Kernel density estimation is a statistical technique which allows a continuous probability distribution to be estimated from empirical data. It is based on the concept of a kernel function, which is a continuous normalised functional form that is peaked (or ‘centred’) at a defined point. To produce an estimate from a set of data points, one such kernel function is centred at each data point, and the sum of these is taken. This essentially produces a ‘smoothed’ version of the data, in which each point is replaced by a ‘bump’ and these are superimposed upon each other. For spatial data, this produces a surface in which the influence of each point is distributed across its immediate vicinity, while for temporal data the influence is spread through time. In the context of crime prediction, these effects can be taken to reflect spatio-temporal propagation of risk, and it is clear that the ProMap approach is an informal version of this principle.