Main content area

Castles built on sand or predictive limnology in action? Part A: Evaluation of an integrated modelling framework to guide adaptive management implementation in Lake Erie

Arhonditsis, George B., Neumann, Alex, Shimoda, Yuko, Kim, Dong-Kyun, Dong, Feifei, Onandia, Gabriela, Yang, Cindy, Javed, Aisha, Brady, Meghan, Visha, Ariola, Ni, Felicity, Cheng, Vincent
Ecological informatics 2019 v.53 pp. 100968
Cyanobacteria, Soil and Water Assessment Tool model, adaptive management, basins, business enterprises, case studies, chlorophyll, diagenesis, dissolved oxygen, ecological models, ecosystems, fractionation, limnology, model validation, nutrient content, organic matter, philosophy, phosphorus, pollution load, prediction, rivers, sand, temperature, total phosphorus, water pollution, water quality, watersheds, zooplankton, Lake Erie
We present a technical analysis of all the recent modelling work that has been conducted to support the adaptive management process in Lake Erie; the most biologically productive system of the Great Lakes. With a wealth of models developed, Lake Erie represents a unique case study where an impressive variety of data-driven and process-based models have been developed to elucidate the major watershed and aquatic processes underlying the local water quality problems. In the Maumee River watershed, the primary contributor of total phosphorus loading (~30%) into Lake Erie, the modelling work is based on five independent applications of the same process-based model, i.e., the Soil and Water Assessment Tool (SWAT). The five SWAT models showed nearly excellent goodness-of-fit against monthly flow rates and phosphorus loading empirical estimates based on a single downstream station, but little emphasis was placed on evaluating the robustness of the hydrological or nutrient loading predictions with a finer (daily) temporal resolution, and even less so in capturing the impact of episodic/extreme precipitation events. The multi-model ensemble for the Lake Erie itself has been based on a wide range of data-driven and process-based models that span the entire complexity spectrum. Consistent with the general trend in the international modelling literature, the performance of the aquatic ecological models in Lake Erie declined from physical, chemical to biological variables. Temperature and dissolved oxygen variability were successfully reproduced, but less so the ambient nutrient levels. Model performance for cyanobacteria was inferior relative to chlorophyll a concentrations and zooplankton abundance. With respect to the projected responses of Lake Erie to nutrient loading reduction, we express our skepticism with the optimistic predictions of the extent and duration of hypoxia, given our limited knowledge of the sediment diagenesis processes in the central basin and the lack of data related to the vertical profiles of organic matter and phosphorus fractionation or sedimentation/burial rates. Our study also questions the adequacy of the coarse spatiotemporal (seasonal/annual, basin- or lake-wide) scales characterizing the philosophy of both the modelling enterprise and water quality management objectives in Lake Erie. We conclude by arguing that one of the priorities of the local research agenda must be to consolidate the ensemble character of the modelling work in Lake Erie. The wide variety of models that have been developed to understand the major causal linkages/ecosystem processes underlying the local water quality problems are a unique feature that should be cherished and further augmented.