Nonlinear variation of stream–forest linkage along a stream-size gradient: an assessment using biogeochemical proxies of in-stream fine particulate organic matter

Authors

  • Takashi Sakamaki,

    Corresponding author
    1. Department of Forest and Conservation Sciences, University of British Columbia, Vancouver, BC, Canada
    • Transdisciplinary Research Organization for Subtropics and Island Studies, University of the Ryukyus, Okinawa, Japan
    Search for more papers by this author
  • John S. Richardson

    1. Department of Forest and Conservation Sciences, University of British Columbia, Vancouver, BC, Canada
    Search for more papers by this author

Correspondence author. E-mail: sakamaki@lab.u-ryukyu.ac.jp

Summary

  1. Spatial scale of processes is necessary information to design appropriate ecosystem research and management. In-stream ecological processes are tightly linked with their adjacent terrestrial environment and impacted by modifications of riparian environments, particularly in small, forested streams. Quantitative understanding of how the dynamic coupling of these adjacent ecosystems varies along stream continua is necessary to enhance stream research and management.
  2. In temperate, forest-dominated watersheds of British Columbia, Canada, we examined the stream–forest linkage along a stream-size gradient based on biogeochemical analysis of fine particulate organic matter retained in streambed sediment (FPOMS). We analysed FPOMS at 48 stations in 14 cobble-bed streams from 2·2 to 39 m bankfull widths during summer low-flow conditions and tested their relationships with local, forest-related, environmental factors (light penetration, benthic coarse POM biomass, riparian-soil properties) along the stream-size gradient.
  3. Carbon/nitrogen ratio of FPOMS (FPOMS-C/N) was distinctly lower in stream bankfull widths larger than an estimated threshold around 7·6 m and corresponded with decrease in N/chlorophyll a of FPOMS and inorganic nitrogen concentration. We found significant linkages of FPOMS-C/N with local streamside environments in streams narrower than the threshold (but not in widths <~3 m). For wider streams, reduced canopy cover was likely to enhance in-stream nitrogen cycling by algae and consistently lowered FPOMS-C/N regardless of streamside environment. These indicate a nonlinear attenuation of the strength of linkage between in-stream biogeochemical processes and forest-related environments along the stream-size gradient.
  4. Synthesis and applications. The relationship of biogeochemical proxies of in-stream fine particulate organic matter with local environments can be a practical tool for measuring the strength of stream-forest linkage. Furthermore, regional- and stream-specific threshold stream widths dividing in-stream biogeochemical regimes can be an important landmark for ecologists and managers to refine designs for stream research and management at appropriate spatial scales. For stream widths narrower than a threshold, reach-scale linkages between stream and terrestrial environments can be a major focus, and riparian conditions need to be carefully managed to protect stream habitats. For stream widths wider than a threshold, catchment impacts on streams at larger spatial scales would be a more important focus.

Introduction

Many ecological processes and structures depend on location and context due to environmental heterogeneity. Since ecosystem research and management needs to be sensitive to the local governing processes, identifying thresholds for spatial shifts between process regimes is essential to determining the appropriate tactics. In streams, fluvial networks comprise diverse reach types in terms of environment and in-stream biological processes (Gomi, Sidle & Richardson 2002). Understanding of those spatial arrangements along stream continua is fundamental for stream ecology and management. Many conceptual models have been proposed to describe spatial structures and processes of stream ecosystems, for example river continuum concept (Vannote et al. 1980), riverine productivity model (Thorp & Delong 2002), process domain concept (Montgomery 1999) and patch dynamics theory (Pringle et al. 1988). However, those models differ in focal environmental characteristics, spatial scale and processes (Thorp, Thoms & Delong 2006). A lack of explicit scaling measures hampers application of existing conceptual models and synthetic, quantitative understanding.

