Eighty years of food-web response to interannual variation in discharge recorded in river diatom frustules from an ocean sediment core.

Little is known about the importance of food-web processes as controls of river primary production due to the paucity of both long-term studies and of depositional environments which would allow retrospective fossil analysis. To investigate how freshwater algal production in the Eel River, northern California, varied over eight decades, we quantified siliceous shells (frustules) of freshwater diatoms from a well-dated undisturbed sediment core in a nearshore marine environment. Abundances of freshwater diatom frustules exported to Eel Canyon sediment from 1988 to 2001 were positively correlated with annual biomass of Cladophora surveyed over these years in upper portions of the Eel basin. Over 28 years of contemporary field research, peak algal biomass was generally higher in summers following bankfull, bed-scouring winter floods. Field surveys and experiments suggested that bed-mobilizing floods scour away overwintering grazers, releasing algae from spring and early summer grazing. During wet years, growth conditions for algae could also be enhanced by increased nutrient loading from the watershed, or by sustained summer base flows. Total annual rainfall and frustule densities in laminae over a longer 83-year record were weakly and negatively correlated, however, suggesting that positive effects of floods on annual algal production were primarily mediated by "top-down" (consumer release) rather than "bottom-up" (growth promoting) controls.

Tracking food-web responses to environmental changes over long timescales (1–3) can be difficult, as species interactions do not generally preserve well (4, 5). Where records can be recovered, as from varved sediments, proxy indicators can reveal changes in density or biomass, which suggest that food-web structures have changed over time. In the absence of experimental manipulations or prolonged, direct observations of processes, however, inferences of ecological causes for recorded change remain uncertain. For example, to what extent are changes in biomass through time a result of changes in environmental conditions or resource availability, versus altered impacts of consumers or other natural enemies? Species interactions, in particular top-down (trophic) interactions, are often difficult to observe even in contemporary time, and hence are commonly underestimated as drivers of ecological change (6).

In the Eel River of northern California, 28 y of ecological research has linked annual hydrologic regimes to alternative food-web structures with contrasting algal abundance (7–10). Under winter-rain, summer-drought Mediterranean seasonality, the Eel shows striking year-to-year variation in algal accrual during its biologically productive summer low flow period. Large proliferations of attached green macroalgae, with average filament lengths peaking in midsummer at >50 cm, often follow winters with at least one bed-scouring flood. In summers following winters without scouring floods, river substrates remain relatively barren over the summer, and attached algal filaments are generally <5- to 15-cm long. These green versus barren years could occur because of either “bottom-up” or “top-down” effects of winter floods (9). Floods extirpate large overwintering grazers, releasing spring and early summer algal growth from consumer control. Wetter winters could also sustain higher nutrient fluxes or flows and temperatures that are more favorable for algal growth longer into the summer. Tentative support favoring the top-down hypothesis (that floods released algae from grazer control) came from the partial recovery of algal biomass following experimental removal of large grazing caddisflies from instream enclosures during one summer that followed a flood-free winter (11) (Fig. S1), but little evidence is available to evaluate the relative importance, over the scale of decades, of top-down versus bottom-up food web controls in regulating variation in annual algal production.

Fig. S1.

Twenty-eight years of surveys and experiments in the upper South Fork Eel River have identified a bimodal food-web regime driven by variation in winter hydrology. When dense, an armored grazer that is invulnerable to most local predators (the caddisfly Dicosmoecus gilvipes) can suppress summer algal growth. This relatively barren algal state occurs during summers following dry winters that lack large, bed-mobilizing floods (Right) (9, 11). During winters with one or more bed-mobilizing floods, however, greatly reduced Dicosmoecus densities during the following summer release early-summer proliferation of the dominant macroalga Cladophora glomerata algae, increasing abundance of its dominant epiphytes, including the diatom genus Epithemia (Left). As animal densities build up over the summer low flow season, algal energy flows up food chains to sustain predators like salmonids. When diatoms are consumed, cell contents are assimilated, but frustules often pass through grazer guts and are excreted intact (10). Big algae summers that follow winter bed scour therefore produce more diatom frustules that are exported by subsequent winter flows, and collect in deposits in the deep marine canyons that receive sediments from the Eel River.

