Research Article |
|
Corresponding author: Mathias Kuemmerlen ( mathias@escazu.co.cr ) Academic editor: Paul B. Frandsen
© 2025 Mathias Kuemmerlen, Wolfram Graf, Johann Waringer, Simon Vitecek, Mladen Kučinić, Ana Previšić, Lujza Keresztes, Miklós Bálint, Steffen U. Pauls.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Kuemmerlen M, Graf W, Waringer J, Vitecek S, Kučinić M, Previšić A, Keresztes L, Bálint M, Pauls SU (2025) Higher predicted climate-change vulnerability for spring-dwelling freshwater biota. In: Ríos-Touma B, Frandsen PB, Holzenthal RW, Houghton DC, Rázuri-Gonzales E, Pauls SU (Eds) Proceedings of the 18th International Symposium on Trichoptera. ZooKeys 1263: 289-315. https://doi.org/10.3897/zookeys.1263.148253
|
Environmental change threatens freshwater biodiversity through altered temperature and precipitation patterns. Available data is frequently insufficient to determine impacts at the species level leading to misinterpreted species’ vulnerability. Conversely, phylogenetic relationships, current distributions and ecological traits of the caddisfly subfamily Drusinae are well known. Thus, species distribution models (SDMs) were set up for 47 Drusinae species to assess individual and trait-specific climate change (CC) vulnerability. Species were grouped by larval feeding guild, stream zonation preference and level of endemism. Models were calibrated with predictors describing climate, topography and geology at a spatial resolution of 1 km2 and were projected for five general circulation models under four future climate scenarios. To limit dispersal, distribution projections were restricted to a maximum of 500 km until the year 2080. Relative predicted range change fluctuated between -100% and 197%, with extinction predicted for five species. Altitudinal shifts varied between -2% and +15%, with distribution centroids shifting between 28 km and 119 km. Our results identify stream zonation, a non-phylogenetic trait, as the best indicator of CC vulnerability. Furthermore, two important conclusions are highlighted: monitoring is best done at the species level while the biodiversity of springs and low order streams requires considerably more attention.
Caddisflies (Trichoptera), climate change, distribution change, Drusinae, endemism, range shifts, trait evolution
Direct anthropogenic impacts on the environment have been widely documented and discussed, but continue to magnify, causing significant biodiversity loss and ecosystem degradation across all biomes (
The potential effects of global environmental change are frequently forecasted through species distribution models (SDMs), based on future environmental scenarios (
In freshwater ecosystems, for example, benthic macroinvertebrates influence energy flows and nutrient cycling (
In the present study, we projected the distributions of 47 caddisfly species of the Drusinae BANKS (1916) subfamily using SDMs and analysed their individual responses to expected future changes in climate. We account for ecological traits, as the effects of CC on modelled species’ distributions have shown to be highly variable in freshwater ecosystems (
Phylogenetic tree of Drusinae caddisflies showing ecological traits: larval feeding guild, stream zonation preference and level of endemism. Modified from
The caddisfly subfamily Drusinae (Limnephilidae) is a clearly defined evolutionary group, with a well-studied phylogeny and abundant occurrence records at the species level (
Predicted future climatic conditions in Europe foresee increased air temperatures ranging from +1.6 to +4.6 °C and an altered precipitation of -10 to +22% when comparing the time periods 1971–2000 and 2071–2100 (
The goal of the present study was to use distribution predictions to evaluate the species-specific vulnerability of the Drusinae caddisflies to CC and, in a second step, identify which ecological traits are indicative of CC vulnerability. We evaluate three hypothesis: (i) that Drusinae caddisflies are highly vulnerable to CC, losing significant areas of their range and experiencing considerable altitudinal shifts (e.g. uphill) under future climate scenarios; (ii) that species grouped by differing ecological traits have similar distribution patterns (e.g. spring species have smaller and higher ranges than stream species); and (iii) that ecological trait groups with smaller ranges at higher elevations are more vulnerable to CC, with phylogenetic constraints playing a subordinate role in CC responses. We also compare the CC vulnerability stemming from the analysis of our distribution predictions with the CCVS, an existing classification regarding CC threats for individual species (
