• Loading metrics

Who will be where: Climate driven redistribution of fish habitat in southern Germany


To improve the robustness of projections of freshwater fish distributions under climate change, species distribution models (SDMs) were calculated for six fish species in southwestern Germany with different ecological requirements along an upstream-downstream gradient in a multi-general circulation model (GCM) approach. Using the maximum entropy (Maxent) algorithm and a high number of occurrence records (N = 4684), species distributions were projected to future climate conditions derived from 13 GCMs across the most likely representative carbon pathways (RCP4.5 and 8.5) and two time spans (near future 2050, and far future 2070), resulting in 104 distribution maps per species that were then used for the statistical analysis of future trends. Climate change is likely to affect the distribution of four of the six fish species. The potential ranges of salmonids are predicted to decline by up to 92% (brown trout) and 75% (grayling). In contrast, habitat suitability for perch and roach is predicted to increase by up to 108% and 53%, respectively. Even when accounting for broad variation in GCMs and realistic RCPs, these results suggest climate change will drive a significant redistribution of fish habitat. Salmonid-dominated communities in headwaters seem more sensitive to climate change than the fish communities of downstream sections. Because headwaters are a prevailing element of the hydrographic network in southwestern Germany, such changes may result in large-scale regressions of valuable fish communities that currently occupy broad geographic niches.


Anthropogenic climate change is a major threat challenging the stability of the world’s economic and natural systems [1] and has already been identified as a major driver of biodiversity loss [2], with species across diverse taxonomic groups responding to recent climate alterations by shifting their ranges or becoming locally extinct [3,4]. Accelerated climate change may even threaten species that show a high potential for adaption or dispersal, for instance when these capabilities cannot compensate for the high rate of habitat loss in their current distribution area [5,6]. This threat is particularly pronounced in freshwater ecosystems [7], that already experience massive ongoing biodiversity losses [810] through multiple stressors [11,12], including changing catchment land-use [13], habitat fragmentation by hydropower [14], and the introduction of invasive non-native species [15,16]. Climate changes are expected to drive further alterations in water temperature and streamflow that will severely affect the character and biota of freshwater habitats [17,18].

Freshwater fish provide essential ecosystem services [19] and, with an estimated number of 18,260 species and high levels of endemism, significantly contribute to vertebrate species diversity [20,21]. In many locations, such as southwestern Germany, they are also of primary socio-economic importance via recreational or commercial fishing of wild stocks [22] or via aquaculture [23]. As ectothermic organisms, the life cycles of fish are closely linked to local climatic conditions [24] and their zoogeographies are strongly influenced by the range of temperatures to with they are adapted [25].

Species distribution models (SDMs), also known as ecological niche models or climate envelope models, represent a primary tool to assess impacts of climate change on fish species [26,27]. SDMs relate species occurrence data to environmental data and use these relationships to predict the habitat suitability of any location [2831]. They have been applied extensively to project future species distributions under climate change and to guide conservation and management measures [27,32,33].

A major source of uncertainty in SDM projections arises from the choice of global climate models (GCMs) and representative carbon pathways (RCPs) [34,35]. A wide variety of GCMs has been published that make projections for future climates and the choice of GCMs can introduce a high variance in projections of climate suitability. Between-GCM variability of SDM projections may even exceed the variability arising from the choice of RCP [35]. Studies of climate change effects on fishes and other freshwater biota typically include climate projections from only a few GCMs (cf. [18,26,36,37]). Pre-selection of particular GCMs is valid when there is evidence that these make the most appropriate representation of variation at a local geographical scale. However, given the broad range of available GCMs [38], this approach fails to capture the variability or similitude resulting from including the full range of future climate projections [39]. Furthermore, and unfortunately, it has become evident that projections calculated from the frequently used low-emission scenario (RCP2.6) that keeps global temperature rise below 2 °C [40,41] are unrealistic and potentially misleading, as humanity remains set on a path towards mid- to high-emission scenarios (RCP4.5 and 8.5) that leads towards a global warming of approx. 3 °C by the end of the century [42].

The aim of this study was to take the wide variety of GCMs into account to make robust predictions for the impact of climate change on habitat suitability for a typical assembly of temperate zone freshwater fishes in southwestern Germany. The underlying hypothesis is that there will be different reactions to the expected change among the native fish species. SDMs were calculated for six species with different ecological requirements and distributions along the upstream–downstream gradient (Table 1) and projected to future climate conditions derived from 13 GCMs across two RCP scenarios (4.5 and 8.5) and two time spans (near future 2050, and far future 2070). The RCPs used reflect moderate (RCP4.5) and high emission (RCP8.5) scenarios within the likely trajectory of the Earth system given current global climate policies [5,43,44]. The results may prove indicative for other geographic regions with comparable habitat conditions.

Table 1. Primary habitat, thermal guild, distribution and angling interest of the study species in southwestern Germany.


Study area, species selection and occurence data

The study was carried out in the federal state of Baden-Württemberg, south-western Germany, a Central European region with an exceptional diversity of freshwater habitats and fish diversity (58 native species, [46]). The state is rich in mid-elevation mountainous areas with rhithral rivers characterized by swift currents, cool water and high levels of dissolved oxygen [47], which drain into the catchment areas of the rivers Rhine and Danube (Fig 1). The river network is largely composed of headwater streams and small rivers but also comprises low-gradient potamal sections of the Rhine and Danube Rivers.

Fig 1. Study area (federal state of Baden-Württemberg; southwestern Germany), including elevation gradients, main drainage areas, and classification of lotic water bodies (Base layer of the map, drainage areas and water bodies are from

For the purpose of this study, a set of six fish species with different niche requirements along the upstream–downstream gradient and a wide spectrum of angling interest was selected (Table 1). All of selected species can be considered to exhibit equilibrium distribution within the study area, i.e., they are commonly present in environmentally suitable areas and absent from unsuitable ones, which is a major assumption for SDMs [30,46], and are present in all major catchments, i.e., forecasted range shifts are unlikely to be restricted by major watersheds. Brown trout (Salmo trutta Linnaeus 1758) and grayling (Thymallus thymallus Linnaeus 1758) represent stenothermic cold-adapted specialists, while perch (Perca fluviatilis Linnaeus 1758), roach (Rutilus rutilus Linnaeus 1758) and wels catfish (Silurus glanis Linnaeus 1758) can be considered as eurythermic habitat generalists [48]. To highlight potential climate-change consequences for recreational fisheries, the assessed species cover a full spectrum of angling interest, from zero interest in the case of stone loach (Barbatula barbatula Linnaeus 1758), to popular game fish, such as grayling, trout, and perch [49].

