Case study of monthly regional rainfall evaluation by spatiotemporal geostatistical method

The rainfall time series as a spatiotemporal process requires suitable tools for prediction. In this paper, the application of a kriging (geostatistic) method in modeling the point rainfall time series is presented in time and space. The two components of rainfall time series-deterministic trends and random components-are modeled using the kriging method. Sequential Gaussian and LU (lower and upper triangular matrix decomposition) simulation are used to simulate the random process of each station, both in space and time. Finally, simulated random components and deterministic trends are used to generate different realizations of rainfall time series at each grid point. Thirty-four years of monthly data of 34 rain gauges in the Zayandeh-rud river basin in the central part of Iran are utilized in this study to model and simulate rainfall data in space and time. A network of 8 by 8 km grids is used to represent the region of approximately 232 by 224 km. The results will be useful in regional studies of climatic and hydrologic events such as droughts, as well as for assessment of exceedance probability of rainfall in time and space. This information will be a basis for probabilistic forecasting of rainfall in this region.