Stream water is a part of the hydrosphere that includes many interactions between natural and anthropogenic effects, namely rock weathering and water contamination, which sometimes overlap or even interfere with each other [1
]. With social and demographic expansion, many threats to water quality have arisen and reached worrying levels around the globe. Many studies have made inventories of pollution causes and have quantified the roles of pollutants at a catchment scale [8
]. Effluent discharge is the most unsettling point source pressure that leads water to degrade [12
]. Discharge of urban effluents is known to contain higher organic loads, comparatively to industrial effluents [13
]. The composition of industrial discharge is highly variable based on the type of industry and degree of treatment. Typically, high loads of heavy metals are released from industrial activities such as mining [14
] or metallurgic activities [15
]. However, even in industrial effluents, it is possible to find high concentrations of organic compounds, such as from industries related to food production [16
]. Livestock production [18
] and agriculture [19
] exert diffuse threats on water quality, which are mostly associated with the transport of pesticides and nutrients through infiltration into groundwater and runoff into rivers. The consequences of these pressures can be moderated by the environment during the course of processes such as phytoremediation [20
] or natural aeration induced by turbulence [21
], which promotes the self-depuration capacity of rivers [22
]. Riparian vegetation has high potential to improve and sustain water quality because it comprises a vegetation barrier against the propagation of contaminants [23
] and promotes a natural environment for aquatic lifeforms, thereby improving biodiversity [24
]. The benefits of riparian vegetation can be accessed through landscape metrics [25
], such as connectance, density, and contrast (e.g., [26
]). For other types of land use patches, these metrics reveal an intrinsic relationship with water quality [28
]. For this reason, several authors have used landscape metrics to explain the degradation of water quality by establishing relationships with different types of contaminants [30
]. When such interactions are studied, it must be accessed a suitable spatial extent for statistical sampling [32
]. In general, the scale can be comprised of buffers, riparian extents, and catchments [33
]. By comparing studies that accessed different types of scales, is found some inconsistency in the results, as some authors argue that the use of whole watersheds is more feasible [32
] while others report that the riparian scale is the more appropriate [25
]. These conflicts in opinion are due to variations in study design, study areas, topographic aspects [37
] and, in particular, the season [38
] at which measurements were made.
Another relevant issue in water quality assessments is the relationship between land use and land capability (natural use). When actual land uses differ from land capability, an environmental land use conflict develops [39
]. This analysis settles on the principle that agricultural areas and livestock production can place heavy demands on soils, while forested areas conserve and protect the soil [39
], looking forward to a sustainable land use policy [41
]. Many authors have studied the consequences of land use conflicts, such as the increase in flood vulnerability [42
], decrease in soil fertility [40
], and the increase of soil erosion [43
]. Land use conflicts can be a significant threat to water quality and ecological integrity because they tend to amplify harmful effects on groundwater [45
], surface water [47
], and the biodiversity of riverine ecosystems [48
] in rural catchments dominated by agriculture and pasture.
One of the most common methods to reveal the effects of pollution is by measuring contaminant concentrations in surface waters. By comparing measurements from previous periods, or even, with legislative limits, it is possible to determine if these are looming concentrations. The point is that damage is directly felt in lifeforms living in surface waters and can occur even when concentrations are within legal limits [49
]. For such reasons, ecological integrity should be encompassed in environmental studies, which is the ecosystems ability to persist when natural or anthropogenic changes occur [51
]. When pressures on freshwater resources are excessive, damage to ecological integrity can be accessed through bio-indicators such as IBMWP [53
] or MELI [55
To fully understand the interactions between pollution sources, water contamination, and ecological integrity, it is worthwhile to use statistical models such as eigenvector based models [56
] or structural equation models (SEM). The most common types of SEM are CB-SEM (covariance-based SEM) and SEM-PLS (SEM-Partial Least Squares), also termed PLS-PM (PLS-Path Modeling) [58
]. In the first case, the estimation of model parameters is based on maximum likelihood, while for PLS-PM, the estimation procedure resorts to ordinary least squares regressions [59
]. SEM models have been widely used in the social sciences, but nowadays, the method applies to other areas of scientific knowledge. One reason for this expansion is due to the current need to address complex interactions among a large number of parameters. In previous applications of SEM to water quality assessments, some interactions between surface water parameters have been exposed, but the specific roles of anthropogenic pressures and natural processes have not been quantifies [60
]. In a study addressing groundwater quality, SEM successfully explained the chemical composition of phreatic aquifer groundwater samples and the origin of water mineralization [61
]. In the Feitsui Reservoir Watershed, SEM was used as a mathematical tool for the evaluation of harmful effects resulting from three types of pollution: organic, sediment, and eutrophication [62
]. From a social standpoint, SEM was used to understand citizen’s awareness about water quality [62
A recent application of SEM explored interactions between anthropogenic pressures, water contamination, and ecological integrity in two very different river basins, one urbanized and densely populated and another rural and sparsely populated [64
]. Following that work, the present study aimed to discover seasonal and scale effects potentially affecting the aforementioned interactions using PLS-PM, because those effects can be substantial and have not yet been investigated. This study presents itself as an innovative analysis since there are few environmental studies using structural equation models. Another breakthrough aspect is that the effects of landscape metrics are studied in conjunction with contaminant emissions. These two types of variables were included in the same model, accessing the interplay with ecological integrity, along with impact changes across different scales and seasons.