Species occurrence data were extracted from the database of the Federal State of Baden‐Württemberg (FiAKa), which contains more than 25,000 georeferenced records of fish occurrence. For the purpose of the present study, occurrence points for brown trout (n = 1,645), grayling (n = 167), stone loach (n = 1,331), perch (n = 345), roach (n = 788) and wels catfish (n = 408) were extracted for the years 1996 to 2018 (supplementary S1 Data). Most records arise from electro fishing surveys that were carried out in summer and autumn. To minimize possible bias from misidentified specimens or erroneous records, only records with secure species identification were considered. To prevent pseudo‐replication, records that fell within the same 30 arcsec grid cells (approximately 0.625 km2 in the research area) were entered as one case in the SDM model.

Niche models

For each of the six study species, potential distribution, i.e., habitat suitability, was assessed using niche-based SDMs constructed in Maxent (version 3.4.1., [50]). Maxent makes use of a machine-learning maximum entropy framework to relate environmental conditions at species occurrence sites to the background environment of the study area, which is obtained by sampling 10,000 random locations. The established relationships are then used to predict habitat suitability across the landscape. In contrast to most other SDM algorithms, Maxent does not require a priori assumptions about species absence, thereby allowing for seamless application to the presence-only dataset used in this study [30,5052]. Bioclimatic predictors were obtained from WorldClim (, [53]), at a resolution of 30 arcsec. Environmental data included slope and Strahler order as a measure of stream size. Slope was calculated from elevation maps using the gdal slope plugin in QGIS (v. 3.4) and Strahler order numbers were extracted from the European Catchments and Rivers Network System provided by the European Environment Agency (Ecrins, version 1, Jun. 2012, Denmark). Values range from 1 (small rhithral headwaters) to 8 (large rivers). Similar sets of candidate predictors have been previously used in ecological niche models for freshwater fish and macroinvertebrates in Baden-Württemberg [15,54,55].

The following variables were selected to represent variation in temperature (see supplementary Table A in S1 Text for justification): mean diurnal range (MDR, bio2), isothermality (ISO, bio3), temperature seasonality (TS, bio4), mean summer temperature (SuT, bio10), and mean winter temperature (WiT, bio11). Values for mean summer precipitation (SuP, bio18) and mean winter precipitation (WiP, bio19) were retained as predictors of precipitation.

The Maxent training and prediction area was limited to grid cells covering lotic surface waters. To correct for uneven sampling effort across the investigation area, a sampling bias grid was specified in Maxent, which indicated relative sampling effort based on the number of sampling sites per cell [30]. For each species, 20 replicate models were fitted with cross-validation [30,50,55]. For this, occurrence data were randomly split into 20 equal-sized folds, and models were run leaving out one fold at a time. In each model iteration, the left-out fold of occurrence data was then used for model evaluation. Using this approach, each occurrence point was used at least once for training and evaluation of the models. Final projections were obtained by averaging the 20 replicate models. Models were run using the default settings, except for the regularization multiplier, which was set to 1.5 (cf. [54,56]). For perch and wels catfish, the regularization multiplier was adjusted to 3 to eliminate overfitting and implausible response curves encountered in preliminary model runs with the multiplier set at 1.5.

For further analysis, the logistic model outputs were converted to binary presence-absence maps using the ‘10th percentile training presence’ threshold (10P) to assess the ‘core distribution’ areas with suitable environment for the assessed fish species [30]. In addition, the ‘equate entropy of thresholded and original distributions’ threshold (EET) was used, which generates less conservative predictions than the 10P (data for EET are presented as supplementary Table F in S1 Text).

Distribution projection

To forecast future occurrence probabilities for the six fish species, final models were projected to future climate conditions derived from 13 GCMs on basis of the CMIP5 data set (supplementary Table B in S1 Text; across two representative carbon pathways (RCP4.5 and 8.5) and two time spans (near future, ‘2050’, average for 2041–2060, and far future, ‘2070’, average for 2061–2080). RCP4.5 and RCP8.5 correspond to ‘moderate’ and ‘high’ emission climate change scenarios, respectively, and were used to cover the most likely outcomes of future global change [40,57]. The moderate scenario envisages a ‘emissions-reduction‘ future (RCP4.5) that might be more appropriate if climate mitigation policies are put into place and tipping points are not reached. The high emission scenario represents climate change under a ‘fossil fuel intensive’ future with highest changes in temperature and precipitation (RCP8.5).

This multi-GCM approach resulted in a total of 104 binary distribution maps per species. Forecasted changes in habitat suitability were subsequently presented either as change in suitable area (given as percentage of the total area). Ternary plots were drawn to illustrate the dynamic changes i.e. loss vs. gain of suitable habitat and potential migrations of fish species over time. When comparing forecasted habitat suitability between present and future conditions, three potential outcomes can be assumed for each grid cell: 1) `stay´ –the cell is predicted as suitable under present and future conditions, 2) `loss´ –the cell is predicted as suitable under current but not future conditions, and 3) `gain´ –the cell is predicted as unsuitable under current conditions but will probably become suitable in the future. These potential outcomes were taken as endpoints for triangular plots to visualize the relative contribution of each process (stay, loss, and gain) to the predicted change in habitat suitability. This was done for all future predictions, whereby hulls were drawn for each combination of time span and RCP, connecting the extreme coordinates of the 13 different climate models.

Management and statistical analysis of georeferenced data

Data preparation, analyses and visualizations were performed in QGIS (version 3.10.3) and R (version 3.6.3 running in Rstudio Version 1.4.1103), while ecological niche modelling was done using Maxent (version 3.4.0). To assess the accuracy of the Maxent models, the area under the receiver operating characteristic curve (AUC) is given. AUC values range between 0.5 for random prediction to 1.0 for perfect prediction [30], values below 0.5 are worse than random guess, and values lower than 0.7 indicate low accuracy of the model, while values higher than 0.7 represent adequate discrimination capacity [58,59]. In addition, response curves of the predictor variables were checked for ecologically plausible behavior and signs of potential overfitting, e.g., random curve behavior. Distribution graphs were drawn with R (packages: ggplot2, ggtern). Statistics were calculated with R (packages: emmeans, gtools, PMCMRplus, stats) and Past (version 4.01, [60]). Effects of the two main factors, i.e. time span (year) and RCP, on species habitat suitability (percentage of total area predicted as suitable) and on the median elevation at which the species is expected to be distributed were examined using least square linear models (ANOVA with Tukey´s HSD post hoc test)


The average SDM outputs for current distributions of the six fish species are shown in Fig 2 and the respective model statistics are summarized in Tables 2 and 3 presenting the average AUC scores, 10th percentile training presence thresholds, and variable contributions. Presence probabilities were mostly determined by stream size, followed by temperature and precipitation variables. Individual variable contributions and response curves are summarized by species in supplementary (Table C in S1 Text).

