Biodiversity is not evenly distributed across our planet. Freshwater ecosystems hold a disproportionate amount of biodiversity when compared with other biomes, though it only covers a small portion of our planet surface. Because water is essential for human activities, population growth and economic development put an enormous pressure on freshwater ecosystems. Besides the direct anthropogenic pressures, such as over exploitation, damming, habitat modification and pollution, the freshwater species usually present restricted distributions, to a watershed or a region, and face the threat of hundreds of invasive species that have been introduced to freshwater ecosystems. Due to all these factors, freshwater biodiversity is among the most threatened of our planet.
Stream networks deserve special attention because they are particularly threatened and rich in biodiversity. Streams networks are linear bodies of water with a dendritic, or tree shape, organization, flowing from the headwaters to a single outlet. The distribution of organisms in stream networks are not random, resulting from several processes that work at different scales, like climate, hydrology and biotic interactions. The diversity and abundance of fish and many other organisms are usually associated with an increase in stream order, but there are also dispersal processes that should be taken into account. The distribution of some species, like invasive species, is often more a reflection of spatial processes, such as multiple introductions and posterior expansion, than environmental filters that limit the distribution. Headwaters can function as refuges from adverse biotic interactions for species that support water intermittency. Stream communities, like fish, usually persist in time in a state of dynamic equilibrium, varying between alternate states with no discernible direction of change. Deviations from this equilibrium may reflect disturbances to the community from natural states, like droughts or floods, or from anthropogenic sources. For proper conservation and management of stream networks, it is essential to understand the drivers of the spatial patterns and dynamics of stream biodiversity.
Species distribution models (SDM’s) are the set of tools used to derive spatially-explicit predictions of environmental suitability, by relating species occurrences to relevant environmental data. Due to their nature and the nature of the stream network habitats, the development of SDM’s for organisms that are associated with streams is challenging. Aquatic organisms are rarely available for direct observation, and even with the help of standard techniques, like electrofishing, it is fair to assume that we will fail to detect some of the species present at any given location. This issue, known as imperfect detectability, is a common source of bias in SDM’s, and tends to be ignored by freshwater researchers. Accounting for spatial autocorrelation (SAC) improves SDM performance, but the dendritic structure of stream networks, together with strong environmental gradients, create spatial dependences with complex structures that are not completely described by Euclidean distances. Biotic interactions, such as competition or predation, are also a potential source of mismatch between the actual and the predicted distribution of species, particularly if the interactions are between invasive species and native species. Long term monitoring of communities is essential to understand the impact of anthropogenic pressures in stream ecosystems, but usually rely on data collected on any given number of discrete locations. A spatial continuous view of the temporal dynamics would be essential to study such pressures, and of value to plan conservation and management actions.
The main aim of this thesis is to develop new tools and frameworks to help ecologists and conservationists to obtain a more realistic depiction of the distribution of species, and the temporal dynamics of communities at the riverscape scale. I mainly focused on solutions to the issues related to dealing with imperfect detectability, accounting for SAC in stream networks, accounting for biotic interactions, and extrapolating the community temporal dynamics to a continuous spatial prediction. To address these issues, I have collected data on the distribution of fish, crayfish, and amphibians on a specific study system, the Sabor River, a Mediterranean watershed in the Northeast of Portugal.
To describe the distribution of fish species with data collected in a comprehensive electrofishing survey, while accounting for imperfect detectability, we extended the time-to-detection occupancy-detection model to include interval-censored observations, because it is difficult to ascertain the exact time-to-detection of a species when sampling fish with electrofishing techniques. Using a Bayesian hierarchical framework, we modelled the probability of water presence in stream segments, and the probability of species occupancy conditional on water presence, in relation to environmental and spatial variables. We also modelled time-to-first detection conditional on occupancy in relation to local factors, using a modified interval-censored exponential survival models. To account for SAC, we included a spatial autocovariate term in the estimation of the probability of water presence and the probability of species occupancy. Species occupancies were consistently affected by stream order, elevation and annual precipitation, while species detection rate was primarily influenced by depth and, to a lesser extent, stream width.
The assumption of equilibrium between organisms and their environment is a standard working postulate in SDM’s that is seldom met, particularly for species that are expanding their range like invasive species. Furthermore, for species invading river systems, the dendritic structure of the stream network will constrain the patterns of the expansion. In this thesis, I addressed these issues by describing the distribution of two invasive crayfish in the Sabor river stream network, using a class of geostatistical models developed to deal with SAC in stream networks, known as spatial stream network models (SSNM). Accounting for SAC greatly improved model performance, evidencing that the distribution of these invasive crayfish was more of a product of spatial process than environmental filtering.
Biotic interactions are important drivers of species distributions. When native species are displaced from part of their distributional range, they may persist in ecological refuges. These refuges may be patches of habitat that are unsuitable for invasive species or areas where invasive species have not reached due to distance, physical barriers or time lags in the expansion. Identifying the distribution and the environmental drivers of these refuges is of conservation concern. We modelled the distribution of amphibian ecological refuges in the Sabor river catchment, by including as predictor variables the probability of presence of the two invasive crayfish, among other environmental and spatial predictors. We found that the refuges of amphibians are located mainly in the headwaters, and that, under plausible expansion scenarios of the crayfish species, these refuges are likely to contract in the future.
Management of stream networks is usually planned at the river basin scale, and as such, it is important to develop frameworks that allow the extrapolation of the community dynamics observed at discrete segments of rivers to a continuous spatial view of the entire river basin. We collected stream fish data on 30 locations on the Sabor river basin, between 2012 and 2019, and used a novel framework to describe and compare the trajectories of the fish communities using their geometric properties in a given dissimilarity space. We computed the mean velocity and the overall directionality of change of the fish community, and used the SSNM framework to relate these metrics to environmental drivers and extrapolate the community dynamics to the entire watershed. We found no evidence of directionality in the change of the Sabor fish communities, supporting the hypothesis that these communities exist in a loose equilibrium state. However, the rate of change was higher in streams draining into the hydroelectric reservoir located near the mouth of the Sabor River. These streams are likely under increased stress from the reservoir, due to alterations of the flow regime and/or expansion of alien species from the reservoir.
Overall this thesis advances our understanding of the drivers that govern the distribution of species in stream networks, providing key information for the conservation of these ecosystems. The new set of tools presented here can aid ecologists and conservationists to obtain a more realistic depiction of species distribution and their temporal dynamics at the riverscape scale.