Ecology Knowledge-Guided Machine Learning
Process models based on empirical data have powerful predictive abilities when it comes to lake dynamics. Machine learning models have even more powerful predictive abilities, but only when high volumes of empirical data are available. Combining both process-based and machine learning approaches in a singular ecological model has been shown to generate more skillful predictions of a target variable than either of the two approaches alone. This paradigm of Knowledge-Guided Machine Learning (KGML) has been developing rapidly and has pushed the fields of ecology and limnology forward. The Eco-KGML group is dedicating to advancing this technology and utilizing it to answer ecological questions by fostering a collaboration between ecologists and computer scientists.
Related Reading:
Water Quality Modeling
Water quality is integral to the ecosystem services of lakes, but its drivers are complex and change temporally and spatially even within one system. We can gain a greater understanding of this variability and its source through modeling. Modular Compositional Learning (MCL) is a framework within the Ecology Knowledge-Guided Machine Learning (Eco-KGML) paradigm that allows ecosystem processes to be segmented and individually modeled using either process-based models or machine learning. Lake metabolism, one lens through which to understand water quality, is a prime candidate for modeling with MCL. We are developing a one-dimensional model of water quality using MCL, capturing the physical and biological processes that control lake thermal structure and water quality to estimate the organic carbon and dissolved oxygen along the water column. By swapping out different modules with process-based and machine learning versions, we can gain a greater understanding of the processes that control water quality.
Related Reading:
|