Basic ensemble Kalman filter algorithms are extremely powerful tools for atmospheric data assimilation. Ensemble Kalman filters are based on theory that assumes that all probability distributions are Gaussian and that the relation between observations and model state variables is linear. However, neither of these assumptions is valid for data assimilation applications in the atmosphere or other parts of the Earth system. In particular, the relation between observations and state variables can be distinctly nonlinear, especially for new high-resolution remote sensing observations that are crucial for analyzing and predicting convective-scale extreme events. This session will explore ensemble filtering algorithms that provide better results than basic ensemble Kalman filters for problems with strong nonlinearity. Because of the complexity of understanding these algorithms, presentations that analyze the capabilities of novel methods in simple idealized or low-order models are encouraged. Results for large atmospheric model applications where nonlinearity is a challenge are also appropriate. Ensemble algorithms and hybrids of ensembles with variational or particle methods are all valid approaches for this session. Contributions from mathematicians or statisticians working on data assimilation applications are encouraged, as the frontiers of nonlinear ensemble data assimilation clearly require interdisciplinary collaboration between geoscientists and other researchers.