Interactive effects of multiple stressors vary with consumer interactions, stressor dynamics and magnitude.

Predicting the impacts of multiple stressors is important for informing ecosystem management but is impeded by a lack of a general framework for predicting whether stressors interact synergistically, additively or antagonistically. Here, we use process-based models to study how interactions generalise across three levels of biological organisation (physiological, population and consumer-resource) for a two-stressor experiment on a seagrass model system. We found that the same underlying processes could result in synergistic, additive or antagonistic interactions, with interaction type depending on initial conditions, experiment duration, stressor dynamics and consumer presence. Our results help explain why meta-analyses of multiple stressor experimental results have struggled to identify predictors of consistently non-additive interactions in the natural environment. Experiments run over extended temporal scales, with treatments across gradients of stressor magnitude, are needed to identify the processes that underpin how stressors interact and provide useful predictions to management.

Spatiotemporal dynamics drive synergism of land use and climatic extreme events in insect meta-populations.

Ecosystems are increasingly threatened by co-occurring stressors associated with anthropogenic global change. Spatial stressor patterns range from local to regional to global, and temporal stressor patterns from discrete to continuous. To date, most multiple stressor studies covered short periods and focused on local effects and interactions. However, it remains largely unknown how stressors with different spatiotemporal patterns interact in their effects over longer periods. In particular, at higher spatial scales, biotic dynamics in ecological networks complicate the understanding of stressor interactions. We used a spatially explicit meta-population model for a generic freshwater insect, parameterized based on traits of the European damselfly Coenagrion mercuriale, to simulate scenarios of discrete climatic extreme events and continuous land use-related stress. Climatic extreme events were modeled as recurring mortality in all patches, whereas land use permanently influenced meta-populations via patch qualities and network connectivity. We found that the risk of discrete climatic extreme events to meta-populations depended strongly on the proportion of land use types, with effects ranging from negligible to extinction. Land use-related stress limited recovery in meta-populations from effects of climatic extreme events, resulting in synergistic stressor interactions. Moreover, the spatial configuration of land use type influenced the combined stressor effects with clustered configurations resulting in lower effects compared to a random configuration. Finally, we found that combined stressor effects can vary with the time point at which they were determined, indicating that inconclusive results in multiple stressor research can partly be due to differences in the time of determination. We conclude that conservation should focus on regional landscape management to mitigate risks on meta-populations from future, intensified extreme climate events. Reducing land use effects, thus improving patch quality and network connectivity, can compensate for effects of additional discrete stressors and, in turn, synergistic interactions.

Invertebrate turnover along gradients of anthropogenic salinisation in rivers of two German regions.

Rising salinity in freshwater ecosystems can affect community composition. Previous studies mainly focused on changes in freshwater communities along gradients of absolute levels of electrical conductivity (EC). However, both geogenic and anthropogenic drivers contribute to the EC level and taxa may regionally be adapted to geogenic EC levels. Therefore, we examined the turnover in freshwater invertebrates along gradients of anthropogenic EC change in two regions of Germany. The anthropogenic change of EC was estimated as the difference between the measured EC and the modeled background EC driven by geochemical and climate variables. Turnover in freshwater invertebrates (ß-diversity) was estimated using the Jaccard index (JI). We found that invertebrate turnover between EC gradient categories is generally greater than 47%, with a maximum of approximately 70% in sites with a more than 0.4 mS cm-1 change compared to the baseline (i.e. no difference between predicted and measured EC). The invertebrates Amphinemura sp., Anomalopterygella chauviniana and Leuctra sp. were reliable indicators of low EC change, whereas Potamopyrgus antipodarum indicated sites with the highest EC change. Variability within categories of EC change was slightly lower than within categories of absolute EC. Elevated nutrient concentrations that are often linked to land use may have contributed to the observed change of the invertebrate richness and can exacerbate effects of EC on communities in water. Overall, our study suggests that the change in EC, quantified as the difference between measured EC and modeled background EC, can be used to examine the response of invertebrate communities to increasing anthropogenic salinity concentrations in rivers. However, due to the strong correlation between EC change and observed EC in our study regions, the response to these two variables was very similar. Further studies in areas where EC change and observed EC are less correlated are required. In addition, such studies should consider the change in specific ions.

