﻿Methods 
Experimental design 
In this study, we tested the interactive effects of shading and drought on algal biomass and ecosystem metabolism. We used 12 outdoor, artificial channels with shading tarps to mimic differences in canopy cover, and in addition, we induced a two-week drought. The experiment ran from August 4 to October 3, 2023, with the drought implemented between September 5 and 18, and the post drought recovery between September 19 and October 3. We consider August as summer and September and October as autumn months. 
 
The artificial channels (or flumes, Figure 1) are located at Svartberget field station in northern Sweden (Laudon et al., 2021). There are 12 individual flumes, and they are all 15 m long and 20 cm wide and during the experiment, water depth varied from 3 to 10 cm (top to bottom), with a slope of ~0.5 cm per meter. Water discharge was constant at 1-2 L s-1 and water velocity was 0.1 m s-1. Water to the flumes was continuously pumped from an adjacent forest stream using a bilge pump (Flygt KS 2610) to a 3000 L water collection tank, from which the water is led to four 1000 L boxes, each feeding three of the channels (Figure 1). The residence time in the boxes is less than 30 min. This setup ensures that water is well mixed in the tank and boxes so that all flumes have similar chemistry. The bottom substrates in all flumes were made up of 1-2 cm layer of finer sediments (majority being sand) on the bottom, with one layer of coarser substrates on top (pebbles, cobbles) from a nearby quarry, representing local, natural material, dominated by gneiss and granite. The substrates were added without a colonization phase. 
 
We created a gradient of shading using tarps with 0, 4, 19, 39, 43 and 74% shading. This shading was supposed to mimic different canopy cover; with the highest shading (74%) representing a stream surrounded by mature forest dominated by spruce and the lowest shading represents a stream in a clear-cut. The gradient in between reflects different levels of riparian protection (buffer widths, Chellaiah & Kuglerová, 2021). Each shading treatment was replicated twice to facilitate drying of one channel per shading treatment during the second half of the experiment. Shading and drought treatments were assigned randomly (Figure 1). We started the experiment on August 4 and added the shading tarps on August 7. We ran the experiment with shading tarps for five weeks and on September 5th; we gradually turned off the water inlet to half of the channels (one replicate of each shading level) for two weeks, until September 18. During the first week of the drought, water was decreased by 90% and during the second week of drought, we turned the water off completely to mimic a gradually occurring drought. During the week when the water was fully turned off, there was still moisture on the bottom of the channel bed, but elevated rocks were completely dry (Figure 2). On September 18, water to the dry channels was turned back on and we continued the experiment for another 2 weeks to capture the recovery period.   
 
 
Water temperature, light and chemistry 
We recorded incident light in lux and water temperature in o C every 15 min using two Hobo pendant loggers (Onset Computer Corporation, Borne, U.S.A.) per channel. We did not estimate lux or temperature in the dry channels during the drought (as this would have recorded air temperature and lux). Lux was converted to photosynthetically active radiation (PAR), using a conversion factor of 0.0185 according to Thimijan & Heins (1983) and presented as daily photon flux (DPF, mol photons m-2 day-1). Loggers were cleaned from algal growth and debris 1-2 times per week. We sampled water from the boxes and from the downstream end of each channel three times to measure concentrations of nitrate (NO3-), ammonium (NH4+), phosphate (PO43-) and dissolved organic carbon (DOC). This was done at the start of the experiment (2023-08-07), pre-drought (2023-09-05) and immediately after turning the dry channels back on (2023-09-18). Dissolved inorganic nitrogen (DIN) hereafter represents the sum of NO3- and NH4+. Water was filtered (0.45 µm) on site and kept refrigerated (DOC) or frozen (DIN and PO43-) before colorimetric analysis of NO3--N (Method G-384-08), NH4+-N (Method G-171-96), and PO43- (Method G-297-03) with an Auto Analyzer 3 Spectrophotometer (Omniprocess).  
 
Algal biomass 
Periphytic algal biomass was estimated on the natural bottom substrate (cobbles and sand) of the channels as total chlorophyll a (chl-α) using a BenthoTorch (Moldanke BBE). The BenthoTorch is a handheld instrument that analyses chl-α pigments in-situ by fluorescence. The BenthoTorch measures an area of 1.1 cm2 and we systematically measured each channel once per week using established transects from top to bottom with ten measurements per channel and sampling occasion. The BenthoTorch enables frequent and non-destructive estimates of chl-α, which we could not accomplish using ceramic tiles or other sub-sampling of the channel bottom.  
 
Ecosystem metabolism 
Details on metabolic calculations in this facility are presented in Myrstener et al., (2023), in short, one miniDOT logger (Precision Measurement Engineering Inc., USA) was placed at the downstream end of each channel to record dissolved oxygen at 10 min intervals. We recorded oxygen throughout the experimental period (August 7 to October 4), except in the dry channels during the drought, as these did not have water during that time. Ecosystem metabolism was estimated using the single-station diel oxygen method where gross primary production (GPP) and Ecosystem Respiration (ER) was estimated using Bayesian inverse modelling. We used time series of dissolved oxygen (DO), water temperature, light (from lux loggers), as well as a constant gas exchange rate coefficient (K) and channel depth (z). The main equation for GPP and ER was: 
 
dDO/dt=(GPP+ER)/z+K(DOsat-DO) 
 
The change in oxygen over time (O2 m-3) equals all oxygen produced by photosynthesis (GPP, g O2 m-2 d-1) minus all oxygen consumed by respiration of both autotrophs and heterotrophs (ER, g O2 m-2 d-1) and the rate of gas exchange between the water and air (K, d-1). We modelled GPP and ER and used constant K based on nighttime regression analysis. Usually, K is modelled based on a K~discharge relationship because these variables co-vary. As Q is stable in the channels, we used the same K for the whole period and this K was based on nighttime regression from the channels. We divided the experimental period into four periods and presented the metabolic rates as the average daily rates over each period. As water is well mixed in the boxes and in the inlet of the channels, we expect little influence of the box and upstream water on the channel metabolism. In any case, all channels have the same background input. 
 
Statistical analysis 
Light, temperature and water chemistry  
Incident light and temperature in the flumes are described using summary statistics. The significance of light for the response variables will be analysed as shading level and is described further down below. Water chemistry, more precisely DOC, phosphate, ammonium, and nitrate, is presented for all three sampling rounds. Averages in water chemistry are summarised for dry and wet channels per sampling round and the differences in the post-drought measurements are compared using t-tests for phosphate (log-transformed), ammonium, and nitrate and a U-test for DOC. 
 
Effects of shading level and drought 
The effects of shading level and drought on chl-α, GPP, and ER was analyzed using a linear mixed-effects ANOVA model. The models included chl-α, GPP, and ER respectively, as the dependent variable, and drought (dry vs. wet) and shading (gradient from 0 to 74%) as the fixed factors, with their interaction (drought * shading) modelled to assess their combined effect. Shading was treated as a continuous variable because we saw a linear relationship between light and the dependent variables. Channel ID was included as a random effect to account for channel-specific variability in measurements. The model was fitted using the lme function from the nlme package (Pinheiro et al., 2021) in R. Specifically; the model was expressed as (example with GPP): 
GPP~Drought*Shading1+Drought*Week+Shading1*Week, random=~1|Channel 
 
To assess the effects of drought and shading for each week of the experiment, we used estimated marginal means (emmeans package, Lenth, 2023) and performed pairwise contrasts within each week. Statistical significance was determined using F-tests from joint tests and the multiple comparisons were adjusted using Tukey’s. 

