Testing the Sensitivity and Limitations of Frequently Used Aquatic Biota Indices in Temperate Mountain Streams and Plain Streams of China

Abstract

Different biological groups show biased responses to similar or different environmental stressors on different scales. The selection of bioindicators based on pressure characteristics is the basis for accurately assessing ecological quality. In this study, we investigated the responses of common bioindicators, namely, macroinvertebrates and fish, to multiple stressors in temperate mountain and plain streams of northeast China. We used 56 indices, including the single biological evaluation index and biological evaluation index system, to analyze and compare characteristic response to different scales under varying environmental stressors. The principal component analysis (PCA) showed that PCA axis 1 in the catchment scale explained 83.6% and 96.1% of the variance in mountain and plain rivers, respectively, which characterized the comprehensive pressure gradient integrated by land-use development and water pollution. PCA axis 1 explained 40.7% and 53.9% of variance in mountain and plain rivers on the reach scale and 63.1% and 61.8% of variance on the site scale. The correlation analysis showed responses of different indices to abiotic variables which did not overlap. Macroinvertebrate and fish indices successfully explained the change in water chemistry on a small scale, whereas fish indices additionally explained the change in land use on a large scale. Macroinvertebrate and fish indices were recommended because of their rich responses to environmental stressors, particularly in plain rivers. For mountain stream biomonitoring programs, especially in the Taizi River of northeastern China, we suggest that macroinvertebrates and fish should be used separately or jointly according to the actual capacity and cost, Moreover, compared with the possible differences in the evaluation results of different single biological evaluation indexes, the biological evaluation index system shows more stable monitoring results, and the single sensitivity index is more significant in biological evaluation, and more sensitive to some special environmental factors.

1. Introduction

Clean and healthy rivers and streams greatly enhance quality of life. Over the centuries, however, global river ecosystems have undergone tremendous changes due to human activities [1]. The question of measuring or quantifying the changes in river ecological status, or evaluating risk based on change has always been a challenge faced by the ecologists [2]. As the main component of material and energy exchange, aquatic organisms play an important role in stabilizing structure and function of river ecosystems [3]. Moreover, changes in biological assemblages can modulate effects of multiple stressors and environmental variables over time; thus, aquatic bioindicators have attracted extensive attention around the world [4,5]. Currently, there are hundreds of indices and methods for assessing river ecological equality using biological community data [6], from single index to multimetric index (MMI) to predictive models, which provide effective evaluation tools for river managers across a wide range of ecosystems.

Macroinvertebrates and fish have been widely used in river and stream ecological quality assessment [6,7]. They are easily sampled using established methods, and have predictable responses to environmental interference. However, different groups of aquatic organisms show unique characteristics in actual biomonitoring due to differences in their life cycle and response sensitivity across environmental pressure gradients in river ecosystems, primary producers always have a short life cycle and can rapidly colonize under good water quality and substrate conditions, as well as respond rapidly to environmental change [8]. It is transient and changes rapidly with time and space. In contrast, fish, as a top-level assemblage species with strong movement and feeding ability, show longer generation time and are more indicative of past accumulated stress [9]. For macroinvertebrates, generation time and dispersal potential, which fall under primary producers and fish, are useful in reflecting medium-duration environmental stress [10]. During biomonitoring, multiple taxonomic assemblages are typically combined, as each assemblage offers specific responses to different stressors [11,12].

Biological evaluation indices have been extensively deepened and developed, ranging from simple to complex and comprehensive indices. Simple indices, such as richness, abundance, and biodiversity, originated and gradually developed from saprobic systems in the 1900s and reflect a specific aspect of community structure [13]. By the 1980s, a multi-metric evaluation system, known as macroinvertebrate-based multimetric index MMI or index of biotic integrity (IBI) was developed by Karr [4]. Unlike the simple indices, this system integrates different response characteristics of each metric to environmental stressors, and shows a more stable response to pressure [4]. In the following decades, various countries applied different types of indices according to their use habits and monitoring capabilities [14,15,16]. For example, the Water Framework Directive (WFD) requires European Union member states to monitor and assess the ecological quality of rivers and streams. While the WFD proposes the characteristics of biotic quality elements that should be assessed, it does not require specific indices [17]. Therefore, much research has focused on the applicability, stability, and uncertainty of indices to different systems [6,18].