Sampling rates for passive samplers exposed to a field-relevant peak of 42 organic pesticides.

Pesticide concentrations in agricultural streams are often characterised by a low level of baseline exposure and episodic peak concentrations associated with heavy rainfall events. Traditional sampling methods such as grab sampling, which are still largely used in governmental monitoring, typically miss peak concentrations. Passive sampling represents a cost-efficient alternative but requires the additional determination of sampling rates to calculate time-weighted average (TWA) water concentrations from the accumulated pesticide mass in the sampler. To date, sampling rates have largely been determined in experiments with constant exposure, which does not necessarily reflect field situations. Using Empore styrene-divinylbenzene (SDB) passive sampler disks mounted in metal holders, we determined sampling rates for 42 organic pesticides, of which 27 sampling rates were lacking before. The SDB disks were in an artificial channel system exposed to a field-relevant pesticide peak. We used an open-source algorithm to estimate coefficients of equations for the accumulated pesticide mass in disks and to determine exposure time-dependent sampling rates. These sampling rates ranged from 0.02 to 0.98 L d-1 and corresponded to those from previous studies determined with constant exposure. The prediction of sampling rates using compound properties was unreliable. Hence, experiments are required to determine reliable sampling rates. We discuss the use of passive sampling to estimate peak concentrations. Overall, our study provides sampling rates and computer code to determine these under peak exposure designs and suggests that passive sampling is suitable to estimate peak pesticide concentrations in field studies.

Preparing GIS data for analysis of stream monitoring data: The R package openSTARS.

Stream monitoring data provides insights into the biological, chemical and physical status of running waters. Additionally, it can be used to identify drivers of chemical or ecological water quality, to inform related management actions, and to forecast future conditions under land use and global change scenarios. Measurements from sites along the same stream may not be statistically independent, and the R package SSN provides a way to describe spatial autocorrelation when modelling relationships between measured variables and potential drivers. However, SSN requires the user to provide the stream network and sampling locations in a certain format. Likewise, other applications require catchment delineation and intersection of different spatial data. We developed the R package openSTARS that provides the functionality to derive stream networks from a digital elevation model, delineate stream catchments and intersect them with land use or other GIS data as potential predictors. Additionally, locations for model predictions can be generated automatically along the stream network. We present an example workflow of all data preparation steps. In a case study using data from water monitoring sites in Southern Germany, the resulting stream network and derived site characteristics matched those constructed using STARS, an ArcGIS custom toolbox. An advantage of openSTARS is that it relies on free and open-source GRASS GIS and R functions, unlike the original STARS toolbox which depends on proprietary ArcGIS. openSTARS also comes without a graphical user interface, to enhance reproducibility and reusability of the workflow, thereby harmonizing and simplifying the data pre-processing prior to statistical modelling. Overall, openSTARS facilitates the use of spatial regression and other applications on stream networks and contributes to reproducible science with applications in hydrology, environmental sciences and ecology.

The Minimum Detectable Difference (MDD) Concept for Establishing Trust in Nonsignificant Results: A Critical Review.

Current regulatory guidelines for pesticide risk assessment recommend that nonsignificant results should be complemented by the minimum detectable difference (MDD), a statistical indicator that is used to decide whether the experiment could have detected biologically relevant effects. We review the statistical theory of the MDD and perform simulations to understand its properties and error rates. Most importantly, we compare the skill of the MDD in distinguishing between true and false negatives (i.e., type II errors) with 2 alternatives: the minimum detectable effect (MDE), an indicator based on a post hoc power analysis common in medical studies; and confidence intervals (CIs). Our results demonstrate that MDD and MDE only differ in that the power of the MDD depends on the sample size. Moreover, although both MDD and MDE have some skill in distinguishing between false negatives and true absence of an effect, they do not perform as well as using CI upper bounds to establish trust in a nonsignificant result. The reason is that, unlike the CI, neither MDD nor MDE consider the estimated effect size in their calculation. We also show that MDD and MDE are no better than CIs in identifying larger effects among the false negatives. We conclude that, although MDDs are useful, CIs are preferable for deciding whether to treat a nonsignificant test result as a true negative, or for determining an upper bound for an unknown true effect. Environ Toxicol Chem 2020;39:2109-2123. © 2020 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Optimisation Model of Dispersal Simulations on a Dendritic Habitat Network.