Small, forested streams have a unique environment, biota and ecological processes which are distinct from those of large streams (Gomi, Sidle & Richardson 2002; Meyer et al. 2007; Finn et al. 2011). Along small forested streams, riparian areas provide adjacent streams with terrestrial-origin organic matter and nutrients and support biological production (Wallace et al. 1999; Nakano & Murakami 2001; Richardson, Zhang & Marczak 2010). The canopy of riparian vegetation also greatly suppresses light penetration and water temperature (Kiffney, Richardson & Bull 2003). Both quality and quantity of local riparian vegetation greatly affect in-stream biogeochemical and biological processes (Kiffney, Richardson & Bull 2003; Kominoski, Marczak & Richardson 2011; Sakamaki & Richardson 2011). Therefore, the linkage between the local forest-related environment and adjacent stream ecosystems is an important perspective for ecological research and management in both stream and riparian subsystems of forest-dominated watersheds. In management sectors, maintaining a forested riparian buffer is today becoming a common strategy to protect adjacent stream habitats (e.g. Lee, Smyth & Boutin 2004; Richardson, Naiman & Bisson 2012). In relatively wider streams, the decrease in canopy cover over a stream channel is expected to diminish its control of local physical environment and material input from riparian vegetation to the adjacent stream. In addition, relatively wider reaches with mild gradient are expected to retain material transported from upper reaches. Thus, the influences of the local forest-related environment on adjacent streams are predicted to weaken with stream size. In this case, upstream cumulative effects at catchment scales may have greater impact on a focal stream reach (Dodds & Oakes 2008; Zhang, Richardson & Pinto 2009; Beechie et al. 2010). However, due to the complex nature of fluvial systems, we do not understand quantitatively the attenuation of the strength of the stream-forest linkage along a stream-size gradient or the relative importance of local, reach-scale and catchment impacts in streams.

Fine particulate organic matter (organic particles <1 mm) is a major component of organic matter retained and transported in streams (Kiffney, Richardson & Feller 2000; Hall & Tank 2003; Thompson & Townsend 2005). Biogeochemical properties of FPOM, such as carbon/nitrogen ratio (C/N) and carbon stable isotopic signature, reflect mixing of different organic matter sources and in-stream biogeochemical processes (Hellings et al. 1999; Hein et al. 2003). Our previous work has shown that FPOM properties are significantly related to local, forest-related environmental factors, such as irradiance and input of terrestrial organic matter, in small forested streams (Sakamaki & Richardson 2011). Furthermore, elemental composition of FPOM can be a measure of its quality as a food resource for consumers (Enríquez, Duarte & Sand-Jensen 1993; Cross et al. 2005). Resource use by some stream consumer species has shown significant relationships with properties of FPOM (Sakamaki & Richardson 2011). Overall, FPOM properties reflect some aspects of in-stream biogeochemistry and food webs which are essential for stream ecosystems. Biogeochemical properties of FPOM have potential to be a practical proxy to characterize stream ecosystems at a reach scale and also assess stream-forest linkage from a biogeochemical perspective.

We analysed FPOM retained in streambed (FPOMS) for biogeochemical properties at 48 stations along a stream-width gradient (2·2–39 m) in temperate, forest-dominated watersheds. We further tested relationships of C/N of FPOMS (FPOMS-C/N, the same notation rule for other indicators hereinafter) with local, forest-related environmental factors along a stream-width gradient to measure the stream-forest linkage. Based on previous studies, we predicted that FPOMS-C/N declines with increasing stream width due to weakened forested-related effects, specifically decreasing input of terrestrial organic matter and easing photic-limitation of algal production. We specifically aimed to (1) determine a scale of spatial variation of in-stream biogeochemical condition and (2) test whether the strength of stream–forest linkages weaken linearly or nonlinearly with increasing stream size.

Materials and methods

Study sites

We sampled at 48 stations in 14 cobble-bed-dominant streams in the Fraser Valley and Vancouver Island of southwestern British Columbia, Canada (49°12′–26′N, 122°24′–124°53′W). Nine streams were in the Greater Vancouver Area, and the other five streams were on the east coast area of Vancouver Island. All the streams flow through mountainous, forested areas. The vegetation in the study region is predominated by conifer species, such as western hemlock Tsuga heterophylla, western red cedar Thuja plicata and Douglas-fir Pseudotsuga menziesii. The major riparian species are red alder Alnus rubra, vine maple Acer circinatum and salmonberry Rubus spectabilis. The bankfull width and watershed area of the sampling stations ranged from 2·2 to 39 m and from 0·18 to 320 km2, respectively. The bankfull width (Y m) was significantly related to watershed area (X km2) (Y = 1·92 X0·5 + 2·80, R= 0·90, < 0·0001). For each stream, 3–5 sampling stations were spaced at the widest intervals possible; c. 0·5 km in a 1st order stream (defined based on observation in the field) to c. 10 km in a 6th order stream. Each station was located within a relatively homogeneous reach of at least a few hundred metres with a specific type of stream and riparian condition. Reaches right below major tributaries were avoided in the selection of sampling stations.