Fig 2. Predicted habitat suitability for six fish species in southwestern Germany under current climatic conditions (a: Brown trout, b: Grayling, c: Stone loach, d: European perch, e: Common roach, and f: Wels catfish).

Areas with high suitability are colored blue (10th percentile cut off), habitats with moderate conditions are green (EET), unsuitable areas are gray. (Base layer of the map is from

Table 2. Species distribution model parameters for six fish species in southwestern Germany and predicted habitat suitability (median and interquartile range, 25%-75%, 10th percentile cut off) under current and future climatic conditions derived from 13 different climate models (see Table B in S1 Text for details) across two RCP pathways.

Table 3. Effects of time (year) and RCP on habitat suitability (percentage of area predicted as suitable) for six different fish species in southwestern Germany (estimated marginal means/least square means).

Suitable existing habitat for brown trout is predicted almost throughout Baden-Württemberg (Figs 2A and 3A), although the predicted core distribution range (based on the 10P threshold) excludes most of the Upper Rhine plain and low-gradient streams north of the Black Forest. Under future climatic changes the area of suitable habitat for brown trout is predicted to decline considerably under both RCP scenarios (F = 109.2, p < 0.001) and time spans (F = 10.14, p = 0.008). Predicted future projections suggest that less than 15% of currently suitable area in Baden-Württemberg will remain suitable for trout populations (Table 2), and only 1–2% of new habitat will be gained. It appears that future brown trout habitat will be concentrated in the elevated mountainous regions of the Black Forrest, Swabian Jura and prealpine regions of the Allgäu, indicated by an upwards shift of the average elevation range of more than 100 m (current: 489 m to future median 645–668 m, see Table D and E in S1 Text, Figs B and C in S1 Text for details). Ternary plots indicate severe range reductions: all but one model showed habitat loss under future conditions (Fig 4A).

Fig 3. Predicted changes in suitable habitat area for six fish species in southwestern Germany under anticipated climate change (a: Brown trout, b: Grayling, c: Stone loach, d: European perch, e: Common roach, and f: Wels catfish).

Data is presented for 13 different climate change models (10th percentile cut off; see Table B in S1 Text for details) as median (points, dashed line) and most likely variation (interquartile range: 25–75 interquartile range % percentiles, shaded area) for different RCP pathways (green: RCP4.5 and red: RCP8.5). Significant changes (One-Way ANOVA with Tukey HSD) between time spans (capital letters, bold) and RCP pathways (lowercase letters, italic) are highlighted.

Fig 4. Triangle plots of predicted distribution changes between current conditions and predicted future conditions (a: Brown trout, b: Grayling, c: Stone loach, d: European perch, e: Common roach, and f: Wels catfish).

Suitable habitat range shifts are displayed for “stay” (suitable under current and future conditions), “loss” (current: Suitable, future: Not suitable) and “gain” (current: Not suitable, future: Suitable) (see Fig A in S1 Text for explanation). Symbols represent distribution changes between current conditions and individual future predictions (13 different climatic models across two RCP scenarios and years; see Table B in S1 Text for details); hulls were drawn for each year-RCP combination connecting the extreme coordinates of the predicted changes.

Under current conditions, suitable habitat for grayling in Baden-Württemberg is mostly predicted in lowland river stretches (Fig 2B). With ongoing climate changes this area of suitable habitat will decline significantly by 2070 (6.5–7.6% reduction, F = 16.31, p <0.001), with minimal predicted area gain (0.1%, Table 2). Most climatic models indicate a decrease in distribution area or describe a stable situation with little changes (Fig 4B).

The habitat suitability model for the stone loach estimates 43.6% suitable habitat of Baden-Württembergs grid cells covering lotic surface waters under current climatic conditions (Fig 2C). Notwithstanding a trend towards range reduction, the models predict no significant changes resulting from future conditions (year: F2,24 = 2.7, p = 0.08; RCP: F1,12 = 0.3, p = 0.58). This is due to the wide variation in predictions based on the RCP8.5 scenario (red shaded area, Fig 3C) in which loach are predicted to experience both habitat loss (9.4–19.8%) and habitat gain (2050: 5.1–5.3%, Table 2). These dynamics are also visualized in the ternary plot, where values range from the remain/increase section (mostly RCP4.5 models, Fig 4C) to the extreme habitat decrease section on the opposite side (mostly RCP8.5 models).

The model for perch predicts 42.7% suitable habitat in Baden-Württemberg and reflects the species wide distribution in particular along the Rhine valley and in northern and southern low mountain ranges (Odenwald, Allgäu). Irrespective of the underlying RCP, perch habitat is likely to increase vastly (28.4–38.6% near future 2050, 40.7–46.3% in the far future, Table 2) under predicted future conditions, a prediction reinforced by colored areas in the ‘increase’ section of the ternary plot (Fig 4D).

Roach habitat in Baden-Württemberg is currently limited to northern and western lowland river stretches (median elevation of 330 m) and is very similar to the distribution of perch, with higher elevations being not suitable (Fig 2E). Under expected climatic changes, roach habitat will increase significantly (year: F2,24 = 6.17, p = 0.007; RCP: F1,12 = 6.26, p = 0.028), with a wide range of variation in the model predictions (Table 2, e.g., habitat gain ranges from 0.3–32.8% interquartile range for 2050 RCP8.5; Fig 3E, shaded area). Hence while most future distribution models for roach show stable distributions (‘stay’), others predict extreme increases or declines (Fig 4E).

Current wels catfish habitat is limited to lowland regions of Baden-Württemberg in areas similar to those occupied by roach and perch but more restricted to big streams and their major tributaries (Fig 2F and Table C in S1 Text: 61% importance by stream size, Table D in S1 Text: median elevation of 277 m). This distribution is unlikely to undergo significant changes under climate change (year: F2,24 = 3.11, p = 0.06; RCP: F1,12 = 0.91, p = 0.36). In detail wels catfish are set to lose very little of their current areas (0.1–0.1% range of means, 0–13.6% IQR loss, Table 3), but appear to gain some areas that are currently not suitable for their distribution (24.4–42.4% range of mean, 0.2–50.5% IQR win). Under severe RCP8.5 scenarios wels catfish appear to gain more habitat than under the moderate RCP4.5 (Table 2), indicated by the line plot showing this indifferent pattern of habitat shift under far future conditions (2070, Fig 4F). A few models predict a decrease in distribution area (right side, Fig 4F), whereas most model runs showed future distribution increase with marginal losses (left side, Fig 4F).