Here we take advantage of the offshore transport, deposition, and storage of freshwater riverine diatoms in coastal marine sediments to reconstruct changes in epiphyte abundance in relation to disturbance, food web, and climatic events. Freshwater diatoms have proven extremely useful as paleoenvironmental indicators in lake and marine depositional environments, because silica in their cell walls (frustules) preserves well in sediments and retains taxonomically diagnostic characters that distinguish taxa with different environmental tolerances (12, 13). Although sedimentary records are frequently absent from erosional riverine environments, many coastal rivers deposit markers of freshwater communities and processes in the near-shore marine environment (14–17). For example, the Eel River, which cuts through steep topography along the tectonically active northern coast of California (18, 19), has few depositional areas in its watershed. Offshore, however, a submarine canyon carved at low sea stands (Fig. 1) provides nearly ideal conditions for river sediment deposition (20). Sediment budget studies using isotopic tracers and fixed ocean-floor monitors in the canyon entrances confirm that the submarine Eel canyon typically receives about half of the Eel River’s annual sediment flux (21–23) at rates high enough to inhibit bioturbation (20). An extensive marine coring program (24) has yielded hundreds of cores of this sediment deposited onto the adjacent continental shelf and slope, including some cores from an adjacent submarine canyon in which 210Pb-estimated sediment age is linearly correlated with burial depth at near annual resolution for nearly a century (20) (Fig. 1). We took advantage of this stable depositional environment and sampled three cores () for identification of freshwater diatoms remains. A fourth core was selected for quantitative estimates of frustule density and to assess potential environmental and ecological changes in the river over the past century.

Fig. 1.

The Eel River watershed and Eel submarine canyon. Draining 9,546 km2, the Eel is the third largest river in California, and is representative of midsized mountain rivers that collectively account for ∼45% of global sediment discharge to the ocean, 18% of global river-to-ocean discharge, and 10% of global drainage area (1). Rivers of this size class are understudied relative to smaller and larger rivers (2). (Left Inset) Shemp Channel of the canyon about 15-km offshore of the Eel River mouth where the core L10C3 used in this study was collected (yellow arrow and circle). (Right Inset) An X-radiograph of the layered sediment and a plot of depth vs. excess 210Pb activity that shows two periods of steady accumulation rate that allowed annual temporal resolution. A change in the rate of sediment accumulation occurs in the mid-1960s, when a superflood destabilized the Eel River banks, and the channel became wider. Thereafter, annual rates of sediment deposition in the canyon increased from 4 (preflood) to 11 (postflood) mm/y. Relationship A1 is used correlating frustule density in cores with algae biomass surveyed along river transects from 1988 to 2001 (Fig. 3), but both lines are used to examine the relationship (or lack thereof) between rainfall at Eureka (near the Eel mouth) from 1918 to 2001 to frustule accumulation in corresponding laminae (Fig. 4). [Figure modified from Drexler et al., (20) © Society for Sedimentary Geology; background Landsat 1 image courtesy of USGS and NASA].

Fig. 3.

Peak modal height of C. glomerata streamers plotted against (A) Rhopalodiaceae frustules and (B) total freshwater diatom frustules from core L10C3. Each point shows data for the algal survey and the corresponding sampled sediment layer from 1988 to 2001 ().

Fig. 4.

(A) Rhopalodiaceae diatom frustules per year and (B) total freshwater diatom frustules per year from core L10C3 plotted against annual total precipitation measured at Eureka, California (NWS Station 42910) from 1918 to 2001.

Summer algal proliferations in the Eel and many similar temperate rivers worldwide are dominated by the green macroalga Cladophora glomerata (L.) Kütz., which becomes heavily overgrown by epiphytic diatoms over time. Two exclusively freshwater diatoms (25) with distinctive, robust frustules in the family Rhopalodiaceae [Epithemia sorex Kütz., Epithemia turgida (Ehrenb.) Kütz.] (Fig. 2) dominate mid- to late-summer epiphyte assemblages on C. glomerata in the Eel River, while other diatoms (e.g., Rhoicosphenia abbreviata Agardh, Cocconeis placentula Ehr., and Cocconeis pediculus Kütz.) are dominant earlier in the low flow season. We developed an 83-y record of freshwater diatoms exported from the Eel River to Eel canyon sediments and evaluated their utility as a proxy for biomass accrual of Cladophora proliferations in the Eel River basin by examining the relationship between freshwater diatom frustule counts in annual laminae and the magnitudes of Cladophora proliferations surveyed in the upper basin from 1988 to 2001 (Tables S1–S3).

