Remote Sensing Reveals Multi-Dimensional Functional Changes in Fish Assemblages Under Eutrophication and Hydrological Stress

Abstract

Understanding how fish communities respond to long-term environmental changes in regulated floodplain ecosystems is essential for managing biodiversity amid increasing anthropogenic and climatic pressures. This study evaluates the spatiotemporal dynamics of functional diversity in juvenile fish assemblages within the Dnipro-Orilskiy Nature Reserve (Ukraine) from 1997 to 2015. By employing a combination of extensive ichthyological field surveys and satellite-derived environmental indices (including NDVI, chlorophyll-a, turbidity, and spectral proxies for algal blooms), we assessed the impacts of eutrophication, hydrological alterations, and climate warming on functional structure. Our results reveal three key responses in fish functional diversity: (1) a decline in functional specialization and imbalance, indicating the loss of unique ecological roles and increased redundancy; (2) a rise in functional divergence, reflecting a shift toward species with outlying trait combinations; and (3) a complex pattern in functional richness, with trends varying by site and trait structure. These shifts are linked to increasing eutrophication and warming, particularly in floodplain areas. Remote sensing effectively captured spatial variation in eutrophication-related water quality and proved to be a powerful tool for linking environmental change to fish community dynamics, not least in inaccessible areas.

1. Introduction

Human-driven changes are increasingly pushing ecosystems beyond historical baselines, leading to novel combinations of environmental conditions and species assemblages. Such ecological novelty, defined as the emergence of unprecedented abiotic and biotic states [1], is particularly prominent in floodplain systems affected by eutrophication, warming, and flow regulation. Floodplain ecosystems are highly dynamic environments where hydrological regimes and nutrient inputs play a crucial role in shaping biodiversity patterns and ecosystem functioning [2,3]. It is estimated that between 70% and 90% of European floodplains have experienced significant environmental degradation due to anthropogenic pressures [4]. This widespread degradation has primarily been driven by hydrological modifications, such as dam construction and flow regulation, as well as land use changes, pollution, and urban expansion [5,6]. These human-induced impacts disrupt natural flood regimes, fragment habitats, reduce biodiversity, and impair essential ecosystem functions, including nutrient cycling and sediment transport [7,8]. Understanding how functional diversity responds under these novel conditions is essential for anticipating future ecosystem trajectories and resilience. Over time, such environmental disturbances lead to a functional reorganization of aquatic communities [9]. Among aquatic taxa, fish assemblages serve as particularly sensitive indicators of changes in functional diversity due to their ecological variability, mobility, and responsiveness to environmental stressors [10,11,12].

Conventional taxonomic metrics, such as species richness and abundance, are often insufficient for detecting or predicting ecosystem responses to disturbances [13]. In contrast, a community-level approach based on functional traits (the biological or ecological characteristics of organisms, such as body size, diet, or habitat use) offers a more sensitive and informative framework for assessing the impacts of environmental stressors. Functional trait analysis can reveal early signs of ecological change, even when taxonomic composition remains stable, thereby providing a potential early warning system before species loss occurs [13]. This functional trait-based methodology can significantly enhance ecological monitoring, deepen our understanding of disturbance-driven dynamics, and facilitate the development of predictive tools for ecosystem management in increasingly novel and rapidly changing environments [14,15]. Functional diversity encompasses multiple aspects of the community trait structure that influence ecological processes. In this study, we focus on three distinct components: functional specialization, which is the extent to which species occupy unique trait positions; functional richness, which is the range of trait space filled by the community; and functional imbalance, which is the unevenness in trait contributions among dominant species [16,17]. Functional imbalance captures how functionally dissimilar traits are distributed relative to species dominance. High imbalance implies that dominant species have unique traits, while low imbalance suggests redundancy among dominants, potentially affecting community resilience [18]. These dimensions provide complementary insights into how fish assemblages respond to environmental change. While the general responsiveness of fish communities to anthropogenic stressors is well documented [19,20,21], our goal is to quantify how specific trait-based metrics respond to environmental parameters (e.g., eutrophication, hydrological alterations, and climate warming).

In this study, we also evaluate the effectiveness of satellite remote sensing as a tool for monitoring and forecasting spatial patterns of functional diversity in floodplain ecosystems [22,23]. Conventional field-based surveys in such systems are often constrained by limited accessibility, high costs, and low temporal resolution [24]. In contrast, remote sensing offers a scalable and non-invasive alternative for capturing ecological patterns across broad and dynamic landscapes [25,26]. Integrating trait-based metrics with remotely sensed data enables more comprehensive and timely assessments of biodiversity trends under ongoing anthropogenic and climatic pressures.

The aim of this study is to assess how long-term environmental changes, specifically eutrophication, hydrological alteration, and climate warming, have affected the spatiotemporal dynamics of functional diversity in juvenile fish assemblages within a regulated floodplain ecosystem, using a combination of field surveys and satellite-derived environmental indicators.

2. Materials and Methods

2.1. Study Sites

Long-term studies (1997–2015) on the community structure of young-of-the-year fish were conducted in the water bodies of the “Dnipro-Orilskiy” Nature Reserve in Ukraine. The aquatic habitats of the reserve encompass the main channel of the Dnipro River, along with various floodplain water bodies, covering a total area of 203 hectares [27]. The aquatic ecosystems exhibit considerable diversity, influenced by factors such as area, vegetation, bottom structure, and distance from the Dnipro River channel. These ecosystems can be classified into five distinct groups [28]. A detailed description of the aquatic ecosystems of the reserve is provided in Supplement 1. A map showing the study area is given in Figure S1.

2.2. Fish Sampling

Fish sampling was performed annually from 1987 to 2015 in the most representative biotopes of the respective water bodies, specifically within coastal zones, in accordance with the sampling design described in [12,29,30]. The flat bottom topography facilitated the effective use of seine nets [31]. Seine sampling is characterised by low selectivity and ease of handling during field operations [32,33,34]. Fish were collected from the coastal zone using a beach seine measuring 15 m in length and 2 m in height, equipped with a bag and a mesh size of 3 mm. One end of the seine net remained on the shore, while the other was deployed perpendicularly into the water. The net was then hauled back to the shore to enclose the sampling area. Each haul covered a surface area ranging from approximately 50 to 300 m². The water depth at the sampling sites did not exceed 1.7 m. In total, 570 samples were collected throughout the study period in the analysed section of the reservoir. Each sample contained both adult and juvenile fish. However, only juvenile fish were selected for analysis because they reflect recent recruitment and are particularly sensitive to short-term environmental changes. Their trait composition may provide early indicators of ecosystem alteration, especially under eutrophication and hydrological shifts. Fish were classified as juveniles until they developed scales, exhibited adult-like morphology, and reached sexual maturity [35]. Sampling was conducted between 8 July and 1 October, with the majority of sampling occurring between 4 August and 14 August. The results are presented as the number of individuals per 100 m². The functional traits of fish are represented by the following groups: traits characterising fish migratory biology [36,37] and habitat preferences [38]; groups distinguished according to the preference of fish species for water flow velocity in the reservoir [39,40], feeding habitat [41,42], and reproduction habitat [43,44,45]; groups of fish distinguished according to the preferred habitat salinity [27,46,47] and feeding diet [48,49]; life span groups based on the approximate maximum age that fish of a given species could potentially reach [50], morphological traits [51,52], and reproductive traits [53]; and tolerance groups reflecting the sensitivity of species to changes in flow regime, nutrient regime, habitat structure, and water chemistry [39,40]. A detailed description of the fishes’ functional traits is provided in Supplement 2. The information about the fish traits was obtained from the freshwaterecology.info database [54].