Taken together, the models seem to better explain the relationship between anthropogenic pressures and water quality than the influences of these latent variables on ecological integrity. Additionally, in winter, the long-scale model seems to behave better than the short-scale model, and in summer, the performance levels of the two models reverses.
Usually, path coefficients linking “Pressures” to “SWP” should be positive, because it is expected that increasing anthropogenic pressures will increase stream water contamination. In turn, path coefficients linking “Pressures” to “Ecological integrity” or “SWP” to “Ecological integrity” should be negative, because increasing human pressure and stream water contamination are expected to cause ecological integrity losses. In the present study, the signs of path coefficients are as expected with a few exceptions. In model M1, the exceptions are the links between “Diffuse Emissions” and “Ecological Integrity (S 16)” (0.158) and between “Land Use” and “Ecological Integrity (W 17)” (0.863). In model M2, the unexpected signs refer to links between “Diffuse Emissions” and “SWP (S 16)” (−0.425) and “Ecological Integrity (W 17)” (0.009), and between “Point Source Emissions” and “SWP (W 17)” (−0.167), “Ecological Integrity (W 17)” (0.163), and “Ecological Integrity (S 16)” (0.107). The unexpected signs of path coefficients may be a consequence of indirect effects between latent variables.
The surface water parameters that seem to have concurrent effects on ecological integrity in the M1 and M2 spatial scale models, while having opposite effects in the seasonal sub-models, are arsenic, manganese, and potassium.
Arsenic appears to be negative for ecological integrity during the winter period. This effect might be related to agricultural practice in the winter [91
]. The use of phosphate fertilizers in agriculture is known to lead to the accumulation of arsenic in soils, water, and crops [92
]. These agrochemicals are toxic, and various recent papers have emphasized this toxicity issue [95
]. However, its use in the Sabor River Basin seems to have continued [68
]. The arsenic concentrations are higher in summer than in winter. Therefore, the harmful effects of arsenic are related to factors other than the concentration that influence the bioaccumulation of this substance [96
Potassium is an essential component of fertilizers used in agriculture, so its adverse effect on the latent variable “Ecological Integrity” during the summer period might be attributed to this activity [97
]. In the Sabor River Basin, the production of olives and grapes occurs on a regular basis. The use of potassium is essential for the growth of these goods [98
]. In the region, fertilization of oliveyards and vineyards occurs in the spring season when rainfall, runoff, and sub-surface flow may be abundant. The leach of potassium in runoff and shallow soil and saprolite waters can justify the high concentrations measured in the forthcoming season (summer).
The effect of manganese on “Ecological Integrity” is adverse during the summer period. The bedrock in the Sabor River Basin comprises manganese bearing rocks such as serpentinites [3
]. In summer, the contribution of groundwater to streamflow is high because rainfall is lacking [100
]. The discharge of groundwater naturally enriched in manganese can then accumulate in the macroinvertebrates reducing the IPtIN
values. The response of IPtIN
to changes in pH, dissolved oxygen, and nitrate concentrations is not consistent across scales or seasons. The results for these parameters are, therefore, difficult to explain.