Fig. 2.

(A) Frustules of E. sorex and C. pediculus from core L10C3. Epithemia spp. often co-occur with and subsequently replace Cocconeis late in the summer growing season. The robust frustule of Epithemia allows better preservation than thinner frustules of Cocconeis or Rhoicosphenia. (B) Scanning electron micrograph (SEM) of an Epithemia frustule showing the distinctive curved raphe and robust silica walls that aid in identification and preservation in sediments. Photo by R.L.L.. (Scale bar, 10 μ.) (C) R. abbreviata (Bottom, Left) is a midsuccessional epiphytic diatom that often co-occurs with Gomphonema acuminatum (Upper Left) and is supplanted by E. sorex (Center), E. turgida (Upper Right), and Rhopalodia gibba (Lower Right) as the growing season progresses. Frustules are from core L10C3 and other Eel submarine canyon sedimentary cores. (D) SEM of a heavily epiphytized strand of C. glomerata. Almost all of the epiphytes are E. turgida and E. sorex. SEM courtesy of R.L.L. and the Society for Freshwater Biology ©2010. (Scale bar, 50 μ.)

Table S1.

Annual frustule counts from deposited sediments in the core and corresponding algal bloom height from annual upstream surveys from 1988 to 2000, the period of overlap of core L10C3 with algae survey record

Depth, mm	Year	Discharge Yr x, x+1, m3/s	Bloom, cm	L10C3 Epithemia (frustules/sample)	L10C3 FW Diat (frustules/sample)
0–11	2000–2001	Drought, drought 83, 53	Big 298	44	109
11–22	1999–2000	Flood, drought 136, 83	Big 108	38	95
22–33	1998–1999	Flood, flood 176, 136	Big 265	40	93
33–44	1997–1998	Flood, flood 418, 176	Big 127	30	70
44–55	1996–1997	Flood, flood 129, 418	Small 17.3	11	56
55–66	1995–1996	Flood, flood 321, 129	Big 177	34	94
66–77	1994–1995	Drought, flood 43, 321	Small 3.5	3	21
77–88	1993–1994	Flood, drought 240, 43	Big 50	19	33
88–99	1992–1993	Drought, flood 45, 240	Small 47	48	99
99–110	1991–1992	Drought, drought 90, 45	Small 25	28	136
110–121	1990–1991	Drought, drought 102, 90	Small 34	14	81
121–132	1989–1990	Flood, drought 157, 102	Big 57	14	36
132–143	1988–1989	Flood, flood 147, 157	Small 24	10	44

Peak daily discharge was measured at the nearest gauging station, former USGS Gage 11475500 (South Fork Eel River at Branscomb, reactivated in 1990 by Angelo Coast Range Reserve researchers). Bankfull discharge at this site is estimated at 120 m3/s. To estimate discharge when records from the reactivated Branscomb gauging station were not available (before April 1990, or during when instruments were down), we developed regressions between the USGS Branscomb record and discharge at the USGS Elder Creek gauging station (USGS 11475560) during 3 y (1967–1979) when both were monitored by the USGS: Branscomb discharge (m3/s) = 0.0886 (Elder discharge, m3/s)1.16, r = 0.98, n = 1,096. Elder Creek is a tributary of the South Fork Eel gaged at a station 4 km from the Branscomb gauging station.

Table S3.

Core L10C3 Rhopalodiaceae frustule proxy

Discharge	Frustules >21	Frustules <21
Flood	35	2
Drought	37	10

Core L10C3 Rhopalodiaceae frustule proxy of >21 frustules per sample for “Big” blooms, corresponding to 50 cm Cladophora bloom height) and Bankfull discharge [using a USGS Scotia gauge threshold of >3,540 m3/s for flood years, (USGS Station 11477000, 1917–2001)]. (P = 0.031 Fisher’s Exact test).