To assure robust distribution predictions for freshwater biodiversity in general, and the Drusinae in particular, SDMs must consider environmental predictors that significantly contribute to shaping freshwater habitats (
Georeferenced occurrences of Drusinae caddisflies were taken from the Distribution Atlas of European Trichoptera (DAET;
Environmental predictors included nine topographical, geological and climatic variables, preselected based on several criteria: (i) low intercorrelation (r < |0.7|;
To forecast distributions for the year 2080, models were projected for all scenarios of the fifth Assessment Report of the Intergovernmental Panel on Climate Change (AR5, IPCC), which are based on four greenhouse gas concentration trajectories known as Representative Concentration Pathways (RCP 2.6, 4.5, 6.0 and 8.5;
RCPs are modelled predictions of global greenhouse gas (GHG) emissions that take socioeconomic drivers into account. Each RCP describes a possible pathway in the development of GHG emissions and atmospheric concentrations, based on different scenarios of air pollutant emissions and land use. RCP 2.6 assumed strict mitigation measures leading to significant reductions in GHG emissions, while RCP 8.5 allowed for very high GHG emissions. RCPs 4.5 and 6.0 were intermediate scenarios. The RCP pathways were implemented in various general circulation models (GCM), which results in different predictions of future climatic conditions. In the present study, five frequently implemented GCMs were projected for each RCP scenario (CSIRO Mk3.6.0, IPSL CM5A-LR, MRI CGCM3, NCAR CCSM4, NCC NorESM1-M). The future climate predictors were acquired from the Downscaled GCM Data Portal of the Consultative Group on International Agricultural Research (CGIAR) Research Program on Climate Change, Agriculture and Food Security (CCAFS;
The spatial resolution of the model was 1 km2 (30 arc-seconds), resulting in 10,649,788 raster grid cells in the extent considered. Models were calibrated with the R package biomod2 using three randomly selected pseudo-absence runs, five algorithms (GLM, GBM, CTA, ANN and MaxEnt) and two repetitions for each combination (
Present and future predicted distributions were first projected for all of Europe and later cropped using an area obtained by buffering the current known occurrences with a radius of 500 km. This buffer was assumed to roughly represent the maximum dispersal of Drusinae caddisflies that could be expected until the year 2080. This maximum dispersal distance was derived from the average dispersal velocity of 7.3 km per year determined for a caddisfly of the same family (Limnephilidae, Potamophylax cingulatus), during its invasion and expansion across Iceland in the last decades (
To identify differences between the current and each one of the RCP scenario distributions in terms of range changes or range shifts, we calculated the mean predicted range size, mean altitude (extracted from DEM) and the centroid of each predicted distribution. This information was also extracted for the ecological trait groups (i) larval feeding; (ii) river zonation preference; and (iii) level of endemism (
To assess the vulnerability of the Drusinae subfamily to CC, present and scenario projections were compared in terms of predicted range size and average altitude using paired Wilcoxon two-sample tests. To compare the present predictions between ecological trait groups, we used Wilcoxon two-sample tests. Paired Wilcoxon two-sample tests were used again to evaluate CC vulnerability within ecological trait groups by comparing present with scenario predictions. To confirm whether particular traits (e.g. microendemism) were truly good indicators of CC vulnerability, we tested whether their relative range changes, relative altitudinal shifts and geographic shifts differed from that of the complementary groups. Tests within groups always compared either grazers vs carnivores or spring species vs stream species. Microendemics and ecoregion endemics, however, were always compared to widespread species. For this purpose, Wilcoxon two-sample tests were used. Finally, a graphical comparison of the CCVS extracted from
Models performed very well on average, with mean values of 0.96 ± 0.06 (mean ± S.D.), 0.99 ± 0.01, 99.1 ± 2.0 and 96.7 ± 4.4 for TSS, AUC, sensitivity and specificity, respectively (Suppl. material
Annual precipitation was, on average, the most important environmental predictor across all modelled species (mean 22% ± 19% S.D.; Fig.
