FIGURE CAPTIONS
Figure 1. Map representing (a) the study area in Amazonia and
Uruguay, (b) intensive land-use cover of Amazon stream catchments, (c)
intesive land-use cover of Uruguay stream catchments, and (d) taxonomic
richness and functional diversity of the three assemblages (fish,
arthropods, macrophytes). Land use information was extracted from
MapBiomas
(https://mapbiomas.org/colecoes-mapbiomas-1?cama_set_lan.).
Importantly, study sites included two river basins in Amazonia and
almost the entire territorial aea of Uruguay. Whereas Amazonia is
dominated by dense tropical forest, Uruguay is dominated by grassland.
Note that both study areas are strongly influenced by human land use,
including agriculture and pasture (strong yellow), urbanization (red),
and afforestation by non-native pine and eucalyptus (faint green).
Finally, taxonomic and functional diversities of the three assemblages
do not differ markedly between the two study regions, although Amazonia
supports slightly higher taxa richness, especially for arthropods and
macrophytes.
Figure 2. Relative importance of intensive human land-use
types, local environmental variables, stream morphology (depth),
climatic variables, and the regional taxa pool in explaining variation
in taxonomic richness and functional diversity of fish, arthropod and
macrophyte assemblages across the Neotropical stream sites studied.
Explained variance (relative effect, % R²) was calculated for each
group of predictors using a model averaging procedure. All predictors
were z-standardized to facilitate interpretation of parameter estimates
on a comparable scale. Importantly, the contribution of stream
morphology is not shown in the graph because depth was removed during
model selection (AICc).
Figure 3. Responses of fish, arthropod and macrophyte
assemblage taxonomic richness and functional diversity to human land-use
types. (a) Effects of best predictors, including land-use types
(agriculture, pasture, and urbanization), climate and local
environmental variables on taxonomic richness and functional diversity
of fish, arthropod, and macrophyte assemblages. Effect sizes were
adjusted using linear mixed-effects models. Colors represent different
assemblages: orange (fish), blue (arthropods), and green (macrophytes).
See Supplementary Table S6 for the model output summaries. Relationships
of the land-use types selected during backward selection with (b)
taxonomic richness and (c) functional diversity of the assemblages.
Lines show model fits and colored shaded areas correspond to the 95%
confidence interval from linear mixed effect models (LMM). Model
predictions were calculated using a model averaging procedure (see
Methods). Land-use types were scaled to interpret parameter estimate on
a comparable scale. P -values of the best predictors for each
model are displayed. Symbols (n = 122) correspond to observed data and
their shape indicates the region: circle (Amazonia) and triangle
(Uruguay).
Figure 4. Relative importance of human land-use types, local
environmental variables, stream morphology (depth), climate variables,
and the regional taxa pool in explaining the diversity trait category
(i.e., recruitment and life-history, resource and habitat use, and body
size) of fish, arthropod, and macrophyte assemblages across the study
sites. Explained variance (relative effect, % R²) was calculated for
each group of predictors, resulting from the model averaging procedure.
All predictors were z-standardized to allow the interpretation of
parameter estimates on a comparable scale. Importantly, the contribution
of stream morphology is not shown in the graph because depth was removed
during model selection (AICc).
Figure 5. Responses of fish, arthropod and macrophyte trait
diversity to human land-use types. (a) Effects of best predictors,
including land-use types, climatic and local environmental variables on
recruitment and life-history, resource and habitat-use, and body size of
fish, arthropods, and macrophytes. Effect sizes were adjusted using
linear mixed-effects models. Colors represent assemblages: orange
(fish), blue (arthropods), and green (macrophytes). See Supplementary
Table S7 for the model output summaries. Relationships of the land-use
types selected during backward selection with (b) recruitment and
life-history, (c) resource and habitat-use, and (d) body size. Lines
show the best model fits and colored shaded areas correspond to the 95%
confidence interval from linear mixed effect models (LMM). Model
predictions were calculated using a model averaging procedure (see
Methods). Land-use types were scaled to interpret parameter estimate on
a comparable scale. P -values of the best predictors for each
model are displayed. Symbols (n = 117) correspond to observed data and
their shape indicates the region: circle (Amazonia) and triangle
(Uruguay).
Figure 6. Structural
equation models (SEMs) showing the overall (Including both Amazonia and
Uruguay streams) direct and cascading effects of intensive land-use
cover on standing fish biomass mediated by (a) taxonomic richness and
(b) functional diversity of fish, arthropod, and macrophyte. Models
accounted for local environmental and climate predictors. Model
selection and simplification steps using Akaike Information Criteria
(AIC) are available in Supplementary Information, Table S8. The full
model fitted well to the data for both (a) species richness (Fisher’s C
= 1.739, P = 0.419) and (b) functional diversity (Fisher’s C =
3.307, P = 0.508) models. Results for the multi-group approach
(i.e., Amazonia and Uruguay separately) are provided in Supplementary
Information, Table S10. Solid black arrows are significant pathways
(P ≤ 0.05, piecewiseSEM), whereas the thickness of the arrows
represent the magnitude of the standardized regression coefficient.
Numbers in the arrows are the standardized path coefficients of the
relationship, and R2 values for each model are given
in the boxes of the variables. Significance levels of each predictor are
*P < 0.05, **P < 0.01, ***P< 0.001. (c) and (d) show the standardized indirect effects of
the ILUC on fish standing biomass mediated by taxonomic richness and
functional diversity, respectively. Effects are derived from the SEMs,
and standardized effect is computed based on multiplication of
coefficients. Significance of indirect effects is calculated based on
significance of direct effects. Importantly, water quality parameters
(nutrient, oxygen, and conductivity) and stream morphology (depth) were
removed during model selection (AICc; see Table S8). Although
precipitation was selected (AICc) for modeling taxa richness,
temperature was selected for modeling functional diversity.