Potentials and limitations of complexity research for environmental sciences and modern farming applications

Open system analysis is prone to the oversimplification of dynamics due to tightly coupled variables and their nonlinear, complex, and often unpredictable behavior. By assessing the combination of different ecosystem variables (structural, chemical, and biological) and their dynamic states in time and space, individual complexity measurements can capture phase changes of ecosystem stability and enhance efficiency, disease detection, and ecosystem understanding. This article summarizes the latest developments in complexity research and investigates the potential of metrics to assess and predict the sustainability and resilience of ecosystems, with a particular focus on farming systems. It provides an outlook on improving machine learning approaches by considering the system's complexity and the necessary data requirements. A GitHub repository [1] is presented that enables practitioners to use complexity applications (e.g. entropy metrics and reconstructed phase spaces). This research provides a deeper understanding of the connections between data complexity, machine learning algorithms, and environmental modeling.