Assessing pelagic and benthic metabolism using free water measurements

Abstract

Automated in situ sensors for measuring changes in dissolved oxygen (DO) at high frequency have facilitated estimates of gross primary production (GPP) and respiration (R) in aquatic systems. Lake researchers usually rely on a single sensor for these estimates, but such point measurements may miss important spatial heterogeneity in within-lake processes and may not accurately represent systemwide values of metabolism. Here we combine simultaneous measurements of metabolism using DO sensors along transects from the shore to the center of a lake with a spatial model to better understand the underlying heterogeneity in processes contributing to whole-lake epilimnetic metabolism. We use this model to achieve better estimates of epilimnetic GPP and R and to determine the relative contributions of benthic-littoral vs. pelagic processes to these estimates. We compared the spatially explicit process-based model to estimates of metabolism from both a single sensor at the lake's center and a spatially explicit averaging of multiple sensor sites. Estimates of both GPP and R varied on average 2.5-to 3.2-fold from site to site within the same lake, whereas variations were sometimes as high as 6-to 7-fold. Estimates of GPP and R near the perimeter of lakes were on average greater than measurements in the middle of the lake. Our model estimates that benthic-littoral processes accounted for ~40% of epilimnetic GPP and R. A single, centrally located sensor often misses a significant component of this benthic metabolism and accounts for only ~81% of lakewide GPP and R.

Intra- and inter-annual variability in metabolism in an oligotrophic lake

“…However, even if the residuals are autocorrelated, the maximum likelihood technique produces the same parameter estimates as least-squares methods if the residuals are assumed to be normally distributed with a mean of 0 (McNair et al 2013). Additionally, modeling an autocorrelation term from the residuals resulted in non-autocorrelated residuals but did not change the estimates of metabolic rates (Van de Bogert et al 2007). All analyses were conducted in the R statistical environment (v. 3.0.2, R Development Core Team 2013).…”

“…Open-water modeling techniques fit a model with daily metabolic rates as parameters to observed curves of diel oxygen concentrations (Van de Bogert et al 2007), which are usually small in oligotrophic lakes. Therefore, we modified existing modeling techniques to calculate metabolic rates for Lake Sunapee (see below).…”

“…2) with a maximum likelihood method (Van de Bogert et al 2007;Hanson et al 2008;Solomon et al 2013). We acknowledge the potential for autocorrelated residuals because a smoothed model often cannot closely follow the irregular variability in a time series (see ''Results'' below for examples).…”

“…We also compared the modified modeling techniques to the traditional bookkeeping techniques. DO dynamics were calculated for each day using a simple model (Odum 1956;Van De Bogert et al 2007;Solomon et al 2013):…”

“…Traditionally, metabolism has been estimated directly from bottle incubations or diel oxygen fluctuations using a 'bookkeeping' calculation (Odum 1956;Cole et al 2000;Staehr et al 2010). Advances in high-frequency dissolved oxygen sensor technology (e.g., Staehr et al 2012;Weathers et al 2013), increased metabolism model complexity, and use of parametric statistical techniques (van de Bogert et al 2007;Hanson et al 2008) have improved the spatial and temporal resolution of metabolism estimates. These advances have enabled new discoveries about the role of lakes in the global C cycle (e.g., Solomon et al 2013), controls on metabolism (Hoellein et al 2013), and spatial variability of metabolism within lakes (Van de Bogert et al 2012).…”