Significant improvements have been made in estimating gross primary production (GPP), ecosystem respiration (R), and net ecosystem production (NEP) from diel, “free‐water” changes in dissolved oxygen (DO). Here we evaluate some of the assumptions and uncertainties that are still embedded in the technique and provide guidelines on how to estimate reliable metabolic rates from high‐frequency sonde data. True whole‐system estimates are often not obtained because measurements reflect an unknown zone of influence which varies over space and time. A minimum logging frequency of 30 min was sufficient to capture metabolism at the daily time scale. Higher sampling frequencies capture additional pattern in the DO data, primarily related to physical mixing. Causes behind the often large daily variability are discussed and evaluated for an oligotrophic and a eutrophic lake. Despite a 3‐fold higher day‐to‐day variability in absolute GPP rates in the eutrophic lake, both lakes required at least 3 sonde days per week for GPP estimates to be within 20% of the weekly average. A sensitivity analysis evaluated uncertainties associated with DO measurements, piston velocity (k), and the assumption that daytime R equals nighttime R. In low productivity lakes, uncertainty in DO measurements and piston velocity strongly impacts R but has no effect on GPP or NEP. Lack of accounting for higher R during the day underestimates R and GPP but has no effect on NEP. We finally provide suggestions for future research to improve the technique.

In the 1980s, Danish coastal waters suffered from eutrophication and several nutrient management plans have been implemented during the years to improve ecological status. This study aims at giving a holistic ecosystem perspective on 25 years of mitigation measures. We report trends of nutrient inputs and the responses to these in various chemical and biological components. Nutrient inputs from land were reduced by ~50 % for nitrogen (N) and 56 % for phosphorus (P) since 1990. These reductions resulted in significant and parallel declines in nutrient concentrations, and initiated a shift in the dominance of primary producers towards reduced phytoplankton biomass (chlorophyll a concentration) and increased cover of macroalgae in deeper waters. In the last 5 years, eelgrass meadows have also expanded towards deeper waters, in response to improving water clarity. An expected improvement of bottom water oxygen conditions has not been observed, presumably because more frequent stratification and higher water temperatures have counteracted the expected positive effects of reduced nutrient inputs. The biomass of the benthic macrofauna decreased as expected, but it was composed of a drastic decline of filter feeders paralleled by a more moderate increase of deposit feeders. This shift was most likely induced by increasing stratification. The reduced benthic filtration along with the limited eelgrass cover probably kept relatively more particles in suspension, which can explain why improvements in the Secchi depths were modest. Overall, several ecosystem components demonstrated clear signs of improvement, suggesting that at least partial recovery is attainable. On this basis, we propose a conceptual scheme for recovery of shallow coastal ecosystems following marked reductions in nutrient inputs.