2.3. Measurements of Environmental Properties

The satellite-derived measures of the Normalised Difference Vegetation Index (NDVI), Surface Algal Bloom Index (SABI), Chlorophyll-a, suspended particulate matter (SPM), Floating Algal Index (FAI and FAI_RHOT), Normalised Difference Turbidity Index (NDTI), absorption coefficient of coloured dissolved organic matter (aCDOM), light attenuation in the photosynthetically active radiation (PAR) domain (Kd(PAR)), and water turbidity were considered to be predictors of fish abundance. These indices were selected because they capture ecological parameters relevant to fish habitat quality, including algal blooms, turbidity, and primary productivity, which influence oxygen availability, visibility, and food web structure. These environmental indicators were averaged within a 60-metre radius centred on the fish sampling points throughout the study period using remote sensing data [55]. The collection 2 Level 2 products (Landsat 5/TM, Landsat 7/ETM, and Landsat 8 OLI) were retrieved from the USGS website by using the EarthExplorer tool [56]. Data from 1997 to 2011 were obtained using Landsat 5/TM, data for 2012 were obtained using Landsat 7/ETM, and data from 2013 to 2015 were obtained using Landsat 8 OLI. The geometrically corrected and radiometrically calibrated images used for the analysis were atmospherically corrected to surface reflectance using the Dark Spectrum Fitting (DSF) algorithm in ACOLITE (L2R products) [57]. Only scenes with minimal cloud cover acquired in the second half of summer or early autumn were selected [57]. The range of image dates was 167 to 266 days from the beginning of the year.

The Normalised Difference Vegetation Index (NDVI) and Floating Algae Index (FAI) were calculated using surface reflectance and top-of-atmosphere reflectance products generated in ACOLITE. The NDVI was computed using atmospherically corrected surface reflectance (rhos) according to the standard formula: NDVI = (ρ_NIR − ρ_Red)/(ρ_NIR + ρ_Red), where ρ_NIR and ρ_Red correspond to the reflectance values in the near-infrared and red bands, respectively. For Sentinel-2 imagery, these are typically Bands 8 (842 nm) and 4 (665 nm), and for Landsat 8, they are typically Bands 5 (865 nm) and 4 (655 nm).

The FAI (specifically FAI_RHOT) was calculated from Rayleigh-corrected top-of-atmosphere reflectance (rhot) values. The index enhances detection of floating algae by removing background water reflectance through linear baseline correction between the red and shortwave infrared (SWIR) bands. It is calculated as follows in Equation (1):

FAI_RHOT = ρ_NIR − [ρ_Red + (ρ_SWIR − ρ_Red) × (λ_NIR − λ_Red)/(λ_SWIR − λ_Red)]

(1)

where λ denotes the central wavelength of each band, and ρ represents Rayleigh-corrected reflectance values. All calculations were performed within the ACOLITE processor using the Dark Spectrum Fitting (DSF) atmospheric correction and exported from L2R output files. The bands used for FAI_RHOT calculations in Sentinel-2 are B4 (665 nm), B6 (740 nm), and B11 (1610 nm), and for Landsat 8, they are Bands 4 (655 nm), 5 (865 nm), and 6 (1610 nm), based on the spectral configuration of each sensor.

The Surface Algal Bloom Index (SABI) was calculated as follows: SABI = (ρNIR − ρRed)/(ρGreen + ρBlue), where ρNIR represents surface reflectance in the near-infrared band, ρRed in the red band, ρGreen in the green band, and ρBlue in the blue band. For Landsat 5 TM and Landsat 7 ETM+, these correspond to Bands 4 (NIR), 3 (Red), 2 (Green), and 1 (Blue), respectively. For Landsat 8 OLI, the corresponding bands are Band 5 (NIR), Band 4 (Red), Band 3 (Green), and Band 2 (Blue). This index was developed to detect surface algal blooms, especially floating aggregations, based on the spectral behaviour of algae, which exhibit high reflectance in the NIR region and low reflectance in the visible spectrum. SABI normalises these spectral differences to enhance bloom detection and minimise atmospheric and glint-related noise [58]. Chlorophyll-a concentrations were estimated using a regression model. The square root of the chlorophyll-a concentration was calculated using the following equation:

(Chl-a)1/2 = 4.778 − 45.5·d₄ − 123.03·d₅ − 26.33·ρ(B/λ_B)

(2)

where d₄ = (ρ_SWIR1 − ρ_NIR)/(λ_SWIR1 − λ_NIR), d₅ = (ρ_SWIR2 − ρ_SWIR1)/(λ_SWIR2 − λ_SWIR1), and λ_B, λ_SWIR1, and λ_SWIR2 represent the central wavelengths of the corresponding blue, SWIR1, and SWIR2 bands, respectively. This model utilises spectral derivatives and reflectance values to enhance sensitivity to chlorophyll concentrations, especially in turbid inland waters [59].

Suspended particulate matter (SPM) concentrations were estimated using sensor-specific exponential models based on surface reflectance values in the red, near-infrared (NIR), and green spectral bands. For Landsat 5 TM and Landsat 7 ETM+, the following Equation (3) was applied:

SPM_TM = 1.663 × exp [2.906 × (ρ_red + ρ_NIR)/ρ_green].

(3)

For Landsat 8 OLI, Equation (4) was used:

SPM_OLI = 2.016 × exp [2.993 × (ρ_red + ρ_NIR)/ρ_green],

(4)

where ρ_red, ρ_NIR, and ρ_green represent the surface reflectance values in the red, near infrared, and green bands, respectively. These equations were derived empirically and validated for use in estimating SPM concentrations in optically complex inland waters [60].

The Floating Algae Index (FAI) was used to detect and differentiate surface cyanobacterial blooms in inland waters based on Landsat TM and ETM+ imagery. The FAI is computed using the following linear baseline algorithm applied to top-of-atmosphere reflectance values from the red (B3), near-infrared (NIR; B4), and shortwave infrared (SWIR; B5) bands [61] (5):

FAI = R_rc,B4 − [(R_rc,B3 + (R_rc,B5 − R_rc,B3) × (λ_B4 − λ_B3)/(λ_B5 − λ_B3))]

(5)

where R_rc,Bi is the Rayleigh-corrected reflectance in band i, and λ_B3, λ_B4, and λ_B5 are the central wavelengths of Bands 3 (red), 4 (NIR), and 5 (SWIR), respectively. This index highlights floating algae by enhancing the NIR reflectance signal relative to a baseline interpolated between red and SWIR bands, thus reducing background effects from water and atmosphere. The threshold used for lake water extraction was FAI > 0.05, as established and validated in Oyama et al. [61].

The Normalised Difference Turbidity Index (NDTI) is a spectral index used to estimate water turbidity using satellite imagery. It is calculated using surface reflectance values from the red and green bands, according to Formula (6):

NDTI = (ρ_red − ρ_green)/(ρ_red + ρ_green),

(6)

where ρ_red and ρ_green represent the surface reflectance in the red and green spectral bands, respectively. This index enhances the contrast between turbid and clear water, as turbid water typically shows higher reflectance in the red band compared to the green band. The NDTI has been widely applied for monitoring turbidity in inland and coastal waters [62].

