Integration of lake modeling and paleolimnological records to perform long-term simulations of water quality in Lake Geneva over 250 years (1850–2100 period)

Lake systems are facing long-term (>150 years) changes around the world acting on multi-decadal to centennial scales. Historic temperature warming at global scales, projected to continue by the end of the century, acting concomitant with eutrophication has modified ecosystem functioning in complex ways. Process-based lake models have emerged as powerful tools to assess the effects of climate and human activities on ecosystems, as well as the responses under future scenarios since they take into account the processes in the boundaries lake-catchment and lake-atmosphere. Most of these models are constrained by short-term monitoring limnological records, traditionally ranging from days to a few decades, potentially limiting the robustness of long-term reconstructions. The integration of lake modeling and paleolimnological records can overcome the short-term monitoring data temporal scale, thereby providing a long-term perspective on lake ecosystem dynamics related to climate variability and human pressures. The present study develops a methodological framework using paleolimnological records from well-dated lake sediment records to constrain, validate and model temporal changes in water quality over a period of 250 years (1850–2100). Lake Geneva (France, Switzerland) was selected as a case study in face of its similarity with other peri-alpine lakes and its representativeness as it is one of the most studied and well-known lentic ecosystems in the world. The 1D hydrodynamic-biogeochemical GLM-AED2 model was applied to simulate dissolved oxygen, nutrients, and chlorophyll-a concentrations along the water column. Pluri-decadal series of limnological data monthly collected by the French Observatoire des LAcs (OLA database) were used to calibrate and validate the model. In addition, model outputs were further validated with published paleolimnological records for the past 170 years. Preliminary results of the calibration procedure show that the GLM-AED2 model accurately predicts the magnitude and seasonal dynamics of the state variables with goodness-of-fit metrics under the literature range (e.g. RMSE = 0.96 mg L^–1 and RRMSE = 25% for dissolved oxygen; RMSE = 6.53 ug L^–1 and RRMSE = 37% for chlorophyll-a, both in the epilimnion). The integration of a one-dimensional lake model, paleolimnological records, and in situ measurements supports a better understanding of the historical dynamics and provides more robust long-term hindcast/forecast simulations to elucidate the impacts of climate change and critical implications for lake management and planning.