Combined ecological risks of nitrogen and phosphorus in European freshwaters

ABSTRACT


29
The "limiting nutrient" concept, following Liebig's Law of the Minimum, was based on 30 the effects of added nutrients on crop performance (van der Ploeg et al., 1999). Later, the concept 31 was extended to productivity-based experiments for eutrophication, such as those testing the 32 effects of nutrient surplus (mainly nitrogen -N -and phosphorus -P) on chlorophyll 33 concentration or biomass productivity (Allgeier et al., 2011;Elser et al., 2007). Despite the 34 benefit prompted by the increase in the availability of a resource, such as increase in 35 productivity, a further increase in the same resource level could cause ecosystem damage, such 36 as a shift in species composition (Odum et al., 1979). 37 Freshwater eutrophication is triggered by agricultural and urban discharges of N and P as production and, thus, the availability of food to planktivores and herbivores (Carpenter et al.,41 1985). On the other hand, it may also lead to increased predation by secondary consumers and 42 decreases in food quality (Carpenter et al., 1985;Grimm and Fisher, 1989), water transparency decomposition of nuisance algae and macrophytes may generate hypoxic or (in extreme cases) 47 anoxic conditions in aquatic systems (Carpenter et al., 1998). Ultimately, the presence of oxygen 48 depleted conditions, exposure to toxins released by phytoplankton, and shifts in food availability 49 may be harmful to invertebrates (Camargo and Alonso, 2006;Correll, 1998) (Figure 1). 50 Therefore, the same nutrient stimulating autotrophic productivity and food availability may, in 51 turn, instigate ecosystem damage at increasing concentrations. Accordingly, defining the nutrient 52 as a resource or as a stressor depends as to whether its concentration prompts a benefit or 53 damage to ecosystems. 54 Ecological theory models detect this dual aspect of N and P. The intermediate disturbance 55 hypothesis (IDH) conveys that species richness is maximized at intermediate levels of stress and 56 minimized at the two extremes (Grime, 1973). Underlying the IDH, the physiological tolerance 57 hypothesis (Currie et al., 2004) conveys that species richness is the upshot of the tolerance of 58 each individual species to specific local conditions. Currie et al. (2004) use the hypothesis to 59 explain species tolerance to climatic variables and we expand it so as to describe species 60 tolerance to the upper end of nutrient levels, i.e. the level of the stressor which triggers species 61 loss. 62 Eutrophication is a complex issue as it encompasses potential feedback mechanisms (van 63 Donk and van de Bund, 2002), non-linear responses of primary production to trophic conditions 64 (Genkai-Kato and Carpenter, 2005), and synergistic effects of N and P on primary production 65 (Elser et al., 2007). The extent to which they drive primary productivity can be examined by 66 4 analyzing past nutrient level patterns (Anderson, 1998) or nutrient stoichiometry changes 67 (Glibert, 2012), ecological modeling (Genkai-Kato and Carpenter, 2005), or via nutrient addition 68 experiments (Schindler, 1977). Nonetheless, the development and the application of 69 eutrophication models which include all the various pathways through which N and P influence 70 individual invertebrate species occurrence may be troublesome due to lack of data and of insights 71 on all relevant mechanisms of impact. 72 Alternatively to mechanistic models, statistical models coupled with available monitoring 73 data of water bodies may be used to underpin biodiversity effects of eutrophication and provide 74 environmental protection agencies with guidelines for the improvement and the maintenance of 75 water quality . We circumvent the uncertainties within each of the different 76 ecological mechanisms by developing a probabilistic model of invertebrate species occurrences 77 with the upper observed stressor tolerance in field observations ( Figure 1). 78 Eutrophication indicators based on the performance of invertebrates may be less certain 79 than those on autotrophs since consumers are not directly affected by N and P concentrations as 80 are photosynthesizing organisms (Johnson et al., 2014). However, invertebrates are convenient to 81 environmental agencies because they are extensively monitored (Growns et al., 1997) and their 82 monitoring can be easily employed as water quality indicators, such as the ecological quality 83 ratio (EQR). In the case of the EQR, the composition of invertebrates is compared with a 84 reference representing minimum impairment (Clarke, 2013). Nevertheless, indicators usually do 85 not detect the main stressor driving the eutrophication impact. 86 In the case of eutrophication, the estimation of the overall health quality of freshwater 87 needs also to uncover what the main cause of impairment is. Therefore, an ecological indicator 88 5 that allows for estimation of the ecosystem health as well as for identification of the driving 89 stressor of eutrophication impairment may provide environmental agencies with the tools to 90 recognize impaired areas and to target the stressor of concern. In this study, we propose the 91 ecological risk (ER) to identify the areas and the main drivers of eutrophication impairment. This 92 framework is compatible with risk assessments proposed for toxicants (Beketov et Figure 3). The ER can be interpreted as the 103 probability that an invertebrate species within a community in a river-basin and in a given year is 104 exposed to a stressor level above its threshold of occurrence in the environment. 105 The dual effects of the two stressors can be combined in order to estimate the total ER to 106 species as  Underlying data for temperate heterotrophic species are reported by Azevedo et al. (2013a) . and shown in appendix S1.