The absorption coefficient of coloured dissolved organic matter (aCDOM) was estimated using an empirical model. Reflectance values at Bands 1 (blue, 0.45–0.52 μm) and 4 (NIR, 0.76–0.90 μm) were used to calculate a band ratio (B4/B1), which showed the highest correlation with in situ aCDOM measurements. The best-performing model was a simple linear regression, as follows in Equation (7):

aCDOM(485) = −0.5986 + 5.5510 × (B4/B1),

(7)

where B4 and B1 are reflectance values from TM Bands 4 and 1, respectively. The model was applied to atmospherically corrected and radiometrically normalised imagery to retrieve spatial and temporal patterns of CDOM in the reservoir [63].

The diffuse attenuation coefficient for photosynthetically active radiation (Kd(PAR)) was estimated using Landsat surface reflectance data through an empirical regression model developed by Song et al. [64]. The model was based on atmospherically corrected reflectance in the red and blue bands, using their difference as the main predictor. The final regression equation was Kd(PAR) = −199.75 × (Red − Blue) + 11.702, where Red and Blue represent the surface reflectance values in the corresponding Landsat bands. The model demonstrated a strong relationship with in situ measurements, achieving a coefficient of determination (R²) of 0.821 and a root mean square error (RMSE) of 0.994 m⁻¹, based on 158 water samples. The method proved suitable for regional-scale assessment of underwater light availability and water clarity in optically complex inland waters [64].

Turbidity was estimated using a multiple regression model based on spectral derivatives derived from surface reflectance. The model was selected based on its statistical performance (R² = 0.60, RMSE = 1.54 NTU) and had the following form: SQRT(Turb) = 4.67 + 9.86·d₂ + 41.70·d₂₄, where d₂ = (ρ_R − ρ_G)/(λ_R − λ_G) and d₂₄ = (d₅ − d₄)/(λ_SWIR2 − λ_NIR), with d₄ = (ρ_SWIR1 − ρ_NIR)/(λ_SWIR1 − λ_NIR) and d₅ = (ρ_SWIR2 − ρ_SWIR1)/(λ_SWIR2 − λ_SWIR1). In these expressions, ρ represents the surface reflectance in the respective spectral band, and λ denotes the central wavelength of that band. This model effectively captures the influence of suspended particulate matter on turbidity in reservoirs located within agricultural watersheds [59].

The results of ecological property assessments of water bodies based on remote sensing data were adjusted relative to the date of the annual maximum temperature to ensure data comparability. The adjustment was performed according to the procedure described in [30]. Data on average daily atmospheric air temperature in the city of Dnipro were obtained from NOAA Weather Data [65].

2.4. Statistical Analysis

A principal component analysis (PCA) was applied to explore the multivariate structure of both environmental variables and functional diversity indices. For spectral indices derived from satellite data, the PCA was performed using the correlation matrix to extract orthogonal axes summarising the main gradients in water quality variability. Components with eigenvalues greater than one were retained for interpretation. These principal components (PC1–PC3) were then used to assess spatiotemporal trends and their relationship with environmental factors. In the analysis of functional diversity, PCA was conducted on the residuals of diversity indices after regressing out the effect of species richness to eliminate collinearity. Five principal components (PC1–PC5) with eigenvalues greater than one were extracted and interpreted based on variable loadings. These components captured distinct dimensions of functional specialisation, imbalance, richness, and trait structure. Scores of species and environmental predictors were projected into the principal component space to facilitate the interpretation of temporal and spatial dynamics. The contribution of functional traits to the derived components was evaluated using correlation analysis. All PCA computations were conducted in R (Version 1.2.0.) using the prcomp functions, with standardisation (z-scores) applied before analysis.

Species diversity was quantified using the Gini–Simpson index, which reflects both the richness and evenness of species in each fish assemblage. Functional diversity was assessed using a comprehensive set of multidimensional indices that describe various aspects of trait composition and distribution. These included functional richness (FRic), which quantifies the volume of trait space occupied by the community; functional evenness (FEve), which describes the regularity of species abundance distribution in the functional space; functional divergence (FDiv), which indicates the extent to which species with extreme trait values dominate [66]; and functional dispersion (FDis), which measures the mean distance of species to the community-weighted trait centroid. Rao’s quadratic entropy was used as a combined measure of trait dissimilarity and species abundance. In addition, more specialised indices were calculated to capture functional distinctiveness and redundancy: functional specialisation (mean distance from the trait centroid), functional originality (mean trait distance to the nearest neighbour), and mean nearest neighbour distance as a proxy for functional redundancy. Further, multidimensional trait-based indices (Fide 1 to Fide 5) were used to capture internal trait dissimilarity structure. Functional imbalance was assessed using three indices: the correlation-based index CorB, the standardised effect size SES B, and QB, a normalised measure of trait asymmetry. Trait data were compiled from the freshwaterecology.info database and primary literature sources, encompassing categorical, binary, and continuous variables. Gower dissimilarity was used to construct species-by-trait distance matrices. All diversity indices were computed in R using the FD, adiv, and mFD packages. To remove the confounding effect of species richness, each functional index was regressed against species richness, and the residuals of these regressions were used in the further analyses. This step allows us to isolate functional trait structure from species richness, which can otherwise dominate variation in functional diversity metrics. By controlling for richness, we focus on how trait distribution and functional roles shift regardless of species count.

Relationships between species traits and the principal component scores derived from the analysis of functional and species diversity indices were assessed using fourth-corner analysis [67]. This method evaluates the significance and direction of the association between environmental gradients or synthetic ordination axes (in this case, PCA scores) and biological traits via species composition as the linking matrix. The analysis was implemented using the fourth-corner function from the ade4 package in R, specifying model type 6 and no p-value adjustment (p.adjust.method.G = “none”, p.adjust.method.D = “none”). The number of permutations was set to nrepet to assess the statistical significance of trait–gradients relationships. This approach enabled the identification of functional traits most strongly associated with variation in community-level diversity structure.

Generalised linear models (GLMs) were used to quantify the effects of temporal dynamics, spatial heterogeneity, and environmental gradients on the principal components extracted from the functional diversity indices. The dependent variables were the scores of principal components (PC1–PC5), which captured the main axes of variation in functional structure across fish assemblages. Each component was modelled separately.

The models included the year of sampling (time, which was treated as a continuous temporal variable), location (a categorical spatial factor), and their interaction (time × location) to capture spatially structured temporal trends. Additionally, environmental predictors (ENV1, ENV2, and ENV3), derived from a principal component analysis (PCA) of water quality variables, were incorporated to assess the influence of abiotic conditions. Because the same sites were monitored annually, we modelled site-level variation through a fixed site × time interaction rather than as a random effect. Model 8 has the following structure:

PCi ~ Time + Location + Time × Location + ENV1 + ENV2 + ENV3,

(8)

where PCi denotes the scores of the i-th principal component representing functional diversity; Time refers to the sequence of years from 1997 to 2015; Location is a categorical variable indicating the sampling site, with one site used as the reference level; and ENV1, ENV2, and ENV3 are the scores of environmental gradients obtained via PCA from spectral indices. A Gaussian error distribution was assumed for normally distributed scores. Model performance was evaluated using F-tests, and the significance of individual predictors was assessed using type III sums of squares. All models were fitted in R using the glm() function, with residual diagnostics performed to confirm model assumptions.