For most of the developing countries, river bioassessment methods are relative recent [19]. In China, bioassessment was initially incorporated into river basin management in 2020 (the 14th 5 Year Plan). Until now, river biomonitoring programs have faced strong subjectivity and difficulty due to many management departments in China work protocols highlighting a knowledge gap, and the need for more basic research on comparing response s of different taxonomic assemblages to multiple stressors.

The present study focuses on frequently used indices of macroinvertebrates and fish in temperate mountain and plain streams in China. We examined the relationships of indices with environmental pressure by investigating the response of various bio-indices for each taxonomic group (macroinvertebrates and fish) to different scales of environmental stressors, including water chemistry and sediment type at the site scale, fluvial physical parameters in the reach scale, and land use patterns at the catchment scale, to build comprehensive pressure aimed to explore better bio-predictors of disturbance in the temperate mountain and plain streams in China.

2. Methods

2.1. Study Area

The Taizi River basin is located in northeastern China (122°26′–124°53′ E, 40°29′–41°39′ N) with a catchment area of 1.39 × 10⁴ km². The Taizi River is a typical temperate river with a total length of 413 km. The annual average precipitation is 778.1 mm, that is unevenly distributed, mostly throughout the flood season from June to September. The multiyear average temperature is 6.2 °C with a huge temperature difference between winter and summer. In addition, the obvious environmental interference gradient in the basin is the key reason for selecting this study area [20]. The geomorphological features of the basin include the upper middle mountainous and hilly forest region and the lower plain agricultural region. The mountain region has a high percentage of natural vegetation cover with little human disturbance, whereas the plain region is mainly agricultural and urban land. The anthropogenic activities gradually increase from upstream to downstream.

A total of 32 sampling sites located at first-, second-, third-, and fourth-order streams in the mountain region were chosen for environmental condition monitoring and biological assemblage collection between August 2009 and October 2010 [21] (Figure 1). The selected sampling sites were located on tributaries, mainly because our study focused on low-level stream assessments, as there are assessments that are more suitable for high-level stream. The monitoring of environmental factors and collection of biological assemblages were according to the following methods.

2.2. Environmental Factors

For each site, 32 environmental factors of three different scales indicating water quality, such as pH, electric conductivity (EC), total dissolved solids (TDS), and dissolved oxygen (DO), were measured in situ using the YSI Pro 2030 multiparameter instrument. Further, five water quality parameters, including, total phosphorus (TP), total nitrogen (TN), ammonia-nitrogen (NH₄-N), permanganate index (COD_Mn), and 5-day biological oxygen demand (BOD₅), were determined in the laboratory according to national standard methods [22]. Hydrological indicators such as flow velocity and depth were measured in situ using a direct reading current meter (FP-201) (

Table 1. Description of environmental variables of all sampling sites.