Figure 1.

Relationships between biogeochemical properties of FPOM and stream bankfull width. (a, b) The open circles and grey circles show FPOMS and FPOMW, respectively. (c) The open circles and crosses show FPOMS and bedrock biofilm, respectively. (d) The concentration of dissolved inorganic nitrogen in stream water is shown. For panel (a), the thick and thin lines indicate bootstrap means and 95% confidence intervals for 10 stations with consecutive stream widths, respectively. FPOMS and FPOMW, fine particulate organic matter in streambed sediment and stream water, respectively.

Sample collection and chemical analyses

We sampled FPOMS, FPOM suspended in stream water (FPOMW), biofilm and riparian soil at all the sampling stations to determine their biogeochemical properties. All field sampling was conducted during summer low-flow periods in June to August of 2007 and 2008. Each type of sample was collected at each sampling station once during the 2-year study period. All stations within a stream were sampled in the same day. Samples and field data were collected only during fair weather conditions, and sampling within a few days after rain was avoided.

A 16-L water sample was collected at the downstream limit of the study reach in each station for sampling FPOMW. To sample FPOMS, sediment was collected from the streambed of five pools along a 50–100 m reach at roughly equal intervals (c. 10–20 m) in each station. The 0–3 cm layer of sediment was collected using a core sampler with diameter of 4 cm. To sample biofilm, several cobbles were collected from three pools and three riffles along a 50–100 m reach at roughly equal intervals (c. 7–15 m) in each station. In addition, to estimate the chemical properties of terrestrial POM entering adjacent streams from riparian areas, riparian soil was collected from three plots with an area of approximately 20 m2 along a 50–100 m reach at roughly equal intervals in each station. Six 4-cm diameter core samples of the 0–4 cm layer soil were randomly collected in each plot and then mixed to make a composite sample.

Coarse-size POM (CPOM; >1 mm) retained on streambeds was collected using a Surber sampler with 30 × 30 cm frame from three pools and three riffles with roughly equal intervals along a 50–100 m reach at each station. In addition, to determine dissolved nutrient concentrations, we took 50-mL subsamples from the 16-L stream water samples for FPOMW.

Fine particulate organic matter retained in streambed sediment, FPOMW, biofilm and riparian-soil samples were analysed for stable isotope ratios of carbon and nitrogen (δ13C, δ15N) and C/N. FPOMS, FPOMW and biofilm samples were also analysed for chlorophyll a (chl. a). C/N and N/chl. a for organic material samples were calculated as mass ratios. Streambed CPOM was analysed for ash-free dry mass to determine their biomass. Concentrations of dissolved inorganic nutrients in water samples were measured using an auto-analyser (Bran+Luebbe TRAACS-800).

For more details of sample collection and chemical analyses, refer to Sakamaki & Richardson (2011).

Field measurements of physical environment

For each station, the stream's bankfull width was measured at 4–6 cross-sections along a 50–100 m reach at roughly equal intervals. The irradiance was measured at six points along a 50–100 m reach at roughly equal intervals in each station using a quantum meter (Apogee BQM-S), which detects the photosynthetically active radiation, that is, 400–700 nm. The irradiance was also measured at unshaded reference sites <10 min prior to light measurement at each sampling station. The percentage of relative light intensity was calculated, defined as (the irradiance of station/the irradiance of unshaded reference site) × 100 (%).

Data analyses

All data analyses were station-based; multiple data obtained in each station were pooled and averaged to represent the station. We focused mainly on properties of FPOMS rather than FPOMW, since in our previous study, biogeochemical properties of FPOMS generally were more tightly linked with local forest-related environments compared with those of FPOMW (Sakamaki & Richardson 2011). We also focused on three environmental indicators which link with riparian forests and potentially affect FPOMS properties: riparian-soil-C/N (as an indicator for quality of terrestrial POM provided to adjacent streams), irradiance (as a factor linked with canopy closure and controlling in-stream primary productivity) and streambed CPOM biomass (as an indicator of quantity of terrestrial POM input and potential production of FPOM through breakdown). For streambed CPOM, its log(10) values were used in the data analyses.

