Water Research

Antibiotic resistance genes (ARGs) travel from cities into rivers and coastal waters through wastewater treatment plants. Monitoring them normally requires metagenomic sequencing, which is sufficiently costly that most utilities can sample only occasionally. Weather and location data are freely available on a daily basis for virtually any location worldwide, making them attractive predictors for identifying where limited sequencing resources should be prioritized. However, whether such models generalize to previously unseen cities remains largely untested, as most published studies train and evaluate models within the same catchments, thereby assessing interpolation rather than geographic transferability. To address this gap, we paired 235 wastewater metagenomes collected from five European cities with 23 abiotic predictors spanning geospatial, meteorological, hydrological, radiative, and temporal domains. Model performance was evaluated using leave-one-group-out (LOGO) cross-validation, in which all samples from one city were withheld for testing while the remaining cities were used for training. CatBoost achieved a median LOGO ROC-AUC of 0.929 and a median F1-score of 0.750. Using freely available environmental reanalysis predictors (meteorological, hydrological, radiative, geospatial, and temporal) - without any metagenomic sequencing - CatBoost achieved a mean ROC-AUC of 0.722, recovering 78% of the predictive performance of the full omics-integrated model. Removing latitude and longitude reduced ROC-AUC by only 0.003, whereas replacing random cross-validation with city-wise validation reduced ROC-AUC by 0.052. Predictive performance varied across ARG classes, ranging from a ROC-AUC of 0.981 for {beta}-lactam resistance genes to 0.762 for glycopeptide resistance genes. These findings demonstrate that freely available environmental reanalysis predictors - spanning meteorological, hydrological, radiative, geospatial, and temporal domains - recover 78% of the predictive signal for antibiotic resistance gene occurrence in urban wastewater, allowing scarce sequencing capacity to be directed to the catchments where it changes a decision.

Bench-scale microbial fuel cells (MFCs) treating mining wastewater frequently exhibit operational variability and uncharacterized degradation that obscure true biocatalytic performance. To decouple genuine biological treatment effects from mechanical failures, this paper presents an integrated diagnostic framework validated on two bench-scale systems treating heavy-metal-rich gold mine tailings. The first system evaluates Micractinium inermum algal bio-augmentation (System 1), while the second compares Psychrobacter alimentarius- and Trichococcus patagoniensis-dominated anodic consortia (System 2). To overcome single-reactor constraints, the framework integrates paired time-series statistical modeling, an adaptive percentile-floor change-point detector, equivalent-circuit modeling, and baseline-corrected spectroscopy (XRD/FTIR). Applying the framework to these systems uncovers previously masked dynamics: statistical analysis demonstrates that algal biocatalysis provides no voltage advantage under stable operation (+0.17%) but increases output by 27.54% under diurnal perturbation, while periodicity analysis links these diurnal shifts to the chamber photoperiod. Furthermore, heavy-metal remediation (up to 97.7%) is governed by system-level physicochemical mechanisms rather than algal-specific processes. The change-point detector successfully isolates distinct failure modes, distinguishing a recoverable excursion from terminal structural collapse. Finally, equivalent-circuit modeling reveals that the superior power density of Trichococcus consortia is driven by combined improvements in internal resistance and open-circuit voltage. Ultimately, pairing statistical controls with automated fault detection resolves operational ambiguity, offering a scalable baseline for health monitoring in bio-electrochemical wastewater treatment.

