Supervised machine learning improves general applicability of eDNA metabarcoding for reservoir health monitoring

Effective and standardized monitoring methodologies are vital for successful reservoir restoration and management. Environmental DNA (eDNA) metabarcoding sequencing offers a promising alternative for biomonitoring and can overcome many limitations of traditional morphological bioassessment. Recent attempts have even shown that supervised machine learning (SML) can directly infer biotic indices (BI) from eDNA metabarcoding data, bypassing the cumbersome calculation process of BI regardless of the taxonomic assignment of eDNA sequences. However, questions surrounding the general applicability of this taxonomy-free approach to monitoring reservoir health remain unclear, including model stability, feature selection, algorithm choice, and multi-season biomonitoring. Here, we firstly developed a novel biological integrity index (Me-IBI) that integrates multitrophic interactions and environmental information, based on taxonomy-assigned eDNA metabarcoding data. The Me-IBI can better distinguish the actual health status of the Three Gorges Reservoir (TGR) than physicochemical assessments and have a clear response to human activity. Then, taking this reliable Me-IBI as a supervised label, we compared the impact of selecting different numbers of features and SML algorithms on the stability and predictive performance of the model for predicting ecological conditions in multiple seasons using taxonomy-free eDNA metabarcoding data. We discovered that even with a small number of features, different SML algorithms can establish a stable model and obtain excellent predictive performance. Finally, we proposed a four-step strategy for standardized routine biomonitoring using SML tools. Our study firstly explores the general applicability problem of the taxonomy-free eDNA-SML approach and establishes a solid foundation for the large-scale and standardized biomonitoring application.

Factors affecting the transferability of bioindicators based on stream fish assemblages

The development of multimetric indices (MMIs) to measure the biotic condition of aquatic habitats is based on metrics derived from biological assemblages. Considering fish assemblages, the inconsistencies in metrics responses outside of the places where they were developed limit MMI transferability and applicability to other locations, requiring local calibration. The factors behind the low transferability of these MMIs are still poorly understood. We investigated how environmental dissimilarity and spatial distance influence the transferability of metrics generated from local stream fish assemblages to other regions. We also tested whether functional and taxonomic metrics respond differently to the spatial distance. We used data from 239 fish assemblages from streams distributed across a Brazilian, the upper Parana basin and characterized each site according to the level of anthropogenic disturbance at the landscape scale using an Anthropogenic Pressure Index (API). We divided the upper Parana basin into sub-basins and used two of them to create template response models of the metrics in relation to the API. We used these response models to predict the responses outside the template sub-basins. Our response variable representing a metric of transferability was the absolute difference between metric's predicted and observed value for each site (prediction error). We thus modeled the prediction error in relation to the predictor variables that were i) the environmental dissimilarity between each site with the average of the sites from template sub-basins (climatic, topographic and soil type variables) and ii) the spatial distance (overland and watercourse distance) between each site and the center of the template sub-basin. We found that errors in metric predictions were associated with both environmental dissimilarity and spatial distance. Furthermore, functional and taxonomic metrics responded equally to spatial distance. These results indicate the need for local calibration of metrics when developing MMIs, especially if the protocols already available come from distant and environmentally dissimilar places.

Eutrophication dangers the ecological status of coastal wetlands: A quantitative assessment by composite microbial index of biotic integrity

The intertidal wetland ecosystem is vulnerable to environmental and anthropogenic stressors. Understanding how the ecological statuses of intertidal wetlands respond to influencing factors is crucial for the management and protection of intertidal wetland ecosystems. In this study, the community characteristics of bacteria, archaea and microeukaryote from Jiangsu coast areas (JCA), the longest muddy intertidal wetlands in the world, were detected to develop a composite microbial index of biotic integrity (CM-IBI) and to explore the influence mechanisms of stresses on the intertidal wetland ecological status. A total of 12 bacterial, archaea and microeukaryotic metrics were determined by range, responsiveness and redundancy tests for the development of ba-IBI, ar-IBI and eu-IBI. The CM-IBI was further developed via three sub-IBIs with weight coefficients 0.40, 0.33 and 0.27, respectively. The CM-IBI (R² = 0.58) exhibited the highest goodness of fit with the CEI, followed by ba-IBI (R² = 0.36), ar-IBI (R² = 0.25) and eu-IBI (R² = 0.21). Redundancy and random forest analyses revealed inorganic nitrogen (inorgN), total phosphorus (TP) and total organic carbon (TOC) to be key environmental variables influencing community compositions. A conditional reasoning tree model indicated the close associating between the ecological status and eutrophication conditions. The majority of sites with water inorgN<0.67 mg/L exhibited good statuses, while the poor ecological status was observed for inorgN>0.67 mg/L and TP > 0.11 mg/L. Microbial networks demonstrated the interactions of microbial taxonomic units among three kingdoms decreases with the ecological degradation, suggesting a reduced reliability and stability of microbial communities. Multi-level path analysis revealed fishery aquaculture and industrial development as the dominant anthropogenic activities effecting the eutrophication and ecological degradation of the JCA tidal wetlands. This study developed an efficient ecological assessment method of tidal wetlands based on microbial communities, and determined the influence of human activities and eutrophication on ecological status, providing guidance for management standards and coastal development.

