Forecasting the Incidence and Duration of Harmful Algal Blooms (HABs) at Daily, Weekly and Seasonal Scales

Objective:

In year 1, NWM hydrological hindcasts and forecasts will be used to drive an already validated process-based lake model (AEM3D) to generate CyanoHABs hindcasts and forecasts at daily (1-2 days), weekly (7-10 days), and seasonal (30 days) lead times in two bays of Lake Champlain, Missisquoi Bay and St. Albans Bay to demonstrate the ability of the NWM output to be used as input in integrated water resource prediction systems.

Abstract:

Despite significant advancements in satellite monitoring of Harmful Algal Blooms (HABs), the "accurate" forecasting of HABs and development of "real-time" HABs Early Warning Systems (HABEWS) at finer spatial (200 m to 500 m) and temporal (daily to seasonal) scales still requires a considerable amount of basic and applied research. In this sub-project, we will advance the National Water Model (NWM’s) predictive intelligence for early warnings of HABs at daily, weekly, and seasonal lead time scales by applying machine learning, process-based modeling, and hybrid frameworks that couple the two. This project will scale and test an approach to forecasting HABs in freshwater lakes and estuaries that leverages hydrological forecasts derived from NWM and existing Earth Observation datasets currently being produced in real time through satellites and in situ monitoring systems and sensors. In year 1, WRF-Hydro derived NWM hydrological hindcasts and forecasts will be embedded in an existing Integrated Assessment Model (IAM) computational workflow to drive an already calibrated and validated process-based lake model (AEM3D). The IAM also uses weather data that can be derived from National Weather Model and/or IBM weather forecast products. This workflow will produce high resolution HAB hindcasts and forecasts in two bays of Lake Champlain (Missisquoi Bay and St. Albans Bay). In year 2, we will develop and test a self-learning AI-HABEWS by identifying a best-fitting deep neural network to update AEM3D by validating HAB forecasts through community science monitors, and in situ & satellite sensors. In year 3, we will conduct a sensitivity analysis of HAB forecast accuracy (as predicted by machine-learning, process based, and hybrid forecast models) to hydrological forecasts based on NWM (both WRF-Hydro and Topmodel/NextGen versions), SWAT, and RHESSys models. This sub-project will advance NOAA's mission to understand and predict the effects of changing climate, weather, and socio-environmental factors on marine ecosystems. This may, in turn, help conserve marine ecosystems and further advance NOAA's vision of building resilient and healthy ecosystems.