Data-driven soft sensors can estimate fermentation product concentrations from routinely recorded process variables, but strong performance during normal operation does not establish reliability during process deviations. This study evaluated current-time penicillin-concentration estimation using 100 simulated IndPenSim batches comprising 90 normal-operation batches and ten documented deviation batches, with 113,935 observations in total. Initial normal-trained models and five follow-up experiments examined complete-batch validation, dependence on batch-progress features, phase-specific error, model-family comparisons, early out-of-distribution warnings and empirical prediction ranges. These analyses motivated a matched comparison between normal-only and fault-inclusive HistGradientBoosting regressors using 36 current and causal history-based inputs. Normal performance was evaluated in five regime-balanced complete-batch folds, and deviation performance by leave-one-fault-batch-out evaluation; every tested deviation batch was excluded from its own model fit. Batch-balanced sample weights were used, with a factor of three assigned to permitted deviation batches in fault-inclusive fitting. On held-out deviation batches, pooled RMSE decreased from 3.195 to 2.564 g/L, a reduction of 19.73%, while MAE decreased from 2.076 to 1.441 g/L and R2 increased from 0.8580 to 0.9085. Normal-operation RMSE was nearly unchanged at 1.981 and 1.983 g/L. Eight of ten deviation batches improved. The mean paired fault-batch RMSE difference was -0.7063 g/L, with a descriptive 95% batch-bootstrap interval of -1.2844 to -0.2194 g/L. Improvement was largest within the first-to-last recorded fault-reference window, but late-stage errors persisted. Batch 100 remained poorly predicted, with fault-inclusive RMSE of 6.631 g/L and R2 of -2.4837. A fault-risk classifier, Isolation Forest OOD detector and empirical error ranges offered incomplete reliability information. Fault-inclusive training therefore improved this benchmark on average, but did not establish generalization to unseen fault mechanisms, calibrated safety warnings or deployment in physical fermentation.

Controlling pathogenic microorganisms during water treatment processes is a major public health concern. However, the impact of different water sources and treatment processes on the drinking water microbiome and opportunistic bacterial pathogens remains poorly understood. Here, we report the analyses of two water systems: a river and groundwater source, and their corresponding treated waters, sampled bi-weekly for 6 months. Microbial communities were analysed through a metabarcoding and dPCR workflow coupling universal prokaryotic and eukaryotic primers with genus-targeting primers for Legionella and Pseudomonas to achieve a deep taxonomic resolution. We show that microbial communities in groundwater were more diverse than those in river water and that water treatment affected -diversity. Water treatment had a stronger influence on community composition than water system or seasonal variation. Despite lower bacterial loads, the relative abundance of Legionella, Pseudomonas, and non-tuberculous Mycobacteria (NTM) remained stable or even increased after water treatment, suggesting that conventional treatment processes promote enrichment of specific taxa. Excitingly, the genus-targeting analyses revealed a remarkable and previously uncharacterized diversity of Legionella species. L.pneumophila, pathogenic Pseudomonas clades and P. aeruginosa were detected in all environments, emphasizing their ability to adapt to diverse ecological conditions. Analysis of the eukaryotic fraction emphasized the persistence and emergence of eukaryotic hosts associated with these waterborne opportunistic pathogens. Our findings offer important insights into how water source characteristics and treatment processes influence the composition of the drinking water microbiome, while the high taxonomic resolution achieved for waterborne opportunistic bacterial pathogens underscores their ecological adaptability.

Cyclosporiasis is a foodborne illness caused by the parasite Cyclospora cayetanensis. Clinical surveillance often lags infection by weeks because diagnosis depends on a clinician ordering a test that is not part of routine parasite panels. Wastewater metagenomic sequencing does not depend on care seeking, which may be useful for early detection and community monitoring. We designed 1,670 31-mers from C. cayetanensis mature ribosomal RNA and retained 1,464 without an exact non-target match in the 28 July 2025 NCBI core_nt release. By analyzing 293 public wastewater metagenomes, we selected a threshold of 24 distinct diagnostic 31-mers on one read. Applying that threshold to 2,333 public runs from BioProject PRJNA1247874 yielded putative C. cayetanensis reads in 81 runs, or 3.5%, and in 24 of the 30 longitudinally sampled sewersheds. Detections rose through the summer in two consecutive years, matching the known seasonality of cyclosporiasis. The most parsimonious explanation is that these reads come from parasites shed into wastewater, although undescribed Cyclospora species absent from reference databases cannot be excluded. The same approach can be applied to other pathogens in existing wastewater sequencing data.

