Climate change is affecting the phenology of organisms and ecosystem processes across a wide range of environments. However, the mechanisms linking organismal to ecosystem process change in complex communities are uncertain. Here we examined how earlier snowmelt will alter the phenology of stream organisms and ecosystem processes, via a large-scale field experiment in outdoor stream channels. Extended low flows increased water temperature, reducing production-to-respiration ratios by 32%. The stream invertebrate community shifted due to phenological shifts in two-thirds of the taxa, and emergent flux pulses of the dominant insect group (Chironomidae) almost doubled, benefitting a generalist riparian predator. Our study shows that climate change in mountain streams is poised to alter the dynamics of stream food webs via fine-scale changes in phenology—leading to novel predator-prey ‘matches’ or 'mismatches’ even when community structure and ecosystem processes appear stable at the annual scale.

Time frame: Begin date 4/21/2019 - End date 8/25/2019.

General study design- We subjected nine large-scale, flow-through outdoor stream mesocosms in California’s Sierra Nevada to three flow regime treatments: a flow regime based on historic average conditions (current treatment), a mitigated climate change scenario where low flow begins three weeks earlier than currently (3-week treatment), and an unmitigated climate change scenario where low flow begins six weeks earlier than currently (6-week treatment). Over the course of a season, we regularly measured primary production; community composition, production, and emergence of benthic and emergent stream invertebrates; and Brewer’s Blackbird (Euphagus cyanocephalus) feeding activity. We tested for immediate vs. delayed effects of advanced low flows by combining the period of the study (i.e., start, middle, and end) with the treatment, creating a variable that captures both timing and treatment effects (i.e., period-treatment). We ran a piecewise structural equation model to elucidate and compare mechanisms driving low-flow effects on stream invertebrate production and emergence.

Methods description- We used nine channels that are 50 m long by 1 m wide, consist of six pools connected by long riffle sections in a meandering fashion, and are fed by the adjacent Convict Creek. We assigned each channel to one of three treatments (with three replicate channels each) in a block design. The three treatments were: (1) current hydrologic conditions based on the historic (long-term) hydrograph at Convict Creek (based on US Geological Survey gage 10265200), with a flow regime that reaches baseflow conditions around August 3rd (i.e., current treatment); (2) hydrologic conditions under a mitigated climate change scenario, where the stream would return to baseflow conditions three weeks earlier than it currently does (i.e., 3-week treatment); and (3) hydrologic conditions under unmitigated climate change, where the stream would return to baseflow six weeks earlier than it currently does (i.e., 6-week treatment). We regulated discharge by controlling sluice gates at the head of each channel. Flows in the channels differed by one order of magnitude between high-flow and low-flow conditions (i.e., 15 L/s and 1.5 L/s, respectively), following a typical Sierra Nevada stream hydrograph for a small stream. Channels were inspected and maintained daily, were heavily instrumented (see next section), and were monitored and sampled for several responses: primary production, secondary production, benthic and emerging stream invertebrates (composition and abundance), and visitation by riparian birds. The three periods we designated in the study are start (5/11/2019 - 6/10/2019), middle (6/11/2019 - 8/2/2019), and end (8/3/2019 - 8/21/2019).

We measured water depth and water temperature every five minutes throughout the experiment (4/21/2019–8/25/2019) with replicated pressure transducers (HOBO U20L-04, Onset). We placed a pressure transducer in the fifth pool downstream in each channel and two emerged sensors on land to correct data for fluctuations in atmospheric pressure, and thus calculate water level (i.e., pool depth). Water level series were subsequently transformed into discharge series via channel-specific rating curves. Rating curves were developed for each channel by estimating discharge manually using channel depth and velocity measurements taken with a Marsh-McBirney Flo-Mate 2000 current meter throughout the summer (17-26 repeated estimates per channel). We measured water temperature using the same HOBO U20L-04 sensors that recorded data every five minutes in pools. We averaged discharge and water temperature to hourly values, which we then used to calculate daily metrics (e.g., daily mean, minimum, maximum, and diel range).

