1. Submerged macrophytes are important foundation species that can strongly influence the structure and functioning of aquatic ecosystems, but only little is known about the temporal variation and the timescales of these effects (i.e. from hourly, daily, to monthly).

2. Here, we conducted an outdoor experiment in replicated mesocosms (1000 L) where we manipulated the presence and absence of macrophytes to investigate the temporal variability of their ecosystem effects. We measured several parameters (chlorophyll-a, phycocyanin, dissolved organic matter [DOM], and oxygen) with high-resolution sensors (15 min intervals) over several months (94 days from spring to fall), and modelled metabolic rates of each replicate ecosystem in a Bayesian framework. We also implemented a simple model to explore competitive interactions between phytoplankton and macrophytes as a driver of variability in chlorophyll-a.

3. Over the entire experiment, macrophytes had a positive effect on mean DOM concentration, a negative effect on phytoplankton biomass, and either a weak or no effect on mean metabolic rates, DOM composition, and conductivity. We also found that macrophytes increased the variance of DOC composition and metabolic rates, and, at some times of the observed period, increased the variance of phytoplankton biomass and conductivity. The observation that macrophytes decreased the mean but increased the variance of phytoplankton biomass was consistent with the model that we implemented.

4. Our high-resolution time series embedded within a manipulative experiment reveal how a foundation species can affect ecosystem properties and processes that have characteristically different timescales of response to environmental variation. Specifically, our results show how macrophytes can affect short-term dynamics of algal biomass, while also affecting the seasonal buildup of DOM and the variance of ecosystem metabolism.

The dataset was assembled by means of four autonomous multi parameter instruments (EXO2 modular sensor platform [YSI-WTW]). The sondes were placed approximately at the center of the mesocosm (~0.5 m depth), away from the walls and outside of patches of macrophytes. Additionally, we measured photosynthetically active radiation (PAR) in 15 min intervals using a quantum sensor (Li-Cor) installed onsite to estimate surface light irradiance. PAR and temperature data (Fig. S2) were used together with the dissolved oxygen data to calculate metabolic rates.

Prior to the experiment, we performed a 48h cross-comparison trial where we installed all the sondes in a single mesocosm, enabling us to correct for differences among sensors and calibrate them against each other. During the cross-comparison trial we also quantified chlorophyll-a concentration by analyzing water samples with high performance liquid chromatography (HPLC, Jasco), and calibrated the optical sensors installed on the sondes in accordance with the manufacturer’s manual (YSI-WTW). Hence, we report chlorophyll-a as µg * L^-1, Phycocyanin and fDOM as raw fluorescence units.  The oxygen sensors were calibrated against water-saturated air.

Prior to the statistical analysis we removed incomplete days at the beginning and end of each measurement period (five time series: t1-t5). After this, each of the five time series had 864 data points (15 min interval = 96 data points per day = 9 days) for t1-3 and 576 data points (= 6 days) for t4 and t5. In a second step, we identified residuals of the detrended data that were outside 2.5 times the interquartile range as outliers and removed them from the data set. Finally, we used sliding windows with a size of 96 timepoints (= 1 day) to calculate time series of mean and cv, resulting in 768 data points for t1-t3 and 480 data points for t4-t5 (8 and 5 days, respectively).

Here we provide both the corrected/calibrated, full data set and the processed data (sliding windows, sliding window summaries, metabolism modelling, manual DOC measurements). We also provide a script from which the figures can be reproduced, as an entry point to the comprehensive dataset. See the readme on how to get started.