Wastewater-based epidemiology uses both targeted and non-targeted bioanalytical approaches to detect and track disease-causing microbes in wastewater. Targeted approaches such as quantitative polymerase chain reaction (qPCR) remain widely used while sequencing-based approaches have slowly gained popularity. However, concordance between established targeted molecular assays and sequencing approaches remains incompletely characterized. To assess the validity of sequencing as an alternative to qPCR, we compared outcomes of both methods targeting five respiratory viruses in wastewater samples collected across 16 Texas cities over 4 years (May 2022-July 2026) of continuous sampling. We found positive correlations between qPCR and sequencing measurements, with both methods independently correlating with statewide clinical emergency department visits. qPCR positivity also increased with genomic sequencing coverage, and sequencing demonstrated greater potential than qPCR as an early respiratory virus outbreak detection method. Our findings support metagenomic sequencing as a sensitive and highly specific alternative to targeted qPCR for wastewater surveillance, and reinforce its utility towards long-term longitudinal monitoring of respiratory viruses.

Wastewater surveillance has become a prominent part of public health efforts to track circulating pathogens and has been incorporated into disease forecasting models. We compared two recent wastewater forecasting models for forecasting COVID-19 hospitalizations using SARS-CoV-2 concentrations in wastewater: a generalized linear mixed model (GLMM) and a Bayesian mechanistic model known as wwinference. We retrospectively produced 1-week ahead forecasts using these two conceptually different models across 10 regions in New York state, from September 2022 to mid-April 2024. We compared the performance of forecasts produced from the two models fit with wastewater and clinical data to a version of each model fit to only clinical data. Models were scored against observations using Continuous Ranked Probability Scores for each forecast (n = 363 forecasts). Of the two models, the GLMM showed improved forecast performance across space and time when evaluated on a natural scale compared to the wwinference model, but no significant difference in forecast performance was found on a log-scale, which evaluates the relative rather than absolute error. Consequently, either model could be used to forecast COVID-19 hospitalizations. Including wastewater concentrations in these models at these spatial and temporal scales did not provide additional forecasting benefit beyond just clinical data. The performance of a wastewater-only model relative to simple null models; however, demonstrates that there is a clear forecasting signal present in wastewater, thus we think the lack of change is likely due to the strong signal from the clinical measures included in the model at this spatial and temporal scale.

Wastewater contains a complex mixture of nucleic acids derived from human cells, microorganisms, diets, and agricultural or industrial inputs. Metatranscriptomics has the potential to broaden the understanding of wastewater viromes, but its utility is constrained by high-abundance RNA molecules that prevent the detection of lower-abundance signals, including human pathogens. Two major sources of this interfering background signal are ribosomal RNAs (rRNAs) from various organisms, and plant-infecting tobamoviruses from human diets, which are often the most abundant RNA viruses in wastewater. These rRNA molecules and RNA genomes can account for upwards of 95% of total RNA sequences, thereby obscuring the detection of lower-abundance viruses of public health importance. This study evaluated whether combined depletion of ribosomal RNAs and genomic RNA from abundant plant tobamoviruses can improve the recovery of underrepresented viral sequences from wastewater. RNA extracts from a large municipal wastewater treatment plant in Calgary, Canada, were processed with or without probe-based depletion of rRNAs and eight abundant tobamoviruses. Metatranscriptomes revealed that depletion reduced rRNA reads by 90%. Reads assigned to tobamoviruses, which accounted for up to 25% of viral reads in untreated samples, were reduced to <2% of viral reads after depletion. This decrease in undesired RNA signal led to an order of magnitude increase in viral read signals and improved assembled viral genome quality. Furthermore, the number of reads assigned to human and animal pathogens increased 42-fold after depletion, enabling the detection of important pathogens like influenza, RSV, and hepatitis A, which were undecteable in untreated samples. Together, these findings demonstrate that targeted depletion of abundant background RNAs improves the detection of low abundance viruses, and enhances the utility of metatranscriptomic approaches for wastewater viral surveillance.