SI Methods

Details on Eel Canyon Core Sites, Core L10C3, and Core Selection.

Cores were sampled with a remotely operated vehicle, enabling sample collection among precisely located morphological features of the canyon (20). Larry is the northernmost entrant channel (sites L1–L5), just south of the Eel River mouth, 4-km wide by 6-km long with a dendritic configuration of side gullies. Wall gradients vary in the channels from common gentle slopes (<5%) to vertical walls exhibiting exposed strata. Shemp (sites L10, L13) is a very narrow, meandering entrant with steep walls, and joins Moe at ∼300-m water depth. Shemp has thick beds of shell hash and wood debris in cores at several locations.

After exploratory sampling revealed abundant, clearly recognizable freshwater and marine diatom frustules in 10 samples from cores L1C5, L1C8, L1C9, and L1C13 (2), collected by C.A.N. and T.M.D. We chose an intact core half from site L10, core 3 for further analysis on the basis of two criteria: (i) maximum core strata preservation and (ii) precisely constrained sediment accumulation rates. In the Eel Canyon, as in many submarine canyons incised into continental shelves, these two goals are often conflicting. Bioturbation is the main cause of core strata degradation in these regions (1) and in nearby shelf sediments, marine polychaete worms rapidly degrade event strata within a timescale of 5–10 y (3). In certain settings, such as the Eel Canyon, however, high sediment accumulation rates suppress the worms and inhibit bioturbation. These areas, typically shallower canyon thalwegs, show high preservation of physical stratification because of low bioturbation, but poorly constrained sediment accumulation rates. This may be because of episodic increases in sediment deposition during storm events (22, 23). Steady (and easily constrainable) sediment accumulation rates are generally low (<4 mm/y) because the transport mechanisms (e.g., hemipelagic sedimentation) are slow and steady, but low rates permit benthos to thrive. Rapid depositional and frequent erosional events are typically episodic (e.g., turbidity flows down thalwegs) and suppress benthic habitation. Pulsed or irregular sedimentation make it very hard to constrain sediment accumulation rates, unlike hemipelagic sedimentation, which is considerably easier to constrain.

Core L10C3 was collected from the Shemp Channel thalweg (see Site L10, marked on Fig. 1). Here, locally steep canyon walls confine the thalweg to a width of ∼40 m (20). Bioturbation is minimal, with greater than 90% preservation of physical structures. The sediment accumulation rate is well constrained, approaching steady-state, with a rate change occurring part way down the core from 4 mm/y (r2 = 0.82) below 45 cm to 11 mm/y (r2 = 0.73) above a 45-cm core depth. This unusual combination of high preservation of strata and well-constrained sediment accumulation rates made core L10C3 an ideal candidate for analysis of freshwater diatom accumulation rates.

Details on Methods.

Following Abrantes (15) and Smol et al. (13), we made slides from core sediments by adding 1 g of sediment to 20 mL deionized water in a 20-mL glass scintillation vial, stirring or mechanically agitating for 1 min to fully suspend the sediment, immediately withdrawing 0.2 mL of suspension and placing it on coverslip. We dried coverslips overnight and mounted them on microscope slides, using Naphrax (Brunel Microscopes, www.brunelmicroscopes.co.uk). For the intact L10C3 core, care was taken to withdraw the 1 g of sediment evenly from the full depth of the annual lamina interval, 11 mm (measured using a micrometer), to match the annual sedimentation rate. Care was also taken to ensure that the sediment was fully suspended and disaggregated. As this core has a high fraction of fine sand (>17%) and a low fraction of clay (<28%) compared with other STRATAFORM cores (5), simple stirring and agitation homogenized the sample sufficiently. Before placing our suspension sample on the coverslip, we preweighed each coverslip and weighed it again after sample drying, so that we could measure frustules counted per milligram of dry sediment. This also enabled us to convert to frustules deposited per square centimeter per annual lamina (or year) (16).

