Context

Scale dependence of bat habitat selection is poorly known with few studies evaluating relationships among landscape metrics such as class versus landscape, or metrics that measure composition or configuration. This knowledge can inform conservation approaches to mitigate habitat loss and fragmentation.

Objectives

We evaluated scale dependence of habitat associations and scaling patterns of landscape metrics in relation to bat occurrence or capture rate in forests of southwestern Nicaragua.

Methods

We captured 1537 bats at 35 locations and measured landscape and class metrics across 10 spatial scales (100–1000 m) surrounding capture locations. We conducted univariate scaling across the 10 scales and identified scales and variables most related to bat occurrence or capture rate.

Results

Edge and patch density, at both landscape and class levels, were the most important variables across species. Feeding guilds varied in their response to metrics. Certain landscape and configuration metrics were most influential at fine (100 m) and/or broad (1000 m) spatial scales while most class and composition metrics were influential at intermediate scales.

Conclusions

These results provide insight into the scale dependence of habitat associations of bat species and the influence of fine and broad scales on habitat associations. The effects of scale, examined in our study and others from fine (100 m) to broad (5 km) indicate habitat relationships for bats may be more informative at larger scales. Our results suggest there could be general differences in scale relationships for different groups of landscape metrics, which deserves further evaluation in other taxonomic groups.

Context

The roosting habits of many temperate zone bats are well documented at microhabitat scales, but fewer studies have included multi-scale assessments of landscape patterns in bat roost site selection.

Objectives

To identify and assess at the landscape-scale the location of spring and early season maternity roosts of female northern long-eared bats (Myotis septentrionalis) from 2015 to 2016 at Mammoth Cave National Park (MACA), Kentucky, USA.

Methods

We used mist-nets and radiotelemetry to catch and track bats to roost trees across the landscape of MACA. Data on roosting sites were evaluated using spatial point pattern analysis to examine distributional trends of roosts. A variety of spatial covariates were used to model the effect of landscape pattern, including: forest type, elevation, and proximity to hibernacula, water, and road corridors.

Results

Data indicate that roost locations of female northern long-eared bats in MACA were typically situated within 2000 m of known winter hibernacula, occurring more often at higher elevations in mesic upland deciduous forests, and in close proximity to water sources and roads. We present hypotheses to account for the patterns observed in relation to landscape features and habitat resources in the Park.

Conclusions

Our data indicate that a more comprehensive understanding of habitat requirements which includes empirically-based, landscape-scale patterns, and not solely considerations at stand or local levels, could lead to better informed management policies targeting conservation of maternity habitat of forest-dwelling bats, including the northern long-eared bat, a species in decline throughout much of its distribution in North America.

Context

In heterogeneous landscapes, habitat complementation is a key process underlying the distribution of mobile species able to exploit non-substitutable resources over large home ranges. For instance, insectivorous bats need to forage in a diversity of habitat patches offering varied compositions and structures within forest landscape mosaics to fulfill their life cycle requirements.

Objectives

We aimed at analyzing the effects of forest structure and composition measured at the stand and landscape scales on bat species richness, abundance and community composition in pine plantation forests of south-western France.

Methods

We sampled bat communities at different periods of the summer season using automatic ultrasound recorders along a tree composition gradient from pine monocultures to pure oak stands. We analyzed bat species activity (as a proxy for bat abundance) and species richness with linear mixed models. Distance-based constrained ordinations were used to partition the spatio-temporal variation in bat communities.

Results

Deciduous tree cover increased bat activity and modified community composition at both stand and landscape scales. Changes in bat communities were mostly driven by landscape-scale variables while bat activity responded more to stand-scale predictors.

Conclusions

The maintenance of deciduous trees at both stand and landscape scales is likely critical for bat communities living in fast-growing conifer plantations, by increasing the availability and diversity of prey and roosting sites. Our study suggests that bats respond to forest composition at both stand and landscape scales in mosaic plantation landscapes, mainly through a resource complementation process.

Context

Context Bats are considered as an ecological indicator of habitat quality due to their sensitivity to human-induced ecosystem changes. Hence, we will focus the study on two indicator species of bats as a proxy to evaluate structure and composition of the landscape to analyze anthropic pressures driving changes in patterns.

Objectives

This study develops a spatially-explicit model to highlight key habitat nodes and corridors which are integral for maintaining functional landscape connectivity for bat movement. We focus on a complex mountain landscape and two bat species: greater (Rhinolophus ferrumequinum) and lesser (Rhinolophus hipposideros) horseshoe bats which are known to be sensitive to landscape composition and configuration.