Over the past 20 years, the Integrated Constructed Wetland (ICW) concept has been applied in Ireland for wastewater treatment, offering a nature-based solution for reducing pollutants and antimicrobial resistance genes (ARGs) in effluents and receiving environments. However, wastewater microbial communities remain inadequately characterized and their dynamics across treatment largely unexplored. Here, we present a culture-independent investigation of four Irish ICWs, aiming to advance our understanding of microbial dynamics in wastewater treatment. Shotgun metagenomic sequencing was conducted on influent and effluent samples collected in biological triplicates from four ICWs treating agricultural, companion animal, industrial, and municipal wastewater, alongside eight positive and ten negative controls for filtering and sequencing. End-to-end metagenomic analysis was performed with SqueezeMeta. Taxonomic placement of high-quality metagenome-assembled-genomes (HQ-MAGs) was confirmed with GTDB-Tk. Decontamination and statistical analyses were conducted in R. Coassembled contigs were screened for ARGs, virulence- and plasmid-associated sequences. Results revealed statistically significant shifts in taxonomic and functional profiles across treatment. Effluent populations exhibited generally higher richness than corresponding influents, yet were more similar to one another across locations. Multi-log-scale increases were observed in the relative abundance of environmental genera (Legionella, Methylotenera, Thiothrix), alongside reductions in common faecal/human indicators (Bacteroides, Lactococcus, Prevotella), and up to 80% ARGs removal. Collectively, our findings provide insights into ICW efficacy, showing that they can drive marked shifts in wastewater microbiome and ARG reduction. Our culture-independent approach retrieved 34 HQ-MAGs, including 19 potentially novel taxa, uncovering previously uncharacterized microbial diversity that is potentially unique to the Irish environment and warrants further investigation.

The COVID-19 pandemic remains a global concern, partly due to the rapid mutation rate of SARS-CoV-2 and the emergence of new variants. Wastewater surveillance has proven effective in estimating infection incidence and detecting variants earlier than clinical testing. Its importance has grown as testing rates decline due to milder disease progression. However, current methods typically rely on the prior classification of SARS-CoV-2 lineages or their signature mutations, which may delay detection. We present an alternative method that identifies changes in the viral genetic population over time without requiring prior lineage classification. This population-based approach was applied to sequencing data from wastewater samples, which are generally noisier than clinical samples. We analyzed publicly available sequencing samples from wastewater plants covering Swiss catchments in Altenrhein, St. Gall, Geneva, and Zurich. To address noise, only samples with read depths above 40 and genome coverage of at least 90% were included. Genetic diversity within pooled populations over two time periods was compared to assess changes in viral composition. We demonstrate that SARS-CoV-2 variants can be detected in wastewater sequencing data without prior lineage classification. Our method successfully detected shifts in genetic populations that corresponded to the emergence of known variants of concern (VOCs) in the analyzed regions. Notably, it also revealed the rising prevalence during the first surges of the Omicron variant. Despite the increased noise in wastewater compared to clinical samples, our approach remains effective. However, achieving reliable predictions depends on high sequencing depth, broad genome coverage, and frequent sampling.

Antimicrobial resistance poses a global health threat, yet the fate of antimicrobial resistance genes (ARGs) and microbial communities from wastewaters in receiving rivers is rarely quantified. This study investigated the degradation kinetics of ARGs in sewage-polluted river water mesocosms and under sulfamethoxazole and copper stress, separately and in combination. Seven clinically relevant ARGs (blaCTX-M, blaNDM, qnrS, sul2, tetW, ermF, and aph(3'')-Ib), a mobile genetic element marker (intI1) and bacterial marker genes (uidA, 16S rDNA) were quantified by qPCR. Bacterial community changes were monitored using 16S rDNA amplicon sequencing. All target genes decayed with approximately first-order kinetics in all conditions, with qnrS (half-life: 8.0 h) and blaCTX-M (8.4 h) declining most rapidly and intI1 (32.5 h) and sul2 (46.5 h) declining most slowly. Ordination showed that ARG composition shifted primarily over time rather than by antibiotic/metal treatment. Bacterial communities shifted from initially Campylobacterota dominated ([&ge;] 90%) to Pseudomonadota, Bacillota, and Actinomycetota dominated communities over 168 h. Quantitative microbiome profiling showed a 95% reduction in ASV richness. PICRUSt2 analysis suggested progressive enrichment of aerobic and several metabolic pathways, as the mesocosms transitioned from highly polluted anoxic conditions to oxygenated conditions. Shifts in microbial community composition were coupled with changes in the resistome as identified by Procrustes analyses (r = 0.841, p = 0.001) and other analyses. Together, results show that river self-purification can rapidly reduce ARG loads, at least in warmer climates. This study provides approximate kinetic parameters for mathematical models to predict the fate of ARGs in rivers.

