Data presented in the previous section suggest that water column N and P should be considered when characterizing the autotrophic state of rivers and streams and perhaps when determining heterotrophic state. The relative trophic state should be based on the frequency distribution of relatively pristine lotic waters because anthropogenic inputs change over time, as will trophic boundaries. Whereas Dodds et al. (1998) considered total N, total P, and benthic chlorophyll across a wide variety of streams, they did not account for streams that are naturally heterotrophic and did not attempt to use only reference streams to create an expected distribution in the absence of anthropogenic effects.
Reference nutrient data can be used to establish rough limits on the autotrophic state of streams with regard to nutrients; I present one possible approach. Reference nutrient Table 2. Lower one-third and upper one-third of the distribution of stream total N and total P pooled across 14 ecoregions according to reference values determined for each individual ecoregion by Smith et al. (2003), 13 ecoregions for total P, and 12 ecoregions for total N from Dodds and Oakes (2004) and the relationship of the boundary numbers from Smith et al. (2003) data to cumulative frequency distribution of benthic chlorophyll (Chl) as a function of total N or total P (Fig. 1) expressed as the percentage of benthic chlorophyll mean or maximum values exceeding 100 mg m22 when nutrient values were less than the boundary value. For example, when seasonal mean of total N was ,714 mg m23, then 10% of the streams had mean benthic chlorophyll values exceeding 100 mg m22 and 29% had maximum values exceeding that amount. concentrations from modeling, including a correction for atmospheric N deposition, have been proposed for 14 nutrient ecoregions across the United States (Smith et al. 2003). I ranked the median values (one for each ecoregion), and the distribution was divided into the lower, middle, and upper one third (oligotrophic, mesotrophic, and eutrophic, respectively, following limnological convention) of the reference nutrient values (Table 2). The distribution of reference nutrient values roughly agreed with those provided by Dodds and Oakes (2004), who corrected for anthropogenic influences (as represented by human population density and land use characteristics) on stream nutrient concentrations with analysis of covariance across the same ecoregions (Table 2). There is a positive correlation between autotrophic activity and benthic chlorophyll concentrations in rivers and streams (Fig. 1). Therefore, I initially base autotrophic boundaries on standing stocks of algal biomass, as is the convention in lakes. To accomplish this, the reference nutrient values from Smith et al. 2003 are applied to observed frequency distributions of seasonal mean and maximum benthic chlorophyll, plotted against water column nutrients (Fig. 2). These frequency distributions are used to calculate the probability that a stream will have a given amount of chlorophyll at a specific level of nutrients (Table 2). Relationships derived from those developed by Dodds et al. (2002, corrected for errors Dodds made when entering data from Lohman) also can be used to calculate expected mean and maximum values for benthic chlorophyll on the basis of the nutrient boundaries presented in Table 2 (Table 3). Benthic chlorophyll values .100 mg m22 previously have been considered a nuisance (Welch et al. 1988). This analysis suggests that a mean value of 100 mg m22 of chlorophyll is attained in ,7% of oligotrophic streams and in 10–13% of eutrophic systems. The regression analyses also suggest that oligotrophic systems should exhibit maximum benthic chlorophyll values .100 mg m22 only 27% of the time. Other approaches are possible (e.g., Dodds et al. 1998), but the method presented in this paper considers the dynamic nature of chlorophyll in streams and is reference based. A similar approach to determining reference trophic state can be taken with regard to planktonic chlorophyll in rivers and streams. A large data set (n 5 292) of lotic planktonic chlorophyll and water column total P was assembled for temperate rivers and streams, and associated regression equations can be used to link nutrients and phytoplankton biomass (Van Nieuwenhuyse and Jones 1996). A smaller data set from 31 rivers in southern Ontario and western Quebec related total N (mg m23) and total P to planktonic chlorophyll (mg m23; Basu and Pick 1996). This paper presented a regression equation for total P, but regression of their raw data yielded the following relationship.