To predict and mitigate biodiversity loss, a better understanding of species distribution and reliable dispersal models are required. A promising approach in dispersal simulation is the method of spatially explicit graph-based analysis. While graph theory is strongly connected to the field of optimisation in a variety of disciplines, the potential of optimisation has not yet been exploited in dispersal models. We introduce an optimisation model built on a graph-based dispersal simulation of an aquatic invertebrate species with a terrestrial life stage. The model simulates a directed dispersal process and investigates the fastest route to colonise predefined vacant habitat patches. The optimisation model run-time is in general an order of magnitude faster than the underlying simulation and provides the minimum time until the considered habitat patches are colonised under the given landscape structure. These results can then be used to estimate how fast newly formed habitat patches can be reached and colonised. Our model can in principle be adapted to other simulation models and can thus be seen as a pioneer of a new set of models that may support landscape conservation and restoration.

Predicting current and future background ion concentrations in German surface water under climate change.

Salinization of surface waters is a global environmental issue that can pose a regional risk to freshwater organisms, potentially leading to high environmental and economic costs. Global environmental change including climate and land use change can increase the transport of ions into surface waters. We fit both multiple linear regression (LR) and random forest (RF) models on a large spatial dataset to predict Ca2+ (266 sites), Mg2+ (266 sites), and [Formula: see text] (357 sites) ion concentrations as well as electrical conductivity (EC-a proxy for total dissolved solids with 410 sites) in German running water bodies. Predictions in both types of models were driven by the major factors controlling salinity including geologic and soil properties, climate, vegetation and topography. The predictive power of the two types of models was very similar, with RF explaining 71-76% of the spatial variation in ion concentrations and LR explaining 70-75% of the variance. Mean squared errors for predictions were all smaller than 0.06. The factors most strongly associated with stream ion concentrations varied among models but rock chemistry and climate were the most dominant. The RF model was subsequently used to forecast the changes in EC that were likely to occur for the period of 2070 to 2100 in response to just climate change-i.e. no additional effects of other anthropogenic activities. The future forecasting shows approximately 10% and 15% increases in mean EC for representative concentration pathways 2.6 and 8.5 (RCP2.6 and RCP8.5) scenarios, respectively.This article is part of the theme issue 'Salt in freshwaters: causes, ecological consequences and future prospects'.

Assessing recovery of stream insects from pesticides using a two-patch metapopulation model.

Pesticides can exert lethal and sublethal effects on streams organisms. Field studies have shown that non-polluted upstream patches promote population recovery from such effects. Nevertheless, the dynamics and potential carryover effects on the upstream patch are largely unknown. We used a metapopulation model with 2 patches to simulate lethal and sublethal effects on the downstream population of an insect with one generation per year, which was structured into early and late instars aquatic life stages, and an adult terrestrial life stage. We examined the implications for the recovery time of a range of scenarios covering different pesticide effect combinations, migration and exposure types. We found that recovery time responded most strongly to a reduction in reproduction in terms of the reduction of the intrinsic growth rate during the early instar aquatic life stage. For 60 of 96 scenarios with pesticide exposure in consecutive years, no recovery occurred within one year if the intrinsic growth rate was reduced by 50% or more. Without migration between patches (32 scenarios), the polluted downstream population went extinct in 5 of these scenarios. Migration lead overall to slightly faster recovery, albeit this was scenario dependent, but also to a carryover of the pesticide effect from the polluted downstream to the non-polluted upstream patch (up to 25% reduction in the minimal population size). A sensitivity analysis revealed that recovery time was most sensitive to the parameters length of the intrinsic growth phase during early instar aquatic life stage and to migration mortality of the late instar aquatic larvae, and least sensitive to the adult emigration rate and timing of pesticide application. Our study highlights the important role of sublethal effects for population responses to pesticides and that migration buffers against effects, but also carries effects over to non-polluted patches.

Contribution of waste water treatment plants to pesticide toxicity in agriculture catchments.