Niche-based SDMs are a powerful tool in understanding species’ environmental niches and potential ranges [61,62] and are widely used in freshwater ecology to relate fish community structure to environment [31,63,64] and to predict future distributions under changing conditions [18,37,53]. In southwestern Germany, SDMs were recently used to underpin risk assessment and conservation planning for endangered freshwater crayfish [15,54] and to assess the risk of fish disease spread under projected climate change in a single-GCM approach [55]. This study went beyond previous SDM applications by integrating projections from 13 different GCMs across two realistic RCP scenarios in an effort to account for the inherent variability of climate models and to arrive at robust predictions [35,36]. This multi-GCM approach is likely to better capture trajectories of future species distributions than SDM projections based on single or only few GCMs [35].

The developed SDMs were able to capture the known ranges of all of the six fish species and the response curves exhibited ecologically plausible behavior (Figs C-H in S1 Text) in accordance with existing knowledge of the modelled species and landscapes [46,48]. The AUC values of the SDMs also indicate good model performance (AUC 0.7–0.9 for all species except brown trout; (cf. [65]). The lower AUC score of the SDM for brown trout (0.6) is probably related to its exceptionally wide distribution in the study area, since great niche breadth is known to decrease AUC [54,62]. The set of predictor variables might also miss environmental variables important to the occurrence of brown trout, although this is deemed unlikely based on the well-known environmental requirements of this species [46,47]. Notwithstanding the low AUC score, the brown trout model shows plausible behavior and realistic predictions, consistent with understanding of the species’ biology [48,66,67]. The presented models are therefore considered indicative of future fish distributions under anticipated climate change.

Climate change effects on fish distributions

Anticipated climate change, as assessed by the 13 GCMs, is likely to affect the distribution area of four out of the six modelled fish species. Specifically, the potential ranges of grayling and brown trout are predicted to decline over time (up to 75 and 92% under RCP8.5, far future conditions), whereas habitat suitability for roach and perch is predicted to increase (up to 53 and 108% respectively). The potential distribution area of wels catfish and stone loach meanwhile are not predicted to undergo significant change, although the extent of suitable habitat for the former is predicted to increase in most climate models (increases of up to 163%). These findings are generally consistent with previous research predicting climate `losers´ or `winners´ and those less sensitive to climate change [18,26,68].

The cold-adapted stenothermic such as brown trout and grayling are consistently predicted to be losers to climate change, while perch and roach, both eurythermic habitat generalists, are predicted to be winner. This distinct pattern provisionally supports the use of species-level traits for quick assessment of the climate‐change susceptibility of fishes (cf. [69]) and hints at differential susceptibility of fish communities along the upstream–downstream gradient under climate-change. Salmonid-dominated communities in the upper stream reaches seem particularly sensitive when compared to the cyprinid-dominated fish communities in downstream sections (cf. [26,70]). The hydrographic network of Baden-Württemberg is dominated by headwaters and small rivers, which are highly sensitive to climate change. This may result in large-scale regression of fish species that are currently occupying broad geographic niches, exemplified by the brown trout in this study.

The predicted changes in habitat suitability in face of expected global warming result from a composite of loss areas and gain areas, and areas that maintain suitability over time (stay areas). Considering the relative contributions of these processes can help to identify range shifts that may go unnoticed if only the combined picture is considered. For instance, about 10% of the potential future range of stone loach is associated with gains that are outweighed by losses. This indicates that the predicted range of loach is geographically shifting over time, despite no significant overall quantitative change in suitable habitat is predicted for the study area. Notably, brown trout is also predicted to gain habitat in high-elevation areas as part of an uphill shift; but this gain cannot even remotely compensate for the losses of suitable habitat predicted in lower-elevation areas. With regard to potential gain areas, it is also important to note that the predicted changes in habitat suitability will probably occur faster than fish populations can adapt and disperse [18]. Furthermore, potential gain areas may not be accessible to fish because of artificial migration barriers, such as river dams, or natural barriers to dispersal, such as catchment boundaries or unfavorable hydromorphological conditions. Predicted gain areas should therefore be interpreted with care, as opposed to loss areas, which are likely to manifest directly.

The magnitude of forecasted climate change, i.e. the underlying RCP scenario, had a distinct effect on the potential ranges of the two salmonid species and of roach, with stronger predicted range loss (salmonids) or gain (roach) occurring under the high emission scenario. This finding re-emphasizes the notably high sensitivity of salmonids to climate-change effects (cf. [26,55,7072]) and the fact that detrimental effects of global warming scale with the magnitude of climate change [1,73]. In this regard, it is important to highlight that this study contrasts two realistic emission scenarios rather than best- and worst-case scenarios, which typically span a wider spectrum of potential future climates. The scenarios used in this study fall within the likely trajectory already set by humanity [7375] and findings shown here indicate that the moderate emission scenario (RCP4.5) holds considerably less drastic implications for fish.

Regardless of the scenarios used, cold water adapted species such as trout and grayling will decline in their distribution in the future, with expected consequences for freshwater habitats, their conversation purposes and value for fisheries. Warm-water adapted or generalist species like wels or perch on the other hand can benefit from anticipated changes, provided that potential new habitat is also accessible to the fish. In addition to range expansion, these species may further increase in abundance within their range, as currently reported for wels [46,68].

Implications for species management

Anticipated climate change will inevitably impact fish communities (this study, [6]) and exacerbate existing stressors [11,14,16,55,76], with cascading effects on ecosystem functioning and services like shifts in body size and biomass fluxes, decoupling of food web interactions or even partial collapses of food webs when predators decline [7780]. Besides functioning, ecosystem services will as well face climate induced changes. For instance, a decline in salmonid fisheries under future climate change, as forecasted in this study, is likely to carry negative economic implications (cf. [81,82]), even if future angling interests shift towards climate-change winners, such as perch or wels catfish (this study, [26,68]). Furthermore, current conservation policies have virtually no experience with wide range shifts of fish communities and their ecosystems. Finally, species may venture into new habitats where they might have to establish new inter-species relationships that may become competitive or mutualistic, parasitic or pathogenic. This may require a paradigm shift in how to deal with migrations of species to localities where these were rarely seen and the way we think about protective areas [8385].

Efforts to preserve functioning and ecosystem services of freshwater ecosystems must therefore account for climate change, and mitigation actions to restore and increase the resilience of fish habitats to higher ambient temperatures and extreme weather effects are urgently needed [8688]. Salmonid fish, for example, would greatly benefit from shading of headwaters by fostering native riparian woods [89,90], restauration of habitat connectivity [88,9193] and preservation of cold-water refugia [94,95], in addition to measures that secure natural surface water runoffs, in particular during drought conditions [86,88]. Finally, and above all, limiting global warming, as pledged under the Paris Agreement [96], is imperative in maintaining native fish communities and their ecosystem services, especially since ecosystem disruption may occur abruptly [97].