Variable	First Quartile	Median	Third Quartile	Mean	SD	Range
Watershed characteristics
Catchment area (km2)	52.94	95.36	326.98	3243.67	597.32	17.26–3407.75
Water area (m2)	0	0	0.0043	0.0088	0.0173	0–0.05
Forest area (m2)	0.58	0.71	0.78	0.64	0.24	0–0.91
Construction area (m2)	0.0038	0.01023	0.05508	0.036	0.047	0–0.20
Crop area (m2)	0.188	0.234	0.377	0.0306	0.209	0.018–0.891
Grass area (m2)	0	0	0.015	0.011	0.019	0–0.069
Water physicochemical conditions
pH	8.0	8.3	8.5	8.2	0.4	7.0–8.8
EC (μS/cm)	189.5	281.0	424.5	313.2	195.4	89.0–1133.0
TDS (mg/L)	169.0	280.0	335.3	277.1	156.1	51.0–746.5
DO (mg/L)	6.0	6.9	7.8	7.1	1.7	3.9–13.5
BOD5 (mg/L)	2.7	4.0	5.3	5.3	4.9	1.9–28.7
CODMn (mg/L)	1.8	2.4	4.0	3.1	1.7	1.4–8.3
TN (mg/L)	1.5	2.1	3.2	3.1	3.0	0.8–17.0
TP (mg/L)	0.0	0.1	0.2	0.2	0.3	0.0–1.6
NH4-N (mg/L)	0.1	0.1	0.5	0.7	2.2	0.03–13.2
SS (mg/L)	23.75	69	119.25	108.49	152.11	11.5–884
Cobble	0.114	0.164	0.200	0.18	0.11	0–0.57
Pebble	0.196	0.289	0.368	0.28	0.12	0.04–0.52
Gravel	0.069	0.109	0.196	0.14	0.10	0.01–0.45
Sand	0.021	0.043	0.102	0.076	0.085	0.003–0.389
Hydrological characteristics
Depth (cm)	13.7	18.3	24.3	19.5	9.6	5.0–52.0
Velocity (m/s)	0.25	0.37	0.44	0.37	0.17	0.0–0.8
Altitude (m)	149.5	256	385	263.86	139.03	8–546
Slope (%)	3.66	7.05	13.48	9.13	7.37	0–29.92
Sinuosity	1.09	1.24	1.42	1.29	0.27	1–2.12
Stream order	1	2	3	1.97	0.81	1–3
Length (m)	4.90	7.16	17.36	11.75	9.84	0.72–35.1
Distance from mouth	292,352	34,535	455,650	346,843.2	105,337.4	130,574–504,743
Average temperature (°C)	4.92	5.83	7.15	5.86	1.89	3.22–9.01
Average rainfall (mm)	835.2	899.3	950.65	876.10	81.37	650.9–954.6
Longitude	123.48	123.75	124.44	123.85	0.62	122.68–124.79
Latitude	40.92	41.21	41.36	41.15	0.27	40.62–41.60
Water temperature (°C)	19.2	21.5	24.1	21.03	3.37	14–26.5

).

2.3. Biological Assemblages

2.3.1. Macroinvertebrates

Firstly, habitat types were assessed using combinations of flow velocity, water depth, and substrate type within a 100 m reach [23]. Then, three replicate samples in different habitat types were collected using a Surber net (0.09 m² in area, with 500 µm mesh size) by the same person to guarantee consistency of the sampling. Each sample was separately put into a 500 mL bottle and preserved in alcohol (70%). In the laboratory, all individuals were sorted, counted, and identified to the lowest feasible taxonomic level.

2.3.2. Fish

A 300 m survey reach containing different types of microhabitats was delimited for fish collection. In areas that could be accessed by wading, fish were sampled by electrofishing for 30 min following a zig-zag path within the survey reach. In areas that were too deep for wading, electrofishing was carried out from an inflatable boat. Once collected, fish were identified at the species level, and biological information was recorded.

2.4. Land Use and Patterns at Various Scales

Using ENV14.4 software, the Landsat Thematic Mapper image (track Numbers P118R31, P119R31, and P119R32) was classified into six land-use types: forest, pasture, farmland (paddy land and dry land), wetland, urban land, and unused land. The man–machine interactive interpretation method combining manual interpretation and proximity classification was used to extract information. Preliminary interpretation results were then modified to form the final interpretation results. Four main land-use types—forest, pasture, farmland, and urban land—were used for analysis (Table 1). Each site was divided into two categories: plain rivers (altitude < 200 m) and mountain and hilly rivers (altitude > 200 m) [24].

2.5. Data Analysis

2.5.1. Biological Indices Selected

A total of 56 biological indices, including species richness, Shannon diversity index, percentage of sensitive assemblages, IBI, BI, and BMWP indices were calculated for each biological groups (macroinvertebrates and fish). We used the following formula to calculate Shannon–Wiener diversity:

$$H^{\prime }=-{\displaystyle \sum }_{n=1}^S\left(P_i\right)\left(lo{g}_2{P}_i\right),$$

where H′ is the diversity index, S is the species number, and P_i is the proportion of the number of individuals of i to the total number of individuals.