Background. Correct and accurate deconvolution of SARS-CoV-2 lineages from wastewater sequencing data is a challenging task, given the intricacy of wastewater amplicon-sequencing data and the ever-growing complexity of the lineage classification. Existing benchmarking studies made use of artificial spike-in compositions, thereby falling short of reflecting the prevailing complexity of wastewater samples. Results. We present a modular, expandable, simulation-based benchmarking framework as a reproducible Snakemake workflow, named Laemple, to evaluate the performance of virus lineage deconvolution tools. Using in silico simulated data sets of varying complexity and sequencing quality, we demonstrate its utility by evaluating seven publicly available tools, based on their precision, sensitivity, and reproducibility. Freyja showed robust sensitivity and consistent performance across diverse data set complexities, alongside user-friendly installation and documentation, while VaQuERo demonstrated the highest precision. Conclusions. Our results reveal substantial variation in tool performance across conditions, emphasizing the need to benchmark with diverse and complex scenarios. This framework enables informed tool selection for researchers and public health agencies and allows developers to stress-test their tools during development and maintenance, i.e., updating the lineage definition for newly emerging virus lineages. To this end, Laemple was designed in a modular fashion for customization and future expansion to support ongoing software development across all stages of the application life-cycle management.

Electrochemical advanced oxidation processes (EAOPs) are increasingly investigated for tertiary wastewater treatment, yet their short-term effects on real wastewater effluent remain poorly constrained. This study investigated real wastewater treatment plant (WWTP) effluent following short-term boron-doped diamond (BDD) electrochemical treatment using an interdigitated double-diamond electrode (iDDE), combined with hydrochemical, stable-isotope and shotgun metagenomic analyses. After 10 min at 7.0 V and 1.02 A, physicochemical parameters and bulk carbon pools remained largely unchanged. In contrast, d18O-DO decreased by 31.3 +/-0.28 permille alongside a small increase in dissolved oxygen (0.16 +/-0.03 mmol/L), indicating substantial oxygen consistent with electrochemical oxygen generation. Shotgun metagenomics revealed pronounced restructuring of the recovered microbial DNA pool, including increases in Comamonadaceae (~12.3% vs. ~25.1%) and Zoogloeaceae (~3.3% vs. ~8.4%), whereas the oxidative-stress-defence gene repertoire showed no coherent response. These taxonomic changes cannot be interpreted as selective survival, as metagenomics do not resolve viability or intra-/extracellular DNA contribution. Overall, short-term iDDE treatment affected the microbial component of real WWTP effluent despite comparatively limited changes in bulk chemistry, highlighting that EAOP effects extend beyond conventional disinfection.

Mainstream anammox implementation for nitrogen removal is constrained by unstable nitrite supply and organic carbon requirements for nitrate byproduct removal. This study evaluated integrated fixed-film activated sludge (IFAS) biofilms to enhance anammox activity by coupling low-oxygen ammonium oxidation with volatile fatty acid (VFA)-driven nitrate reduction. Nanopore long-read metagenomic assembly recovered a high-quality, circular single-contig Candidatus Brocadia sapporoensis metagenome-assembled genome (MAG) from full-scale IFAS biofilms. This MAG encodes complete anammox metabolism, dissimilatory nitrate reduction to ammonium (DNRA) and acetate/propionate carbon transformation pathways. Metatranscriptomics showed that low dissolved oxygen (DO) upregulated Ca. B. sapporoensis genes involved in anammox, nitrate reduction, and carbon metabolism. Microaerobic assays established a DO level of 0.7 mg/L as optimal for sustaining near-maximal ammonium oxidation alongside anammox-driven total inorganic nitrogen (TIN) loss. Anoxic tests conducted in secondary effluent indicated that external acetate amendment promoted greater partial nitrate reduction and TIN loss than additional propionate amendment. Integrating this dissolved oxygen concentration with external acetate amendment in a two-stage microaerobic-anoxic system successfully achieved sequential ammonium oxidation, partial nitrate reduction, and anammox-mediated TIN removal. Stage-specific expression suggested Ca. B. sapporoensis could contribute to nitrite self-supplementation via nxrAB-mediated nitrate reduction. Overall, microaerobic ammonium oxidation and Ca. B. sapporoensis-driven partial nitrate reduction jointly sustain mainstream anammox activity. Furthermore, this study demonstrates that successful metabolic synergy depends fundamentally upon precise dissolved oxygen control and effective external acetate amendment.