Predicting bend-induced heterogeneity in sediment microbial communities by integrating bacteria-based index of biotic integrity and supervised learning algorithms

Prioritizing the relationship between heterogeneity of sediment habitats and river bends is critical when planning and reconstructing urban rivers. However, the exact relationship between ecological heterogeneity and river bends remains ambiguous. Therefore, this research proposed a new approach to quantify and predict bend-induced ecological heterogeneity, incorporating the bacteria-based index of biotic integrity (Ba-IBI), path model, and random forest regression model. The developed Ba-IBI quantified heterogeneity in sediment microbial communities, ranging from low (1.40) to high (3.97). A path model was developed and validated in order to further investigate the relative contributions of environmental factors to the Ba-IBI. The established path model, which was considered acceptable with a CMIN/df = 1.949 < 4, suggested that primary environmental factors affecting the sediment bacterial communities were flow velocity and ammonium concentration in sediment. To further characterize the relationship between environmental factors and the Ba-IBI, a function was constructed using the random forest regression model that predicts the responses of sediment bacterial communities to environmental factors with R² = 0.6126. The proposed approach and prediction tools will provide knowledge to improve natural channel design and post-project evaluations in river restoration projects.

The use of multiscale stressors with biological condition assessments: A framework to advance the assessment and management of streams

Incorporating information on landscape condition (or integrity) across multiple spatial scales and over large spatial extents in biological assessments may allow for a more integrated measure of stream biological condition and better management of streams. However, these systems are often assessed and managed at an individual scale (e.g., a single watershed) without a larger regional multiscale context. In this paper, our goals were: (1) To develop a conceptual framework that could combine stream biological condition to abiotic landscape integrity (or, conversely, stressor) data at three spatial scales: watershed, catchment and stream-reach scale, to enable more targeted management actions. Measures of landscape integrity and stressors are negatively related, i.e., integrity on a 0-1 scale is equal or equivalent to stressors on a 1-0 scale. (2) To develop the framework in such a way that allows operational flexibility, whereby different indicators can be used to represent biological condition, and landscape integrity (or stressors) at various scales. (3) To provide different examples of the framework's use to demonstrate the flexibility of its application and relevance to management. Examples include stream biological assessments from different regions and states across the U.S. for fish, macroinvertebrates and diatoms using a variety of assessment tools (e.g., the Biological Condition Gradient (BCG), and an Index of Biotic Integrity (IBI)). Landscape integrity indicators comprise U.S. EPA's nationally available Index of Watershed Integrity (IWI) and Index of Catchment Integrity (ICI), and state and regional derived watershed and stream-reach scale integrity indicators. Scatterplots and a landscape integrity map were used to relate samples of stream condition classes (e.g., good, fair, poor) to watershed, catchment and stream-reach scale integrity. This framework and approach could provide a powerful tool for prioritizing, targeting, and communicating management actions to protect and restore stream habitats, and for informing the spatial extent at which management is applied.