The CDF for NO 3 and TP is shown in Fig. 2b and 2c, respectively.
The underlying data a c (i,f,r), β c (i,f,r) are shown in appendix S2.
An example of the PDF is shown in Fig. 2a.
Step 1 Step 2 Step 3 Step 4 ER T (f,y) Total ecological risk freshwater type f in river basin r, determined as response addition of ER i .
S(x,f) Slope of linear regression of changes in Er i and ER T with time in freshwater f river basin r.
The changes in ER i and ER T with time (the slope S) are shown in Fig. S2.2.
Step 5 The spatial and temporal variability of ER i and ER T is shown in Fig. 4.
Step 6 ER_CA(f,y) Comparison between the total ecological risk derived under the assumption of response addition (ER T _RA) and of concentration addition assumptions (ER T _CA).
The results of the comparison are shown in appendix S3. to high stressor levels). The CDF was determined for the two stressors (i.e. NO 3 and TP) in two 12 We use the same CDF expressing the vulnerability of species towards high nutrient levels 161 (i.e. α coefficient) and their sensitivity to changes (i.e. β coefficient) across the years and across 162 the same freshwater type. We expect that this function is unlikely to change in such a short 163 period of time as its parameters describe characteristics inherited during years of evolution and 164 are driven by differentiated exposures to hydrological and biogeochemical patterns in the two 165 freshwater types (Azevedo et al., 2013a). 166 Probability density function (PDF) 167 The probability of a stressor being found at a 10 log concentration x can be described by a 168 PDF of a logistic curve as  (Tables S1.1a and S1.2a) and from 0.01 to 17 mg P/L for 214 TP (Tables S1.1b and S1.2b). Our results show that the tolerance to N and P levels is lower in 215 lakes than in streams ( subjected to a decrease in lake ER P but the same was only observed in 29 of 79 basins for stream 229 ER P (Figure S2.2b). 230 The ER N was predominantly higher than ER P in streams (Figure 4a,b). For example, from 231 2001 to 2011, 46 to 77% of river basins comprised ER N higher than ER P in a given year. 232 However, the opposite pattern is observed in streams. Over the same period, 11 to 52% of river 233 basins in a given year comprised ER N above ER P . We also found a strong variability in the ER  Stream fauna appeared to be less vulnerable to high nutrient levels (α Lake < α Stream ). This 264 also corresponds with the lower N and P levels defining trophic state thresholds for lakes than for 265 streams (Smith et al., 1999). Additionally, lake invertebrates are more sensitive to increasing 266 nutrient levels compared to streams (β Lake < β Stream ). Here, we propose biogeochemical and The ecological risk posed by N stress is estimated to be considerably higher than the 290 ecological risk of P stress in both streams and lakes (Figure 4a,b). Here, we propose two reasons 291 19 for this trend. In low primary production systems, nutrient demand by autotrophs is reduced. The 292 lessened uptake by N by autotrophs may cause accumulation of N. High NO 3 values were 293 associated to low primary production rates in a low productivity temperate lake (Sterner, 2011). 294 Second, the excess of N supply compared to P was also identified in stoichiometric analysis of 295 tissues of herbivores and of tissues of herbivore food supplies (Elser et al., 2000). In their study,  Because the monitoring of biodiversity shifts can be costly, ecological indicators of water 387 quality impairment should be an available tool for environmental agencies (Johnson et al., 2014). 388 Here, we estimated the ecological risks in lakes and streams due to dual N and P stress to 389 invertebrates as this species group is frequently monitored (Growns et al., 1997)