Combined Earth observations reveal the sequence of conditions leading to a large algal bloom in Lake Geneva.

Freshwater algae exhibit complex dynamics, particularly in meso-oligotrophic lakes with sudden and dramatic increases in algal biomass following long periods of low background concentration. While the fundamental prerequisites for algal blooms, namely light and nutrient availability, are well-known, their specific causation involves an intricate chain of conditions. Here we examine a recent massive Uroglena bloom in Lake Geneva (Switzerland/France). We show that a certain sequence of meteorological conditions triggered this specific algal bloom event: heavy rainfall promoting excessive organic matter and nutrients loading, followed by wind-induced coastal upwelling, and a prolonged period of warm, calm weather. The combination of satellite remote sensing, in-situ measurements, ad-hoc biogeochemical analyses, and three-dimensional modeling proved invaluable in unraveling the complex dynamics of algal blooms highlighting the substantial role of littoral-pelagic connectivities in large low-nutrient lakes. These findings underscore the advantages of state-of-the-art multidisciplinary approaches for an improved understanding of dynamic systems as a whole.

Physico-chemical and high frequency monitoring dataset from mesocosm experiments simulating extreme climate events in lakes.

We present two datasets composed of high frequency sensors data, vertical in situ profiles and laboratory chemical analysis data, acquired during two different aquatic mesocosm experiments performed at the OLA ("Long-term observation and experimentation for lake ecosystems") facility at the UMR CARRTEL in Thonon les Bains, on the French shore of Lake Geneva. The DOMLAC experiment lasted 3 weeks (4-21 October 2021) and aimed to simulate predicted climate scenarios (i.e. extreme events such as storms and floods) by reproducing changes in quality and composition of lake subsidies and runoff by increased inputs of terrestrial organic matter. The PARLAC experiment lasted 3 weeks (5-23 September 2022) and aimed to simulate turbid storms by light reduction. The experimental setup consisted of nine inland polyester laminated tanks (2.1 m length, 2.1 m width and 1.1 m depth) with a total volume of approximately 4000 L and filled with water directly supplied from the lake at 4m depth. Both experimental design included three treatments each replicated three times. The DOMLAC experiment involved a control treatment (no treatment applied) and two treatments simulating allochthonous inputs from two different dissolved organic matter (DOM) extract from peat moss Sphagnum sp. (Peat-Moss treatment) and Phragmites australis (Phragmite treatment). The PARLAC experiment involved a control treatment (no treatment applied) and two treatments simulating two different intensity of light reduction. In the Medium treatment transmitted light was reduced to 70% and in the High treatment transmitted light was reduced to 15%. The datasets are composed of: 1. In situ measures from automated data loggers of temperature, conductivity, dissolved oxygen and CO2 acquired every 5 minutes at 0.1, 0.5 and 1 m depth (DOMLAC) and 0.5m (PARLAC) for the entire period of the experiment. 2. In situ profiles (0-1 m) of temperature, conductivity, pH, dissolved oxygen (concentration and saturation) acquired twice a week during the experiment. 3. In situ measures of light spectral UV/VIS/IR irradiance (300-950 nm wavelength range) taken in the air and at 0, 0.5 and 1 m twice a week on the same day of the profiles at point 2. 4. Laboratory chemical analysis of integrated samples taken twice a week on the same day of the in situ profiles at point 2 and 3 of conductivity, pH, total alkalinity, NO3, total and particulate nitrogen (Ntot, Npart), PO4, total and particulate phosphorus (Ptot, Ppart), total and particulate organic carbon (TOC, POC), Ca, K, Mg, Na, Cl, SO4 and SiO2. Only for DOMLAC also analyses of NH4, NO2 and dissolved organic carbon (DOC). 5. Laboratory analysis of pigments (Chla, Chlc, carotenoids, phaeopigments) extracted from samples collected at point 4. 6. Only for DOMLAC, specific absorbance on the range 600-200nm of DOM (i.e. <0.7 µm) measured on samples collected at point 4. This dataset aims to contribute our understanding of how extreme climate events can alter lake subsidies and affect the regulation of ecosystem processes such as production, respiration, nutrient uptake and pigment composition. The data can be used for a wide range of applications as being included in meta-analysis aiming at generalising the effect of climate change on large lakes including simulating future scenarios in a broad range of geographical areas as we used different inputs of DOM leached from litters reproducing catchments characteristics typical of different latitudes, such as mostly dominated by large leaf forests and phragmites at middle latitude, and coniferous forests rich of peat mosses that spread along the water surface typical of Northern regions.

The value of human data annotation for machine learning based anomaly detection in environmental systems.

