highest levels of ARGs.

Coastal Georgia's (USA) salt marshes function as soil organic carbon (SOC) sinks because of their high primary productivity and slow decomposition rates. Quantifying SOC stocks is essential to establish a baseline for long‐term SOC trends and to highlight their capacity for long‐term carbon (C) sequestration. From 2021 to 2023, 28 soil cores (25–165 cm deep) were collected from six Georgia marshes and analyzed for SOC%, total nitrogen (N), bulk density (BD), and particle size by horizon. Stable C isotopes were examined in three marshes to identify C sources, and a small subset of samples was analyzed for radiocarbon. Surface SOC (0–5 cm) ranged from 1.4% to 8.5% (SE = 0.28), while SOC across all depths ranged from 0.08% to 8.8% (mean = 3.7%, SE = 0.07). Mainland sites exhibited significantly higher surface SOC than island sites and were dated to be younger with radiocarbon testing. Stable isotopes indicated shifts in C ₃ and C ₄ vegetation dominance over time. Georgia salt marshes store an estimated 16,777,000 Mg C to 70 cm depth. Calculated to 60 cm, the mean SOC density was 88.5 Mg C ha ⁻¹ , approximately 55% of the total U.S. national average for tidal salt marshes. Variations in SOC content were primarily explained by clay content, depth, and BD. These results provide SOC estimates for Georgia's salt marshes and help address a regional data gap for southeastern U.S. coastal wetlands, underscoring the importance of broad scale assessments for improving national C accounting and understanding spatial variability in C storage.

We present our methodology and data for science ready vertical carbon fluxes from a Spartina alterniflora tidal salt marsh as part of the Georgia Coastal Ecosystems Long Term Ecological Research (GCE-LTER) site on Sapelo Island, Georgia, USA. Vertical carbon fluxes were measured using the eddy covariance (EC) method from 2014 to 2024. The EC flux tower was located on the western side of Sapelo Island bounded by the Duplin River and Barn Creek. The proportional influence of marsh habitats on the flux measurements were 4% tall, 38% short, and 58% medium height form Spartina alterniflora. We provide the net ecosystem exchange (NEE), ecosystem respiration (ER), and gross primary production (GPP) at 30-minute fluxes (μmol CO₂ m⁻²s⁻¹), daily averages (μmol CO₂ m⁻²s⁻¹) and totals (g C m⁻² day⁻¹), and annual (g C m⁻² year⁻¹) quantities. We estimated uncertainty for each flux at each integrated timescale as 95% confidence intervals. Providing open access to 10-year carbon flux datasets can facilitate collaboration for advancing regional and global blue carbon synthesis and scale-up studies.

Neural Radiance Fields (NeRFs) have demonstrated remarkable capabilities in 3D reconstruction and novel view synthesis. However, most existing NeRF frameworks require complete retraining when new views are introduced incrementally, limiting their applicability in domains where data arrives sequentially. This limitation is particularly problematic in satellite-based terrain analysis, where regions are repeatedly observed over time. Incremental refinement of NeRFs remains underexplored, and naive approaches suffer from catastrophic forgetting when past data is unavailable. We propose $Δ$-NeRF, a unique modular residual framework for incremental NeRF refinement. $Δ$-NeRF introduces several novel techniques including: (1) a residual controller that injects per-layer corrections into a frozen base NeRF, enabling refinement without access to past data; (2) an uncertainty-aware gating mechanism that prevents overcorrection by adaptively combining base and refined predictions; and (3) a view selection strategy that reduces training data by up to 47\% while maintaining performance. Additionally, we employ knowledge distillation to compress the enhanced model into a compact student network (20\% of original size). Experiments on satellite imagery demonstrate that $Δ$-NeRF achieves performance comparable to joint training while reducing training time by 30-42\%. $Δ$-NeRF consistently outperforms existing baselines, achieving an improvement of up to 43.5\% in PSNR over naive fine-tuning and surpassing joint training on some metrics.

Environmental remote sensing (ERS) has been a cornerstone technology for decades, allowing for comprehensive monitoring and management of ecosystems. Recent advancements in artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), have transformed the field of ERS by offering robust, data-driven solutions for analyzing complex, large-scale remote sensing datasets and answering important scientific questions. This intervention explores the diverse applications of AI in ERS, focusing on key domains such as water, agriculture, urban, and wetland monitoring. This work also includes discussions on crucial aspects such as the impact of foundation models in ERS and outlines key points on the ethical and responsible use of AI in ERS. The intervention is concluded by identifying promising next steps for future research.