Pesticide residues are frequently found in water bodies and may threaten freshwater ecosystems and biodiversity. In addition to runoff or leaching from treated agricultural fields, pesticides may enter streams via effluents from wastewater treatment plants (WWTPs). We compared the pesticide toxicity in terms of log maximum Toxic Unit (log mTU) of sampling sites in small agricultural streams of Germany with and without WWTPs in the upstream catchments. We found an approximately half log unit higher pesticide toxicity for sampling sites with WWTPs (p < 0.001). Compared to fungicides and insecticides, herbicides contributed most to the total pesticide toxicity in streams with WWTPs. A few compounds (diuron, terbuthylazin, isoproturon, terbutryn and Metazachlor) dominated the herbicide toxicity. Pesticide toxicity was not correlated with upstream distance to WWTP (Spearman's rank correlation, rho = - 0.11, p > 0.05) suggesting that other context variables are more important to explain WWTP-driven pesticide toxicity. Our results suggest that WWTPs contribute to pesticide toxicity in German streams.

Modeling Macroinvertebrate Community Dynamics in Stream Mesocosms Contaminated with a Pesticide.

Modeling community dynamics of aquatic invertebrates is an important but challenging task, in particular in ecotoxicological risk assessment. Systematic parameter estimation and rigorous assessment of model uncertainty are often lacking in such applications. We applied the mechanistic food web model Streambugs to investigate the temporal development of the macroinvertebrate community in an ecotoxicological mesocosm experiment with pulsed contaminations with the insecticide thiacloprid. We used Bayesian inference to estimate parameters and their uncertainty. Approx. 85% of all experimental observations lie within the 90% uncertainty intervals indicating reasonably good fits of the calibrated model. However, a validation with independent data was not possible due to lacking data. Investigation of vital rates and limiting factors in the model yielded insights into recovery dynamics. Inclusion of the emergence process and sub-lethal effects turned out to be potentially relevant model extensions. Measurements of food source dynamics, individual body size (classes), and additional knowledge on sub-lethal effects would support more accurate modeling. This application of a process-based, ecotoxicological community model with uncertainty assessment by Bayesian inference increased our process understanding of toxicant effects in macroinvertebrate communities and helped identifying potential improvements in model structure and experimental design.

Pesticide runoff from energy crops: A threat to aquatic invertebrates?

The European Union aims to reach a 10% share of biofuels in the transport sector by 2020. The major burden is most likely to fall on already established annual energy crops such as rapeseed and cereals for the production of biodiesel and bioethanol, respectively. Annual energy crops are typically cultivated in intensive agricultural production systems, which require the application of pesticides. Agricultural pesticides can have adverse effects on aquatic invertebrates in adjacent streams. We assessed the relative ecological risk to aquatic invertebrates associated with the chemical pest management from six energy crops (maize, potato, sugar beet, winter barley, winter rapeseed, and winter wheat) as well as from mixed cultivation scenarios. The pesticide exposure related to energy crops and cultivation scenarios was estimated as surface runoff for 253 small stream sites in Central Germany using a GIS-based runoff potential model. The ecological risk for aquatic invertebrates, an important organism group for the functioning of stream ecosystems, was assessed using acute toxicity data (48-h LC50 values) of the crustacean Daphnia magna. We calculated the Ecological Risk from potential Pesticide Runoff (ERPR) for all three main groups of pesticides (herbicides, fungicides, and insecticides). Our findings suggest that the crops potato, sugar beet, and rapeseed pose a higher ecological risk to aquatic invertebrates than maize, barley, and wheat. As maize had by far the lowest ERPR values, from the perspective of pesticide pollution, its cultivation as substrate for the production of the gaseous biofuel biomethane may be preferable compared to the production of, for example, biodiesel from rapeseed.

Modeling global distribution of agricultural insecticides in surface waters.

Agricultural insecticides constitute a major driver of animal biodiversity loss in freshwater ecosystems. However, the global extent of their effects and the spatial extent of exposure remain largely unknown. We applied a spatially explicit model to estimate the potential for agricultural insecticide runoff into streams. Water bodies within 40% of the global land surface were at risk of insecticide runoff. We separated the influence of natural factors and variables under human control determining insecticide runoff. In the northern hemisphere, insecticide runoff presented a latitudinal gradient mainly driven by insecticide application rate; in the southern hemisphere, a combination of daily rainfall intensity, terrain slope, agricultural intensity and insecticide application rate determined the process. The model predicted the upper limit of observed insecticide exposure measured in water bodies (n = 82) in five different countries reasonably well. The study provides a global map of hotspots for insecticide contamination guiding future freshwater management and conservation efforts.