Spatial models were constructed in ArcGIS [68] to explore and visualise the geographic patterns of variation in ecological indices and principal component scores derived from species and functional diversity metrics. The sampling points were georeferenced using the WGS 84 coordinate system, and all spatial operations were performed in ArcGIS Pro 3.1.

3. Results

The species composition and functional traits of fish larvae have been comprehensively documented in our previous studies [12,30], which provide the basis for the results presented below.

3.1. Dynamics of Water Quality Indicators

The principal component analysis of spectral indices facilitated the extraction of three principal components with eigenvalues exceeding one (

Table 1. Principal component analysis of spectral index variability (only statistically significant loads for p < 0.001 are presented).

Variable	PC 1, λ = 6.0, Explained Variance = 54.4%	PC 2, λ = 1.4, Explained Variance = 12.7%	PC 3, λ = 1.1, Explained Variance = 9.2%
NDVI	0.96	−	−
FAI	0.93	−	−
SABI	0.94	−	−
Chl-a	0.94	−	−
SPM	−0.74	−	−
FAI	−0.93	−	0.24
NDTI	−	0.51	−
CDOM	0.89	−	−
Kd(PAR)	0.40	0.72	−
TURBIDITY	−	0.58	0.77
Maximal temperature	−	−0.47	0.50

).

The first principal component accounted for 54.4% of the total variance and primarily represented the variability of indicators characterising aquatic environmental properties influenced by algal blooms. Negative scores along this component indicated higher bloom intensity. Year, location, and their interaction explained 70% of the variability in PC1 (F = 44.7, p < 0.001). The variable “Year” had a statistically significant beta regression coefficient of 0.075 ± 0.023 (t = 3.23, p = 0.001), indicating an overall increasing trend in bloom intensity during the study period. This general trend exhibited spatial variation. At two sites, the local trend in bloom intensity aligned with the overall trend; at 16 sites, the increase in bloom intensity over time was slower than the general trend; and at 12 sites, the increase was more rapid. In the main channel of the Dnipro River, the trend in bloom intensity during the study period was either neutral or decelerating, whereas in the floodplain ecosystems, the trend was accelerating (Figure 1).

The second principal component (PC2) accounted for 12.7% of the total variance and was most sensitive to light attenuation in the photosynthetically active radiation (PAR) domain, turbidity, and the Normalised Difference Turbidity Index (NDTI), all of which were negatively correlated with the annual maximum temperature. Therefore, PC2 can be interpreted as reflecting variability in turbidity that is largely driven by factors also responsible for changes in light attenuation within the PAR domain. Positive scores on this component indicate increased water turbidity. Year, location, and their interaction explained 19% of the variability in PC2 (F = 5.5, p < 0.001). The variable “Year” had a statistically significant beta regression coefficient of −0.323 ± 0.037 (t = −8.88, p < 0.001), indicating a decreasing trend in water turbidity over time, which is associated with light attenuation in the PAR domain. At 20 sites, the temporal trend in PC2 did not differ from the overall trend; at 4 sites, the decline was more pronounced than the general trend; and at 6 sites, the decline was less pronounced than the general trend.

The third principal component (PC3) accounted for 9.2% of the total variance. This component was also sensitive to water turbidity and the Floating Algae Index (FAI). PC3 increased with rising temperatures, and its positive scores indicated higher water turbidity. This component can be interpreted as representing a distinct aspect of turbidity variability, primarily driven by the intensity of algal development in the water. Year, location, and their interaction explained 13% of the variability in PC3 (F = 3.9, p < 0.001). The variable “Year” exhibited a statistically significant beta regression coefficient of 0.117 ± 0.038 (t = 3.05, p = 0.002), indicating an increasing trend in water turbidity over time, associated with the Floating Algae Index. At 22 sites, the temporal trend in PC3 did not differ from the overall trend; at 4 sites, the increase was slower than the general trend; and at 4 sites, the increase was more rapid than the general trend. The greatest deviations from the general trend, both in terms of acceleration and deceleration, were observed in floodplain ecosystems.

This study used a principal component analysis (PCA) of spectral indices to identify three major dimensions of variability in aquatic environmental conditions, primarily related to algal blooms and water turbidity. The first component reflected changes in bloom intensity and showed an overall increasing trend over time, especially in floodplain ecosystems. The second component captured turbidity linked to light attenuation, showing a general decline over time. The third component represented turbidity associated with floating algae, also showing an increasing trend. These patterns varied across locations, highlighting spatial differences in environmental changes over the study period.

3.2. Correlation Between Taxonomic and Functional Diversity Indices and Species Richness

Species richness accounted for between 1% and 67% of the variation in the Gini–Simpson index and the functional diversity indices (

Table 2. Principal component analysis of the variability in species, functional diversity indices, and identity score factors (correlation coefficients statistically significant at p < 0.001 are presented). The largest (by absolute value) loadings on the principal components are highlighted in bold.

Diversity Index	Parameters of Linear Regression Analysis of Functional Diversity Indices Regarding Species Richness			Principal Components
	R2	K (Regression Slope)	p-Value	PC1, λ = 4.1, 24.1%	PC2, λ = 3.2, 19.0%	PC3, λ = 2.3, 13.6%	PC4, λ = 1.7, 10.0%	PC5, λ = 1.4, 8.3%
Species diversity indices
Species richness	−	−	−	−	−	−	−	−
Gini–Simpson (Gsimpson)	0.36	0.11	<0.001	0.39	−	−	−0.36	0.15
Functional diversity indices
Functional mean pairwise distance (Fmpd)	0.01	0.01	0.050	−0.31	−	−0.30	−0.20	0.15
Functional divergence (Fdiv)	0.02	0.03	0.001	−	0.22	−	0.34	0.44
Functional redundancy (FunRedundancy)	0.05	0.01	<0.001	−	−0.31	−	−	0.24
Functional specialization (Fspe)	0.01	−0.01	0.018	−0.40	−	−	−0.28	0.26
Functional dispersion (Fdis)	0.16	0.08	<0.001	−	0.33	−	−0.29	0.32
Rao’s quadratic entropy (FunRao)	0.38	0.10	<0.001	0.37	0.18	−	−0.35	−
B correlation coefficient (CorB)	0.01	0.03	0.074	−	0.35	0.28	−	−
Standardized effect size (SESB)	0.00	−0.16	0.093	−	0.39	−	−	−0.15
Normalized Q (QB)	0.01	−0.04	0.041	−	0.35	−	−	−0.16
Functional originality (Fori)	0.01	0.02	0.022	0.27	0.21	−0.16	−	−0.28
Functional mean nearest neighbour distance (Fnnd)	0.10	−0.08	<0.001	−	0.26	−0.17	−	−0.38
Functional evenness (Feve)	0.00	−0.01	0.450	0.13	−	−	−	0.16
Functional richness (Fric)	0.67	0.29	<0.001	−	−	−0.51	−0.38	−0.22
Functional identity axes
Fide 1	0.01	−0.01	0.083	−	−0.26	0.31	−0.29	−0.20
Fide 2	0.07	0.04	<0.001	0.24	0.22	−	0.35	−
Fide 3	0.11	0.11	<0.001	0.37	−	−0.24	−	−
Fide 4	0.00	0.00	0.733	−	0.17	0.40	−	−
Fide 5	0.01	0.01	0.069	−	−	0.41	−0.17	−0.31

). Statistically significant correlations with species richness were not observed for functional evenness in Fide 1, Fide 4, and Fide 5. In most instances, the correlation between diversity indices and species richness was positive. However, exceptions included functional specialisation and functional mean nearest neighbour distance, which exhibited negative correlations with species richness. To isolate the component of diversity indices attributable to species richness, a linear regression analysis was conducted. The residuals from these regressions, which represent the variation in diversity indices independent of species richness, were utilised in subsequent analyses.