Associations between Biotic Integrity and Sport Fish Populations in Upper Midwest, USA Rivers, with Emphasis on Smallmouth Bass

Indices of biotic integrity (IBIs) are used to assess ecosystem health of streams and rivers. Streams and rivers with high IBI scores should support abundant and healthy populations of recreationally important sport fishes. However, the fundamental assumption that IBI scores and sport fish populations are associated needs to be examined. To verify this assumption, we tested associations between IBI scores and relative abundance of all sport fishes targeted by anglers, with emphasis on relative abundance of four size groups of smallmouth bass Micropterus dolomieu at 54 stream and river reaches in 2012 and 2013. We also tested for associations with smallmouth bass body condition and growth. A total of 13,708 fishes representing 85 species were captured including 11 sport fish species that included 571 smallmouth bass. We found that the maximum potential relative abundance of sport fishes and smallmouth bass size classes, as well as body condition of bass between 180 and 279 mm, could be predicted by IBI scores. We did not observe significant relationships with body condition of other bass size classes or with growth. Whereas abundance patterns were variable at reaches with higher IBI scores, abundance of larger, quality-sized sport fishes were more limited at reaches with IBI scores <30 that were classified as having poor biotic integrity. Maximum potential body condition was predicted to exceed 95, a condition value indicative of healthy fish, at IBI scores exceeding 50, reflective of reaches being classified as either fair, good, or excellent. These results confirm that management activities that enhance or maintain biotic integrity also support high-quality habitat for sport fish. While our findings support using IBIs as an indicator of the fishable goal specified in the United States Clean Water Act, managers should recognize that other factors not necessarily represented by the index can also limit sport fish populations.

Reduction of bacterial integrity associated with dam construction: A quantitative assessment using an index of biotic integrity improved by stability analysis

Rivers are extensively regulated by damming, yet the effects of such interruption on bacterial communities have not been assessed quantitatively. To fill this gap, we proposed a bacteria-based index of biotic integrity (Ba-IBI) by using bacterial community dataset collected from the Three Gorges Reservoir and its upper reaches. Stability analysis based on bacterial resistance (RS) and resilience (RL) to external disturbance was conducted to improve the performance of the index. Four core metrics, i.e. the ratio of Bacilli, Bacteroidetes and Clostridia to Alphaproteobacteria (BBC/A), Oxalobacteraceae, Methanotrophs and Thermophiles were selected after range, responsive and redundancy tests. The improved Ba-IBI, ranging from 1.04 to 4.10, was better at distinguishing sites with or without direct dam effects compared with the unimproved one. The index values maintained high in the riverine sites while reducing in the reservoir, demonstrating the negative influence of dam construction on bacterial integrity. Based on the assessment results, 23.1%, 46.2% and 30.8% sampling sites were large, moderately and little affected by damming, respectively. A Random Forest (RF) regression model was trained and tested, offering a valid prediction of the input Ba-IBI and environmental parameters. Sensitivity analysis revealed the significant contributions of flow velocity towards the predicting process performance, indicating the importance of hydrodynamic conditions on determining the spatial variability of bacterial communities. This study provides not only a first quantitative insight for assessing bacterial response to damming, but also a guideline for applying the improved index in the dam regulation and ecological protection.

Development of a bacteria-based index of biotic integrity (Ba-IBI) for assessing ecological health of the Three Gorges Reservoir in different operation periods

It is urgently needed to quantitatively assess ecological health of the Three Gorges Reservoir (TGR) when considering its special environmental conditions and temporal variations caused by reservoir operation. This study developed a bacteria-based index of biotic integrity (Ba-IBI) based on sediment samples collected along the TGR in low water level period, impoundment period and sluicing period, respectively. Reference conditions were defined using 8 ecological variables describing the hydromorphology and anthropogenic disturbances around the sites. Five core metrics, including % Acidobacteria, % Gemmatimonadetes, % Geobacter, Methanotroph and Phototroph, were selected after the screening processes. The developed index could clearly discriminate reference and impaired conditions and exhibited significant relationship with environmental parameters according to the redundancy (p < 0.01) and multivariable linear regression analysis (R² = 0.76). By implementing Ba-IBI in the TGR, the ecological health of the sampling sites was defined as "Excellent" (25%), "Good" (50%) and "Fair" (25%) separately. The spatial variation of biotic integrity was closely associated with environmental and ecological changes, especially the increase of nutrient concentrations. This study revealed a significant tendency that the ecological health in the low water level and sluicing periods was better than that in the impoundment period, which could be attributed to the hydrodynamic changes due to water level fluctuation. This study provides a comprehensive understanding of ecological health of the TGR in different operation periods and the index offers a guideline for the reservoir regulation in the similar areas.