Some diffuse emissions revealed concordant effects on water quality at the seasonal level but opposite effects at the spatial scale level. Agricultural areas and water body connectivity were shown to contribute to water quality decline in the summer and winter periods, but only on the short spatial scale. On the long scale, the effect seemed to be neutral (agricultural areas) or even positive (water body connectivity), especially in winter. Adverse effects of agriculture patches have been reported elsewhere [38
]. However, to the best of our knowledge, there have been no previous reports on negative effects of water body connectivity on water contamination. Eventually, disconnected water bodies are not able to dilute contaminants brought from upstream areas. This could be a reliable setting on a short spatial scale. Besides, water body connectivity increases the water quality on the long scale, a result that corroborates and completes this reasoning.
The landscape metrics “npc” (number of patches with a land use conflict of class 3) and “ci cp” (connectivity of pasture areas) were associated with adverse effects on water quality on the long spatial scale. In both cases, these areas are scattered within the Sabor River Basin and are characterized by small patch areas. For that reason, only the long scale analysis was able to trace them.
The metric “edge density” revealed a general negative effect on the ecological integrity in both seasons and spatial scales. The complexity of a landscape is intrinsically connected to the “edge density” metric. Higher values indicate a wider variety of land uses and irregular patches [101
]. According to some authors, the increase in edge density is a consequence of human intervention in the landscape [31
], which transforms natural habitats into agricultural and urban areas and influences the transport of nutrients and contaminants from soil to water [102
]. In the Sabor River Basin, the “edge density” metric seems to be the prominent threat to ecological integrity, because it is independent of season and scale. This study revealed that the use of two PLS-PM models was enough to compare the seasonal and spatial scale effects of anthropogenic pressures on water quality and ecological integrity. Frequently, the assessment of water contamination and ecological integrity involving landscape metrics uses a very large number of these indices [30
]. The present study was based on five landscape metrics, because multicollinearity is a crucial issue in PLS-PM and many landscape metrics are collinear. In spite of using just a few landscape metrics, this study had the advantage of combining these important variables with other parameters relevant for water quality, such as point source and diffuse emissions. This combination is rarely seen in other studies focused on relationships between anthropogenic pressures, water quality, and ecological integrity.
One weak point of this study is the small sample size (just 11 entries). We note that PLS-SEM can handle efficiently small samples with a large number of variables, which is preferable over other structural equation (SEM) models (e.g., CB-SEM) [59
]. A larger number of samples would increase parameter estimates (e.g., R2
), which are already satisfactory or even good in most cases. Further studies on this or other river basins should, therefore, be based on a larger number of sampling sites.
4.1. Scope of Study and Environmental Management Guidelines
As the measurements of IPtIN
revealed, the ecological status of Sabor River Basin is under acceptable conditions. Compared to other river basins under poor ecological status, such as the Ave River Basin, it is eminent to mitigate pollution effects looking forward to achieving a better ecological status [103
]. Anyhow, it is important to preserve such conditions in Sabor River Basin at points where the ecological status is under “good” status and improve some sampling sites where an ecological status is below the “good” classification. For such reasons, the present work fits within the scope of 2030 Agenda Sustainable Development Goals, announced by the United Nations, by adopting goal 15, called “Life On Land”. From an overall perspective, this goal attracts worldwide attention to preserving natural resources and promote natural biodiversity in all ecosystems. In addition, it is given special attention to terrestrial ecosystems, including inland freshwater, for the purpose not only of restoration but also of preservation.
With regard to river basin management, this study contributes to such cause, as the pressures that degrade ecological integrity are identified, as well as the most vulnerable season. It was also identified the scale at which pressures should be addressed, advising that, for other studies of the same scope, the scales where there is a negative effect of the pressures, should be applied. In this context, the results obtained from Figure 5
were compiled into Table 3
, where are shown the scales and seasons to be given particular focus by decision-makers are identified.