Methods

Species distribution models are used to delineate high-quality foraging habitat for each species using opportunistic ultrasonic bat data. We then performed connectivity analysis combining (modelled) suitable foraging habitat and (known) roost sites. We use graph-theory and the deviation in the probability of connectivity to quantify resilience of the landscape connectivity to perturbations.

Results

Both species were confined to lowlands (<1000 m elevation) and avoided areas with high road densities. Greater horseshoe bats were more generalist than lesser horseshoe bats which tended to be associated with broadleaved and mixed forests.

Conclusions

The spatially-explicit models obtained were proven crucial for prioritizing foraging habitats, roost sites and key corridors for conservation. Hence, our results are being used by key stakeholders to help integrate conservation measures into forest management and conservation planning at the regional level. The approach used can be integrated into conservation initiatives elsewhere.

Context

The pasture-woodlands of Central Europe are low-intensity grazing systems in which the structural richness of dynamic forest-grassland mosaics is causal for their high biodiversity. Distinct mosaic patterns in Picea abies- and Fagus sylvatica-dominated pasture-woodlands in the Swiss Jura Mountains suggest a strong influence of tree species regeneration ecology on landscape structural properties. At the landscape scale, however, cause-effect relationships are complicated by habitat selectivity of livestock.

Objectives

We asked which tree species regeneration traits and what kind of feedbacks among local-scale vegetation dynamics and landscape-scale herbivore behavior are causal for the contrasted landscape structural characteristics of Picea- and Fagus-dominated pasture-woodlands.

Methods

We performed simulation experiments of mosaic pattern formation in both pasture-woodland types. The regeneration traits, namely dispersal distance, resistance to browsing and tolerance to shade, and the rules for habitat selection of cattle were modified and the corresponding shifts in landscape structure were analyzed.

Results

Dispersal distance showed a significant, but only local, effect promoting forest fringe formation. Saplings’ resistance to browsing mainly determined overall tree cover, but did not influence landscape structure. At the landscape scale, both shade tolerance of saplings and selective habitat use by cattle were responsible for forest-grassland segregation: high shade tolerance triggered segregation, whereas non-selective habitat use hindered it.

Conclusions

Existing local-scale theory on pasture-woodland dynamics is complemented by an herbivore-vegetation feedback among spatial scales. In low-intensity pastures, where large herbivores are preferentially “grazers” and trees form dense canopies, an intrinsic trend towards forest-grassland segregation at the expense of forest-grassland ecotones is predicted.

Context

The ability to detect ecological networks in landscapes is of utmost importance for managing biodiversity and planning corridors.

Objectives

The objective of this study was to evaluate the information provided by a synthetic aperture radar (SAR) image for landscape connectivity modeling compared to aerial photographs (APs).

Methods

We present a novel method that integrates habitat suitability derived from remote sensing imagery into a connectivity model to explain species abundance. More precisely, we compared how two resistance maps constructed using landscape and/or local metrics derived from AP or SAR imagery yield different connectivity values (based on graph theory), considering hedgerow networks and forest carabid beetle species as a model.

Results

We found that resistance maps using landscape and local metrics derived from SAR imagery improve landscape connectivity measures. The SAR model is the most informative, explaining 58% of the variance in forest carabid beetle abundance. This model calculates resistance values associated with homogeneous patches within hedgerows according to their suitability (canopy cover density and landscape grain) for the model species.

Conclusions

Our approach combines two important methods in landscape ecology: the construction of resistance maps and the use of buffers around sampling points to determine the importance of landscape factors. This study was carried out through an interdisciplinary approach involving remote sensing scientists and landscape ecologists. This study is a step forward in developing landscape metrics from satellites to monitor biodiversity.

Context

The conversion of natural environments into agricultural land has profound effects on the composition of the landscape, often resulting in a mosaic of human-altered and natural habitats. The response to these changes may however vary among organisms. Bats are highly vagile, and their requirements often imply the use of distinct habitats, which they select responding to both landscape and local features.

Objectives

We aimed to identify which features influence bat richness and activity within Baixo Vouga Lagunar, a heterogeneous landscape located on the Central-North Portuguese coast, and to investigate if that influence varies across a gradient of focal scales.

Methods

We sampled bats acoustically, while simultaneously sampling insects with light traps. We assessed the relationships between species richness, bat activity, and activity of eco-morphological guilds with landscape and local features, across four scales.