3.3. Principal Component Analysis of Diversity Indices

A principal component analysis identified five components with eigenvalues greater than one (Table 2).

The sixth principal component had an eigenvalue equal to one and a loading of one from species richness, while all other variables exhibited zero loadings on this component. Conversely, species richness had zero loadings on principal components 1 through 5. This indicates that the variability explained by species richness is completely independent of the other variables and is solely represented by principal component 6, which exclusively reflects species richness. Principal component 1 is primarily influenced by functional specialisation. Its correlation with species relative abundance is highlighted by the significant loading of the Gini–Simpson index. The substantial loadings of functional specialisation, functional originality, and Rao’s quadratic entropy further reinforce the interpretation of PC1 as a comprehensive measure of community functional specialisation. Principal component 2 was significantly influenced by the B correlation coefficient, the standardised effect size, and the normalised Q index, enabling it to be interpreted as a comprehensive indicator of functional imbalance. Functional richness exhibited the highest loadings on principal components 3 (PC3) and 4 (PC4). PC3 can be interpreted as an indicator of “pure” functional richness. In contrast, PC4 displayed high loadings from multiple indices, complicating its interpretation. However, its sensitivity to species relative abundance is indicated by the significant loading from the Gini–Simpson index. Furthermore, strong contributions from functional specialisation, functional dispersion, and Rao’s quadratic entropy suggest that PC4 reflects a dimension of functional richness variation associated with shifts in species specialisation. PC5 exhibited an unusual pattern, characterised by opposing signs for the loadings of functional mean nearest neighbour distance and functional divergence. When these indices share the same sign, as observed in PC2, the component can be interpreted as functional divergence. In the case of PC5, however, the most appropriate interpretation is that of functional packing density.

3.4. Interpretation of Principal Components About Functional Traits

PC1 (functional specialisation) exhibited a negative correlation with the proportion of non-migratory and limnophilic fish species within the assemblage (

Table 3. Relationships between all traits and the scores principal components extracted after analysis of the functional and species diversity indices. Pearson r-correlation coefficients and standard deviations were derived after the fourth-corner analysis. Only those statistically significant at p < 0.05 are presented.

Trait *		Principal Components Extracted Based on the Analysis of Variability in Functional Indices					Species Richness
Category	Level	PC1	PC2	PC3	PC4	PC5
M	dia	−	−	−	0.07 ± 0.03	−	0.09 ± 0.05
	nom	−0.25 ± 0.11	−	−	−	−	−
	pot	−	−	−	−	−	−
H	bpl	−	−	−	−	−	−
	dem	−	−	−	−	−	−
	pel	−	−	−	−0.06 ± 0.03	−	−
R	eur	−	−	−	−	−	−
	lim	−0.19 ± 0.11	−	0.27 ± 0.10	−	−0.12 ± 0.06	0.09 ± 0.05
	rhe	−	−	−0.37 ± 0.02	−	0.13 ± 0.06	−
FH	ben	−	−	−	−	−	−
	wat	−	−	−	−	−	−
RH	lit	−	−	−0.25 ± 0.10	−	0.12 ± 0.06	−
	oth	−	−	−	−	−	−
	phy	−	−	−	−	−	−
	pli	−	−	−	−	−	−
S	fbm	−	−	−	−	−	−
	fbr	−	−	−	−	0.13 ± 0.06	−
	fre	−	−	−	−	−	−
FD	car	−	−	−	−	−	−
	inv	−	−	−	−	−	0.11 ± 0.05
	omn	−	−	−	−	−	−0.10 ± 0.05
	pis	−	−	−	−	−	−
LS	ls1	−	−	−	0.06 ± 0.03	−	−
	ls2	−	−	−	−	−	−
	ls3	−	−	−	−0.07 ± 0.03	−	−
BL	bl1	0.20 ± 0.11	−	−	0.08 ± 0.01	−	0.10 ± 0.05
	bl2	−0.19 ± 0.11	−	−	−0.07 ± 0.01	−	−0.10 ± 0.05
	bl3	−	−	−	−	−	−
SH	sh1	−	−0.14 ± 0.07	−	−	−	−
	sh2	−	−	−	−0.11 ± 0.03	−	−
	sh3	−	−	−	−	−	−
	sh4	−	−	−	−	−	−
SF	sw1	−	−	−	−	−	−
	sw2	−	−	−	−	−	−
	sw3	−	0.16 ± 0.06	−	−	−	−
FM	ma1	−	−	−	−	−	−
	ma2	−	−	−	−	−	−
	ma3	−	−	−	−	−	−
	ma4	−	−	−	−	−	−
	ma5	−0.31 ± 0.10	−	−	−	−	−0.13 ± 0.05
ST	st1	−	−	−	−	−	−
	st2	−	−	−	−	0.11 ± 0.06	−
	st3	−	−0.15 ± 0.06	−	−	−0.13 ± 0.06	−
Ip	ip1	−	−	−	−	−	−
	ip2	−	−	−	−	−	−
	ip3	−	−	−	−	−	−
Fc	fe1	−	−	−	−	−	−
	fe2	−	−	−0.25 ± 0.09	−	−	−
	fe3	−	−	−	−	−	−
rFc	fr1	−	−	−	−	−0.16 ± 0.06	−
	fr2	−	−	−	−	−	−
	fr3	−	−	−	−	0.12 ± 0.06	−
Eg	ed1	−	−	−	−	−	−
	ed2	0.25 ± 0.01	−	−	−	−	−
	ed3	−0.30 ± 0.02	0.15 ± 0.06	−	−	−	−0.10 ± 0.05
LL	ll1	−	−	−	−	−	−
	ll2	−	−	−	−	−	−
	ll3	−0.18 ± 0.11	−	−	−	−	−
PC	nnh	−0.24 ± 0.10	−	−	−	−	−0.13 ± 0.05
	nop	−	−	−	−	−	−
	pnh	−	−	−	−	−	0.10 ± 0.05
LD	ld1	−	−	−	−	−	−
	ld2	−	−	−	−	−	−
	ld3	−	−	−	0.06 ± 0.03	−	−
HD	im	−	−	−	−	−	−
	intol	−	−	−0.19 ± 0.10	−	−	−
	tol	−	−	−	−	−	−
QT	im	−	−	−	−	−	−
	intol	−0.30 ± 0.02	0.15 ± 0.06	−	−	−	−0.10 ± 0.05
	tol	−	−0.16 ± 0.06	−	−	−	−
OT	im	0.20 ± 0.11	−	−	−	−	0.10 ± 0.05
	intol	−	−	−	−	−	−
	tol	−0.20 ± 0.11	−	−	−	−	−0.10 ± 0.05
TT	im	−	−	−	−	−	−
	intol	−	−	−	−	−	−
	tol	−	−	−	−	−	−
TrL		−	−	0.23 ± 0.10	−	−	−
Res		−	−0.13 ± 0.06	−	−0.07 ± 0.03	−	−
Vuln		−	−	−	−0.08 ± 0.01	−	−

* Legend to trait abbreviations. M is the migration type; H is the habitat preference; R is the rheophily; FH is the feeding habitat; RH is the reproduction habitat; S is the salinity tolerance; FD is the feeding diet; LS is the life span; BL is the body length; SH is the shape factor (body length/body depth); SF is the swimming factor (caudal peduncle length/depth); FM is the female maturity (age at first reproduction); ST is the spawning time; Ip is the incubation period; Fc is the absolute fecundity (number of oocytes); rFc is the relative fecundity (oocytes per 100 g); Eg is the egg diameter; LL is the larval length at hatching; LD is the larval duration; PC is the parental care; HD, QT, OT, and TT refer to tolerance traits related to habitat degradation, water quality, oxygen availability, and toxicity, respectively; Res is the resilience level; and Vuln is the vulnerability to disturbance. For a detailed description of the traits, see Supplement 2.