Since the COVID-19 pandemic, wastewater-based surveillance (WBS) has emerged as a key approach to assess community-level health and the evolution of pathogens. To date, most established WBS systems focus on polymerase chain reaction (PCR) based detection and targeted sequencing of known pathogens because these approaches are well-accepted and include amplification of pathogen target sequences of interest thereby enabling lower limits of detection. Metagenomic next-generation sequencing (mNGS) is a promising approach to enable pathogen detection and surveillance beyond predefined pathogen lists, but its regular application to WBS has not been yet widely adopted because many key performance characteristics are not well-understood, including limit of detection (LOD) and false positive/negative rates. This paper describes a computational analysis to estimate the operational LOD of various sequencing approaches using a simplified model of a local wastewater (WW) system involving a military base. This paper also presents findings from two types of experiments: 1) laboratory-spiked, those for which Atlantibacter subterraneus (Asub) is introduced into real-world WW samples in a laboratory setting, and 2) system-spiked, those for which Asub is introduced at a source location of a real-world WW system. Findings indicate that mNGS detection performance varies with sequencing method and the data analysis process. In addition, findings indicate that site-specific method characterization should be used when implementing mNGS for WBS because sites can have different WW system configurations, background organisms and sequencing inhibitors.

The sewer microbiome is known to be the main source of process-critical bacteria involved in nitrogen and phosphorus removal in wastewater treatment plants (WWTPs) and is thus crucial for the assembly of activated sludge (AS) microbial communities. However, the extent of these bacterial loads arriving at WWTPs currently lacks scientific attention. We provide a quantitative description of the biomass loads from sewer systems to WWTPs by coupling microbial community data by 16S rRNA amplicon sequencing from the Aalborg municipality sewer catchment with mass balances and biomass production across gravity and pressure sewer systems. Accounting for microbial transformations in sewer wastewater, biofilm and sediments under steady state dry weather flow conditions, gravity sewers were found to produce higher biomass loads than pressure systems, with loads of 1.57 and 0.27 kgCOD m-3 d-1 respectively. A key feature of our sewer model is its ability to estimate biomass loads from specific taxa, such as polyphosphate accumulating organisms (PAOs). Using Aalborg West WWTP as a case study, model estimations correlated well with figures calculated based on influent wastewater data monitoring. The model-estimated loads for the PAOs Azonexus and Ca. Accumulibacter were 2.10 kg day-1 and 1.80 kg day-1, compared to the influent wastewater monitoring-based loads of 1.77 kg day-1 and 1.20 kg day-1, respectively. While providing a framework for further research on sewer microbial dynamics, this study offers a tool for gaining insights into microbial immigration, generating valuable information for improving microbial communities management in WWTPs, optimizing nutrient removal, and enhancing WWTP design.