How stressor specific are trait-based ecological indices for ecosystem management?

Using macroinvertebrates as ecological indicators for different stressors has a long tradition. However, when applied to field data, one often observes correlations between different macroinvertebrate indices that can be attributed to both correlations of stressors and inherent correlations due to the sensitivity of taxa to different stressors. Ignoring the source of any given correlation leads to ambiguous conclusions about the impact of different stressors. Here, we demonstrate how to distinguish the causes of correlation by means of Monte Carlo simulations. We assessed to which degree trait-based indices are stressor-specific and whether this depends on the pool of taxa and its taxonomic resolution. Therefore, we (1) analysed the frequencies of "sensitive" and "insensitive" taxa for pairwise combinations of different indices, (2) analysed the inherent correlation of indices with random samples from different taxon pools derived from field samples and from a complete species list of a whole ecoregion, and (3) compared this inherent correlation with the actual correlation of the field samples. We exemplified this approach by analysing two existing indices (SPEARpesticides, Saprobic Index) and new indices for temperature, flow and pH stress. We used these new indices to illustrate our approach while in-depth testing of their applicability was not the focus of our study. We found strong correlations between several indices in our study area at the Swiss Plateau. The probability that this correlation is only due to inherent correlation in the taxa sensitivities was low (maximum of 0.34). The problem of inherent correlation between indices is more severe for the smaller taxon pool with lower taxonomic resolution. Correlation in the sensitivity of different taxa to different stressors leads to an inherent correlation in trait-based indices, which weakens their explanatory power. Our results highlight the importance of correlation analyses when using trait-based indices to guide ecosystem-management, especially in regions with reduced biodiversity.

Landscape parameters driving aquatic pesticide exposure and effects.

Pesticide contamination is considered one of the reasons streams fail to achieve good ecological and chemical status, the main objectives of the Water Framework Directive. However, little is known on the interaction of different pesticide sources and landscape parameters and the resulting impairment of macroinvertebrate communities. We evaluated the potential effects of diffuse and point sources of pesticides using macroinvertebrate monitoring data from 663 sites in central Germany. Additionally, we investigated forested upstream reaches and structural quality as landscape parameters potentially mitigating or amplifying the effects of pesticides. Diffuse pesticide pollution and forested upstream reaches were the most important parameters affecting macroinvertebrate communities (pesticide-specific indicator SPEARpesticides). Our results indicate that forested upstream reaches and riparian buffer strips at least 5 m in width can mitigate the effects and exposure of pesticides. In addition, we developed a screening approach that allows an initial, cost-effective identification of sites of concern.

Competition matters: species interactions prolong the long-term effects of pulsed toxicant stress on populations.

Recent empirical studies have revealed the importance of species competition for the effects of toxicants on populations. In the present study, the authors applied a generic individual-based simulation model of 2 competing species to analyze the consequences of interspecific competition for population dynamics under pulsed contamination. The results indicated that competition that causes a density-dependent decrease in reproduction can substantially prolong the long-term effects of the toxicant. In the example investigated, population recovery time increased from approximately 1 generation time without competition to more than 3 generation times under competition. In particular, species with low reproductive capacity exhibited a strongly prolonged recovery time when interspecific competition was included in the model. The authors conclude that toxicant concentrations derived from risk assessments for pesticides that do not consider competition might be under-protective for populations in real-world systems. The consideration of competition is especially relevant for species with low reproductive capacities to enable a realistic estimation of recovery pace.

Statistics matter: data aggregation improves identification of community-level effects compared to a commonly used multivariate method.