).

As the PC1 scores increased, the proportion of fish with an adult body length of less than 20 cm also increased, while the proportion of fish with adult body lengths ranging from 20 to 39 cm decreased. Additionally, PC1 was negatively correlated with the proportion of fish whose female maturity age typically exceeds 5 years. It demonstrated a positive correlation with the proportion of fish possessing egg diameters between 1.35 and 2 mm, and a negative correlation with the proportion of fish with egg diameters greater than 2 mm. Furthermore, PC1 was negatively correlated with fish whose larval length at hatching exceeded 6 mm. An increase in PC1 was also associated with a decrease in species that exhibit no protective strategies, such as nest or egg hiding. Lastly, PC1 showed a negative correlation with species intolerant of poor water quality and with species tolerant of low oxygen availability, while demonstrating a positive correlation with species that are intermediately tolerant of oxygen availability.

PC2 (functional imbalance) exhibited a positive correlation with species that are intolerant of poor water quality and a negative correlation with species that are tolerant of such conditions. Additionally, this component was negatively correlated with resilience levels. PC2 was sensitive to body shape, as evidenced by its negative correlation with the proportion of fish exhibiting a shape factor of less than 4.35, which corresponds to relatively deep-bodied, short fish. It was also influenced by swimming ability, demonstrating a positive correlation with fish whose swimming factor exceeded 0.43. These species possess narrow, elongated caudal peduncles, adaptations that facilitate fast, energy-efficient swimming over long distances, as well as an active lifestyle in flowing waters or open habitats. Furthermore, PC2 displayed a negative correlation with species that spawn during the summer.

PC3 (functional richness) exhibited a positive correlation with limnophilic fish species and a negative correlation with rheophilic and lithophilic species. It also demonstrated a negative correlation with fish species that produce between 5001 and 30,000 oocytes, as well as with species that are intolerant of habitat degradation. Furthermore, PC3 was positively correlated with the trophic levels of fish.

PC4 (changes in functional richness driven by shifts in species specialization) exhibited a positive correlation with diadromous fish species and a negative correlation with pelagic species. This component was positively associated with fish that typically have a lifespan of less than 8 years and a body length under 20 cm, while it was negatively correlated with fish whose lifespan exceeds 15 years and body length is greater than 40 cm. PC4 was characterised by a negative correlation with fish possessing a shape factor between 4.36 and 4.78. Additionally, it showed a positive correlation with fish whose larval duration exceeds 25 days. This component was negatively correlated with resilience levels and vulnerability to disturbance.

PC5 (functional packing density) exhibited a negative correlation with limnophilic fish species and a positive correlation with rheophilic and lithophilic species. Additionally, it demonstrated a positive correlation with freshwater–brackish water fish species. PC5 was positively correlated with fish that spawn in spring and negatively correlated with those that spawn in summer. Furthermore, it showed a negative correlation with species whose relative fecundity is less than 600 oocytes per 100 g of body weight, while positively correlating with species whose relative fecundity exceeds 1200 oocytes per 100 g.

Changes in species richness were accompanied by shifts in the functional structure of fish assemblages. Increases in species richness were associated with a higher abundance of diadromous, invertivorous, and limnophilic species, as well as a lower abundance of omnivorous species. Additionally, an increase in the number of larger-bodied fish was correlated with greater species richness. However, species with a female maturity age exceeding 5 years and an egg diameter greater than 2 mm were linked to lower species richness. A higher abundance of fish lacking protective nesting or egg hiding strategies was associated with reduced species richness, whereas species employing protective strategies exhibited a positive correlation with species richness. Furthermore, species richness was sensitive to the tolerance of fish species to water quality and oxygen availability.

3.5. Temporal Trends and the Influence of Environmental Factors on Functional Diversity

PC1 (functional specialisation) exhibited a decreasing trend throughout the study period (

Table 4. GLM of the effect of time, interaction of location and time, and environmental factors on the principal components of functional diversity variation (N = 589; only statistically significant β-coefficients at p < 0.05 are displayed).

Effect	Sum of Squares	Degrees of Freedom	Mean Square	F-Statistic	p-Value	β-Coefficient ± Std. Error
F1 (Radj2 = 0.54, F = 21.2, p < 0.001)
Intercept	43.4	1	43.4	23.3	<0.001	−
Time	43.3	1	43.3	23.3	<0.001	−0.14 ± 0.03
N*Time	1073.8	30	35.8	19.3	<0.001	− *
ENV1	40.2	1	40.2	21.6	<0.001	−0.24 ± 0.05
ENV2	6.2	1	6.2	3.3	0.07	−
ENV3	2.6	1	2.6	1.4	0.24	−
Error	1030.0	554	1.9	−	−	−
F2 (Radj2 = 0.37, F = 11.2, p < 0.001)
Intercept	212.0	1	212.0	105.8	<0.001	−
Time	212.0	1	212.0	105.8	<0.001	−0.36 ± 0.04
N*Time	503.8	30	16.8	8.4	<0.001	− *
ENV1	16.2	1	16.2	8.1	0.005	−0.16 ± 0.06
ENV2	0.3	1	0.3	0.2	0.69	−
ENV3	0.1	1	0.1	0.1	0.82	−
Error	1110.3	554	2.0	−	−	−
F3 (Radj2 = 0.68, F = 37.6, p < 0.001)
Intercept	8.9	1	8.9	12.1	0.001	−
Time	8.9	1	8.9	12.1	0.001	0.08 ± 0.02
N*Time	744.9	30	24.8	33.5	<0.001	− *
ENV1	0.4	1	0.4	0.5	0.46	−
ENV2	0.0	1	0.0	0.0	0.97	−
ENV3	2.0	1	2.0	2.7	0.10	−
Error	410.0	554	0.7	−	−	−
F4 (Radj2 = 0.32, F = 9.0, p < 0.001)
Intercept	9.0	1	9.0	7.8	0.005	−
Time	9.0	1	9.0	7.8	0.005	−0.10 ± 0.03
N*Time	326.9	30	10.9	9.4	<0.001	− *
ENV1	23.7	1	23.7	20.5	<0.001	0.28 ± 0.06
ENV2	11.0	1	11.0	9.5	0.002	0.12 ± 0.04
ENV3	0.3	1	0.3	0.2	0.618	−
Error	640.9	554	1.2	−	−	−
F5 (Radj2 = 0.47, F = 16.2, p < 0.001)
Intercept	15.2	1	15.2	20.5	<0.001	−
Time	15.2	1	15.2	20.5	<0.001	0.14 ± 0.03
N*Time	333.2	30	11.1	15.0	<0.001	− *
ENV1	3.4	1	3.4	4.6	0.033	−0.12 ± 0.06
ENV2	19.0	1	19.0	25.6	<0.001	0.17 ± 0.05
ENV3	6.0	1	6.0	8.1	0.005	−0.10 ± 0.03
Error	411.5	554	0.7	−	−	−
Species richness (Radj2 = 0.47, F = 16.2, p < 0.001)
Intercept	290.3	1	290.3	94.4	<0.001	−
Time	324.5	1	324.5	105.5	<0.001	0.33 ± 0.03
N*Time	1111.6	30	37.1	12.1	<0.001	− *
ENV1	18.0	1	18.0	5.9	0.016	−0.13 ± 0.05
ENV2	26.3	1	26.3	8.6	0.004	0.10 ± 0.03
ENV3	15.8	1	15.8	5.1	0.024	0.08 ± 0.03
Error	1703.5	554	3.1	−	−	−

Note: * —shown in Figure 2.