Univariate linear regression analyses were conducted to test relationships of biogeochemical indicators and environmental factors with stream bankfull width. When they were significant, mean and 90% bootstrap confidence interval of threshold stream widths above and below which those indicators/factors distinctly changed were estimated using nonparametric changepoint analysis (Qian, King & Richardson 2003; Qian 2009). Furthermore, variances of the indicators/factors belonging to stream width and streams were compared using linear mixed-effects model (LMM) with random slope, where fixed and random factors were stream bankfull width and streams, respectively. The ratio of variance explained by the fixed and random factors to total variance (R2F and R2R) was estimated based on a method proposed by Nakagawa & Schielzeth (2013). In addition, the variances of stream bankfull width due to the effects of station and stream were compared using a nested-anova where the stations were nested in the streams.

To explore stream width ranges in which FPOMS-C/N has significant, negative relationships with stream bankfull width, 50 000 combinations of ten stations were randomly selected from all the study stations, and bootstrap linear regression analyses with 1000-case resamplings were conducted. Means of coefficients of determination (R2) and 95% confidence intervals of slopes were determined. To visualize the effect of range of stream width on the relationship, R2 values were plotted against the mean stream widths and standard deviations of the combinations with 10 randomly selected sampling stations.

We used bootstrap regression analyses to test relationships between FPOMS-C/N and the three local forest-related environmental factors (i.e. irradiance, riparian-soil-C/N, streambed CPOM) along the stream-size gradient. All combinations of data from 10 stations with consecutive stream bankfull widths were analysed. Likewise, we also tested the relationships for FPOMS13C vs. biofilm-δ13C, and FPOMS15N vs. biofilm-δ15N. Means and 95% confidence intervals of slopes were determined, and the slopes were plotted against the mean stream widths.

For the bootstrap regression analyses, the significance of relationship between two variables was based on assessing whether the range between 2·5% and 97·5% quantiles of slope did not overlap zero. All statistical analyses were conducted using the statistical software R ver. 2.15, and ‘lme4’ package was used for LMM.

Results

Fine particulate organic matter retained in streambed sediment-C/N, FPOMW-C/N, FPOMS-N/chl. a, FPOMW-N/chl. a C/N and their variance generally decreased along the range of stream widths (Fig. 1, Table 1). For inorganic nutrients, nitrate was the predominant form of inorganic nitrogen in all the stations, whereas the concentrations of nitrite, ammonium and phosphate were lower than detection limits in many stations. The concentrations of dissolved inorganic nitrogen (DIN; nitrate + nitrite +ammonium) also generally decreased with stream width. The means of threshold stream bankfull width for FPOMS-C/N, FPOMW-C/N, FPOMW-N/chl. a and DIN were estimated between 7·1 and 7·6 m, and their 90% confidence intervals were relatively narrow (e.g. < ~6 m); those chemical indicators distinctly changed above and below the thresholds. The estimated mean of threshold width for FPOMS-N/chl. a was 2·2 m, but was 7·8 m when excluding a station with a remarkably high N/chl. a of 1315 at width 2·2 m. Those chemical indicators with a threshold stream width around 7–8 m generally showed modest fixed effect of stream width and a relatively larger random effect of streams in LMM. The other chemical indicators generally had a weak or no trend along the stream-size gradient and unclear thresholds with wide confidence interval. Some of these indicators, for example FPOMS13C, FPOMW13C, FPOMW15N, showed very weak fixed effect of stream widths, but relatively larger random effect of streams. In the nested-anova, the ratios of variance of stream widths explained by streams and stations nested in streams to total variance were 82% (F13,20 = 24·2, < 0·0001) and 13% (F14,20 = 3·43, = 0·006), respectively; the effect of stream identity on stream width was greater than that of station within streams.

Table 1. Results of (a) regression analysis for in-stream chemical and environmental indicators against stream bankfull width, (b) nonparametric changepoint analysis to determine threshold stream width (mean and 90% confidence interval) for variations of indicators showing significant relationship with stream bankfull width, and (c) comparison of variance explained by fixed factor (stream bankfull width, R2F) and random factor (streams, R2R) by LMM. The results in parenthesis indicate the case where one station with notably high FPOMS-N/Chl. a (ratio of 1315) at 2·2 m width was excluded from analysis
 (a) Relationship with bankfull width(b) Stream bankfull width of changepoint (m)(c) LMM variance comparison
TrendPMean(5%, 95%)R2FR2R
  1. FPOMS and FPOMW, fine particulate organic matter in streambed sediment and stream water, respectively; CPOM, coarse particulate organic matter on streambeds.