Supporting information

S1 Data. Fish occurrence points used for SDM analysis.


S1 Text. Tables (A-F) and Figs (A-H) with additional information for the study.

Table A: Predictor variables of fish species presence probability used in the species distribution models, including justification for inclusion and data source. The range and average values of under current climate conditions are given for study area. Table B: Thirteen different climate change data sets were used to generate future predictions for fish species. All data were obtained from the database website. Table C: Estimates of relative contributions of the environmental variables to the Maxent model for different fish species under current conditions in Baden-Württemberg. Table D: Elevation range of Maxent model for fish species under current and future conditions in Baden-Württemberg (average median and interquartile range, 25%-75%, 10th percentile cut off). Table E: Analysis of Variance Model with Post Hoc Test (estimated marginal means/least square means) for elevation distribution of six different fish species in southwestern Germany under current and future climatic conditions. Significant differences to current distribution are highlighted in bold. Table F: Species distribution model parameters for fish species in Baden-Württemberg. Suitable range (median and interquartile range, 25%-75%, ‘equate entropy of thresholded and original distributions threshold’ cut off) is given for current and future climatic conditions for 13 different models (see Table B in S1 Text for details), two timespans and two different RCP pathways. The values correspond to percentage of total assessed area. Maxent AUC for training and test are given as mean ± SD. [Compare with Table 2 for 10th percentile cut off values for more conservative/restrict future estimations]. Fig A: Triangle plot of theoretical species distribution changes between current conditions and predicted future conditions. X axis (“new habitat”) illustrates a gain of new suitable habitat, y axis (“old habitat”) represents percentage of area remaining suitable for the species whereas z-axis (“lost habitat”) highlights a reduction in suitable habitat. The population “remains” in its original space, when more than 60% of habitat stays suitable. If the portion of lost habitat is greater than the portion gained the overall population will “decrease” (loss>gain, right sector), if vice versa the population will “increase” (gain>loss, left sector). If the proportions of loss and gain are almost equal, a species migration will occur (“migrate”, lower middle sector). Fig B: Elevation range of Maxent models for fish species (A: Trout, B: Grayling, C: Stone loach, D: Perch, E: Roach, F: Catfish) under current (solid black line) and future conditions in Baden-Württemberg (light blue 2050, RCP 4.5; blue: 2050, RCP 8.5; orange: 2070 4.5, red: 2070 RCP8.5). Gray bars represent the elevation range of the study area in southern Germany. Fig C: Species occurrence and distribution map of brown trout (Salmo trutta) under current and future climatic conditions in the study area of Baden-Württemberg. Species occurrence (blue points) results in a Maxent binary map under current conditions (10P cut off, blue area), future graphs display gradient maps for 13 different climatic models for near (2050) and far future (2070) under two different RCP scenarios. White area contains no surface water bodies; grey indicates unsuitable habitat speciconditions. The more models predict suitability, the darker the colors (Base layer of the map is from Fig D: Species occurrence and distribution map of grayling (Thymallus thymallus) under current and future climatic conditions in the study area of Baden-Württemberg. Species occurrence (blue points) results in a Maxent binary map under current conditions (10P cut off, blue area), future graphs display gradient maps for 13 different climatic models for near (2050) and far future (2070) under two different RCP scenarios. White area contains no surface water bodies; grey indicates unsuitable habitat conditions. The more models predict suitability, the darker the colors (Base layer of the map is from Fig E: Species occurrence and distribution map of stone loach (Barbatula barbatula) under current and future climatic conditions in the study area of Baden-Württemberg. Species occurrence (blue points) results in a Maxent binary map under current conditions (10P cut off, blue area), future graphs display gradient maps for 13 different climatic models for near (2050) and far future (2070) under two different RCP scenarios. White area contains no surface water bodies; grey indicates unsuitable habitat conditions. The more models predict suitability, the darker the colors (Base layer of the map is from Fig F: Species occurrence and distribution map of perch (Perca fluviatilis) under current and future climatic conditions in the study area of Baden-Württemberg. Species occurrence (blue points) results in a Maxent binary map under current conditions (10P cut off, blue area), future graphs display gradient maps for 13 different climatic models for near (2050) and far future (2070) under two different RCP scenarios. White area contains no surface water bodies; grey indicates unsuitable habitat conditions. The more models predict suitability, the darker the colors (Base layer of the map is from Fig G: Species occurrence and distribution map of roach (Rutilus rutilus) under current and future climatic conditions in the study area of Baden-Württemberg. Species occurrence (blue points) results in a Maxent binary map under current conditions (10P cut off, blue area), future graphs display gradient maps for 13 different climatic models for near (2050) and far future (2070) under two different RCP scenarios. White area contains no surface water bodies; grey indicates unsuitable habitat conditions. The more models predict suitability, the darker the colors (Base layer of the map is from Fig H: Species occurrence and distribution map of catfish (Silurus glanis) under current and future climatic conditions in the study area of Baden-Württemberg. Species occurrence (blue points) results in a Maxent binary map under current conditions (10P cut off, blue area), future graphs display gradient maps for 13 different climatic models (10P cut off) for near (2050) and far future (2070) under two different RCP scenarios. White area contains no surface water bodies; grey indicates unsuitable habitat conditions. The more models predict suitability, the darker the colors (Base layer of the map is from



We thank S. Blank (Fisheries Research Station Baden-Württemberg) for his invaluable assistance on handling the FiAKa database and A.J. Beer for language correction.