Because freshwater diatoms were less abundant in cores than marine diatoms, we counted all freshwater diatoms on each entire slide and subsampled the common marine taxa (15, 16). Total diatoms counted were at least 350. Counts of freshwater diatoms ranged from 100 to 300 for 54 y and from 45 to 99 for 29 y. In 2 y, 1994 and 1996, 21 and 37 freshwater diatoms were counted, restricting our ability to do more detailed taxonomic analysis. Diatoms were counted as columns covering each slide were systematically scanned at 400×, using the movable stage of the Olympus microscope. Frustules were identified to species (most freshwater diatoms) or genus (most marine diatoms; for <5% of marine diatoms, cells could only be identified to family). We measured cell diameters using an ocular Whipple grid. When needed for identification, a 1,000× oil-immersion eyepiece was used. Because of the paired nature of diatom frustules and their tendency to separate during transport, we counted each frustule separately, then divided by two to estimate the total number of diatoms. We converted our total counts from each slide to the estimated numbers of diatoms per lamina as follows: we counted all frustules in the subsample, divided by two, estimated the proportion of the total lamina weight per square centimeter represented by the subsample (typically ∼1/400), and divided the subsample total diatom count by this proportion to get the totals for each lamina (12).

Diatom frustule densities in cores were compared with the magnitude of algal proliferations repeatedly surveyed during summer by M.E.P. across four permanent cross-stream transects from 1988 through 2001 (Table S2). These transects were spaced at 1- to 1.5-km intervals along the upper South Fork Eel River within the Angelo Coast Range Reserve, about 210-km upstream from the mouth of the Eel River. To survey algae, we stretched a meter tape across the channel between nails in trees benchmarking each transect (nail-to-nail distance varied <2 cm between surveys), and visually assessed dominant and subdominant macroscopic algal taxa using a diving mask or Plexiglas view box. We recorded the modal height of attached filaments (or the length from point of attachment of strands of algae floating on the water surface, if these obscured view of the bed); algal dry mass (g cm−2) = 0.026 algal field height (cm) −0.295, (r2 = 0.80, n = 36, P < 0.0001) (erczo.berkeley.edu/datasets). We also estimated the percentage cover and condition of algae within an estimated area of ∼100 cm2 under each transect point (9, 41, 42).

Table S2.

Frustule counts from core L10C3 Rhopalodiaceae

Discharge	Frustules >21	Frustules <21
Flood	15	1
Drought	11	9

Frustule counts from core L10C3 Rhopalodiaceae above and below a proxy threshold of >21 frustules per sample, which corresponds to “Big” Cladophora blooms, exceeding 50 cm in surveyed height) and Bankfull discharge [using a USGS Leggett gauge threshold of >565 m3/s for flood years, (USGS Station 11475800, 1965–2001)]. (P = 0.011 from a Fisher’s Exact test).

Details on Core Dates.

Each frustule “year” is defined as the year of deposition, measured from August 1 to July 31 to match the core collection date. We expect that most of a given growing season’s frustules will be deposited in the offshore Eel canyon during the deposition year immediately following the preceding growing season, as under the low baseflow of the growing season, few riverine diatom frustules from that season would have reached the canyon before higher winter flows. So we assume, for example, that frustules produced during the May–August 2000 growing season frustules would be found in the August 2000–August 2001 lamina. Geochronological techniques using 7Be and 210Pb document that Eel fluvial sediments are rapidly (weeks to a few months) deposited during winter following the onset of high river discharge (20, 21, 23). Peak discharges are based on water years measured from October 1 to September 30.

Overwinter Storage Analysis.

If summer diatom production were stored in the river over the subsequent winters in deep pools, relating the record to a given year’s environmental regimes or surveyed algal biomass would become more difficult. As a canyon-bound, incised river, the Eel lacks major off-channel storage areas, but some of its pools are rather deep (5–7 m), particularly where flow is impeded by large bedrock formations. The evidence presented here, however, suggests that even with subsequent low-flow winters, multiyear retention does not occur.

The regression of either Rhopalodiaceaen or total freshwater diatom frustules with the discharge of floods following a given summer season yielded weak negative relationships: Total Freshwater Diatoms vs. Scotia Peak Subsequent Discharge (y = 57,029–0.0435x; r2 = 0.048, P < 0.05, n = 83); and Rhopalodiaceae vs. Scotia Peak Subsequent Discharge, y = 26,665–0.0264x; r2 = 0.051, P < 0.04, n = 83) (Fig. S2).