The percentage of sensitive assemblages was calculated on the basis of the identification of sensitive species or taxon identification. Sensitive periphyton were screened according to classification by Wang et al. and expert opinions [25]. Sensitive macroinvertebrates taxa were selected from Ephemeroptera, Plecoptera, and Trichoptera, which have been recognized by many bioassessment studies. Finally, sensitive fish species were identified at the species level according to the expert’s experience and historical documentation [26].

Our research group recently conducted relevant studies on development of IBI indices for macroinvertebrates and fish in the Taizi River basin [27,28]. Therefore, we used the core metrics directly from these previous studies to calculate IBI for macroinvertebrates and fish, respectively, in this study (

Table 2. Correlation coefficients between bioindicators and PCA axis 1.

Abbreviation	Index Parameter	Mountain and Hilly Rivers			Plain Rivers
		PCA Axis 1 (Catchment Scale)	PCA Axis 1 (Reach Scale)	PCA Axis 1 (Site Scale)	PCA Axis 1 (Catchment Scale)	PCA Axis 1 (Reach Scale)	PCA Axis 1 (Site Scale)
F1	Number of fish species	0.355	0.474 *	−0.163	−0.197	−0.41	−0.464
F2	Diversity index	0.242	0.533 **	0.192	0.192	−0.232	−0.088
F3	Percentage of Gobiaceae	0.458 *	0.302	0.242	0.017	0.207	−0.066
F4	Percentage of Cyprinidae	0.093	0.196	−0.385	0.668 *	0.813 **	0.904 **
F5	Percentage of Cobitidae	0.09	−0.31	0.07	−0.381	−0.625 *	−0.63 *
F5	Percentage of Cobitidae	0.121	−0.352	0.116	−0.371	−0.642 *	−0.655
F6	Percentage of Leuciscinae	−0.283	−0.279	−0.315	−0.515	−0.557	−0.654 *
F7	Percentage of Gobiidae	0.061	−0.064	0.009	−0.166	0.009	0.192
F8	Percentage of pelagic fish	−0.058	0.393	0.364	0.249	−0.15	0.241
F9	Percentage of bottom-dwelling fish	0.156	−0.175	0.039	−0.463	−0.232	−0.411
F10	Percentage of lower- and middle-class fish	−0.154	−0.078	−0.286	0.09	0.266	0.107
F12	Percentage of herbivorous fish	−0.285	0.435 *	−0.23	−0.038	0.413	0.209
F13	Percentage of omnivorous fish	−0.043	−0.069	0.031	−0.223	−0.526	−0.326
F14	Percentage of benthic feeders	−0.102	−0.263	−0.049	0.559	0.316	0.363
F15	Percentage of tolerant fish	0.235	0.243	−0.013	0.336	0.632 *	0.625 *
F16	Percentage of sensitive fish	0.026	0.613 **	−0.32	−0.49	−0.566	−0.698 *
F17	Percentage of pelagic egg fish	−0.28	−0.305	−0.401	0.710 **	0.507	0.721 **
F18	Percentage of demersal egg fish	0.177	0.351	0.153	−0.065	−0.369	−0.077
F19	Percentage of viscid egg fish	−0.06	0.389	0.362	−0.519	−0.413	−0.547
F20	Percentage of fish with special spawning methods	−0.125	−0.525 *	−0.35	0.082	0.623 *	0.405
F21	Individual number	0.097	0.144	−0.311	−0.337	−0.462	−0.668 *
F22	Percentage of cold-water fish	0.275	−0.272	−0.062	−0.506	−0.640 *	−0.716 **
F24	Percentage of widely distributed species (frequency >50%)	−0.163	−0.537 **	−0.226	−0.506	−0.557	−0.716 **
M1	Total taxa	−0.108	0.454 *	−0.530 *	−0.741 **	−0.517	−0.649 *
M2	EPT	−0.215	0.531 *	−0.453 *	−0.607 *	−0.569	−0.61 *
M3	Ephemeroptera	−0.469 *	0.595 **	−0.453 *	−0.599 *	−0.589 *	−0.573
M4	Plecoptera	0.106	−0.149	−0.354	0.199	0.458	0.395
M5	Trichoptera	0.088	0.430 *	−0.273	−0.544	−0.505	−0.612 *
M6	Amphipoda + Mollusca	0.226	−0.162	−0.23	−0.446	−0.204	−0.126
M7	Pleccoptera %	−0.075	−0.259	−0.252	0.199	0.458	0.395
M8	Ephemeroptera %	−0.442 *	0.171	−0.037	−0.233	−0.337	−0.168
M9	Trichoptera %	0.016	0.167	0.351	−0.472	−0.222	−0.373
M10	EPT %	−0.356	0.236	0.188	−0.357	−0.317	−0.263
M11	Chironomidae %	0.314	−0.195	−0.299	0.04	−0.46	−0.356
M12	Diptera %	0.358	−0.08	−0.255	−0.002	−0.505	−0.401
M13	Amphipoda + Mollusca %	0.167	−0.347	0.144	−0.498	−0.304	−0.44
M14	Oligochaeta %	0.06	−0.316	−0.022	0.642 *	0.826 **	0.835 **
M15	Intolerant taxa	−0.231	0.393	−0.571 **	−0.538	−0.521	−0.461
M16	Relative abundance of species number of fouling-tolerant groups	0.085	−0.207	0.018	0.607 *	0.598 *	0.703 *
M17	Relative abundance of the most dominant taxa	0.4	−0.179	0.017	0.453	0.354	0.334
M18	Filterer %	0.02	0.024	0.318	0.579 *	0.864 **	0.829 **
M19	Scraper %	−0.357	0.077	−0.04	−0.207	−0.249	−0.134
M20	Collector/Gatherer %	0.324	0.068	−0.123	−0.557	−0.904 **	−0.863 **
M21	Predator %	−0.352	−0.28	−0.313	−0.038	0.151	0.117
M22	Shredder %	0.082	−0.319	−0.215	−0.227	0.068	−0.109
M23	Clinger %	−0.403	0.245	0.046	−0.424	−0.249	−0.375
M24	Clinger taxa	−0.214	0.488 *	−0.381	−0.556	−0.504	−0.512
M25	Shannon	−0.433 *	0.311	−0.124	−0.525	−0.462	−0.465
M26	Margalef	−0.121	0.442 *	−0.338	−0.663 *	−0.558	−0.625 *
M27	Evenness	−0.426 *	0.13	0.179	−0.401	−0.448	−0.405
M28	Simpson	−0.394	0.209	−0.021	−0.543	−0.445	−0.459
M29	B-IBI	−0.166	0.383	−0.315	−0.737 **	−0.417	−0.516
M30	BMWP	0.095	0.069	−0.555 **	−0.571	−0.386	−0.45
M31	FBI	−0.205	0.063	0.174	0.561	0.774 **	0.821 **
M32	BI	0.468 *	−0.284	0.012	0.762 **	0.785 **	0.876 **