  1. 1. Steffen W, Rockström J, Richardson K, Lenton TM, Folke C, Liverman D, et. al. Trajectories of the Earth System in the Anthropocene. Proc Natl Acad Sci USA. 2018;115(33):8252–9. pmid:30082409
  2. 2. Sala OE, Chapin FS, Armesto JJ, Berlow E, Bloomfield J, Dirzo R, et. al. Biodiversity—Global biodiversity scenarios for the year 2100. Science. 2000;287(5459):1770–4. pmid:10710299
  3. 3. Parmesan C, Yohe G. A globally coherent fingerprint of climate change impacts across natural systems. Nature. 2003;421(6918):37–42. pmid:12511946
  4. 4. Hickling R, Roy DB, Hill JK, Fox R, Thomas CD. The distributions of a wide range of taxonomic groups are expanding polewards. Glob Change Biol. 2006;12(3):450–5.
  5. 5. Román-Palacios C, Wiens JJ. Recent responses to climate change reveal the drivers of species extinction and survival. Proc Natl Acad Sci USA. 2020;117(8):4211–7. pmid:32041877
  6. 6. Woolway RI, Maberly SC. Climate velocity in inland standing waters. Nat Clim Chang. 2020;1–6. pmid:33262808
  7. 7. Woodward G, Perkins DM, Brown LE. Climate change and freshwater ecosystems: impacts across multiple levels of organization. Philos Trans R Soc Lond B Biol Sci. 2010;365(1549):2093–106. pmid:20513717
  8. 8. Xenopoulos MA, Lodge DM, Alcamo J, Märker M, Schulze K, Van Vuuren DP. Scenarios of freshwater fish extinctions from climate change and water withdrawal. Glob Change Biol. 2005;11(10):1557–64.
  9. 9. Petsch DK. Causes and consequences of biotic homogenization in freshwater ecosystems. Int Rev Hydrobiol. 2016;101(3–4):113–22.
  10. 10. Reid AJ, Carlson AK, Creed IF, Eliason EJ, Gell PA, Johnson PT, et. al. Emerging threats and persistent conservation challenges for freshwater biodiversity. Biol Rev. 2019;94(3):849–73. pmid:30467930
  11. 11. Stendera S, Adrian R, Bonada N, Cañedo-Argüelles M, Hugueny B, Januschke K, et. al. Drivers and stressors of freshwater biodiversity patterns across different ecosystems and scales: a review. Hydrobiologia. 2012;696(1):1–28.
  12. 12. Birk S, Chapman D, Carvalho L, Spears BM, Andersen HE, Argillier C, et. al. Impacts of multiple stressors on freshwater biota across spatial scales and ecosystems. Nat Ecol Evol. 2020;4(8):1060–8. pmid:32541802
  13. 13. Weijters MJ, Janse JH, Alkemade R, Verhoeven JT. Quantifying the effect of catchment land use and water nutrient concentrations on freshwater river and stream biodiversity. Aquat Conserv. 2009;19(1):104–12.
  14. 14. Brinker A, Chucholl C, Behrmann‐Godel J, Matzinger M, Basen T, Baer J. River damming drives population fragmentation and habitat loss of the threatened anube streber (Zingel streber): Implications for conservation. Aquat Conserv. 2018;28(3):587–99.
  15. 15. Chucholl C. The bad and the super-bad: prioritising the threat of six invasive alien to three imperilled native crayfishes. Biol Invasions. 2016;18:1967–88.
  16. 16. Gallardo B, Clavero M, Sánchez MI, Vilà M. Global ecological impacts of invasive species in aquatic ecosystems. Glob Change Biol. 2016;22(1):151–63. pmid:26212892
  17. 17. Buisson L, Grenouillet G, Villéger S, Canal J, Laffaille P. Toward a loss of functional diversity in stream fish assemblages under climate change. Glob Change Biol. 2013;19(2):387–400. pmid:23504778
  18. 18. Radinger J, Essl F, Hölker F, Horkỳ P, Slavík O, Wolter C. The future distribution of river fish: the complex interplay of climate and land use changes, species dispersal and movement barriers. Glob Change Biol. 2017;23(11):4970–86. pmid:28500795
  19. 19. Holmlund CM, Hammer M. Ecosystem services generated by fish populations. Ecol Econ. 1999;29(2):253–68.
  20. 20. Balian EV, Segers H, Martens K, Lévéque C. The freshwater animal diversity assessment: an overview of the results. In: Balian EV, Lévêque C, Segers H, Martens K (eds) Freshwater Animal Diversity Assessment. Developments in Hydrobiology, vol 198. Springer, Dordrecht. 2007;627–37.
  21. 21. Fricke R, Eschmeyer WN, Van der Laan R (eds). Eschmeyer´s Catalog of fishes: genera, species, references. Internet, 1. March 2022.
  22. 22. Cooke SJ, Arlinghaus R, Johnson BM, Cowx IG. Recreational fisheries in inland waters. In: Craig JF (eds.) Freshwater Fisheries Ecology. Wiley-Blackwell Oxford, UK; 2016.
  23. 23. Tacon AG. Trends in global aquaculture and aquafeed production: 2000–2017. Rev Fish Sci Aquac. 2020;28(1):43–56.
  24. 24. Comte L, Olden JD. Evolutionary and environmental determinants of freshwater fish thermal tolerance and plasticity. Glob Change Biol. 2017;23(2):728–36. pmid:27406402
  25. 25. Isaak DJ, Rieman BE. Stream isotherm shifts from climate change and implications for distributions of ectothermic organisms. Glob Change Biol. 2013;19(3):742–51. pmid:23504832
  26. 26. Buisson L, Thuiller W, Lek S, Lim PUY, Grenouillet G. Climate change hastens the turnover of stream fish assemblages. Glob Change Biol. 2008;14(10):2232–48.
  27. 27. Pont D, Logez M, Carrel G, Rogers C, Haidvogl G. Historical change in fish species distribution: shifting reference conditions and global warming effects. Aquat Sci. 3. 2015;1–13. pmid:26257502
  28. 28. Guisan A, Thuiller W. Predicting species distribution: offering more than simple habitat models. Ecology Letters. 2005;8(9):993–1009. pmid:34517687
  29. 29. Elith J, Leathwick JR. Species Distribution Models: Ecological explanation and prediction across space and time. Annu Rev Ecol Evol Syst. 2009;40(1):677–97.
  30. 30. Elith J, Phillips SJ, Hastie T, Dudík M, Chee YE, Yates CJ. A statistical explanation of MaxEnt for ecologists. Divers Distrib. 2011;17(1):43–57.
  31. 31. Schmidt H, Radinger J, Teschlade D, Stoll S. The role of spatial units in modelling freshwater fish distributions: Comparing a subcatchment and river network approach using MaxEnt. Ecol Model. 2020;418:108937.
  32. 32. Wenger SJ, Som NA, Dauwalter DC, Isaak DJ, Neville HM, Luce CH, et. al. Probabilistic accounting of uncertainty in forecasts of species distributions under climate change. Glob Change Biol. 2013;19(11):3343–54. pmid:23765608
