This study presents a spatiotemporal Markov-chain mixture distribution model of the clear-sky index for an arbitrary number of locations, and is particularly suited for simulations of small-scale spatial networks with a span of 10 km or less. The model is statistical, but in practice data-driven and based on clear-sky index input from an arbitrary number of locations to generate synthetic time-series for the same locations. When trained on clear-sky index data based on the NREL Hawaii network radiometer solar irradiance data, dispersed within 1 km x 1.2 km, the model is shown to have high goodness-of-fit compared with test data from the network in terms of probability distributions, autocorrelations, location pair-correlations and k-lag correlations between locations. It is also shown to perform comparably to state of the art temporal, spatial and spatiotemporal clear-sky index generators. All measures of model goodness-of-fit are shown to improve with increased number of bins, up to a certain limit of N > 4, where the performance improvements reaches a plateau. The results are also shown to be insensitive with respect to choice of training and test data sets as well as number of output time-steps.