Note: * indicates the significance level of Kolmogorov–Smirnov test, 0.01 < p < 0.05. ** indicates the significance level of Kolmogorov–Smirnov test, p < 0.01.

).

2.5.2. Statistical Analysis

All environmental variables at the catchment scale (land-use characteristics), reach scale (hydrological characteristics), and site scale (physicochemical variables) were integrated into the comprehensive environmental gradient using principal component analysis (PCA) [29]. Prior to PCA, environmental variables that were not distributed (normally according to Kolmogorov–Smirnov test, p < 0.05) were all standardized to have zero mean and variance unity to account for different measurement units. The site scores on the principal component axis (PC1) were used as independent variables in spatial regression models to evaluate the effects of the comprehensive environmental gradient on different biotic indices. PCA were performed with Canoco for Windows 4.5.

A correlation analysis among macroinvertebrate and fish indices and stressors was performed to explain the similarities among different indices and the relationship between biotic indices and abiotic stressors using SPSS 21.

3. Results

3.1. Comprehensive Environmental Gradient

PCA was performed on environmental characteristics in the study area of mountain and hilly rivers (Figure 2 and Figure 3). In this case, the first four principal components accounted for 99.9% of the variance on the catchment scale. The first, second, third, and fourth principal components explained 83.6%, 11.3%, 3.6%, and 1.4% of the total variance, respectively. On the reach scale, the first four principal components accounted for 94.9% of the total variance, explaining 40.7%, 39.3%, 9.4%, and 5.5%, respectively. On the site scale, the first four principal components explained 93.8% of the total variance (63.1%, 20.4%, 5.4%, and 4.9%, respectively).

