The workflow supports analysis and modelling of spatio-temporal data in the form of spatially referenced numeric time series. Raw data are first transformed into spatial time series through spatio-temporal aggregation. The time series are then clustered by temporal similarity, and the resulting grouping is iteratively refined through visual assessment of within-group homogeneity on time graphs and spatial pattern interpretability on maps. For each group of similar time series, the analyst visually identifies temporal variation characteristics—such as periodicity, cycle lengths, and trends—from the time graph. A representative time series is derived from the group, and a suitable modelling method is selected and configured. A statistical model is fitted and iteratively refined by adjusting parameters while comparing the model curve against the data until it captures the group's characteristic temporal variation. Model quality is evaluated by examining distributions of residuals over time and space: randomly distributed residuals indicate that the model captures the essential spatio-temporal variation. If systematic residual patterns emerge, the analyst decides whether to subdivide clusters, adjust the modelling approach, or both, and repeats the analysis cycle. Once all models pass the residual check, their descriptions—including method, parameters, group membership, and distribution statistics—are stored externally for reuse, communication, and prediction. The same workflow structure applies to modelling dependencies between two spatio-temporal variables, using dependency series and regression models in place of temporal models.