Adaptive forecasting of phytoplankton communities

Abstract

The global proliferation of harmful algal blooms poses an increasing threat to water resources, recreation and ecosystems. Predicting the occurrence of these blooms is therefore needed to assist water managers in making management decisions to mitigate their impact. Evaluation of the potential for forecasting of algal blooms using the phytoplankton community model PROTECH was undertaken in pseudo-real-time. This was achieved within a data assimilation scheme using the Ensemble Kalman Filter to allow uncertainties and model nonlinearities to be propagated to forecast outputs. Tests were made on two mesotrophic lakes in the English Lake District, which differ in depth and nutrient regime. Some forecasting success was shown for chlorophyll a, but not all forecasts were able to perform better than a persistence forecast. There was a general reduction in forecast skill with increasing forecasting period but forecasts for up to four or five days showed noticeably greater promise than those for longer periods. Associated forecasts of phytoplankton community structure were broadly consistent with observations but their translation to cyanobacteria forecasts was challenging owing to the interchangeability of simulated functional species.

A near-term iterative forecasting system successfully predicts reservoir hydrodynamics and partitions uncertainty in real time

“…Overall, our study demonstrates the utility of a workflow for lake and reservoir water temperature forecasting that can be applied to other waterbodies. In addition, FLARE builds the foundation for future water quality data assimilation and forecasting because ecosystem models can easily be coupled to the hydrodynamic model, enabling predictions of dissolved oxygen, algal blooms, and biogeochemical cycling with uncertainty [e.g., Hipsey et al 2013, Page et al 2018, Zwart et al 2019 ]. Importantly, FLARE provides a method for partitioning uncertainty in forecasts that identifies how to prioritize future research to increase confidence in forecasts.…”

A Near‐Term Iterative Forecasting System Successfully Predicts Reservoir Hydrodynamics and Partitions Uncertainty in Real Time

“…[ Baracchini et al, 2020a;Clark et al, 2008;Dietze, 2017a;Page et al, 2018]. Our implementation of the EnKF with state augmentation to calibrate parameters (Supporting Information A) follows Zhang et al [2017].…”

“…Real-time forecasts of water temperature with fully-specified uncertainties are particularly valuable for managers that oversee drinking water supply lakes and reservoirs, as waterbody temperatures can be very dynamic due to meteorological forcing, management, the evaluation of both forecast accuracy and the reliability of uncertainty estimation. Despite the importance of quantifying multiple uncertainty sources, few water resource forecasting studies quantify more than one or two sources of uncertainty and when they do, they typically only include initial conditions uncertainty (via data assimilation) and meteorological uncertainty (via ensemble weather forecasts) [e.g., Baracchini et al, 2020b;Komatsu et al, 2007;Ouellet-Proulx et al, 2017a;Ouellet-Proulx et al, 2017b;Page et al, 2018].…”

“…Fourth, FLARE propagates and partitions multiple sources of forecast uncertainty, which are updated over time, allowing for the evaluation of both forecast accuracy and the reliability of uncertainty estimation. Despite the importance of quantifying multiple uncertainty sources, few water resource forecasting studies quantify more than one or two sources of uncertainty and when they do, they typically only include initial conditions uncertainty (via data assimilation) and meteorological uncertainty (via ensemble weather forecasts) (e.g., Baracchini, Wuest, & Bouffard, 2020; Komatsu et al, 2007; Ouellet‐Proulx, Chimi Chiadjeu, et al, 2017; Ouellet‐Proulx, St‐Hilaire, & Boucher, 2017; Page et al, 2018). Furthermore, they rarely partition the relative contributions of the individual sources of uncertainty to the total forecast uncertainty (but see Ouellet‐Proulx, Chimi Chiadjeu, et al, 2017).…”

Long-term influence of climate and experimental eutrophication regimes on phytoplankton blooms

“…The response of lakes to concurrent drivers of eutrophication can be explored using 91 process-oriented models (Couture et al 2018; Page et al 2018;Janssen et al 2019). A variety of 92 lake ecosystem models exist that include physical processes and nutrient dynamics, varying in 93 modeling approach, spatial dimensions, and complexity of process representation (Robson 94 2014).…”