). A more pronounced decline was noted in the eastern and western water bodies of the floodplain lake system (Figure 2). In contrast, the central part of the floodplain system displayed a slower decline compared to the overall trend. ENV1 had a negative impact on PC1. Similarly, PC2 (functional imbalance) also demonstrated a negative temporal trend and a detrimental influence from ENV1. A distinctive characteristic of this principal component was the spatial variability of the temporal trend. An intensification of the temporal trend was observed in areas with higher flow velocities in the channels connecting floodplain lakes or within the main channel of the Dnipro River.

PC3 (functional richness) exhibited a positive temporal trend. In the Dnipro River channel, local temporal trends were significantly slower than those observed in the floodplain water bodies. Environmental factors did not influence this component. Conversely, PC4 (changes in functional richness driven by shifts in species specialization) displayed a negative temporal trend, with the most pronounced decline occurring in the southern and eastern regions of the Dnipro River channel within the study area. ENV1 positively influenced PC4, while ENV2 had a negative impact on this component. PC5 (functional packing density) also demonstrated a positive temporal trend and showed a statistically significant relationship with ENV1, ENV2, and ENV3. Throughout the study, species richness increased, with a more pronounced temporal trend noted primarily in the eastern part of the study area. Species richness was influenced by ENV1, ENV2, and ENV3.

4. Discussion

Functional transformations of communities are indicative of ecological novelty, where ecosystems are widely driven beyond historical baselines due to unprecedented environmental conditions and species reassemblies [1]. Notably, Dnipro floodplain fish community dynamics unfold against a general backdrop of accelerating climate change and nutrient enrichment across European floodplains. Global warming in Europe is accelerating and occurring at a rate that exceeds the global average [69]. Warming intensifies the detrimental effects of other human activities that degrade the living conditions of freshwater ecosystems [70]. Climate change is accompanied by a series of ecological transformations in freshwater ecosystems, initiated by the construction of reservoirs on major European rivers [71]. The building of dams and the creation of reservoirs turned the lotic environment of these rivers into a lentic environment [72]. Monitoring the condition of aquatic ecosystems through Earth remote sensing data has revealed three primary trends in water quality variability. The foremost trend is eutrophication, which has been consistently increasing over the study period. The most significant intensification of eutrophication was observed in floodplain ecosystems. The opposite trend was observed in various aspects of water turbidity. Water turbidity, which affects the attenuation of light in the photosynthetically active radiation (PAR) range (400–700 nm) within the water column, decreased during the study period, while turbidity caused by algal growth increased. The paradoxical dynamics of turbidity in the Dnipro River can be attributed to different sources and mechanisms of the particles that contribute to it. It is likely that under the influence of warming and associated anthropogenic changes, the contribution of inorganic detritus to the total turbidity of the reservoir diminished, while the organic (planktonic) fraction increased. The construction of a cascade of reservoirs and regular purification discharges has reduced seasonal surges of sediment flow from the basin, resulting in the river receiving fewer finely suspended clay and sandy particles [73]. The hypothesis regarding different sources of water turbidity is supported by the spatial variability of the corresponding temporal trends. A certain level of turbidity, which influences the attenuation of light in the PAR range, is maintained in water bodies with relatively high turbulence. The most significant decrease in this indicator is observed in floodplain water bodies, which exhibit the lowest flow velocity and, consequently, the least potential for the removal of inorganic residues.

A general trend of increasing species richness was observed during the study period (1997–2015). However, cascading dams are known to result in a significant decline in fish diversity when compared to historical data [74]. Our study is also consistent with this pattern, as we recorded 38 species of fish in the study area, while before the reservoir was created, the species richness was 52 fish species [75]. This comparison is significant, as it highlights a decrease in gamma diversity resulting from the creation of the reservoir, a finding that is corroborated by other studies [76]. Our findings indicate a trend toward increasing alpha diversity. During the study period, gamma diversity experienced a slight increase due to the presence of invasive species, which resulted in a decrease in beta diversity, indicating environmental homogenization. The homogenization of habitats and an increase in alpha diversity result from species with significant migration potential and the ability to tolerate a wide range of water salinity conditions, which are predominantly limnophilic. Increased eutrophication diminishes the species diversity within fish communities. This result does not align with some findings that indicate that eutrophication has minimal effect on the species diversity of fish communities, while functional diversity appears to be more sensitive to this factor [77].

The distribution of functional diversity into primary components, such as functional richness, functional evenness, and functional divergence, has become nearly canonical [66]. This observation is consistent with the concept of functional imbalance as an alternative to functional evenness [78]. The principal components derived from functional imbalance metrics (CorB, SESB, and QB) exhibit a significantly greater degree of contribution to overall variability in contrast to that demonstrated by the functional evenness index. However, the most salient factor in the variability of functional diversity within fish communities is functional specialisation, a facet that is noteworthy for its absence from the list of primary components of functional diversity. Functional specialisation refers to the degree to which the functional role of a specific species differs from that of other species. This phenomenon is influenced by both the size and the position of a species’ ecological niche in relation to the niches of other species [79]. The increase in eutrophication leads to greater functional specialisation in fish communities [80]. This result aligns with the findings of our study. Specialisation is increasing due to limnophilic fish species, which exhibit a higher body weight in adulthood. The rheophilic and lithophilic fish species are generally more sensitive to environmental impacts than limnophilic and phytophilic fish species [81]. Our results highlight that the impact of eutrophication on fish communities occurs across multiple dimensions. The increase in fish biomass is a well-documented response to heightened eutrophication [77]. Furthermore, the reproductive traits of fish have been demonstrated to be a contributing factor to alterations in the level of specialisation within the community in response to eutrophication. The maturation of females and the size of their eggs are the defining factors that establish a gradient along which species specialisation occurs in fish communities. As eutrophication progresses, water bodies are stratified based on factors such as oxygen content, transparency, bottom structure, and sediment thickness [82]. Site-specific trends in eutrophication variability over time confirm the presence of spatial patterns reflecting complex ecological conditions induced by the eutrophication of water bodies. This environmental stratification can enhance the specialisation of fish species [83]. Some generalist species may struggle to survive the stressful conditions associated with eutrophication, such as hypoxia and reduced water transparency. In contrast, specialised species that are adapted to low oxygen levels, specific food sources, or increased turbidity gain a competitive advantage [84]. As a result, there is a shift in the community composition toward these specialised species. As fish community biomass increases due to eutrophication, species with overlapping ecological requirements engage in direct competition, allowing specialised species with narrower, unique niches to gain a competitive advantage.