Development of a diatom-based multimetric index for acid mine drainage impacted depressional wetlands

Acid mine drainage (AMD) from coal mining in the Mpumalanga Highveld region of South Africa has caused severe chemical and biological degradation of aquatic habitats, specifically depressional wetlands, as mines use these wetlands for storage of AMD. Diatom-based multimetric indices (MMIs) to assess wetland condition have mostly been developed to assess agricultural and urban land use impacts. No diatom MMI of wetland condition has been developed to assess AMD impacts related to mining activities. Previous approaches to diatom-based MMI development in wetlands have not accounted for natural variability. Natural variability among depressional wetlands may influence the accuracy of MMIs. Epiphytic diatom MMIs sensitive to AMD were developed for a range of depressional wetland types to account for natural variation in biological metrics. For this, we classified wetland types based on diatom typologies. A range of 4-15 final metrics were selected from a pool of ~140 candidate metrics to develop the MMIs based on their: (1) broad range, (2) high separation power and (3) low correlation among metrics. Final metrics were selected from three categories: similarity to reference sites, functional groups, and taxonomic composition, which represent different aspects of diatom assemblage structure and function. MMI performances were evaluated according to their precision in distinguishing reference sites, responsiveness to discriminate reference and disturbed sites, sensitivity to human disturbances and relevancy to AMD-related stressors. Each MMI showed excellent discriminatory power, whether or not it accounted for natural variation. However, accounting for variation by grouping sites based on diatom typologies improved overall performance of MMIs. Our study highlights the usefulness of diatom-based metrics and provides a model for the biological assessment of depressional wetland condition in South Africa and elsewhere.

Relationship of fish indices with sampling effort and land use change in a large Mediterranean river

Fish are invaluable ecological indicators in freshwater ecosystems but have been less used for ecological assessments in large Mediterranean rivers. We evaluated the effects of sampling effort (transect length) on fish metrics, such as species richness and two fish indices (the new European Fish Index EFI+ and a regional index, IBICAT2b), in the mainstem of a large Mediterranean river. For this purpose, we sampled by boat electrofishing five sites each with 10 consecutive transects corresponding to a total length of 20 times the river width (European standard required by the Water Framework Directive) and we also analysed the effect of sampling area on previous surveys. Species accumulation curves and richness extrapolation estimates in general suggested that species richness was reasonably estimated with transect lengths of 10 times the river width or less. The EFI+ index was significantly affected by sampling area, both for our samplings and previous data. Surprisingly, EFI+ values in general decreased with increasing sampling area, despite the higher observed richness, likely because the expected values of metrics were higher. By contrast, the regional fish index was not dependent on sampling area, likely because it does not use a predictive model. Both fish indices, but particularly the EFI+, decreased with less forest cover percentage, even within the smaller disturbance gradient in the river type studied (mainstem of a large Mediterranean river, where environmental pressures are more general). Although the two fish-based indices are very different in terms of their development, methodology, and metrics used, they were significantly correlated and provided a similar assessment of ecological status. Our results reinforce the importance of standardization of sampling methods for bioassessment and suggest that predictive models that use sampling area as a predictor might be more affected by differences in sampling effort than simpler biotic indices.

Development and validation of a bacteria-based index of biotic integrity for assessing the ecological status of urban rivers: A case study of Qinhuai River basin in Nanjing, China

With the increasing human disturbance to urban rivers, the extinction and biodiversity losses of some macroorganism species decreased the accuracy of bioassessment. In this study, a novel index of biotic integrity based on bacteria (Ba-IBI) was first developed for Qinhuai River in Nanjing city, China. Thirty-two biofilm samples were collected along the river bank and the bacterial communities were identified by high-throughput sequencing. By the range, responsive, and redundancy tests, four core metrics were selected from the dataset of 78 candidate metrics, including Pielou's evenness index, proportion of Paenibacillus, proportion of OTUs tolerant to organic pollution and proportion of Nitrosomonas. The results showed that the Ba-IBI was able to effectively discriminate different impaired site groups, and had a good correlation with the index of water quality (r = 0.79, p < 0.01) and the qualitative habitat evaluation index (r = 0.51, p < 0.01). Moreover, the Ba-IBI was negatively correlated with the number of population within a 1 km buffer (r = -0.71, p < 0.01). Application of the index showed that most of the sites were in the poor or bad class in the river. Our study revealed that the Ba-IBI is an effective and reliable approach for assessing the ecological status of Qinhuai River basin, which can complement the existing ecological assessment approaches for urban rivers. Meanwhile, repeted surveys and field validations are still needed to further improve the applicability of the index in future studies.