FPOMS
C/N<0·0017·6(5·1, 10·7)0·240·35
N/Chl. a0·0082·2(2·2, 11·9)0·160·27
(N/Chl. a0·0077·9(3·1, 11·9)0·160·41)
δ13C 0·69  0·070·69
δ15N+<0·00112·0(9·2, 19·1)0·140·31
FPOMW
C/N<0·0017·1(4·6, 8·4)0·090·48
N/Chl. a0·0017·6(4·5, 10·6)0·110·79
δ13C 0·72  0·040·61
δ15N 0·29  0·000·60
Biofilm
C/N+0·04612·0(2·8, 24·0)0·000·93
N/Chl. a 0·55   0·000·09
δ13C+0·03010·6(4·8, 19·1)0·080·69
δ15N+0·02610·6(4·0, 25·4)0·120·00
Concentration in stream water
POC 0·69  0·020·00
Chl. a+0·0059·1(7·4, 19·1)0·500·09
DIN0·0207·6(6·5, 8·4)0·080·62
Environmental factors
Riparian-soil C/N 0·89  0·010·10
Streambed CPOM 0·29  0·020·25
Irradiance+0·01318·2(5·1, 19·7)0·120·00

Although FPOMS-C/N generally showed a negative relationship with stream bankfull width, it was dependent on the range of stream size due to the nonlinearity of their relationship (Fig. 2). Particularly when the analysis covered a relatively narrow range of stream widths, there were cases where the distinct contrast of FPOMS-C/N between sites with widths above and below the estimated threshold width was not reflected in the analysis (i.e. the cases of relatively small and large means of stream width with relatively small standard deviation in Fig. 2). In such cases, the negative relationship between FPOMS-C/N and stream bankfull width tended not to be significant.

Figure 2.

The markers other than the lightest grey ones indicate stream width ranges in which C/N of fine particulate organic matter retained in streambed sediment (FPOMS-C/N) has significant, negative relationships with stream bankfull width. The mean coefficients of determination (R2) in the bootstrap regressions are plotted against the mean stream widths and standard deviations of the combinations with randomly selected 10 sampling stations. The lightest grey markers indicate the cases of insignificant relationship, and the blank areas are due to lack of calculation for those particular combinations of stream width.

The relative irradiance had a significant, positive relationship with stream bankfull width, but did not have a clear threshold effect (Fig. 3, Table 1). Riparian-soil-C/N or streambed CPOM abundance was not significantly related to stream bankfull width. In LMM, the random effects of streams on those three environmental factors were generally weak to modest.

Figure 3.

Plots of (a) relative irradiance, (b) riparian-soil-C/N and (c) biomass of streambed coarse-size particulate organic matter (CPOM) along the stream-size gradient. The solid circle indicates the mean at each station. The relative irradiance was positively related to the stream bankfull width (R= 0·13, = 0·013), but riparian-soil-C/N or CPOM biomass had no significant relationships with bankfull width.

Fine particulate organic matter retained in streambed sediment-C/N showed negative relationships with irradiance around mean stream bankfull widths of 4–5 m (Fig. 4a). FPOMS-C/N also showed positive relationships with riparian-soil-C/N at mean stream bankfull width c. 5 m (Fig. 4b) and with streambed CPOM around mean stream bankfull widths of 5–9 m (Fig. 4c). Riparian-soil-C/N was a weak indicator for FPOMS-C/N compared with the other two, since we found only single significant relationship between FPOMS-C/N and riparian-soil-C/N. Those local environmental factors showed no significant effect on FPOMS-C/N in mean widths < ~3 m or >~9 m.

Figure 4.