Functional imbalance is an important aspect of the variability of functional diversity in fish communities, quantifying both the direction and magnitude of the interaction between species abundances and their functional dissimilarities [78]. This aspect of functional diversity exhibited a decreasing temporal trend throughout the study period. It is important to note that other indices of functional diversity provide substantial insights into the functional imbalance of the assemblage. Functional dispersion serves as a highly sensitive indicator that reflects functional imbalance. The decreasing trend in functional imbalance over time reflects the accumulation of anthropogenic changes in the environment, which lead to directional shifts in the functional diversity of fish assemblages. Fish species that exhibit significant differences in their functional traits are more likely to experience a decline in abundance, whereas functionally similar species tend to be among those with higher abundance. The functional differences among the dominant species within an assemblage can be considered a contributing factor to the stability of that assemblage [85]. Our results are consistent with the conclusion that high functional redundancy among dominant species may serve as a buffer, reducing the sensitivity of ecosystems to external stressors [86]. The functional redundancy of dominant fish species can help maintain community stability in the face of anthropogenic pressures. However, the long-term preservation of ecosystem functions also relies on rare species that fulfil unique ecological roles [87]. The anthropogenic changes, particularly alterations in land use and the proliferation of invasive species, should be acknowledged, as they can diminish functional diversity. This reduction highlights the limited capacity of functional redundancy to compensate for the long-term loss of ecological functions [88]. This finding is further supported by a study demonstrating that functional redundancy increases with species richness, particularly in larger, deeper, and warmer water bodies. The dominance of species with similar traits diminishes the community’s capacity to adapt to environmental changes [89]. The progressive degradation of habitat conditions for fish, resulting from the regulation of the Dnipro River’s flow, has heightened the demand for enhanced resilience mechanisms within fish assemblages [90]. Eutrophication can be assumed to be the most significant environmental factor contributing to a decrease in functional imbalance.

Functional richness is regarded as a primary component of functional diversity [66], a finding that is further supported by the results of our study. We have identified a new pattern indicating that functional richness possesses a complex nature, represented by two orthogonal main components. One component reflects an aspect of functional richness that is independent of species diversity, while the other indicates a dependence on the evenness of species abundance distribution within the community. These components exhibited opposing trends over time. Specifically, functional richness independent of evenness demonstrated a decreasing trend, whereas functional richness dependent on evenness showed an increasing trend. This complex variability, characterised by opposing trends, renders functional richness an uncertain indicator for characterising functional diversity. On one hand, these patterns underscore the variability in the functional structure of fish communities in response to anthropogenic influences. The actual patterns may be non-linear and non-monotonic, which is quite natural given the complex nature of anthropogenic pressures. However, for practical application, the index needs to be closely correlated with a specific type of anthropogenic pressure so that this impact can be quantified and assessed. Therefore, the sensitivity of functional richness to a wide range of patterns in functional structure variability can be viewed as an advantage of this index. Findings from various studies highlight the divergent responses of functional richness to increasing anthropogenic pressure. The functional diversity of fish tends to decline under anthropogenic impacts, indicating that elevated human pressure is associated with reduced overall abundance and species richness, as well as an increased proportion of invasive species and accelerated community turnover in river ecosystems [91]. Although the functional richness of fish assemblages in rivers and lakes increases with species richness, it remains vulnerable to the loss of rare or sensitive species [87]. Concurrently, evidence suggests that the functional richness of fish communities may either increase or remain stable under anthropogenic pressures. For instance, while the functional richness of fish communities tends to remain relatively stable despite anthropogenic influences, functional originality declines significantly. This decline indicates the loss of species with unique ecological roles and the potential weakening of ecosystem functioning [92]. Thus, both our findings and existing literature indicate that the functional richness of fish communities may exhibit divergent trends in response to anthropogenic pressure.

Functional divergence is also represented in the canonical triad of primary facets of functional diversity, a finding further supported by the results of our study. Throughout the study period, functional divergence exhibited an increasing trend and was influenced by all three components of environmental variability (ENV1, ENV2, and ENV3). The functional divergence within fish communities has been shown to increase in response to anthropogenic pressures, particularly elevated total phosphorus levels. Eutrophication creates a heterogeneous environment enriched with resources, particularly organic matter and nutrients such as nitrogen and phosphorus, which facilitates the survival of species with extreme or specialised traits [93]. This phenomenon frequently results in the exclusion of less tolerant or functionally similar species, thereby reducing functional redundancy [87]. Consequently, only species with unique or distinct functional traits remain, resulting in increased dispersion within functional space. Eutrophication may also promote the invasion or spread of opportunistic species that occupy unique or marginal ecological niches. These species frequently exhibit unconventional combinations of life–history strategies, further enhancing functional divergence. Overall, increasing eutrophication creates conditions that favour the persistence of species with extreme or unique functional characteristics while displacing less tolerant and functionally similar species.

The results of this study provide a scientific foundation for utilising satellite-derived spectral indices to monitor the condition of aquatic environments as a driver of functional diversity dynamics in fish assemblages. This approach is particularly effective for the floodplains of large rivers, which are characterised by complex geomorphology and limited accessibility for researchers, making remote sensing the most feasible method for assessing the spatial and temporal dynamics of ecological systems. Restricted access is a common condition in protected areas, as well as regions impacted by military activities. The southern section of the Dnipro River falls within such areas. By identifying site-specific trends in turbidity, algal bloom intensity, and trait-based biodiversity metrics, this approach enables spatially explicit assessments of ecological trajectories across extensive and hydrologically complex landscapes. The application of remote sensing allows for the detection of subtle functional changes in fish communities that may go unnoticed when relying solely on taxonomic data. These findings have practical implications for ecological monitoring, conservation planning, and adaptive management of aquatic ecosystems affected by eutrophication, hydrological alterations, and climate variability. Furthermore, the methodological framework developed in this study can be applied to other riverine and lacustrine systems experiencing similar anthropogenic and climatic disturbances.

5. Conclusions

This study demonstrates the effectiveness of integrating satellite-derived water quality indicators with trait-based ecology to assess long-term ecological change in floodplain fish assemblages. Over an 18-year period, the Dnipro-Orilskiy Nature Reserve experienced intensifying eutrophication, especially in floodplain water bodies, driven by climate warming, nutrient inputs, and altered hydrology from river regulation. Remote sensing revealed a paradoxical shift: while turbidity from inorganic particles declined due to reduced erosion and flooding, biological turbidity linked to algal blooms increased sharply.

These environmental changes led to significant shifts in the functional structure of juvenile fish communities. Functional specialization and functional imbalance both declined over time, indicating reduced ecological distinctiveness among dominant species and a trend toward functional redundancy. In contrast, functional divergence and species richness increased, suggesting greater abundance of trait-extreme species and structural reorganization. Functional richness exhibited complex, multidimensional responses, shaped by both trait space volume and species evenness, limiting its value as a standalone indicator.

The trait-based responses observed—particularly those linked to reproduction, tolerance, and habitat preference—highlight how eutrophication and regulation act as strong ecological filters. The spatial heterogeneity in trends underscores the importance of localized monitoring.

Overall, our approach shows that remote sensing can serve as a valuable, non-invasive tool for monitoring biodiversity change and detecting early signs of ecological reorganization in freshwater ecosystems. Tracking functional diversity—especially specialization and imbalance—provides deeper insight into how communities respond to multiple stressors. This knowledge is essential for supporting adaptive conservation strategies and maintaining the resilience of floodplain ecosystems in the face of climate and land use change.