Fig. S2.

(A) Total Rhopalodiaceaen diatom frustules per year from core L10C3 plotted against peak annual discharge during the subsequent winter following the summer growth season, measured at Scotia, California [US Geological Survey (USGS) Station 11477000] from 1918 to 2001. (B) Total Freshwater diatom frustules per year from core L10C3 versus peak annual discharge during the winter following the summer growth season, measured at Scotia, California (USGS Station 11477000) from 1918 to 2001.

These weak negative relationships suggest that frustules are not retained (e.g., in deep pools), even during low flow (drought) winters. In addition, mean frustule counts from flood years preceded by flood years were not different from counts from flood years preceded by drought years (Table S1) (P = 0.5), further suggesting that no significant multiyear storage of frustules occurred.

Spatial Congruence and Requirements for Upscaling.

Algal surveys from 1988 to 2001 in basin of the upper South Fork Eel River basin draining 116–137 km2 correlated well with diatom frustule abundances integrating fluxes from the entire 9,546 km2 Eel River basin over this period. Following a winter with scouring (bankfull) floods, algae are released from suppression by predator-resistant grazers and typically achieve large (>50 cm) proliferation sizes, proliferating before other consumers recover to exert grazing pressure in late June–July. For grazer release of algal biomass, only one peak discharge exceeding the bankfull threshold is required.

For Rhopalodiaceae and other freshwater diatom counts in annual laminae in the deep marine canyons offshore to adequately index basin-wide annual algal production, there would need to be positive correlations of annual production of Cladophora, Epithemia, and other epiphytic diatoms across different basins in the watershed. This suggests that factors that potentially limit growth of these algae (light availability, temperature, nutrient fluxes, water velocity, grazing, substrate stability), while not necessarily similar among subbasins of the river, covary consistently and positively across the years of record. These factors likely include the occurrence throughout the entire basin of at least one scouring flood during the previous winter, to release algae from grazer suppression (Fig. S1).

Methods

Based on the 210Pb geochronology observed for core L10C3, sediment in the upper ∼40 cm accumulated at a relatively steady rate of 11 mm/y. This is based on calculations from the 210Pb profile shown in Fig. 1, and the corresponding regression coefficient (r2 = 0.73) developed by Drexler et al. (20). Following Smol et al. (13) and Abrantes (15), we made slides from core sediments by withdrawing 1 g of sediment from the full depth of the lamina in the upper 40 cm of core L10C3, sampling at 11-mm increments, measured using a micrometer. Care was taken to ensure that the sediment was fully suspended and disaggregated. Because freshwater diatoms were much less abundant in cores than marine diatoms, we counted all freshwater diatoms on each slide and subsampled the common marine taxa (12) at 400–1,000×. We converted our diatom totals from each slide to estimates of frustule accumulation rates for the entire annual lamina following standard procedures () (15).

Diatom frustule densities in cores were compared with the magnitude of algal proliferations surveyed by one of us (M.E.P.) across four permanent cross-stream transects during summers from 1988 through 2001. These transects were established over a 5-km reach along the upper South Fork Eel River within the Angelo Coast Range Reserve, about 210-km upstream from the mouth of the Eel River. Modal height of attached filaments (hereafter “proliferation heights”), the percentage cover, and the condition of algae were all recorded within an estimated ∼100-cm2 area under each transect point (additional methodological details are in ).

Results and Discussion

Annual peaks of spatially averaged Cladophora proliferation heights surveyed along the South Fork of the Eel River near Branscomb, California, for a given year were positively related to the abundance of freshwater Rhopalodiaceae frustules recovered from the corresponding lamina in the marine core (Fig. 3). Total counts of all epiphytic freshwater diatom taxa (including C. placentula, C. pediculus, R. abbreviata, and Gomphonema spp.) were also positively correlated with Cladophora proliferation height during a given year (Fig. 3). The abundance of Rhopalodiaceae diatom frustules in a lamina for a given year explains 47% of the year-to-year variation in surveyed Cladophora peak modal height averaged over surveyed transects, and the total freshwater epiphytic diatom frustules in a lamina for a given year explains 33% of interannual variation in this index of peak Cladophora biomass. Paleolimnological studies of periphyton in the St. Lawrence River have also reconstructed Cladophora biomass from epiphytic diatoms (26), including the same or similar species found in the Eel flora.