(a–c) Relationships for FPOMS-C/N vs. three local environmental factors. The mean coefficient of determination (R2) for bootstrap regression of 10 stations with consecutive stream bankfull widths is plotted against mean stream bankfull width. The significant relationships were negative for vs. irradiance and positive for vs. riparian-soil-C/N and vs. CPOM. (d) Relationships of stable isotopic signatures (SIS: δ13C and δ15N) between FPOMS and bedrock biofilm along the stream-size gradient. The mean slope for bootstrap regression of 10 stations with consecutive stream bankfull widths is plotted against mean stream bankfull width. The circles and triangles indicate δ13C and δ15N, respectively. For all panels, the grey and open symbols indicate significant (< 0·05) and insignificant (> 0·05) relationships, respectively. FPOMS, fine particulate organic matter in streambed sediment; CPOM, coarse-size particulate organic matter on streambeds.

Fine particulate organic matter retained in streambed sediment-δ13C had a narrower range compared with that of bedrock biofilm over the stream-size gradient and also had only weak relationships with biofilm-δ13C (Figs 1c and 4d; for pooled data from all stream widths, [FPOMS13C] = 0·06[biofilm-δ13C] − 25·41, R2 = 0·10, = 0·03). In addition, FPOMS13C (−27·1 ± 0·7‰, mean ± SD) was consistently close to δ13C which are generally reported for terrestrial plant materials. Our previous study in the same region showed −29·1 ± 2·4‰ and −27·4 ± 0·7‰ for riparian C3 plants and soil, respectively (Sakamaki & Richardson 2011). These indicate predominance of terrestrial-origin carbon and minor contribution of in-stream-producer-origin carbon in FPOMS pools over the study range of stream widths. In contrast, FPOMS15N showed tight linkages with biofilm δ15N along most of the stream widths [for pooled data from all stream widths, (FPOMS15N) = 0·38 (biofilm-δ15N) + 0·58, R2 = 0·43, < 0·001]. This indicates that a substantial fraction of nitrogen source contained in FPOMS was identical with that of in-stream-produced POM. Whereas δ13C indicates the predominance of terrestrial source in in-stream FPOM, FPOMS-C/N (9·8–24·4) and FPOMW-C/N (5·7–15·4) were generally lower than C/N reported for terrestrial-origin POM (approximately 15–200) (Enríquez, Duarte & Sand-Jensen 1993; Cross et al. 2005; Sakamaki & Richardson 2011).

Discussion

Nonlinear spatial shift of in-stream biogeochemical regime

The relationship between FPOMS-C/N and stream size gives a quantitative basis for the heterogeneity and continuity of habitat traits which is fundamental information for stream ecology and management (Cooper et al. 1997; Gomi, Sidle & Richardson 2002; Brown 2003). In particular, our results for FPOMS-C/N highlight distinct biogeochemical conditions on either side of the threshold stream bankfull width (i.e. estimated mean of 7·6 m). Based on the insignificant relationships between FPOMS-C/N and forest-related environmental factors in widths >~9 m, the spatial shift of in-stream biogeochemical conditions probably corresponded with the attenuation of forest-related environmental effects in stream widths larger than the threshold stream width.

Fine particulate organic matter retained in streambed sediment-C/N was likely controlled mainly by in-stream biogeochemical processes, since the elemental origins of FPOMS differed between carbon and nitrogen (i.e. terrestrial and stream, respectively) and FPOMS-C/N was also relatively lower despite the predominance of terrestrial-origin organic matter. The tight linkage of FPOMS15N with biofilm-δ15N suggests that nitrogen cycling by microbes contributes to lowering of FPOMS-C/N over the stream-size gradient (Findley & Tenore 1982; Hamilton et al. 2001; Ashkenas et al. 2004). In general, bacteria and fungi are heterotrophic microbial components predominantly developing on FPOM and CPOM, respectively, in forested streams and incorporate substantial nitrogen from ambient water (Peterson et al. 2001; Hall & Tank 2003). The significant, negative relationship between FPOMS-C/N and irradiance in the relatively smaller stream widths suggests that algae also significantly enhanced nitrogen cycling and lowered FPOMS-C/N in FPOMS pools predominated by terrestrial-origin organic matter (Hall & Tank 2003). In addition, the decrease in FPOMS-N/chl. a with increasing stream width suggests the larger contribution of algae to nitrogen incorporation into FPOMS pools in wider streams. The fact that the concentration of in-stream DIN decreased with stream width and had a threshold equivalent to that of FPOMS-C/N (i.e. estimated mean of 7·6 m) supports that microbial incorporation of nitrogen from ambient water into FPOMS pools was more active in the larger streams. Overall, our results suggest that algae enhanced the in-stream nitrogen cycling and greatly contributed to lowering FPOMS-C/N in the larger streams. Reductions of canopy cover enhance light input and decrease provision of fresh terrestrial CPOM which is the primary resource for heterotrophic microbes (Gessner & Chauvet 1994; Findlay et al. 2002; Artigas et al. 2009). We hypothesize that the primary drivers of nitrogen incorporation into FPOMS pools shifted between the heterotrophic microbes and algae around the threshold stream bankfull width due to the reduced canopy cover (Albariño, Villanueva & Canhoto 2008; Finlay et al. 2011) and exhibited the two in-stream biogeochemical regimes differing in stoichiometric composition of FPOMS and rate of nitrogen cycling. Further study needs to examine rates of biogeochemical processes and microbial community compositions driving the processes (e.g. heterotrophic microbes vs. algae) to assert this potential mechanism.