The highest stacked predicted probabilities of occurrence concentrate in the main European mountain ranges, such as the Pyrenees, the Alps, the Apennines, the Carpathians and the Dinaric Alps (Fig.
Average absolute and relative changes in range size, altitude and shifts in space for feeding groups and zonation preferences, for present and four RCP scenario projections. Bold values indicate significant differences.
| Larval feeding guild | Stream zonation preference | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Carnivores | Grazers | Spring species | Stream species | |||||||
| Scenario | Value | Average | SD | Average | SD | Average | SD | Average | SD | |
| Range size (thousand km2) | Present | Abs. | 266.8 | 299.3 | 322.2 | 468.4 | 196.2e | 350.2 | 526.4e | 528.4 |
| RCP 2.6 | Abs. | 292.6 | 331.5 | 328.7 | 505.5 | 200.7 | 372.4 | 545.5 | 583.8 | |
| RCP 4.5 | Abs. | 247.4a | 294.5 | 259.8c | 417.9 | 152.4c | 286.5 | 452.6a | 508.4 | |
| RCP 6.0 | Abs. | 265.6 | 298.4 | 276.0b | 433.9 | 166.4a | 304.2 | 472.5 | 520.1 | |
| RCP 8.5 | Abs. | 225.3a | 253.9 | 228.6c | 379.6 | 130.8c | 255.9 | 408.2b | 461.6 | |
| Change in range size (%) | RCP 2.6 | Rel. | -6.6 | 38.2 | -15.8 | 38.2 | -20.4 | 42.5 | 0.2 | 19.0 |
| RCP 4.5 | Rel. | -32.1 | 34.6 | -34.9 | 44.4 | -37.4 | 48.8 | -24.0 | 23.7 | |
| RCP 6.0 | Rel. | -15.8 | 36.2 | -19.8 | 64.4 | -19.5 | 70.9 | -15.8 | 23.2 | |
| RCP 8.5 | Rel. | -29.2 | 41.2 | -40.7 | 53.5 | -46.6 | 52.5 | -21.3 | 42.1 | |
| Mean altitude (m) | Present | Abs. | 1210.6d | 361.6 | 945.5d | 378.4 | 1109.5e | 428.5 | 796.9e | 274.5 |
| RCP 2.6 | Abs. | 1186.6 | 355.6 | 958.1 | 403.0 | 1131.0 | 434.6 | 780.2 | 280.5 | |
| RCP 4.5 | Abs. | 1231.0 | 469.7 | 1027.5c | 453.5 | 1227.6.0c | 485.7 | 819.3 | 327.8 | |
| RCP 6.0 | Abs. | 1234.0 | 364.3 | 963.4 | 421.0 | 1146.9 | 455.8 | 807.9 | 313.5 | |
| RCP 8.5 | Abs. | 1278.7 | 467.1 | 1073.8c | 560.3 | 1283.0c | 585.6 | 830.3 | 370.8 | |
| Altitudinal shift (%) | RCP 2.6 | Rel. | -2.1 | 3.2 | 0.9 | 10.6 | 1.6 | 10.8 | -2.3 | 6.0 |
| RCP 4.5 | Rel. | -0.4 | 9.3 | 9.2 | 17.9 | 10.5 | 19.4 | 2.1 | 8.8 | |
| RCP 6.0 | Rel. | 1.9 | 7.3 | 1.7 | 9.8 | 2.2 | 9.9 | 0.5 | 7.8 | |
| RCP 8.5 | Rel. | 4.1 | 4.1 | 12.0 | 28.3 | 14.7d | 31.1 | 2.5d | 8.8 | |
| Geographical shift (km) | RCP 2.6 | Abs. | 28.3 | 26.2 | 33.7 | 26.0 | 31.3 | 27.5 | 32.7 | 22.6 |
| RCP 4.5 | Abs. | 54.5 | 56.8 | 75.0 | 60.7 | 66.5 | 78.1 | 78.1 | 40.7 | |
| RCP 6.0 | Abs. | 34.0 | 17.1 | 65.2 | 62.4 | 61.7 | 65.5 | 55.8 | 37.9 | |
| RCP 8.5 | Abs. | 37.6a | 39.8 | 114.5a | 91.6 | 103.1 | 99.3 | 96.2 | 67.1 | |
Stacked predictions for 47 species of the Drusinae, for the present projection (top) and four scenario forecasts (RCPs 2.6, 4,5, 6.0 and 8.5). The five panels on the left depict occurrence probability, the five panels on the right show predicted occurrence. Maps produced in R, v. 3.3.2 (https://www.r-project.org/), using data layers as described in the Methods section.