We estimated epilithic biofilm primary production using the light/dark bottle method at each channel, once every three weeks. We calculated epilithic biofilm respiration (ER), net primary production (NPP), and the sum of their absolute values–gross primary production (GPP). We used three representative cobbles from the streambed for each sample and measured their surface area using aluminum foil to correct for differences in surface area. All primary production measurements were taken during peak sunlight hours between 10 am and 2 pm using two 90-minute incubation periods for light, followed by dark measurements. Benthic stream invertebrates were removed from rocks prior to incubation. We conducted three replicates for each channel at each sampling date (n = 162). Daily epilithic biofilm GPP per channel was estimated by multiplying the channel average hourly rate by the number of sunlight hours at each date (n = 54). We estimated daily epilithic biofilm ER per channel by multiplying the channel average hourly rate by 24 hours at each date (n = 54). Daily primary production was then estimated for the interval between each sampling date by averaging the bookend interval values. We multiplied the average interval value by the number of days in the interval and finally summed these values to generate cumulative seasonal channel estimates (n = 9).

We sampled benthic stream invertebrates using a 500-micron Surber sampler at six visit dates three weeks apart throughout the experiment. Each sample was a composite of three subsamples (two riffles and one pool sample for 0.279 m^2 total) to represent the overall stream community. We took benthic samples for the current and 6-week treatment channels (n = 36) and stored them in 70% ethanol. We then subsampled the composite samples using a rotating-drum splitter in the laboratory to sort and identify at least 500 individuals from each composite sample under a stereomicroscope. All subsamples were completely processed to avoid bias regarding the size of individuals picked and identified. Benthic stream invertebrates were identified to the highest resolution possible, typically genus or species level, and all intact specimens were measured. Benthic stream invertebrate biomass was then estimated using published taxon-specific length-mass relationships. The subsampled community was multiplied by the inverse of the fraction of the total sample that was identified (e.g., if ¼ of the sample was identified to get a count over 500 individuals, then the abundance of each taxon was multiplied by 4). We assigned length values to these extrapolated individuals (and individuals that could be identified but not measured due to damage) using the length values from randomly selected individuals of the same taxon in the sample.

We sampled emergent stream invertebrates using emergence traps, each deployed for 72 hours every three weeks during the experiment. We sampled emergence four additional times halfway between the three-week intervals for every sample visit after the second one when flows began to differ between treatments (n = 90 overall). We deployed emergence traps at the tail of riffles (to capture the influence of both riffle and pool habitat) next to HOBO sensors. We identified emergent insects to genus or family level (depending on taxa), and measured the length of intact specimens. Emergence traps were tent-shaped, covered 0.33 m2 of the stream, and had 2 mm white mesh.

We noticed Brewer's Blackbirds (Euphagus cyanocephalus) feeding in channels at the onset of low flow in the 6-week treatment channels (June 22, 2019). We recorded the feeding behavior of Brewer’s Blackbirds shortly thereafter by observing the time duration that any bird of this species occupied the benthos of the channels over a 30-minute period periodically throughout the remainder of the experiment. We observed all channels every few days initially but switched to weekly observations once Brewer’s Blackbirds fledged and moved to meadow habitat.

Laboratory, field, or other analytical methods- We estimated benthic stream invertebrate secondary production via a combination of three methods. We used the size-frequency method for taxa that were abundant throughout the experiment (i.e., >1% of total abundance) and had known generation times, excluding Chironomidae, Oligochaeta, Turbellaria, and Muscidae. For Chironomids, we used the instantaneous growth rate method. Production was calculated using regression equations for non-Tanypodinae chironomids, which incorporate mean temperature into growth estimates for small, medium, and large chironomids. Finally, we used the production-to-biomass ratio method (P/B) for the remaining taxa, including Tanypodinae, by multiplying seasonal biomass by known P/B ratios in the literature of the closest related taxa possible. Uncertainty in production from P/B ratios is unlikely to affect our results, as taxa in this group comprised <1% of the total assemblage production. We estimated emergent insect biomass using published, taxon-specific length-mass relationships.

*Quality control- Data was recorded in hard copies and digitally to reduce the risk of mistyped data. Data was plotted visually for outliers that were erroneous and paper copies were referenced to ensure values were correct.