Among the total environmental factors on the catchment scale, crop area had a high contribution rate to PCA axis 1, followed by forest area. This principal component (PC1) mainly represented the gradient of land use. It was used as a comprehensive environmental gradient for further analysis. On the reach scale, stream order and slope had a high contribution to PCA axis 1. On the site scale, representing environmental factors of physical and chemical indices and substrate types, SS and TDS had the highest contribution to PCA axis 1.

The same analysis was applied to plain rivers (Figure 3). The first four principal components accounted for 99.8% of the total variance on the catchment scale. The first, second, third, and fourth principal components explained 96.1%, 2.9%, 0.5%, and 0.3% of the total variance, respectively. On the reach scale, the first four principal components accounted for 97.6% of the total variance, explaining 53.9%, 27.3%, 10.7%, and 5.7% respectively. On the site scale, the first four principal components explained 95.0% of the total variance (61.8%, 21.9%, 7.1%, and 4.2%, respectively).

Among the total environmental factors on the catchment scale, forest area had the highest contribution rate to PCA axis 1, followed by crop area. This principal component (PC1) mainly represented the gradient of land use. It was used as a comprehensive environmental gradient for further analysis. On the reach scale, altitude and slope had a high contribution to PCA axis 1, whereas length had a high contribution to PCA axis 2. On the site scale, representing environmental factors of physical and chemical indices and substrate types, TDS, SS, EC, and NH₄-N had the highest contribution to PCA axis 1.

At the three different scales (catchment scale, reach scale, and site scale), a total of 32 environmental factors had significantly correlated relationships according to Spearman correlation analysis (Figure 4). In the mountain and hilly rivers, longitude and latitude were negatively correlated with physical and chemical indices including TDS (correlation coefficient = −0.76 and −0.66, respectively), SS (correlation coefficient = −0.89 and −0.79, respectively), TN (correlation coefficient = −0.59 and −0.57, respectively), and NH₄-N (correlation coefficient = −0.53 and −0.44, respectively). Altitude was correlated with land-use types. Sediment types were also positively correlated with physical and chemical indices. Boulders and pebbles were correlated with TN (correlation coefficient = 0.62 and 0.53, respectively), and gravel was correlated with TN and NH₄-N (correlation coefficient = 0.51 and 0.53, respectively) in this area.

Furthermore, several physical and chemical environmental factors were significantly correlated with each other. NH₄-N had a significant correlation with DO (correlation coefficient 0.60). TP was correlated with TDS (correlation coefficient = 0.58), and SS was positive correlated with TDS (correlation coefficient = 0.75) and TN (correlation coefficient = 0.52).

For the three different scales in the plain river sites, 32 environmental factors had significantly correlated relationships according to Spearman correlation analysis (Figure 5). Longitude was negatively correlated with average temperature (correlation coefficient = −0.76), whereas it was positively correlated with distance from mouth (correlation coefficient = 0.93) and water area (correlation coefficient = 0.78). Altitude was also correlated with land-use type. There was a negative correlation between altitude and crop area (correlation coefficient = −0.87) and a positive correlation between altitude and forest area (correlation coefficient = 0.88). Furthermore, EC was significant correlated with TN and NH₄-N (correlation coefficient = 0.85 and 0.89, respectively).

3.2. Relationships between Biotic Indices and Environmental Variables

The correlation results between bioindicators and PCA axis 1 on three scales showed that the biological groups and indices responded slightly differently to the comprehensive pressure gradient describing the stressors (Table 2). In the mountain and hilly rivers, biological indices had a higher correlation on the reach scale than on the catchment and site scales. PCA axis 1 on the reach scale was positively correlated with F1, F2, F12, F16, M1, M2, M3, M5, M24, and M26 but negatively correlated with F20 and F24 (see details in Table 2). On the site scale, PCA axis 1 was negatively correlated with M1, M2, M3, M15, and M30. On the catchment scale, PCA axis 1 was negatively correlated with M3 and M8 but positively correlated with M32. Overall, PCA axis 1 had a greater correlation with biological indices on the reach scale, and the macroinvertebrate group was better able to respond to environmental stressor gradients in mountain and hilly rivers.