Both Rhopalodiaceaen and total freshwater diatom frustule counts were weakly negatively correlated with precipitation during a given year (Fig. 4). This suggests that top-down controls as a result of release from grazers were more significant than were bottom-up effects linked to increased annual precipitation, such as increased nutrient fluxes, or more prolonged periods of favorable flow velocities or temperatures. In mixed-size gravel-bedded rivers, scour of the bed occurs as a threshold event, when flood discharges reach “bankfull,” estimated as that magnitude with a recurrence interval of ∼1.5 y (27, 28). A flood of this magnitude appears necessary for extirpating predator-resistant overwintering grazers. After this flow threshold is crossed, however, summer algal proliferation magnitude is not correlated with flood magnitude or with the number of subsequent floods (9). In some years after scouring floods, early summer algal growth was exported by late June spates, with modest subsequent recovery, as waning, warming late-summer flows became less favorable for Cladophora proliferation. These complications and nonlinearities added noise to the relationship between flood peak magnitudes and summer algal biomass, which suggested that frustule abundance in an annual lamina was a true paleoproductivity record, rather than a result of river discharge.

It was surprising that frustule counts representing annual flux from the entire 9,546-km2 Eel basin were strongly related to proliferation heights surveyed over a relatively small (5 km) study reach draining 116–137 km2 of the upper South Fork of the Eel River. For successful upscaling, three assumptions would have to be met: (i) annual algal proliferation sizes were positively correlated among subbasins across the Eel River because of correlated or compensatory trends in the annual hydrologic, environmental, and food-web controls that limit accrual of Cladophora, and Epithemia, and other epiphytic diatoms during a given year; (ii) frustules were not stored in depositional riverine environments (deep pools, off channel water bodies or wetlands) even during low flow years, so that most frustules are exported offshore during the year in which they were produced (); and (iii) most exported frustules were deposited in their final canyon repository during the same water year that they were produced. We examine each of these assumptions in turn below.

Hydrologic mediation via food-web interactions has been established as a major mechanism releasing Cladophora proliferations in the Upper South Fork Eel River basin (3), but has not been examined elsewhere in the basin. Counts of Rhopalodiaceae and total epiphytic diatom frustules recovered from the marine core suggest that large algal proliferations tended to follow bed-scouring winter floods across the entire Eel River basin during the interval recorded in the core. This congruence may suggest similar release of Cladophora following flood scour of predator-resistant grazers, but the pattern could also arise from flood-mediated effects that enhanced algal accrual by enhancing environmental conditions or nutrient fluxes. In the eastern portions of the basin, there is less forest cover and a potential for a different combination of hydrologically mediated top-down and bottom-up controls to influence algal accrual during a given year (). The negative correlation of frustule densities in laminae with total annual rainfall, however, favors grazer release rather than bottom-up controls as a basin-wide explanation (Fig. 4). Overwinter storage of a summer’s diatom production in the river channel or basin could produce time lags, complicating the relationship between annual algal production and frustule densities in laminae. If summer algal production were stored over low-flow winters until flushed during a subsequent high flow year, we would expect the abundance of frustules in a lamina to increase with the magnitude of the peak winter flood that follows a given summer growth period. However, the relationship of frustule counts to actual surveyed peak algal biomass surveyed in the upper South Fork Eel watershed in a given year was stronger than the relations of either Rhopalodiaceae frustules, or total freshwater diatom frustules to that water year’s peak hourly maximum discharge during the subsequent winter (Fig. S2). The geomorphology of the Eel, a steep, canyon-bound river, and the apparent lack of frustule storage even over drought winters suggest that the Eel River has only one major repository: its submarine canyon () (29, 30). Storage of frustules in shelf deposits for one or more years before their transport to the canyon would, like river channel storage, complicate the relationship between riverine algal production and frustule density in the lamina assumed to record deposition from a given year. The sinking rates of diatoms in still columns of salt water have been estimated to range from 0.6 to 8 m d−1 in marine diatoms similar in size (long axes 10–50 μ) to the two Epithemia species we studied () (31). If diatom cells or frustules in this size range traveled rapidly (within hours or days) offshore from the Eel mouth with the wash load (finest sediment visible in surface currents) during high river discharges (30), then sank (with no turbulent displacement) 130 m to the sea floor at site L10C3 in thalweg at the head of the Shemp canyon, frustules would arrive on the canyon floor (130 m deep) 16–217 d after their export from the river, permitting deposition within the year following summer production of these diatom cells or frustules. The confirmed presence of 7Be in Eel Canyon sediments supports a rapid (subannual) deposition of Eel River sediments (20–23, 29). If freshwater diatoms aggregated with marine phytoplankton or detritus, they would sink more rapidly (32, 33). We have little information, however, on the degree to which diatoms or frustules might be resuspended or displaced by currents, smaller scale turbulence, or even zooplankton feeding or other food-web interactions in the coastal ocean. All of these events would complicate the trajectories and fates of freshwater diatom cells or frustules on their journey from the river mouth to the canyon floor.