The identification of the effects of toxicants on biological communities is hampered by the complexity and variability of communities. To overcome these challenges, the trait-based SPEAR approach has been developed. This approach is based on (i) identifying the vulnerable taxa using traits and (ii) aggregating these taxa into a group to reduce the between-replicate differences and scattered low-abundance distribution, both of which are typical for biological communities. This approach allows for reduction of the noise and determination of the effects of toxicants at low concentrations in both field and mesocosm studies. However, there is a need to quantitatively investigate its potential for mesocosm data evaluations and application in the ecological risk assessment of toxicants. In the present study, we analysed how the aggregation of the sensitive taxa can facilitate the identification of the effects. We used empirical data from a long-term mesocosm experiment with stream invertebrates and an insecticide as well as a series of simulated datasets characterised by different degrees of data matrix saturation (corresponding to different sampling efforts), numbers of replicates, and between-replicate differences. The analyses of both the empirical and simulated data sets revealed that the taxa aggregation approach allows for the detection of effects at a lower saturation of the data matrices, smaller number of replicates, and higher between-replicate differences when compared to the multivariate statistical method redundancy analysis. These improvements lead to a higher sensitivity of the analysed systems, as long-term effects were detected at lower concentrations (up to 1,000 times). These outcomes suggest that methods based on taxa aggregation have a strong potential for use in mesocosm data evaluations because mesocosm studies are usually poorly replicated, have high between-replicate variability, and cannot be exhaustively sampled due to technical and financial constraints.

Culmination of low-dose pesticide effects.

Pesticides applied in agriculture can affect the structure and function of nontarget populations at lower doses and for longer timespans than predicted by the current risk assessment frameworks. We identified a mechanism for this observation. The populations of an aquatic invertebrate (Culex pipiens) exposed over several generations to repeated pulses of low concentrations of the neonicotinoid insecticide (thiacloprid) continuously declined and did not recover in the presence of a less sensitive competing species (Daphnia magna). By contrast, in the absence of a competitor, insecticide effects on the more sensitive species were only observed at concentrations 1 order of magnitude higher, and the species recovered more rapidly after a contamination event. The underlying processes are experimentally identified and reconstructed using a simulation model. We conclude that repeated toxicant pulse of populations that are challenged with interspecific competition may result in a multigenerational culmination of low-dose effects.

Effects of organic pollutants from wastewater treatment plants on aquatic invertebrate communities.

Pesticides are a major stressor for stream ecosystem health. They enter surface waters from diffuse agricultural sources but also from point sources such as municipal wastewater treatment plants (WWTPs). However, to date, no studies have focused on the ecological effects of pesticide-contaminated WWTP effluent on macroinvertebrate communities. On the basis of governmental monitoring data of 328 sites in Hesse, Germany, we identified insecticidal long-term effects on the structure of the macroinvertebrate community up to 3 km downstream of WWTPs. The effects were quantified using the trait-based SPEAR(pesticides) index, which has been shown to be an effective tool for identifying community effects of pesticide contamination. In addition, based on the German Saprobic Index, we revealed that WWTPs are still an important source of oxygen-depleting organic pollution, despite the extensive technological improvements in wastewater management over several centuries. In general, our findings emphasize the need to take municipal WWTPs into consideration in the management of river basins under the EU Water Framework Directive to achieve good ecological and chemical status for European streams and rivers.

Climate change, agricultural insecticide exposure, and risk for freshwater communities.

Climate change exerts direct effects on ecosystems but has additional indirect effects due to changes in agricultural practice. These include the increased use of pesticides, changes in the areas that are cultivated, and changes in the crops cultivated. It is well known that pesticides, and in particular insecticides, affect aquatic ecosystems adversely. To implement effective mitigation measures it is necessary to identify areas that are affected currently and those that will be affected in the future. As a consequence, we predicted potential exposure to insecticide (insecticide runoff potential, RP) under current conditions (1990) and under a model scenario of future climate and land use (2090) using a spatially explicit model on a continental scale, with a focus on Europe. Space-for-time substitution was used to predict future levels of insecticide application, intensity of agricultural land use, and cultivated crops. To assess the indirect effects of climate change, evaluation of the risk of insecticide exposure was based on a trait-based, climate-insensitive indicator system (SPEAR, SPEcies At Risk). To this end, RP and landscape characteristics that are relevant for the recovery of affected populations were combined to estimate the ecological risk (ER) of insecticides for freshwater communities. We predicted a strong increase in the application of, and aquatic exposure to, insecticides under the future scenario, especially in central and northern Europe. This, in turn, will result in a severe increase in ER in these regions. Hence, the proportion of stream sites adjacent to arable land that do not meet the requirements for good ecological status as defined by the EU Water Framework Directive will increase (from 33% to 39% for the EU-25 countries), in particular in the Scandinavian and Baltic countries (from 6% to 19%). Such spatially explicit mapping of risk enables the planning of adaptation and mitigation strategies including vegetated buffer strips and nonagricultural recolonization zones along streams.