The correlation results between PCA axis 1 and biotic indices in plain rivers showed a better responsiveness of biological communities to environmental stressor gradients. On the catchment scale, PCA axis 1 was positively correlated with F4, F17, M14, M16, M18, and M32 but negatively correlated with M1, M2, M3 and M26. On the reach scale, PCA axis 1 was positively correlated with F4, F20, M14, M16, M18, M31, and M32 but negatively correlated with F5, F6, F22, and M20. On the site scale, PCA axis 1 was positively correlated with F4, F15, F17, M14, M16, M18, M31, and M32 but negatively correlated with F6, F16, F21, F22, F24, M1, M2, M5, M20, and M26. In plain rivers, the macroinvertebrate group also showed a significant response to environmental stressor gradients, especially on the site scale.

4. Discussion

Landscape transition drives global environmental change. For most mountain regions in China, landscape changes are mainly reflected in residential land construction, agricultural production, and logging practices in native forest [30]. This change in land-use type is an important driving force for mountainous aquatic ecosystems [31]. Previous studies indicated that expanding artificial land (urban and farmland) and the shrinking natural land (forest and pasture) has been significant in the mountain area of the Taizi River [32], thus affecting the stream water quality. Our results showed the comprehensive environmental stressor gradient at a watershed scale constructed on the basis that land-use types had a high explanatory power. Landscape changes were significant correlated with physical and chemical indices of water quality and sediment type. Further, a similar relationship between landscape and water quality was also supported by many other studies [33,34]. Our data analysis results showed more correlations between different environmental factors in mountain and hilly rivers than in plain rivers. Physical and chemical indices and sediment types were main constituent of environmental stressor gradients in plain rivers, whereas land-use types predominated in mountain and hilly rivers. These results were caused by the difference in landscape at different elevations.

Our data support the idea that macroinvertebrate and fish show different discriminatory power when detecting comprehensive environmental change owing to variations in taxa life cycle. In this study, multiscale environmental stressors (land use and water chemistry) were integrated into PC1, we discussed the correlation linking fish and macroinvertebrates to the integrated environmental stressor gradient at different scales, finding that macroinvertebrates showed a significant response in mountain and hilly rivers. In plain rivers, macroinvertebrates also exhibited an obvious effect but fish played a major role in response to the environmental stressor gradient at different scales. In general, there is a positive correlation between the number of fish species and river grades at watershed scale, but some indigenous fish species only exist at certain elevations [35,36].

We hypothesized that macroinvertebrate and fish indices would decrease along the comprehensive environmental gradients. Differences in stream fish assemblages among reaches are well explained by habitat structure patterns [37,38]. During a field survey, we found that the channel gradually widened and deepened, whereas hydro-geomorphological types became mor diversified from upstream to downstream. In addition, several habitat-specific fish such as Cottus poecilopus, which prefer a shallow and fast-current habitat [39], were distributed upstream, whereas more species across various habitats were distributed downstream. It is well-known that fish species are always poorly represented in upstream reaches due to harsh environments (e.g., low productivity and severe droughts/floods) [11]. In downstream reaches, although the environmental interference increased moderately, the diversified habitats met the living needs of different fish species and contributed to an increase in fish richness. Therefore, we assume that the change in habitat pattern was responsible for the increase in fish species along the stream longitudinal gradients, which conforms to the intermediate disturbance hypothesis.

The responses of different taxonomic groups to the same stressor or to different stressors may be quite different. As expected, only fish indices responded to large-scale drivers in our mountain streams. Although Johnson et al. found no evidence supporting that fish could integrate effects over larger scales than other organisms [40], fish assemblages frequently show strong relationships with land use [41,42]. For example, Flinders et al. found fish indices strongly responded to land-use gradients at the catchment scale for mid-Atlantic upland streams [43]. Highlighting response variables that represent important attributes of community diversity and composition were suitable for indicating changes in large-scale drivers in mountain streams. In addition, there are similar results in the study of river evaluation in China [44].