The distinct changes in variance of FPOMS-C/N around the threshold stream width 7·6 m suggest that the spatial heterogeneity of in-stream biogeochemical condition also differed between the two biogeochemical regimes which divided around the threshold. The relatively low FPOMS-C/N and its small variance in streams wider than the threshold suggest that the biogeochemical condition was less sensitive to variations of extrinsic environmental factors. Sufficient light penetration and its in-stream control of FPOMS-C/N may have overwhelmed effects of other processes, such as input of terrestrial organic matter. This potentially led to a spatially homogeneous biogeochemical condition within and across the streams. In contrast, the larger variances of FPOMS-C/N in streams narrower than the threshold can be attributable to spatially heterogeneous environments, probably associated with variations of in-stream geomorphology and riparian vegetation (Gomi, Sidle & Richardson 2002; Sakamaki & Richardson 2011). The streams narrower than the threshold were likely susceptible to various, extrinsic environmental factors due to the lack of single mechanisms predominantly controlling FPOMS-C/N (such as the effect of light in larger widths) and less stable biogeochemically compared with the relatively wider streams (Guttal & Jayaprakash 2009; Dakos et al. 2010).

Although we found a biogeochemical regime shift within stream reaches consistently predominated by terrestrial-origin POM modulated by in-stream nitrogen cycling and local environments, mixing and relative abundance of terrestrial-origin POM (originally with high C/N) vs. in-stream-produced algae (originally with low C/N) have been a general explanation for spatial variations of FPOM-C/N in rivers (Hellings et al. 1999; Kendall, Silva & Kelly 2001; Hein et al. 2003). Streams wider than those measured in our study, which have high primary productivity and predominance of algal-origin POM, are expected to exhibit another distinct biogeochemical regime (Hellings et al. 1999; Kendall, Silva & Kelly 2001; Hein et al. 2003). Thus, we can predict one more type of in-stream biogeochemical regime shift which is mediated by relative abundance of terrestrial-origin POM and algal-origin POM in streams wider than our study set. We emphasize that biogeochemical regimes and primary mechanisms of their shifts are variable depending on focal range of stream size.

Although tighter stream-terrestrial linkage is more obvious in smaller streams, we detected no significant relationship of FPOMS-C/N with the study local environmental factors in stream bankfull width <3 m. In general, such small streams have spatially diverse and temporally unstable hydrological processes (Gomi, Sidle & Richardson 2002). Woods, Sivapalan & Duncan (1995) reported that cross-site variation of discharge was generally much greater in headwater streams of <1 km2 drainages, which is equivalent to < ~5 m stream bankfull widths in our study than in larger streams. Thus, transport and retention of in-stream POM are expected to greatly vary across such small streams depending on site-specific hydrological factors. Given that retention time of FPOMS allowing qualitative alteration by in-stream biogeochemical processes greatly differed between the study small streams, it probably perturbed relationships between FPOMS-C/N and the studied environmental factors. Such possible effects of variable hydrological processes on in-stream biogeochemistry is unclear today, so we suggest that future studies should examine them to potentially refine our measurement approach for stream–forest linkage and to better understand biogeochemistry in such small headwater streams.

