The overall morphology and dynamics of magnetospheric substorms is well established in terms of the observed qualitative auroral features seen in ground based magnetometers. This paper focuses on the quantitative characterisation of substorm dynamics captured by ground based magnetometer stations. Network analysis is a well established technique that spans many fields of research. We present the first analysis of substorms using dynamical networks obtained from the full available set of ground based magnetometer observations in the northern hemisphere. The dynamical networks here are comprised of time dependent connections between pairs of stations. The stations are connected when the correlation between vector magnetometer time series from pairs of stations within a running time window exceeds a threshold. Dimensionless parameters can then be obtained that characterise the network and by extension, the spatio-temporal dynamics of the substorm under observation. We analyse four isolated substorm test cases as well as a steady magnetic convection (SMC) event and a day in which no substorms occur. These test case substorms are found to give a consistent characteristic network response at onset in terms of their spatial correlation. Such responses are differentiable from responses to the SMC event and non-subtorm times. We present a method to optimise network parametrisation with respect to the spatial inhomogeneity of stations in the northern hemisphere and the choice of correlation window sizes. Our results suggest that dynamical network analysis has potential to quantitatively categorise substorms.