Our experiment demonstrates that salamander loss can alter the functioning of montane stream ecosystems, mainly through top-down effects that induce changes in both the green and brown food web pathways (López-Rojo et al. 2022). Some of these effects have similarities with those reported for predatory invertebrates such as stonefly or dragonfly larvae (Peckarsky et al. 1993, Werner and Peacor 2006). Such effects could be consumptive –as salamander larvae prey on invertebrates and thus reduce their abundance, thus driving cascading effects on basal resources (Holomuzki et al. 1994, Blaustein et al. 1996, Urban 2013)– and/or non-consumptive –as they can induce antipredator strategies in invertebrates and thus reduce their foraging activity or increase the preference for safer habitats. Both of them might induce significant effects on basal resources, as observed with other freshwater predators (Peckarsky et al. 1993, Wudkevich et al. 1997, Bernot and Turner 2001, Simon et al. 2004, Preisser et al. 2005, Davenport et al. 2020).
Effects of salamander loss on the green pathway were reflected in several variables in our experiment. In enclosures with no salamander, algal abundance and chlorophyll concentration were reduced compared to enclosures with salamander, indicating a decrease in periphyton accrual as a result of higher invertebrate grazing (Kurle and Cardinale 2011). Moreover, the increase in the autotrophic index in the absence of salamander, together with the smaller change in periphyton biomass compared to chlorophyll concentration, indicated that salamander larvae promoted the growth of the autotrophic components of periphyton, maybe due to the release of nutrients by salamander activity. These nutrients could benefit primary producers in these oligotrophic streams, as observed for some anuran tadpoles (Kupferberg 1997), fish (Small et al. 2011), crayfish (Arribas et al. 2014) and dragonfly larvae (Costa and Vonesh 2013). The algal community shifted after salamander loss, with several genera –Achnanthes, Gomphonema and Amphora– particularly affected; this had been previously observed with dragonfly larvae, whose absence induced a reduction in the same genera (Werner and Peacor 2006). The greater effect on these diatom genera could be related to their large size (Ranvestel et al. 2004, Werner and Peacor 2006, Alonso et al. 2022b), which could facilitate their assimilation in invertebrate digestive tubes and make them less likely to pass through the gut and recolonize the substrate compared to green algae, cyanobacteria or smaller diatoms (Blaustein et al. 1996, Peterson and Boulton 1999, Peterson and Jones 2003, Connelly et al. 2008). Another possible explanation is that they may be more affected by salamander bioturbation (Ranvestel et al. 2004, Barnum et al. 2022), but this is unlikely given the absence of differences in sediment accrual between enclosures with and without salamander.
The brown food web pathway was also altered by the loss of salamander, but the effect on leaf litter decomposition was opposite to our expectation. Total decomposition decreased, thus not reflecting the increase in invertebrate shredding activity as a result of release from predation that we expected (Atwood et al. 2013, Rezende et al. 2015). It is possible that antipredator strategies, and thus the release from predation, are more obvious for grazers than for shredders, given that stone (and tile) surfaces are much more exposed areas compared to leaf packs, where invertebrates can hide more easily, as observed with fish predators (Wudkevich et al. 1997, Bernot and Turner 2001). Moreover, pine and willow are more or less recalcitrant leaves and hence a more appropriate shelter (safer habitat) than other, more palatable, leaf types that disappear more rapidly (Jabiol et al. 2014). In the presence of salamander, invertebrates may thus preferentially feed on leaf litter instead of periphyton and thus avoid the risk of being seen and preyed upon by salamanders.
In contrast to total decomposition, microbial decomposition was not affected by salamander loss, neither were aquatic hyphomycete sporulation rate nor community structure. This suggests that microbial decomposers are not affected by any trophic cascade effects of salamander mediated by invertebrates, which has been previously observed in terrestrial ecosystems where fire salamander adults control invertebrate communities but not the mycobiome (Laking et al. 2021). The lack of effect on aquatic hyphomycetes can explain the absence of effect on microbial decomposition, since aquatic hyphomycetes are the main microbial decomposers in streams (Suberkropp 1998). Our results contrast with studies on anurans, where tadpole excretion stimulated microbial decomposition due to nutrient release (Connelly et al. 2011, Rugenski et al. 2012, Ramamonjisoa and Natuhara 2018). In our study, the high contents of tannins and other inhibitory compounds in willow leaves and pine needles (Casas et al. 2013, Martínez et al. 2013b, Alonso et al. 2021) may have been more limiting for fungal decomposers than a moderate increase in nutrient availability due to salamander excretion (Ramamonjisoa and Natuhara 2018). Also, water renewal due to stream flow may have precluded any effects of salamander loss on microbial decomposers related to nutrient availability, which could be expected to be greater in lentic freshwaters (Rubio-Ríos et al. 2021), although for other freshwater predators, such as fish and crayfish, their nutrient excretion has been observed to be relevant also in running waters (Small et al. 2011, Arribas et al. 2014).
Both total and microbial decomposition were affected by leaf type, but in opposite ways. Microbial decomposition was higher in willow than pine, and so were aquatic hyphomycete sporulation rate and taxon richness, as could be expected (Webster and Benfield 1986). The low specific leaf area, low nutrient concentration and high toxic concentration of pine needles often lead to low microbial colonization and decomposition (Martínez et al. 2013b), whereas willow leaves, despite their low nutrient concentration, have been found to support aquatic hyphomycete communities similarly to those of leaves with higher nutrient concentration (Alonso et al. 2021, Alonso et al. 2022a). Total decomposition was higher in pine than willow, contrasting with previous studies where pine needles decomposed very slowly (Martínez et al. 2013a, Martínez et al. 2013b). The high decomposition rate of pine needles in our study probably represented not only consumption, but also losses related to other invertebrate activities. For example, pine needles can be used by caddisflies to build their cases, as these larvae usually prefer recalcitrant materials for their higher protective value (Whiles and Wallace 1997, Casas et al. 2013). Also, needle fragments could be lost due to stream flow even though they were initially tied, because invertebrate shredders may cut them, and their narrow shape makes it easier for big fragments to be removed and flushed out of the bags, compared to willow leaves.
Invertebrate communities were not altered by salamander loss, and abundance was increased in only one of the three streams, suggesting low consumptive effects of salamander on invertebrates. This contrasts with a study conducted in temporary ponds, where salamander presence significantly reduced invertebrate abundance (Blaustein et al. 1996). The higher effect in temporary habitats could be due to an earlier arrival of predators compared to prey, which contrasts with permanent habitats where overwintering salamanders and invertebrate prey always coexist (Blaustein et al. 1996). While invertebrate abundance could also be affected by predator closeness through their chemical cues (Boyero et al. 2006, Boyero 2011, Boyero et al. 2011, Wesner et al. 2012, Correa-Araneda et al. 2017), this is unlikely in our study because enclosures with and without salamander were adjacent and because salamander larvae were present in the stream, so chemical cues could be present regardless of the treatment. The only stream where invertebrate abundance increased as a result of salamander loss was the stream with higher invertebrate abundance and taxon richness, suggesting that a higher consumptive effect could be expected in streams with less limiting environmental conditions (i.e., not as highly oligotrophic as montane streams). In streams with higher invertebrate density, the number of encounters and therefore captures would increase (Kratz 1996, Griffith et al. 2023) and may result in more evident effects of salamander loss on invertebrates.