  33. 33. Kuemmerlen M, Schmalz B, Cai Q, Haase P, Fohrer N, Jähnig SC. An attack on two fronts: predicting how changes in land use and climate affect the distribution of stream macroinvertebrates. Freshw Biol. 2015;60(7):1443–58.
  34. 34. Buisson L., Thuiller W., Casajus N., Lek S., Grenouillet G. Uncertainty in ensemble forecasting of species distribution. Glob Change Biol. 2010; 16(4), 1145–1157.
  35. 35. Goberville E., Beaugrand G., Hautekèete N. C., Piquot Y., Luczak C. Uncertainties in the projection of species distributions related to general circulation models. Ecol Evol. 2015;5(5), 1100–1116. pmid:25798227
  36. 36. Filipe AF, Markovic D, Pletterbauer F, Tisseuil C, De Wever A, Schmutz S, et. al. Forecasting fish distribution along stream networks: brown trout (Salmo trutta) in Europe. Divers Distrib. 2013;19(8):1059–71.
  37. 37. Pletterbauer F, Melcher AH, Ferreira T, Schmutz S. Impact of climate change on the structure of fish assemblages in European rivers. Hydrobiologia. 2015;744(1):235–54.
  38. 38. Navarro-Racines C, Tarapues J, Thornton P, Jarvis A, Ramirez-Villegas J. High-resolution and bias-corrected CMIP5 projections for climate change impact assessments. Scientific Data. 2020;7(1):7. pmid:31959765
  39. 39. Knutti R, Sedláček J. Robustness and uncertainties in the new CMIP5 climate model projections. Nat Clim Chang. 2013;3(4):369–73.
  40. 40. IPCC. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Internet]. Geneva, Switzerland; 2014.
  41. 41. Schellnhuber HJ, Rahmstorf S, Winkelmann R. Why the right climate target was agreed in Paris. Nat Clim Chang. 2016;6:649–653.
  42. 42. Raftery AE, Zimmer A, Frierson DMW, Startz R, Liu P. Less than 2 °C warming by 2100 unlikely. Nat Clim Chang. 2017;7(9):637–41. pmid:30079118
  43. 43. Wang Z, Lin L, Zhang X, Zhang H, Liu L, Xu Y. Scenario dependence of future changes in climate extremes under 1.5 °C and 2 °C global warming. Scientific reports. 2017;7(1):1–9. pmid:28127051
  44. 44. Xu C, Kohler TA, Lenton TM, Svenning J-C, Scheffer M. Future of the human climate niche. Proc Natl Acad Sci USA. 2020;117(21):11350–5. pmid:32366654
  45. 45. Krabbenhoft TJ, Myers BJ, Wong JP, Chu C, Tingley RW, Falke JA, Kwak TJ, et. al. FiCli, the fish and climate change database, informs climate adaptation and management for freshwater fishes. Sci Data. 2020; 7:124. pmid:32317639
  46. 46. Dußling U, Baer J, Gaye-Siessegger J, Schumann M, Blank S, Brinker A. Das große Buch der Fische Baden-Württembergs. Stuttgart: Ministerium für Ländlichen Raum und Verbraucherschutz Baden-Württemberg (eds.); 2018. 372p.
  47. 47. Crisp DT. Environmental requirements of common riverine European salmonid fish species in fresh water with particular reference to physical and chemical aspects. Hydrobiologia. 1996;323:201–21.
  48. 48. Matthews WJ. Patterns in Freshwater Fish Ecology. New York: Chapman & Hall; 1998. 756 p.
  49. 49. Arlinghaus R. Recreational fisheries in Germany: a social and economic analysis-report of the IGB. Berlin: Department of Biology and Ecology of Fishes, Leibniz Institute of Freshwater Ecology and Inland Fisheries; 2004; 18: 168 p.
  50. 50. Phillips SJ, Anderson RP, Schapire RE. Maximum entropy modeling of species geographic distributions. Ecol Model. 2006;190(3–4):231–59.
  51. 51. Wisz MS, Hijmans RJ, Li J, Peterson AT, Graham CH, Guisan A, NCEAS Predicting Species Distributions Working Group. Effects of sample size on the performance of species distribution models. Divers Distrib. 2008; 14(5), 763–773.
  52. 52. Merow C, Smith MJ, Silander JA Jr. A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter. Ecography; 2013; 36(10), 1058–1069.
  53. 53. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. Very high resolution interpolated climate surfaces for global land areas. Int J Climatol. 2005;25(15):1965–78.
  54. 54. Chucholl C. Niche-based species distribution models and conservation planning for endangered freshwater crayfish in south-western Germany. Aquat Conserv. 2017;27(3):698–705.
  55. 55. Ros AFH, Baer J, Basen T, Chucholl C, Schneider E, Teschner R, et. al. Current and projected impacts of the parasite Tetracapsuloides bryosalmonae (causative to proliferative kidney disease) on Central European salmonid populations under predicted climate change. Fresh Biol. 2021;66(6):1182–99.
  56. 56. Gallardo B, Aldridge DC. The „dirty dozen“: socio-economic factors amplify the invasion potential of 12 high-risk aquatic invasive species in Great Britain and Ireland. J Appl Ecol. 2013;50(3):757–66.
  57. 57. Schwalm CR, Glendon S, Duffy PB. RCP8.5 tracks cumulative CO2 emissions. Proc Natl Acad Sci USA. 2020;117(33):19656–7. pmid:32747549
  58. 58. Swets JA. Measuring the accuracy of diagnostic systems. Science. 1988;240(4857):1285–93. pmid:3287615
  59. 59. Pearce J, Ferrier S. Evaluating the predictive performance of habitat models developed using logistic regression. Ecol Modell. 2000;133(3):225–45.
  60. 60. Hammer Ø, Harper DA, Ryan PD. PAST: Paleontological statistics software package for education and data analysis. Palaeontol electron. 2001;4(1):9.
  61. 61. Elith J, Graham CH., Anderson RP., Dudík M, Ferrier S, Guisan A, et. al. Novel methods improve prediction of species’ distributions from occurrence data. Ecography. 2006;29(2):129–51.
  62. 62. Aguirre-Gutiérrez J, Carvalheiro LG, Polce C, van Loon EE, Raes N, Reemer M, et. al. Fit-for-purpose: species distribution model performance depends on evaluation criteria–Dutch hoverflies as a case study. PloS One. 2013;8(5):e63708. pmid:23691089
  63. 63. Smokorowski KE, Pratt TC. Effect of a change in physical structure and cover on fish and fish habitat in freshwater ecosystems–a review and meta-analysis. Environ Rev. 2007;15(NA):15–41.
  64. 64. Buisson L, Blanc L, Grenouillet G. Modelling stream fish species distribution in a river network: the relative effects of temperature versus physical factors. Ecol Freshw Fish. 2008;17(2):244–57.
  65. 65. Araújo MB, Guisan A. Five (or so) challenges for species distribution modelling. J Biogeogr. 2006;33(10):1677–88.
  66. 66. Elliot JM. Quantitative Ecology and the Brown Trout. Oxford: Oxford University Press; 1994. 286 p.