Salt marshes provide valuable ecosystem services but are vulnerable to drowning with accelerated sea-level rise (SLR). Marsh belowground biomass (BGB) production helps avoid drowning by building marsh surface elevation. Reductions in BGB can serve as an early warning sign of marsh deterioration, as they often precede decreases in aboveground biomass (AGB). However, landscape-scale BGB assessments to predict broad trends in marsh deterioration have not been previously available. We applied the Belowground Ecosystem Resiliency Model (BERM) to assess standing stocks and trends in both BGB and AGB over the past decade (2014–2023) across US Georgia coast Spartina alterniflora marshes (691 km ² ). Over this time period, BGB and AGB averaged 841 ± 323 and 221 ± 14 g m ⁻² , respectively, but showed opposite trends. BGB decreased on average by 0.94% per year and over most of the marsh area (72%), while AGB increased on average by 0.66% per year and showed a net increase across most of the marsh area (88%). This disconnect suggests that AGB is not a good indicator of marsh resilience, and we highlight two areas with similar AGB but different BGB. Inundation intensity, an important predictor of BGB, rose through time and was negatively related to BGB. SLR trends suggest continuing increases in inundation, which will result in further declines in BGB followed by widespread marsh drowning. Landscape BGB assessments are a valuable tool to identify ecosystem vulnerability and proactively manage salt marshes and the services they provide under rising sea levels.

Cyanobacterial harmful algal blooms (CyanoHABs) pose significant threats to aquatic ecosystems, public health, and economic sustainability worldwide. This progress report explores recent advancements in CyanoHAB detection, quantification, and monitoring using multi-sensor remote sensing approaches, artificial intelligence (AI) applications, and their integration with health impact studies. We presented the capabilities of various satellite sensors in monitoring CyanoHABs across different spatial and temporal scales, discussing the integration of multiple data sources to overcome individual sensor limitations. The report highlights the promise of AI, particularly machine learning (ML) techniques, in improving CyanoHAB detection and forecasting, demonstrating how ML methods consistently outperformed traditional algorithms in estimating phycocyanin concentrations, a key indicator of CyanoHABs.We examined the development of cloud-based applications for real-time monitoring and public awareness. Furthermore, we explored the health impacts of CyanoHABs on humans and animals, emphasizing the role of remote sensing and ML in mitigating these effects. The economic implications of CyanoHAB-related health issues are discussed, along with the potential for integrating remote sensing data with epidemiological studies. Overall, this report underscores the importance of cross-disciplinary, integrated approaches that combine cutting-edge remote sensing technologies, advanced ML techniques, and health impact assessments to address the complex challenges posed by CyanoHABs in inland waters.

Estuarine wetlands have been shown to be an important source of dissolved organic carbon (DOC) to surrounding waters by consistently exporting optically and compositionally distinctive organic materials with tidal draining. Numerous studies have been conducted in marshes mostly drained by coastal creeks and regional rivers, yet the spatial biogeochemical gradients of DOC developed from coastal marshes influenced by world's mega rivers are little studied. Using high resolution Landsat-8/-9 and Sentinel-2, we focused on characterizing the spatial gradients of marsh-DOC exported with tidal exchanges over a unique estuarine background impacted by a globally significant river. Designated as the ‘Wetlands of International Importance’ by the Ramsar Convention, the Chongming Dongtan Nature Reserve is the largest, representative estuarine wetlands in China, influenced by Yangtze River — the fifth largest river worldwide by discharge. LaSRC and Sen2Cor were found to be the preferred atmospheric correction models for Landsat-8/-9 and Sentinel-2A/B, respectively, for this turbid nearshore environment. An ocean color DOC algorithm based on the strong linear correlation between DOC concentration and the absorption coefficient of its chromophoric component (CDOM, a_CDOM) retrieved from Landsat-8/-9 and Sentinel-2A/B remote sensing reflectance showed good DOC estimates, with a mean absolute percent difference of 11% for a_CDOM(300) and 5% for DOC. Maps of marsh-to-estuary DOC distribution emphasized the predominance of Yangtze River, with less identifiable biogeochemical gradients of DOC in this mega-river-impacted estuary. This implies the impact of lateral carbon transfer from coastal marshes influenced by world's large rivers might be limited in altering coastal biogeochemistry.