Results

Our results revealed both scale- and guild-dependent responses of bats to landscape and local features. At broader scales we found positive associations between open-space foraging bats and habitat heterogeneity and between edge-space foraging bats and greater edge lengths. Woodland cover and water availability at an intermediate scale and weather conditions and insect abundance at a local scale were the factors that mostly influenced the response variables.

Conclusions

Globally, our results suggest that bats are sensitive to local resource availability and distribution, while simultaneously reacting to landscape features acting at coarser scales. Finally, our results suggest that the responses given by bats are guild-dependent, and some habitats act as keystone structures for bats within this mosaic.

Context

Species site-occupancy patterns may be influenced by habitat variables at both local and landscape scales. Although local habitat variables influence whether the site is suitable for a given species, the broader landscape context can also influence site occupancy, particularly for species that are sensitive to land-use change.

Objectives

To examine the relative importance of local versus landscape variables in explaining site occupancy of eight bat species within the Brazilian Cerrado, a Neotropical savanna that is experiencing widespread habitat loss and fragmentation.

Methods

Bats were surveyed within 16 forest patches over two years. We used a multi-model information-theoretic approach, adjusted for species detection bias, to assess whether landscape variables (percent cover and number of patches of natural vegetation within a 2- and 8-km radius of each forest site) or local site variables (canopy cover, understory height, number of trees, and number of lianas) best explained site occupancy in each species.

Results

Landscape variables were among the best models (ΔAIC_c or ΔQAIC_c < 2) for four species (top-ranked model for black myotis), whereas local variables were among the best for five species (top-ranked model for vampire bats). Neither local nor landscape variables explained site occupancy in two frugivorous species.

Conclusion

Species associated with a particular habitat type will not respond similarly to the amount, distribution or relative suitability of that habitat, or even at the same scale. This reinforces the challenge of species distribution modelling, especially in the context of forecasting species’ responses to future land-use or climate-change scenarios.

Context

The patch-mosaic model is lauded for its conceptual simplicity and ease with which conventional landscape metrics can be computed from categorical maps, yet many argue it is inconsistent with ecological theory. Gradient surface models (GSMs) are an alternative for representing landscapes, but adoption of surface metrics for analyzing spatial patterns in GSMs is hindered by several factors including a lack of meaningful interpretations.

Objectives

We investigate the performance and applicability of surface metrics across a range of ecoregions and scales to strengthen theoretical foundations for their adoption in landscape ecology.

Methods

We examine metric clustering across scales and ecoregions, test correlations with patch-based metrics, and provide ecological interpretations for a variety of surface metrics with respect to forest cover to support the basis for selecting surface metrics for ecological analyses.

Results

We identify several factors complicating the interpretation of surface metrics from a landscape perspective. First, not all surface metrics are appropriate for landscape analyses. Second, true analogs between surface metrics and patch-based, landscape metrics are rare. Researchers should focus instead on how surface measures can uniquely measure spatial patterns. Lastly, scale dependencies exist for surface metrics, but relationships between metrics do not appear to change considerably with scale.

Conclusions

Incorporating gradient surfaces into landscape ecological analyses is challenging, and many surface metrics may not have patch analogs or be ecologically relevant. For this reason, surface metrics should be considered in terms of the set of pattern elements they represent that can then be linked to landscape characteristics.

Context

Habitat loss and fragmentation may alter habitat occupancy patterns, for example through a reduction in regional abundance or in functional connectivity, which in turn may reduce the number of dispersers or their ability to prospect for territories. Yet, the relationship between landscape structure and habitat niche remains poorly known.

Objectives

We hypothesized that changes in landscape structure associated with habitat loss and fragmentation will reduce the habitat niche breadth of forest birds, either through a reduction in density-dependent spillover from optimal habitat or by impeding the colonization of patches.

Methods

We surveyed forest birds with point counts in eastern Ontario, Canada, and analyzed their response to loss and fragmentation of mature woodland. We selected 62 landscapes varying in both forest cover (15–45%) and its degree of fragmentation, and classified them into two categories (high versus low levels of loss and fragmentation). We determined the habitat niche breadth of 12 focal species as a function of 8 habitat structure variables for each landscape category.

Results

Habitat niche breadth was narrower in landscapes with high versus low levels of loss and fragmentation of forest cover. The relative occupancy of marginal habitat appeared to drive this relationship. Species sensitivity to mature forest cover had no apparent influence on relative niche breadth.