  67. 67. Jonsson B, Jonsson N. Ecology of Atlantic Salmon and Brown Trout [Internet]. Dordrecht: Springer Netherlands. 2011;
  68. 68. Markovic D, Freyhof J, Wolter C. Where are all the fish: potential of biogeographical maps to project current and future distribution patterns of freshwater species. PLoS One. 2012;7(7):e40530. pmid:22792361
  69. 69. Jarić I, Lennox RJ, Kalinkat G, Cvijanović G, Radinger J. Susceptibility of European freshwater fish to climate change: Species profiling based on life-history and environmental characteristics. Glob Change Biol. 2019;25(2):448–58. pmid:30417977
  70. 70. Walters AW, Mandeville CP, Rahel FJ. The interaction of exposure and warming tolerance determines fish species vulnerability to warming stream temperatures. Biol Lett. 2018 14(9), 20180342. pmid:30185608
  71. 71. Hari RE, Livingstone DM, Siber R, Burkhardt-Holm P, Guettinger H. Consequences of climatic change for water temperature and brown trout populations in Alpine rivers and streams. Glob Change Biol. 2006;12(1):10–26.
  72. 72. Dahlke FT, Wohlrab S, Butzin M, Pörtner H-O. Thermal bottlenecks in the life cycle define climate vulnerability of fish. Science. 3. Juli 2020;369(6499):65–70. pmid:32631888
  73. 73. IPCC. Summary for Policymakers. In: Global Warming of 1.5 °C. An IPCC Special Report on the impacts of global warming of 1.5 °C above preindustrial levels and related global greenhouse gas emission pathways, in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty. Masson-Delmotte, V, Zhai P, Pörtner HO, Roberts D, Skea J, Shukla PR, Pirani A, et. al. (eds.). World Meteorological Organization, Geneva, Switzerland; 2018. 32 pp.
  74. 74. Sherwood SC, Webb MJ, Annan JD, Armour KC, Forster PM, Hargreaves JC, et. al. An assessment of Earth’s climate sensitivity using multiple lines of evidence. Rev Geophys. 2020;58(4):e2019RG000678. pmid:33015673
  75. 75. Wunderling N, Donges JF, Kurths J, Winkelmann R. Interacting tipping elements increase risk of climate domino effects under global warming. Earth Syst Dyn. 2020;1–21.
  76. 76. Horn H, Paul L, Horn W, Uhlmann D, Röske I. Climate change impeded the re-oligotrophication of the Saidenbach Reservoir. Int Rev Hydrobiol. 2015;100(2):43–60.
  77. 77. Petchey OL, McPhearson PT, Casey TM, Morin PJ. Environmental warming alters food-web structure and ecosystem function. Nature; 1999; 402(6757), 69–72.
  78. 78. Perkins DM, Reiss J, Yvon-Durocher G, Woodward G. Global change and food webs in running waters. Hydrobiologia. 2010; 657(1), 181–198.
  79. 79. Comte L, Buisson L, Daufresne M, Grenouillet G. Climate‐induced changes in the distribution of freshwater fish: observed and predicted trends. Fresh Biol. 2013; 58(4), 625–639.
  80. 80. Ledger ME, Brown LE, Edwards FK, Milner AM, Woodward G. Drought alters the structure and functioning of complex food webs. Nat Clim Chang. 2013; 3(3), 223–227.
  81. 81. Jones R, Travers C, Rodgers C, Lazar B, English E, Lipton J, et. al. Climate change impacts on freshwater recreational fishing in the United States. Mitig Adapt Strateg Glob Chang. 2013;18(6):731–58.
  82. 82. DeWeber JT, Wagner T. Predicting brook trout occurrence in stream reaches throughout their native range in the Eastern United States. Trans Am Fish Soc. 2015;144(1):11–24.
  83. 83. Hannah L, Midgley G, Andelman S, Araújo M, Hughes G, Martinez-Meyer E, et al. Protected area needs in a changing climate. Front Ecol Environ. 2007; 5(3), 131–138.
  84. 84. Dudgeon D, Arthington AH, Gessner MO, Kawabata ZI, Knowler DJ, Lévêque C, et al. (2006). Freshwater biodiversity: importance, threats, status and conservation challenges. Biol Rev. 2006; 81(2), 163–182. pmid:16336747
  85. 85. Dobrowski SZ, Littlefield CE, Lyons DS, Hollenberg C, Carroll C, Parks SA, et al. Protected-area targets could be undermined by climate change-driven shifts in ecoregions and biomes. Commun Earth Environ. 2021; 2(1), 1–11.
  86. 86. Palmer MA, Lettenmaier DP, Poff NL, Postel SL, Richter B, Warner R. Climate change and river ecosystems: protection and adaptation options. Environ Manage. 2009;44(6):1053–68. pmid:19597873
  87. 87. Markovic D, Carrizo SF, Kärcher O, Walz A, David JNW. Vulnerability of European freshwater catchments to climate change. Glob Chang Biol. 2017;23(9):3567–80. pmid:28186382
  88. 88. Grantham TE, Matthews JH, Bledsoe BP. Shifting currents: Managing freshwater systems for ecological resilience in a changing climate. Water Secur. 2019; 8:100049.
  89. 89. Bond RM, Stubblefield AP, Van Kirk RW. Sensitivity of summer stream temperatures to climate variability and riparian reforestation strategies. J Hydrol Reg Stud. 2015;4:267–79.
  90. 90. Dugdale SJ, Malcolm IA, Kantola K, Hannah DM. Stream temperature under contrasting riparian forest cover: Understanding thermal dynamics and heat exchange processes. Sci Total Environ. 2018;610:1375–89. pmid:28851157
  91. 91. Merenlender AM, Matella MK. Maintaining and restoring hydrologic habitat connectivity in mediterranean streams: an integrated modeling framework. Hydrobiologia. 2013; 719:509–525.
  92. 92. Harrod C. Climate change and freshwater fisheries. In: Freshwater fisheries ecology. Craig JF (eds). 2015; 641–694.
  93. 93. Flitcroft RL, Arismendi I, Santelmann . A Review of Habitat Connectivity Research for Pacific Salmon in Marine, Estuary, and Freshwater Environments. J Am Water Resour Assoc. 2019; 55 (2): 430–441
  94. 94. Isaak DJ, Young MK, Nagel DE, Horan DL, Groce MC. The cold-water climate shield: delineating refugia for preserving salmonid fishes through the 21st century. Glob Change Biol. 2015;21(7):2540–53. pmid:25728937
  95. 95. White SL, Kline BC, Hitt NP, Wagner T. Individual behaviour and resource use of thermally stressed brook trout Salvelinus fontinalis portend the conservation potential of thermal refugia. J Fish Biol. 2019;95(4):1061–71. pmid:31309548
  96. 96. United Nations. Paris Agreement. L 282/4 2015 4–18.
  97. 97. Trisos CH, Merow C, Pigot AL. The projected timing of abrupt ecological disruption from climate change. Nature. 2020;580(7804):496–501. pmid:32322063