Hospital-based wastewater surveillance may complement community and clinical surveillance data in important ways, and may be useful in jurisdictions without community-based wastewater surveillance. From May 2024-April 2026, we analyzed weekly samples (n=190) from three hospitals in New York City using digital PCR to evaluate the sensitivity, specificity, and positive predictive value (PPV) of wastewater viral detection against facility SARS-CoV-2 and influenza A/B inpatient caseloads. Sensitivity was 38-42% for SARS-CoV-2 and 36-49% for influenza A/B, while specificity exceeded 72% for all pathogens. During respiratory seasons, sensitivity reached 81% for SARS-CoV-2 and 81% for influenza A; both had 100% sensitivity during peak case weeks. Notably, off-peak influenza detections occurred in hospital wastewater at all three hospitals in summer 2024 without corresponding hospital case detection, suggesting the presence of undiagnosed cases. These findings underscore the potential utility of hospital-based wastewater monitoring for tracking respiratory virus activity.

Zoonotic spillover and early human-to-human transmission of emerging viruses are often missed by clinical surveillance. Wastewater surveillance (WS) offers low-cost complementary pathogen detection, but its value for emerging viruses remains uncertain. We reviewed evidence of viral shedding in human waste and combined shedding data, detection models and simulated transmission dynamics into a quantitative framework to identify for which types of emerging viruses WS could add most value. Simulations showed that WS improved probability or speed of outbreak detection for viruses shedding >1/100 to >10 times as much as SARS-CoV-2, depending on probability of clinical symptoms and diagnosis. Value added was highest for transmission scenarios with temporally concentrated infections resulting in spikes in daily shedders. Characteristics of SARS-CoV-2, Mpox, Influenza A, Lassa, and Zaire Ebola viruses appear more favourable for WS than others e.g. chikungunya virus. We anticipate this framework can support more targeted, evidence-based application of WS to emerging threats.

Naegleria fowleri is known to be the causative agent of highly aggressive (> 95 % mortality rate) primary amoebic meningoencephalitis (PAM). N. fowleri can colonize drinking water distribution systems (DWDSs) which place additional pressures on water distribution authorities and public health officials. Chlorination and chloramination are widely adopted to combat N. fowleri in DWDSs but this approach is susceptible to failure in certain situations which makes effective N. fowleri surveillance necessary to protect the public from infection. Based on our previous efforts to develop a rapid N. fowleri detection method focusing on the lab-cultured and field-collected samples, this manuscript presents our recent progress investigating an operational DWDS field site seasonally colonized by N. fowleri on a monthly basis over the course of a year using untargeted metabolomics of the entire microbial ecology. A panel of significant features exhibiting changes between N. fowleri positive and negative samples were found and further correlated with the findings from the previous studies. The chemical identities for a subset of the common significant features were confirmed. A new, seasonal prediction model was built based on the common significant features and its performance was compared with our previous efforts. This study illustrates the further development of a rapid N. fowleri detection approach from a longitudinal perspective.

Accurate ecological forecasting is increasingly essential for understanding and managing ecosystem responses to climate variability and anthropogenic pressures; however, prediction remains difficult in lakes because key biological variables, such as phytoplankton biomass, exhibit nonlinear dynamics, observational noise, data sparsity, and non-stationarity. This study aimed to test whether large pre-trained time-series foundation models can improve forecasts of lake phytoplankton dynamics, focusing on chlorophyll-a (Chl.a) and water-quality variables in two ecologically contrasting Japanese lakes, the deep-stratified Lake Biwa and the shallow nutrient-rich Lake Kasumigaura. Using extensive monthly monitoring records spanning up to 30 years, we benchmarked two time-series foundation models - Transformer-based Chronos-T5 and probabilistic Lag-Llama - against 15 statistical (AR, ARIMA, SARIMA, Prophet), machine learning (Random Forest, XGBoost, KNN, SVR), and deep learning (LSTM, CNN, TCN, and SSA-hybrid) approaches. Chronos-T5 achieved superior reproduction accuracy and stability across diverse environmental variables and outperformed all other tested models. This performance stemmed from its ability to capture long-term dependencies and complex temporal patterns through large-scale pre-training. We further identified an optimal training window of approximately 14 years, balancing data sufficiency with regime stability, beyond which model accuracy diminished owing to ecological regime shifts. Our findings highlight the transformative potential of time-series foundation models for ecological forecasting, providing a scalable, data-driven framework that can underpin robust early warning systems and adaptive management strategies for aquatic ecosystems worldwide.