Modelled species were predicted to occur across a wide altitudinal gradient, from an average altitude of 427 m a.s.l. for Drusus annulatus, to 2,349 m a.s.l. for D. nigrescens. Carnivores, spring species, microendemics and ecoregion endemics were all predicted to occur, on average, at higher altitudes than their counterparts: grazers (p < 0.05, Fig.
Species richness predictions for two feeding groups of the Drusinae, for the present projection (top) and four scenario forecasts (RCPs 2.6, 4,5, 6.0 and 8.5). The five panels on the left depict grazers, the five panels on the right show carnivores. Maps produced in R, v. 3.3.2 (https://www.r-project.org/), using data layers as described in the Methods section.
Distribution predictions for individual species showed high heterogeneity in all four scenarios with some species gaining and some losing range. On average, however, the net effect was a range contraction across all modelled species, which was significant in all, but the RCP 2.6 scenario (p < 0.01 for RCP 6.0, p < 0.001 for RCPs 4.5 and 8.5; Table
Average absolute and relative changes in range size, altitude and shifts in space for level of endemism and all Drusinae, for present and four RCP scenario projections. Bold values indicate significant differences.
| Level of endemism | All Drusinae | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Microendemics | Ecoregion end. | Widespread | ||||||||
| Scenario | Value | Average | SD | Average | SD | Average | SD | Average | SD | |
| Range size (thousand km2) | Present | Abs. | 37.8f | 35.8 | 162.7f | 144.5 | 678.9f | 540.3 | 301.6 | 438.0 |
| RCP 2.6 | Abs. | 32.5 | 37.0 | 164.7 | 140.7 | 708.5 | 597.2 | 310.8 | 472.9 | |
| RCP 4.5 | Abs. | 23.7c | 28.0 | 131.8b | 127.2 | 568.1b | 505.1 | 248.2c | 392.4 | |
| RCP 6.0 | Abs. | 29.5 | 35.7 | 136.0 | 121.7 | 602.2 | 516.5 | 264.1b | 406.8 | |
| RCP 8.5 | Abs. | 24.8a | 36.9 | 103.9b | 103.2 | 506.8c | 461.8 | 219.3c | 355.1 | |
| Change in range size (%) | RCP 2.6 | Rel. | -33.4d | 47.4 | -2.1 | 24.2 | -1.2d | 24.4 | -13.8 | 37.7 |
| RCP 4.5 | Rel. | -48.4d | 49.1 | -23.0 | 42.6 | -24.1d | 31.3 | -33.1 | 42.6 | |
| RCP 6.0 | Rel. | -33.0 | 68.7 | 4.4 | 73.1 | -18.8 | 30.5 | -18.3 | 59.6 | |
| RCP 8.5 | Rel. | 48.9 | 60.4 | -37.1 | 44.7 | -28.6 | 44.2 | -38.6 | 50.8 | |
| Mean altitude (m) | Present | Abs. | 1047.2d | 393.5 | 1249.8e | 471.9 | 800.7e,d | 274.4 | 1009.8 | 410.4 |
| RCP 2.6 | Abs. | 1101.7 | 408.7 | 1226.8 | 490.7 | 783.2a | 263.5 | 1016.6 | 422 | |
| RCP 4.5 | Abs. | 1112.2 | 498.6 | 1315.1a | 540.2 | 893.7c | 334.0 | 1081.8c | 474.9 | |
| RCP 6.0 | Abs. | 1065.0 | 438.0 | 1280.3 | 510.0 | 826.0 | 285.8 | 1031.4 | 440.2 | |