Despite these three potential sources of spatial and temporal noise, our results (Fig. 3) suggest that the signal of interannual variation in riverine algal production remains discernable in the offshore record, as indicated by the significant correlation of peak biomass of riverine algae produced during a given year with the density of freshwater diatom frustules in lamina estimated, from 210Pb dating (20), to correspond to that year.

Freshwater diatom frustules recovered from submarine canyon cores appear to be useful as a paleoproductivity (proliferation size) proxy of algae and algal-based food-web dynamics during the period of overlap between the core and survey record. Our results suggest that we can use Epithemia and other freshwater epiphytic diatoms to extend temporal inferences about hydrologically controlled food-web states (e.g., the alternative food-web structure with contrasting algal abundance as a response to sub- vs. superbankfull discharge during the preceding winter) (Fig. S1) back nearly a century for the L10C3 core (34). The results are also encouraging for spatial upscaling. Similar research with a lower-resolution core from coastal Portugal established a correlation between freshwater diatom frustules and instrumental records of river flood stages (15, 16). Work in the Murray-Darling River basin of Australia, St. Lawrence River of Canada, and the Amazon River has also shown that diatom frustules recovered from freshwater off-channel deposits recorded basin-wide environmental changes (17, 26, 35).

Based on frustule recovery from marine cores, we are able to estimate the size attained by past riverine algal proliferations up until the 2001 core collection, primary production that can support consumers and predators. Paleoproxies that provide such a record of primary production may expand the time domain of food-web analyses beyond not only the period of contemporary research, but also beyond that of the instrumental record. The dynamics of food-web interactions that involve long-lived organisms or slow feedbacks often unfold over decades to centuries or millenia. In such ecosystems, paleoproduction proxies may allow hypotheses about trophic controls to be evaluated over more appropriate time scales. Ecosystem change will also be driven by external forcing processes with long time scales. With the increased understanding of the importance of long-period climate cycles, such as the Pacific Decadal Oscillation for marine and freshwater food webs, the need for long time-series data on responses has become more apparent. The widespread geographic dominance of C. glomerata in temperate freshwater rivers and lakes globally (25, 36, 37) suggests that its epiphytic diatoms could be used as proxies in other temperate fresh waters where highly resolved, integrative depositional records have accumulated: at river mouths, in reservoirs or off-channel lakes, or within depositional subbasins. These records would expand the spatial and temporal scales of inferences linking food webs to environmental change, enhancing our understanding of how freshwater webs may respond to future changes in climate, land cover, biota, and other factors affecting riverine runoff and conditions.

Finally, the development of a paleoproductivity proxy using freshwater diatoms recovered from marine sediments provides a quantitative tool for measuring the transfer of riverine biota, nutrients, and organic matter into marine ecosystems. In rivers that do not retain their sediments for longer than 1 y (e.g., small steep rivers in tectonic settings along colliding continental margins), marine records may contribute to our understanding of interannual variation in the interactions of climate with freshwater and marine productivity (13, 38–40).