Implications for stream ecology and management

The relationships of FPOM properties and stream width which we obtained from multiple streams potentially represent a general pattern of in-stream biogeochemical condition along stream-size gradient in our study region. A recent study in a forested watershed of northern California reported that a nonlinear shift in dissolved nutrient composition occurred around c. 10 km2 watershed area, which is equivalent to several metres of stream bankfull width based on their relationship for watershed area vs. bankfull width, and also that it was mediated by change of light condition (Finlay et al. 2011). This supports generality of the approximate threshold stream size for in-stream biogeochemical regime shift associated with canopy reduction in forested watersheds. However, our findings still should be carefully interpreted and applied particularly taking into consideration potential effects of various site-dependent factors (e.g. hydrology, geomorphology, vegetation, land-use, nutrient input) (e.g. Gomi, Sidle & Richardson 2002). The relatively large cross-stream variations of the values of biogeochemical indicators demonstrated by LMM was likely mainly due to the variation of range of encompassed stream size between the study streams, but cross-stream variability of spatial patterns of the in-stream biogeochemical indicators along streams and their thresholds is still unclear. In addition, temporal variations of environments (e.g. discharge, terrestrial CPOM inputs, irradiance) also potentially affect spatial arrangement of in-stream biogeochemical conditions (e.g. Gomi, Sidle & Richardson 2002; Frost et al. 2009). The patterns that we present may represent a seasonally relatively higher heterogeneity of in-stream condition in the study region, since advective transport and longitudinal mixing of in-stream materials were weak due to summer low-flow conditions. Cross-stream, cross-regional and cross-seasonal comparisons of relationships between FPOMS properties, local environmental factors and stream size would further enhance general understanding of processes controlling spatial structures of in-stream biogeochemical conditions and their threshold dynamics.

Although complex, and potentially interactive, effects of advective transport and local processes complicate processes in fluvial networks, our findings suggest that the relative importance of the advective and local effects in a focal reach is to some extent predictable. In particular, a threshold stream size, which divides the stream continuum in terms of in-stream biogeochemical regime, can be an important landmark for ecologists and managers to build sound designs for research and management at the proper spatial scales. Within stream widths narrower than the threshold, a tight linkage between stream and terrestrial environments at a reach scale need to be noted. In such stream width ranges, for purposes of ecosystem conservation, riparian conditions need to be carefully managed to protect adjacent stream ecosystems, as in-stream biogeochemical and biological processes would respond sensitively to change of the local, riparian environment. For stream widths wider than the threshold, we can predict that a local impact of riparian alteration on in-stream processes in the adjacent stream is of less consequence. Instead, at larger spatial scales one should pay attention to processes such as longitudinal mixing of materials and catchment impacts on stream systems. Specifically, small streams provide downstream systems with various materials, for example organic matter, nutrients, sediment, drifting invertebrates, and are also considered to play important roles in structuring riverine ecosystems and maintaining watershed biodiversity (Gomi, Sidle & Richardson 2002; Meyer et al. 2007; Dodds & Oakes 2008; Finn et al. 2011). Responses of lower reaches to conditions of catchment and upper reaches would be a more important focus in streams larger than the threshold stream size.

Nonlinearity and threshold dynamics are considered to be ubiquitous in natural environmental systems, but more rarely demonstrated empirically (Groffman et al. 2006; Dodds et al. 2010). Our results highlight the nonlinear, spatial variations of stream-forest linkages and stream reach traits which are a rare example of spatial nonlinearity in environmental structure. Our findings imply some important points which would be applicable for general ecosystem ecology and management. As the recognition of spatial arrangement of biogeochemical conditions along the stream-size gradient depended on range of stream size taken into account, nonlinear environmental structures can lead to strong scale-dependency of their spatial patterns. In addition, nonlinear, spatial variation of environmental structure can link with a distinct shift of predominant ecological processes. Without attention to these points, ecological research and management in such environmental systems could be misled. Spatial thresholds of environmental structures and ecological processes need further exploration to enhance understanding of ecosystem dynamics and its applications for management.

Acknowledgements

We thank C. Lovatt, S. Leung and J. Shum for assistance with field sampling and laboratory work. This work was funded by Forest Science Program (The Government of British Columbia, Canada, Y103261), Sumitomo Foundation (Japan) and Funding Program for Next Generation World-Leading Researcher (Japan Society for the Promotion of Science, GR083).

Ancillary