Conclusions

Regional abundance and, in turn, density-dependent spillover into suboptimal habitat appeared to be determinants of habitat niche breadth. For a given proportion of forest cover, fragmentation also appeared to alter habitat use, which could exacerbate its other negative effects unless functional connectivity is high enough to allow individuals to saturate optimal habitat.

Context

Quantitative models of forest dynamics have followed a progression toward methods with increased detail, complexity, and spatial extent.

Objectives

We highlight milestones in the development of forest dynamics models and identify future research and application opportunities.

Methods

We reviewed milestones in the evolution of forest dynamics models from the 1930s to the present with emphasis on forest growth and yield models and forest landscape models We combined past trends with emerging issues to identify future needs.

Results

Historically, capacity to model forest dynamics at tree, stand, and landscape scales was constrained by available data for model calibration and validation; computing capacity; model applicability to real-world problems; and ability to integrate biological, social, and economic drivers of change. As computing and data resources improved, a new class of spatially explicit forest landscape models emerged.

Conclusions

We are at a point of great opportunity in development and application of forest dynamics models. Past limitations in computing capacity and in data suitable for model calibration or evaluation are becoming less restrictive. Forest landscape models, in particular, are ready to transition to a central role supporting forest management, planning, and policy decisions.

Recommendations

Transitioning forest landscape models to a central role in applied decision making will require greater attention to evaluating performance; building application support staffs; expanding the included drivers of change, and incorporating metrics for social and economic inputs and outputs.

Context

The anthropocene is characterised by global landscape modification, and the structure of remnant habitats can explain different patterns of species richness. The most pervasive processes of degradation include habitat loss and fragmentation. However, a recovery of modified landscape is occurring in some areas.

Objectives

The main goal is to know how lichen and bryophyte epiphytic richness growing on Mediterranean forests is influenced not only by fragments characteristics but also by the structure of the landscape. We introduce a temporal dimension in order to evaluate if the historical landscape structure is relevant for current epiphytic communities.

Methods

40 well-preserved forest fragments were selected in a landscape with a large habitat loss over decades, but with a recovery of forest surface in the last 55 years. The most relevant fragment and landscape-scale attributes were considered. Some of the variables were measured in three different years to incorporate a temporal framework.

Results

The results showed that variables at fragment scale had a higher influence, whereas variables at the landscape scale were irrelevant. Among all the historical variables analyzed, only the shift in forest fragment size had influence on species richness.

Conclusions

Mediterranean forests had suffered fragmentation along centuries. Their epiphytic communities also suffer the hard conditions of Mediterranean climate. Our results indicate that Mediterranean epiphytic communities may be in a threshold since it they will never be similar to those communities existing previous fragmentation process even a recovery habitat occur or, they may require more time to response to this habitat recovery.

Context

Golden-cheeked warblers (Setophaga chrysoparia), an endangered wood-warbler, breed exclusively in woodlands co-dominated by Ashe juniper (Juniperus ashei) in central Texas. Their breeding range is becoming increasingly urbanized and habitat loss and fragmentation are a main threat to the species’ viability.

Objectives

We investigated the effects of remotely sensed local habitat and landscape attributes on point occupancy and density of warblers in an urban preserve and produced a spatially explicit density map for the preserve using model-supported relationships.

Methods

We conducted 1507 point-count surveys during spring 2011–2014 across Balcones Canyonlands Preserve (BCP) to evaluate warbler habitat associations and predict density of males. We used hierarchical Bayesian models to estimate multiple components of detection probability and evaluate covariate effects on detection probability, point occupancy, and density.

Results

Point occupancy was positively related to landscape forest cover and local canopy cover; mean occupancy was 0.83. Density was influenced more by local than landscape factors. Density increased with greater amounts of juniper and mixed forest and decreased with more open edge. There was a weak negative relationship between density and landscape urban land cover.

Conclusions

Landscape composition and habitat structure were important determinants of warbler occupancy and density, and the large intact patches of juniper and mixed forest on BCP (>2100 ha) supported a high density of warblers. Increasing urbanization and fragmentation in the surrounding landscape will likely result in lower breeding density due to loss of juniper and mixed forest and increasing urban land cover and edge.

Context

Land-use change impacts biodiversity and ecosystem services, which are intrinsically related. There is a serious lack of knowledge concerning on how land-use change affects this relationship at landscape level, where the greatest impacts have been reported. A proper knowledge of that relationship would provide crucial information for planning conservation strategies. The forest landscape of southern Chile, which includes Valdivian Temperate Forest, has been designated as a hotspot for biodiversity conservation. However, this landscape has been transformed by land-use change.