The identified pressures mainly have a negative effect throughout the seasons. Nevertheless, it should be noted that nitrogen discharges from agricultural areas should be given particular attention during winter. This is probably due to fertilization periods, analyzed on a short scale, the effect is negative. Furthermore, this indicates that when quality measurements were taken nearby, the Directorate-General for Agriculture and Rural Development (DGADR) [104
], in cooperation with the Portuguese Environmental Agency (APA) [73
], should supervise the fertilization period to minimize the effects of excessive fertilizer application [105
Nitrogen discharges from livestock farming should be monitored during winter and summer to ensure that contamination does not spread through surface runoff or groundwater contamination. Regarding this environmental problem, several preventive measures must be taken, aiming at sustainable livestock production [106
]. When livestock production farms are installed, the site selection is a crucial aspect, in order to allocate the farms distant from wells, in areas with low soil permeability and a low slope to control the runoffs of agricultural fertilizers [107
]. The site selection is a crucial aspect, in order to allocate the farms distant from wells, in areas with low soil permeability and a low slope to control the runoffs of agricultural fertilizers [108
The Sabor River Basin contains mostly urban effluent discharge points. In this study, the analysis was done only on a large scale, as there were no discharge points on the short scale. The effect of effluent discharges into surface waters has been found to degrade quality during the summer period, whereas discharges to the soil only have a negative impact during the winter period. This information is useful for monitoring and management purposes. By knowing which is the season that effluent discharges have a negative impact on water quality, environmental entities should enforce inspections, looking forward to overseeing if discharge values are under legal thresholds, even if the treatment station is operating properly.
The effects of land use metrics in ecological integrity shown clearly dependence on scale and season. As other authors have shown is seen that when a scale is accessed, a positive effect can be detected while, on another scale, a negative effect can be detected [32
]. Agriculture (areas) and Water bodies (ci pp) only have a negative impact on a short scale, conflict class 3 (npc) and Pastures (ci cp) have a negative impact on a long scale, and only Edge Density revealed a negative impact in all scales and seasons. Such results show the effort of geospatial data effects, which should be applied at various scales. In terms of river basin management, the landscape variables revealed a stronger effect on ecological integrity when compared to diffuse and point source emissions. Conflict class 3 (npc) revealed the strongest effect in the long-scale (−0.378 in S 16 and −0.880 in W 17) and Water Bodies (ci pp) in the short-scale (−1.150 in S 16 and −0.236 in W 17). In the (PGRH) Portuguese Management Plan of Hydrographic Regions are identified many threats to the ecological integrity of Sabor River Basin, namely the diffuse discharges of livestock and urban effluents. By default, the plan does not include intervention measures based on land use configuration. As shown in this study, land use configuration is a key aspect that overlaps the effect of effluent discharges. In this context, it is crucial that such studies where diffuse and point discharges are analyzed should add the metric effects of land use to access a wide range of pressures and phenomena that affect the ecological integrity. Even more, it is understood that in the PGRH should be implemented territorial planning as a measure to improve water quality at a national level.
Land use changes in the Sabor River Basin would be a step of ecological quality assurance. On a large scale, land use transitions can be expensive and take long periods to implement [110
], but strategic reordering can be a key process. Even so, the places where change can be less expensive are forest and agricultural areas. In the results, it was found that areas with land use conflict class 3, which are agricultural areas, have a strong impact on water quality. One of the measures is to propose the reallocation to areas where the land is suitable for such practices. As it is detected in Figure 3
E, areas with land use conflict are small patches strategically located and easily replaced by forestry land uses.
In the present study, the measured concentration of nitrates fulfilled the Portuguese legislation threshold (minor to 25 mg/L) and subsequently the Nitrates Directive (2006/118/EC) (ND) requirements for drinking purposes (minor to 50 mg/L). However, stricter benchmarks should be addressed, since a concentration minor to 3 mg/L indicates contamination [111
], and the United States Environmental Protection Agency maximum concentration level for nitrate is 10 mg/L in order to protect against blue-baby syndrome [112
]. As it was identified in this study, the nitrogen sources are affecting Sabor River Basin water quality. By identifying the effects of each pressure, the applied methodology revealed itself as an important tool for the protection of aquatic systems, even when contaminant concentration fulfil legal limits. Notwithstanding that the Sabor River Basin is under an acceptable ecological status, the strategies to reduce the diffuse emissions from agriculture and livestock (presented in the ND) and complementary strategies (presented in the WFD) [113
] should be taken into action, in order to preserve water quality.
