There are many different types of m

There are many different types of models (in addition to the level of abstraction or biological organization discussed in Section 3) and ways in which observations can inform them and vice versa. Empirical models, such as regression equations or artificial neural networks, are based or built on observations. Mechanistic models make predictions of the state of the microbial system (e.g., the concentration of cyanobacteria) based on principles of conservation of mass and rate laws or equations. For these models, observations are typically compared with model output during calibration or validation. Observations may also be used less directly by providing information about the ecosystem structure—e.g., observations of certain functional genes may influence state variables and processes included in a model. In addition, observations can be used more directly, as input or via data assimilation, where model predictions are corrected using data. Here, the focus is on mechanistic models and the direct comparison of their predictions with observations.