SPACE-TIME MODELING OF VECTOR HYDROLOGIC SEQUENCES1

Stochastic modeling of vector hydrologic sequences is examined with a general class of space-time autoregressive integrated moving average (STARIMA) models. The models describe spatial and temporal autocorrelatjon, through dependent variables lagged both in space and time. The model structures incorporate a hierarchical ordering scheme to map the vector of observations into a network configuration. The neighboring structure used introduces a physical/geographical hierarchy to enable the model identification procedures to assist in determining appropriate correlative relationships. The three-stage iterative space-time model building procedure is illustrated using average monthly streamfiow data for a four-station network of the Southeastern Hydropower System.