4.2. Study Limitations and Future Recommendations
There are essential aspects for unerring river basin monitoring, such as the number and spatial distribution of sampling sites, the number of measurements made at each point and analysis method [114
], which may consist on a statistical or physical approach [115
]. Such elaborated methodologies are craved by researchers and stakeholders but, nevertheless, it is crucial to have proper funding [116
] for the elaboration of extensive and detailed sampling campaigns, or even to achieve detailed datasets, by the application of in situ automated samplers [117
]. The present study design is quite innovative since it compares simultaneously the scale and seasonal effects in water quality through an explicative statistical approach. Still, it is alerted that some aspects of the present work could not be optimal due to logistic issues, namely the number of sampling sites and sampling frequency. Besides PLS-PM requires a smaller sample size than other statistical methods, such as traditional SEM [118
], is preferable to use a higher number of samples. This aspect could be improved by increasing the number of sampling sites in Sabor River Basin, covering a greater part of the river basin. It is not expected that the cause-effect relationships would change, since the sub-catchment areas are predominantly rural, but it would improve statistical significance [119
]. The present study pictured the interplay among landscape metrics, pollution sources with water quality parameters, through water quality samples that were measured simultaneously with IPtIN
in each sampling site. For such reason, our results can exhibit linkages between variables that were contained in the dataset, but this cannot be seen as undisputed interactions for Sabor River Basin. To achieve such evidences, it would be necessary the sampling frequency by achieving a more detailed timestep, by using seasonal, or even monthly averages, that would give a more detailed picture of Sabor River Basin environmental interactions. For such reasons, some monitoring technical aspects are recommended, in order to enhance the virtue of future studies. By increasing the number of sampling sites it is possible to achieve high reliability when sampling by using at least 20 sampling sites strategically distributed [120
]. To ensure reliability is crucial to have a detailed time step to monitor contaminant concentrations. The adequate sampling frequency is discussed by many authors and is also dependent on the type of contaminant [121
], but sampling frequencies such as weekly to bi-monthly can be suitable for monitoring [122
], supported by other author results [123
] is advised that, for contaminant concentrations, the sampling frequency should be weekly. Since benthic macroinvertebrates communities take time to assemblage, comparing the changes of lifeforms across a short time step might be meaningless [124
]. For such reasons is recommended that the IPtIN
sampling should be done monthly, since it was already revealed that the presence of such lifeforms can change in periods of two months [125
This study, which was based on partial least squares-path modeling (PLS-PM) and conducted in the Sabor River Basin (Northeast Portugal) revealed significant seasonal and spatial scale influences of anthropogenic pressures on stream water quality and ecological integrity (measured by a macroinvertebrate biodiversity index). The most influential pressure was the metric “edge density” because it was associated with ecological integrity declines on both short (within and 250 m around target sites) and long spatial scales (within the drainage area upstream of the target site), as well as in winter and summer. Other pressures were also important but influenced water contamination or ecological integrity differently on the short vs. long scale or in winter vs. summer. For example, agricultural area patches and water body connectivity were shown to be adversely related to the short spatial scale but positively related to the long spatial scale. Land use conflicts (deviations of actual from natural uses) as well as pasture area connectivity were shown to adversely affect the ecological integrity solely on the long scale. Finally, surface water parameters, such as arsenic, potassium, and manganese, were shown to adversely affect the ecological integrity on both the short and long spatial scales, but the effects were different in summer vs. winter. The explanations for the seasonal differences were diverse for the three parameters, for example, seasonality in the application of farmland fertilizers, geology, and the share of surface water and groundwater in stream flow. Overall, the study exposed a complex interplay among anthropogenic pressures, stream water contamination, and ecological integrity, which was uncovered using PLS-PM. This sophisticated statistical model proved to be efficient for the assessment of these intricate relationships, but could also be used for the complex management of multiple land use watersheds such as the Sabor River Basin. As it was identified in another study, Sabor River Basin is under acceptable ecological conditions (according to data collected in 2012) [64
]. However, as it was revealed in this study (with data from 2016 and 2017), it is crucial to monitor point source emissions and to decrease nutrient loads from diffuse emissions and reforest agricultural areas that are under land use conflicts in order preserve such ecological status, or even to improve it.
Our results, therefore, present a challenge to municipal and regional water planners, as well to water management authorities, which are responsible for the safeguarding of water quality.