| RCP 8.5 | Abs. | 1155.3 | 601.0 | 1465.2b | 617.9 | 860.0a | 325.6 | 1125.1c | 560.3 | |
| Altitudinal shift (%) | RCP 2.6 | Rel. | 4.5 | 13.0 | -2.1 | 5.7 | -2.0 | 6.2 | 0.4 | 9.6 |
| RCP 4.5 | Rel. | 4.7 | 12.7 | 4.3 | 7.9 | 12.0 | 22.9 | 7.5 | 16.8 | |
| RCP 6.0 | Rel. | -0.3 | 13.3 | 2.1 | 4.7 | 3.0 | 7.1 | 1.6 | 9.2 | |
| RCP 8.5 | Rel. | 9.9 | 37.0 | 16.5 | 25.8 | 6.7 | 13.0 | 10.4 | 26.1 | |
| Geographical shift (km) | RCP 2.6 | Abs. | 31.0 | 31.3 | 27.4 | 14.1 | 35.6 | 26.8 | 31.8 | 25.8 |
| RCP 4.5 | Abs. | 48.8d | 43.0 | 65.7 | 58.1 | 90.7d | 65.4 | 70.6 | 58.7 | |
| RCP 6.0 | Abs. | 55.8 | 55.6 | 61.3 | 80.0 | 62.0 | 40.8 | 59.7 | 57.2 | |
| RCP 8.5 | Abs. | 70.5d | 90.3 | 115.3 | 99.3 | 115.3d | 77.6 | 100.7 | 88.6 | |
The average altitude based on all Drusinae predictions increased significantly in the RCP 4.5 and 8.5 scenarios (p < 0.001), indicating a general upward shift (Table
Predicted range shifts across geographical space for all modeled species based on centroids were lowest in the RCP scenario 2.6, with values between 4.0 km and 125.3 km (Table
Predicted ranges size for carnivores decreased significantly in the RCP 4.5 and 8.5 scenarios (both p < 0.05, Table
Predicted ranges of spring species significantly decreased in RCP scenarios 4.5, 6.0 and 8.5 (p < 0.001, p < 0.05 and p < 0.001, respectively: Table
Species richness predictions for two stream zonation groups of the Drusinae, for the present projection (top) and four scenario forecasts (RCPs 2.6, 4,5, 6.0 and 8.5). The five panels on the left depict spring species, the five panels on the right show stream species. Maps produced in R, v. 3.3.2 (https://www.r-project.org/), using data layers as described in the Methods section.
Predicted ranges decreased significantly in the RCP scenarios 4.5 and 8.5 for microendemics (p < 0.001 and p < 0.05, respectively; Table
Relative altitudinal shifts were similar within all levels of endemism. Moreover, microendemics experience a significantly smaller geographical shift than the widespread species in RCP scenarios 4.5 and 8.5 (both p < 0.05). Ecoregion endemics experience severe range losses in the Iberian Peninsula, Western Highlands, the Carpathians and the Eastern Balkans, but also range gains in mountain ranges such as the Dinaric Western Balkans (Fig.
Species richness predictions for two levels of endemism of the Drusinae, for the present projection (top) and four scenario forecasts (RCPs 2.6, 4, 5, 6.0 and 8.5). The five panels on the left depict ecoregion endemics, the five panels on the right show microendemics. Maps produced in R, v. 3.3.2 (https://www.r-project.org/), using data layers as described in the Methods section.