log10 (planktonic chlorophyll)
= - 1.247 + 0.676 log10 (total N) r 2 = 0.65
The distribution of reference values from Smith et al. (2003) can then be used to calculate autotrophic categories from these equations (Table 4). These data agree roughly with both the Van Nieuwenhuyse and Jones (1996) and the Basu and Pick (1996) equations for total P, but the total N boundaries derived from the Basu and Pick chlorophyll–total N relationship were substantially lower than those derived for total P from the same data set. The data suggest that planktonic chlorophyll only exceeds values considered typical of eutrophic lakes (8 mg m23; Dodds 2002) when nutrients are abundant relative to the reference condition. The data also are consistent with the idea that the amount of planktonic chlorophyll per unit total N or total P is less in lotic waters than in lentic waters (Søballe and Kimmel 1987).
More limited data are available for whole-stream estimates of autotrophic and heterotrophic state, but some idea of the ranges expected for the trophic states can be gleaned from analysis of the results of a cross-system study (Mulholland et al. 2001). Although this study and an additional data point (P. Mulholland pers. comm.) only covers nine streams, it has three important characteristics. First, all the measurements were done the same way at each site with methods likely to give the best results (two-station diel O2 method, corrected for groundwater influences). Second, all the sites studied but one were relatively pristine small streams, so the data can be used to determine trophic boundaries mostly in the absence of human effects. Third, the streams were located in a variety of biomes, including one desert, one prairie, one tropical, one arid montane, one mesic montane, and four temperate deciduous biomes (Mulholland et al. 2001). Whole-stream autotrophic state varied over 150- fold in this data set (very high rates of GPP were associated with the lighted desert stream), with the central one third of the distribution falling between 0.4 and 1.8 g O2 m22 d21 (Table 5). Heterotrophic state was considerably less variable, ranging about 10-fold with the central one third of the distribution falling between 6.7 and 8.3 g O2 m22 d21 (Table 5). Bott et al. (1985) reviewed studies of ;70 streams with maximum rates of 48 and 50 g O2 m22 d21 for GPP and respiration, respectively. These rates were from streams with human effects and were several-fold higher than the maximum from more pristine streams. This indicates that both autotrophic state and heterotrophic state can be influenced by eutrophication. Maximum rates of GPP are probably limited by light under nutrient-replete conditions, whereas respiration is probably limited by O2 aeration rate in streams with high loading of biochemical oxygen demand. I speculate that light limits autotrophic state of streams (interception by the canopy), but not heterotrophic state, because although light is intercepted by riparian vegetation, it does not substantially influence rates of C input. I predict that the amount of C fixed by the riparian canopy that enters the streams to fuel heterotrophic activity is approximately equal to what would enter by autochthonous production in a lighted stream without canopy cover.
Small streams in forested biomes are shaded, have substantial amounts of organic C input from nearby riparian areas fueling heterotrophic activity, and have minimal autotrophic production (except in deciduous seasonal forests in which light can penetrate the canopy when leaves are not present). Prairie, tundra, or desert streams have limited riparian canopy and substantial autotrophic production fueling heterotrophic activity. An independent measure of total metabolic activity, N uptake rates, also varied little across the range of biomes studied by Mulholland et al. (2001), supporting the concept of relatively constant heterotrophic activity in small pristine streams (Webster et al. 2003). Heterotrophic state might be more variable in rivers; canopy has less of an influence, and turbidity could substantially interfere with riverine C production.
Although the approach taken here might provide useful in setting boundaries for autotrophic and heterotrophic state, more comprehensive measurements of stream metabolism are required. Until such comprehensive measurements are made, the values for boundaries presented here should be used with caution. In addition, whole-river metabolism rates are difficult to measure, and data are difficult to come by for such rivers. Very few large rivers remain in temperate regions that are relatively weakly influenced by humans, so it might not be possible to set definitive autotrophic and heterotrophic state boundaries for larger lotic systems in some regions.
Although determining trophic boundaries could be useful in describing fundamental ecosystem processes, changes in trophic state must be linked to other aspects of stream ecosystems for such boundaries to be relevant. Furthermore, it is important to explore how stream eutrophication is propagated through the food web to influence biotic integrity and community structure.