When the water chemistry was assessed, macroinvertebrate and fish indices performed adequate performance. EPT taxa were found to be strongly related to substrate composition [45], but this was not considered in our study. Schäffer et al. [46] found that fine sediment loading had a strong and significantly negative effects on macroinvertebrate assemblages, particularly the relative abundance of EPT individuals. Substrate composition is similar among mountain streams of the Taizi River and is characterized by boulders, cobbles, and pebbles, which can supply sufficient living support for EPT taxa, including attachments, shelters, and intercepted foods [47]. These characteristics may have contributed to the widespread distribution of EPT taxa in mountain streams and the homogeneity of EPT composition, thus reducing the sensitivity of sensitive macroinvertebrates taxon to water chemistry variables.

We found that evaluating the IBI index system is better than using a single biological index developed at the watershed scale, whereas core metrics exhibited gradient changes across the whole basin including mountain and plain regions. In mountain streams, changes in the IBI range narrowed, which probably led to a poor performance in environmental pressure detection. In other words, the degree of homogenization of periphyton assemblages is relatively increased in mountain streams. Therefore, a new IBI should be redeveloped for mountain regions in future. In our research, system evaluations using indices such as IBI and BMWP were steadier and more sensitive. The single biological evaluation indices were also of special value, with EPT and sensitive species of fish having indicative significance. In recent years, the river health assessment system has been widely used in China. Macrobenthos and fish are also selected as indicator species in Taihu Basin, Yangtze River, Poyang Lake, and other typical basins in China [48,49,50].

Macroinvertebrate and fish assemblages are the most commonly utilized assemblages in bio-monitoring programs. Selection of an ideal bioindicator should be a knowledge-based decision based on an ecological response curve [40]. Johnson and Hering found that macroinvertebrates were more appropriate as a pressure predictor in mountain streams than periphyton and fish [11]. Similarly, Freund and Petty found that macroinvertebrate assemblages were more responsive to water quality than fish [51]. Johnson and Ringler found that macroinvertebrate assemblages were correlated with land use and substrate in a perturbed watershed, while fish assemblages were correlated with water chemistry variables and stream width [12]. Karaouzas et al. found that periphyton assemblages responded to water quality variables on a reach scale, but macroinvertebrate assemblages responded to land use on a catchment scale [14]. The responses of different taxonomic assemblages vary and are, at times, inconsistent; hence, the combined application of different bioindicators is advocated in biomonitoring works [7]. Additionally, a composite index developed using multiple taxonomic assemblages is also an effective and efficient way to reduce bias [45]. Macroinvertebrate and fish indices in this study showed significant ecological dose-response relationships. Moreover, macroinvertebrate index and fish index have different responses to abiotic variables, and have special indicators for specific environmental factors. Therefore, we assumed that the combined application of macroinvertebrate and fish indices in mountain stream bioassessment could provide a comprehensive indication of environmental pressure. The results of this work suggest that biomonitoring and assessment programs in temperate mountain streams with low or moderate interference can rely on the use of macroinvertebrate and/or fish indices according to the management’s monitoring capacity and cost (with the exception of sensitive macroinvertebrate taxa and richness of fish). In addition, our findings suggest that the bioindicators representing community attributes of composition and diversity can be used as a surrogate for a multimetric index, which can be difficult to develop, especially in cases with poor datasets and homogenization interference conditions.

The common biological indices have limitations in practical application and cannot fully indicate the health status of the study area. It is necessary to select appropriate evaluation methods and indicator species for rivers of different grades. Different biological groups and different biological indexes may have different responses to the same environmental pressure gradient. In this study, macrobenthos and fish have different responses to pressure gradients constructed at different scales, which also explains this problem to some extent. Based on our existing research foundation for many years, combined with the two common indicator groups of macrobenthos and fish, the single indicator and the indicator results of the comprehensive evaluation system were compared and analyzed, so as to comprehensively evaluate the health status of the Taizi River Basin.