The vulnerability to CC estimated from SDMs based on future scenarios showed some important differences to those indicated by the climate change vulnerability score (CCVS, ranging from 0 = invulnerable, to 6 = highly vulnerable; Table
Climate change vulnerability for each species assessed by classifying predictions into one of eight classes of relative range change in the four RCP scenarios (2.6, 4.5, 6.0, 8.5) and counting them (scenario frequency). Colours indicate the frequency projections assigned to either positive (range gain) or negative (range loss) quartiles. The highest positive quartile includes predicted range gains above 100%. The climate change vulnerability score (CCVS) by
Distributions for widespread species, as well as for very restricted ones, were predicted with reasonable accuracy. Thus, we assume our model setup to appropriately describe the distributions of species with very specific habitat requirements and high endemism levels. Precipitation was an important predictor for describing species’ distributions, suggesting a strong link to discharge patterns driven by climate. Discharge and other hydrological regime components are among the most important factors determining the distribution of freshwater biodiversity at multiple scales, such as the extent used in this study (
Average predictions for all Drusinae show clear range losses and uphill movements, suggesting significant range changes in the three most severe scenarios (RCP 4.5, 6.0, 8.5) and altitudinal shifts in two of these (RCP 4.5, 8.5). Accordingly, significant changes in Drusinae distributions may be expected if greenhouse gas emissions are not reduced in the near future and CC continues along the current trend. The magnitude of predicted range losses from the present study was lower than that determined by
Although we observed geographic variation in range changes, predicted species richness losses and range contractions were mostly projected south of the Alps, where climatic conditions are generally warmer and drier. This coincides with species-loss predictions in North America (
Overall, our projections confirm our first hypothesis that Drusinae caddisflies are highly vulnerable to CC and give rise to serious concerns on how general community composition in freshwater ecosystems could be altered under future climatic conditions. Even for species with lower predicted vulnerability to CC, the generalised tendency is to disperse into higher elevations at the expense of decreasing their predicted range sizes. This dispersal is likely to lead to novel biological interactions between species with similar ecological traits, which adds to stress induced by CC, resulting in an elevated extinction risk (
Carnivorous Drusinae, although few in number, occur at higher elevations than grazers according to present projections, but the size of their predicted ranges remained similar. Moreover, range losses for carnivores are only projected in two future CC scenarios and altitude shifts were not projected. Conversely, grazers showed some vulnerability to climate change through range losses in three scenarios, altitudinal shifts in two and geographic shifts in one scenario. These results indicate that larval feeding behaviour is not a useful ecological trait to assess CC vulnerability, because (i) carnivores and grazers do not sufficiently differ in terms of range losses and (ii) carnivores, as the arguably most threatened group because of their higher average altitude, show no altitudinal shifts and lesser geographical shifts than grazers under future scenarios. Because larval feeding types did not clearly differ in CC sensitivity, our ability to draw conclusions regarding ecosystem functions (e.g. autochthonous organic matter processing by grazers) is limited. The causes for this poor differentiation may be twofold: the comparatively low number of carnivores limiting statistical power (unrelated to any sampling artefacts; see Fig.
In the present projection, spring species were predicted to have significantly smaller ranges and at higher altitudes than stream species. Spring species were also predicted to lose range and experience uphill shifts in more RCP scenarios than stream species. Conversely, range losses in stream species are only projected in two RCP scenarios. Also, stream species are not predicted to experience altitudinal shifts, indicating lower CC vulnerability. The fact that spring species shift in altitude is particularly important because of the difficulty to track suitable climatic conditions within their habitat (i.e. along the stream network). As spring species inhabit exclusively springs and spring-brooks, they cannot disperse further upstream, while downstream aquatic dispersal requires traversing unsuitable habitats. These species are thus stuck in a ‘dead end’ and their survival will depend on tracking suitable habitats through ‘out-of-network’ (i.e. overland) dispersal. This kind of dispersal has been shown to be reduced beyond distances of 10–20 km for several species of Drusus (
Range predictions suggest micro- and ecoregion endemics to have smaller ranges at higher altitudes than widespread species. Microendemism would initially imply a very high vulnerability to CC as such species lose range in greater proportions than widespread species, even in the two least severe RCP scenarios. Interestingly, neither further range losses nor altitudinal shifts increase with the severity of the scenarios in microendemics, while both increase for ecoregion endemics. The CC effects on relative range losses differentiated from those shown by widespread species, but in the two lower intensity scenarios. Moreover, widespread species experience altitudinal shifts more frequently than micro- or ecoregion endemics. The initial strong negative trend of climate change on the distribution of microendemics, does not continue to increase with the severity of the RCP scenarios. This could result from the relatively large dispersal capacity assumed, allowing any species to track climate across long distances in the modelled predictions. The projections of microendemics also result in a wide variety of reactions to CC across species and scenarios. For example, relative range changes in D. croaticus vary between +157.2% and -4.4% depending on the RCP scenario, while those in D. siveci are projected to vary from +36.5% to -49.8%. Therefore, level of endemism is likely not an appropriate ecological trait to gauge CC vulnerability in an SDM approach because projections may be misleading and need to be assessed carefully species by species.