Anomaly detection is the process of identifying unexpected data samples in datasets. Automated anomaly detection is either performed using supervised machine learning models, which require a labelled dataset for their calibration, or unsupervised models, which do not require labels. While academic research has produced a vast array of tools and machine learning models for automated anomaly detection, the research community focused on environmental systems still lacks a comparative analysis that is simultaneously comprehensive, objective, and systematic. This knowledge gap is addressed for the first time in this study, where 15 different supervised and unsupervised anomaly detection models are evaluated on 5 different environmental datasets from engineered and natural aquatic systems. To this end, anomaly detection performance, labelling efforts, as well as the impact of model and algorithm tuning are taken into account. As a result, our analysis reveals the relative strengths and weaknesses of the different approaches in an objective manner without bias for any particular paradigm in machine learning. Most importantly, our results show that expert-based data annotation is extremely valuable for anomaly detection based on machine learning.

Physico-chemical dataset from an in situ mesocosm experiment simulating extreme climate events in Lake Geneva (MESOLAC).

This dataset complement a previously published dataset [1] and corresponds to the physico-chemical parameters data series produced during the MESOLAC experimental project [2]. The presented dataset is composed of: 1. In situ profiles (0-3m) of temperature, conductivity, pH, dissolved oxygen (concentration and saturation). 2. In situ measurements of light spectral UV/VIS/IR irradiance (300-950 nm wavelength range) taken at 0, 0.25, 0.5, 1, 1.5, 2 and 2.5m. 3. Laboratory chemical analysis of samples collected at 0 and 2 m (conductivity, pH, total alkalinity, NH4, NO2, NO3, total and particulate nitrogen (Ntot, Npart), PO4, total and particulate phosphorus (Ptot, Ppart), total, organic particulate and total particulate carbon (Ctot, Cpart-org, Cpart-tot), Cl, SO4, SiO2. 4. Laboratory analysis of pigments extracted from samples collected at 0 and 2 m (Chla, Chlc, carotenoids, phaeopigments). The experimental design is the same as in Tran-Khac et al [1]. Briefly, it consisted of nine pelagic mesocosms (about 3000 L, 3m depth) deployed in July 2019 in Lake Geneva near the shore of Thonon les Bains (France) aiming to simulate predicted climate scenarios (i.e. extreme events) and assess the response of planktonic communities, ecosystem functioning and resilience. During the experiment, physical parameters were measured twice a week. At the same time, samples were collected at 0 and 2m of depth for subsequent chemical laboratory analyses. These data are presented in the dataset file, ordered by sampling event (numbered from S1 to S8), treatment (Control-C, High-H and Medium-M) and replicates (1 to 3). For each sampling point the measured parameters are listed in columns, missing data and values below the detection limit are marked as NA (not available). This data set aims to contribute to the understanding of the effect of environmental forcing on lake physico-chemical characteristics (such as temperature, oxygen and nutrient concentration) under simulated intense weather events. To a broader extent, the presented data can be used for a wide variety of applications, including monitoring of a large peri-alpine lake functioning under environmental stress and being included in further meta-analysis to generalise the effect of climate change on large lakes. The two complementary dataset differ in the acquired data and methods, temporal and spatial resolution. They complete each other in terms of physico-chemical characterization of the experimental treatments and together can allow comparison of the two different monitoring strategies (continuous vs punctual) during in situ experimental manipulations.

In situ pelagic dataset from continuous monitoring: A mesocosm experiment in Lake Geneva (MESOLAC).

This dataset corresponds to a data series produced from automated data loggers during the MESOLAC experimental project. Nine pelagic mesocosms (about 3000 L, 3 m depth) were deployed in July 2019 in Lake Geneva near the shore of Thonon les Bains (France), simulating predicted climate scenarios (i.e. intense weather events) by applying a combination of forcing. The design consisted of three treatments each replicated three times: a control treatment (named C - no treatment applied) and two different treatments simulating different intensities of weather events. The high intensity treatment (named H) aimed to reproduce short and intense weather events such as violent storms. It consisted of a short-term stress applied during the first week, with high pulse of dissolved organic carbon (5x increased concentration, i.e. total DOC ∼ 6 mg L-1), transmitted light reduced to 15% and water column manual mixing. The medium intensity treatment (named M) simulated less intense and more prolonged exposures such as during flood events. It was maintained during the 4 weeks of the experiment and consisted of 1.5x increased concentration of dissolved organic carbon (i.e. total DOC ∼ 2 mg L-1), 70% transmitted light and water column manual mixing. Automated data loggers were placed for the entire period of the experiment in the mesocosms and in the lake for comparison with natural conditions. Temperature, conductivity, dissolved oxygen and CO2 were monitored every 15 min at different depths (0.15, 0.25, 1 and 2 m). This data set aims to contribute our understanding of the effect of environmental forcing on lake ecosystem processes (such as production, respiration and CO2 exchange) under simulated intense weather events and the ability of the planktonic community to recover after perturbation. To a broader extent, the presented data can be used for a wide variety of applications, including monitoring of lake community functioning during a period of high productivity on a large peri-alpine lake and being included in further meta-analysis aiming at generalising the effect of climate change on large lakes.