Abstract
Studies on historical contingency in community assembly frequently report the presence of alternative stable states as the result of different assembly sequences. If we can observe multiple assembly sequences and resulting community structure, these observations collectively inform the constraints on community assembly dynamics that emerge from various ecological processes. These observations would be a basis for predicting the outcome of assembly processes and understanding the mechanisms that shape species assemblage. However, empirical approaches such as invasion/removal experiments require enormous time and effort and are impossible in some cases. Here, we show that data on multispecies occurrences analyzed using a pairwise maximum entropy model and energy landscape analysis are capable of providing insights into the constraints on community assembly dynamics. This approach is a minimal theoretical framework to systematically and mechanistically study community assembly dynamics. Community assembly has a prominent role in shaping real world ecosystem organization. Our approach provides a new systemic paradigm for developing a predictive theory of community ecology and can have broad impact on ecological studies and its applications including regime shifts and ecosystem management.