Objective

We evaluated the impact of land-use change on the spatial patterns of the diversity of native forest habitat and the influence of these impacts on the provision of the ecosystem services water supply, erosion control, and organic matter accumulation from 1986 to 2011.

Methods

The evaluation, at the landscape level, was carried out using satellite images, landscape metrics, spatially explicit models and generalized linear models. Results: We found that the area loss of native forest habitat was 12%, the number patches of native forest habitat increased more than 150% and the Shannon diversity index decreased by 0.20. The largest decrease in the provision of services was recorded for erosion control (346%), and the smallest for water supply (11%).

Conclusions

The loss of provision of the ecosystem services can be explained by the interaction between the area loss, increase in the number patches and diversity loss. We recommend that the conservation planning strategies should consider the current landscape configuration, complemented with land-use planning.

Context

Jack pine (Pinus banksiana)-dominated ecosystems of northern Lower Michigan are the primary breeding habitat for the federally endangered Kirtland’s warbler (Setophaga kirtlandii, KW). Historically, young stands used by KW were produced by stand-replacing wildfires, but fire suppression has necessitated the management of jack pine plantations for KW habitat since the 1970s. Effects of this long-term management on landscape age heterogeneity have previously not been quantified.

Objectives

We hypothesized that forest management has altered the spatial and temporal distribution of jack pine-dominated ecosystems beyond their historic range of variability.

Methods

By developing a diameter-age relationship for jack pine, we estimated ages of pre-European settlement trees found in General Land Office survey notes. We compared pre-European and current landscapes using geostatistical modeling of survey notes, and landscape metrics to quantify changes in pattern.

Results

Three KW management-based age classes (<20, 21–50, >50 years) are now more evenly distributed (31, 39, and 30 %, respectively) compared to the pre-European distribution (5, 19, 76 %) with little variability over time. Landscape metrics suggest the current landscape is younger and more fragmented than the pre-European landscape. These changes indicate restriction of the historic range of age variability, largely due to conversion of older jack pine stands to young KW habitat plantations.

Conclusions

Management has met KW population objectives, but has altered the temporal variability of the landscape’s age structure. Pre-European settlement patterns of stand-ages may provide a foundation for an ecosystem-based management plan for the region that supports both KW and the ecosystems upon which they depend.

Context

The positive correlation between landscape area of semi-natural habitat and wild pollinator richness and abundance in agroecosystems has been well studied. However, we lack a deep understanding of local scale floral resource and nest provisioning for wild bees necessary to optimize implementation of pollinator conservation practices.

Objectives

The primary objective of this study was to use a spatially interactive landscape pollination model (hereafter, the Lonsdorf model) to represent field scale spatial patterns of wild bee abundance and richness within a heterogeneous landscape in the mid-Atlantic USA.

Methods

We parameterized the Lonsdorf model with high resolution aerial imagery and insight from a previously published floristic study. To test the Lonsdorf model predictions, field studies were conducted to measure wild bee abundance and species richness in apple orchards as a function of distance from a forest edge.

Results

Field measurements indicated apple pollinator abundance and species richness significantly decreased with increasing distance from the forest edge. The Lonsdorf model pollination service score was highly sensitive to changes in resource provisioning in orchard and non-crop areas, and including resource rich forest and forest edge habitats in the model significantly improved pollination service estimates.

Conclusions

We demonstrated a novel application of the Lonsdorf model at a field scale to predict trends in pollination service provisioning as a factor of local habitat features. With sufficiently detailed inputs, the Lonsdorf model is a promising tool to quantify field scale pollination service deficits, guiding more cost effective habitat supplementation and other conservation efforts.

Context

Landscape modification is an important driver of biodiversity declines, yet we lack insight into how ongoing landscape change and legacies of historical land use together shape biodiversity.

Objectives

We examined how a history of agricultural land use and current forest fragmentation influence the abundance of red-backed salamanders (Plethodon cinereus). We hypothesized that historical agriculture and fragmentation cause changes in habitat quality and landscape structure that limit abundance.

Methods

We measured salamander abundance at 95 forested sites in New York, USA, and we determined whether sites were agricultural fields within the last five decades. We used a structural equation model to estimate relationships between historical agriculture and salamander abundance mediated by changes in forest vegetation, microclimate, and landscape structure.