Regarding our trait-based hypotheses, we can summarise that stream zonation is a relevant indicator for CC vulnerability because their predictions consistently exhibited negative influences of CC on their projected distribution patterns. The results also show that there are no evolutionary constraints on the reaction of species to climate change, as there is no signal for common climate change reactions among the three main evolutionary lineages that align with feeding group in Drusinae and as a result feeding groups appear to be poor predictors of climate change sensitivity.
We evaluated distribution predictions under a posteriori consideration of ecological traits and observed that changes in predicted distributions are strongly related to stream zonation preference. Conversely, previous studies have estimated climate change vulnerability directly from traits assigned to species, assuming homogenous responses between ecological trait groups (
A holistic approach including indices based on known species traits and changes in predicted occurrences (e.g. via SDM) could be a suitable solution integrating information provided by both approaches. As the number of SDMs predicting future distributions continues to increase – stimulated by specific calls for an expansion in biodiversity monitoring (
Distribution predictions for Drusinae caddisflies showed a general vulnerability to CC, but responses were highly variable among species. Stream zonation was associated with CC vulnerability, but without links to particular evolutionary lineages of Drusinae (Fig.
More generally, additional anthropogenic impairments to freshwater ecosystems such as habitat destruction, land-use changes, pollution, eutrophication, water abstraction and structural modification, among many others, are not considered here, although they are omnipresent (
Our study on Drusinae caddisflies can serve as an example of the possible outcomes of changing climate conditions on freshwater biodiversity. The need for increased CC research on freshwater ecosystems becomes evident from two main points. First, several species of Drusinae have only been described recently in Europe (
Contributors to the Distribution Atlas of European Trichoptera (DAET; BioFresh) are thanked for permission to use their data on sampling localities. We thank Paul Frandsen and an anonymous reviewer for their constructive comments on a pervious version of the manuscript.
The authors have declared that no competing interests exist.
No ethical statement was reported.
No use of AI was reported.
This study was funded by the Austrian Science Fund (FWF, project number P 23687-B17, PI: J. Waringer).
SUP, JW and WG, conceived the study and SUP and MK designed the experiments. MK performed the experiments, analysed the data, prepared figures and tables. SUP, JW, WG, LK, AP, MK, MB and SV collected and processed data. All authors discussed, wrote and reviewed the manuscript.
Mathias Kuemmerlen https://orcid.org/0000-0003-1362-3701
Wolfram Graf https://orcid.org/0000-0001-6559-0644
Johann Waringer https://orcid.org/0000-0002-0114-1636
Simon Vitecek https://orcid.org/0000-0002-7637-563X
Mladen Kučinić https://orcid.org/0000-0003-2756-2305
Ana Previšić https://orcid.org/0000-0002-0332-7522
Lujza Keresztes https://orcid.org/0000-0002-4609-6496
Miklós Bálint https://orcid.org/0000-0003-0499-8536
Steffen U. Pauls https://orcid.org/0000-0002-6451-3425
The datasets generated during and analysed during the current study are available from the corresponding author on reasonable request. Occurrence data for the Drusinae caddisflies are available upon request from the Distribution Atlas of European Trichoptera (DAET) (https://doi.org/10.15504/fmj.2015.9). Environmental predictor data is freely available from the cited sources. The R package used to build the SDMs, biomod2 is accessible through the CRAN repository.
List of species modeled, with code, number of occurrences used in the model, classifications according to feeding group, endemism and stream zonation, as well as climate vulnerability score (CCVS)
Data type: docx
Model performance indicators and variable importance by species
Data type: docx