Results

Historical agriculture affected salamander abundance by altering forest vegetation at a local scale and forest cover at a landscape scale. Abundance was lowest at post-agricultural sites with low woody vegetation, leaf litter depth, and canopy cover. Post-agricultural sites had limited forest cover in the surrounding landscape historically, and salamander abundance was positively related to historical forest cover, suggesting that connectivity to source populations affects colonization of regenerating forests. Abundance was also negatively related to current forest fragmentation.

Conclusions

Historical land use can have legacy effects on animal abundance on par with effects of ongoing landscape change. We showed that associations between animal abundance and historical land use can be driven by altered site conditions and surrounding habitat area, indicating that restoration efforts should consider local site conditions and landscape context.

Context

The application of regional-level airborne lidar (light detection and ranging) data to characterize habitat patches and model habitat connectivity over large landscapes has not been well explored. Maintaining a connected network of habitat in the presence of anthropogenic disturbances is essential for regional-level conservation planning and the maintenance of biodiversity values.

Objectives

We quantified variation in connectivity following simulated changes in land cover and contrasted outcomes when different conservation priorities were emphasized.

Methods

First, we defined habitat patches using vegetation structural attributes identified via lidar. Second, habitat networks were constructed for different forest types and assessed using network connectivity metrics. And finally, land cover change scenarios were simulated using a series of habitat patch removals, representing the impact of implementing different spatial prioritization schemes.

Results

Networks for different forest structure types produced very different patch distributions. Conservation scenarios based on different schemes led to contrasting changes during land cover change simulations: the scheme prioritizing only habitat area resulted in immediate near-term losses in connectivity, whereas the scheme considering both habitat area and their spatial configurations maintained the overall connectivity most effectively. Adding climate constraints did not diminish or improve overall connectivity.

Conclusions

Both habitat area and habitat configuration should be considered in dynamic modeling of habitat connectivity under changing landscapes. This research provides a framework for integrating forest structure and cover attributes obtained from remote sensing data into network connectivity modeling, and may serve as a prototype for multi-criteria forest management and conservation planning.

Context

Despite continued forest cover losses in many parts of the world, Atlantic Forest, one of the largest of the Americas, is increasing in some locations. Economic factors are suggested as causes of forest gain, while enforcement has reduced deforestation.

Objectives

We examine three aspects of this issue: the relative importance of biophysical versus anthropogenic factors in driving forest dynamics; role of forest mean patch age influencing areas targeted for losses; and what future forest mean patch age mosaic we can expect (more forest cover and full forest maturity?).

Methods

Three land cover maps from 1990, 2000 and 2010, were used in the study. We selected six biophysical and six anthropogenic spatial determinants to analyze by means of weights of evidence, using Dinamica software.

Results

Results show that forest regrowth is influenced by multiple factors, working in synergy. Biophysical variables are related to forest gain while anthropogenic are associated with loss. Clear patterns of regrowth on pasture and sugarcane plantations occurred, especially near rivers and forest patches, on steeper slopes and with sufficient rainfall. Forest loss has targeted both older and newer forests. Future projections reveal forest gain in a slow pace, followed by specific ecosystem service losses, due to continuous trends of older mature forest loss.

Conclusions

Regrowth is linked to land abandonment, and to neighboring environmental conditions. It is important to question which mechanisms will guarantee and potentiate new regrowth, thus contributing to landscape restoration and reestablishment of ecosystem services in the Atlantic Forest.

Context

Land use change and forest degradation have myriad effects on tropical ecosystems. Yet their consequences for low-order streams remain very poorly understood, including in the world´s largest freshwater basin, the Amazon.

Objectives

Determine the degree to which physical and chemical characteristics of the instream habitat of low-order Amazonian streams change in response to past local- and catchment-level anthropogenic disturbances.

Methods

To do so, we collected field instream habitat (i.e., physical habitat and water quality) and landscape data from 99 stream sites in two eastern Brazilian Amazon regions. We used random forest regression trees to assess the relative importance of different predictor variables in determining changes in instream habitat response variables.

Results

Multiple drivers, operating at multiple spatial scales, were important in determining changes in the physical habitat and water quality of the sites. Although we found few similarities in modelled relationships between the two regions, we observed non-linear responses of specific instream characteristics to landscape change; for example 20 % of catchment deforestation resulted in consistently warmer streams.

Conclusions

Our results highlight the importance of local riparian and catchment-scale forest cover in shaping instream physical environments, but also underscore the importance of other land use changes and activities, such as road crossings and upstream agriculture intensification. In contrast to the property-scale focus of the Brazilian Forest code, which governs environmental regulations on private land, our results reinforce the importance of catchment-wide management strategies to protect stream ecosystem integrity.