A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 1990

For the past two decades , point process models have become the focus of many research activities related to the rainfall processes. Until now, however, there has not been much work done on exploring the applicability of such model s on rainfall data associated with monsoon. Three point process model s have been selected for modeling the rainfall process of Rechna Doab, a monsoon affected region in Pakistan Two of these point process models are continuous-time namely, Neyman- Scott Clustering Model (NSCM) and Rectangular Pulses Model (RPM), and the third model, Markov renewal model (MRM) is the discrete- time. All three models were calibrated on seasonal basis using some selected sets of data ( about twenty one years) and then validated with independent sets of data. The data were taken from six stations representing the spatial variations of the rainfall behavior of Rechna Doab. The assessment of goodness of fit of the mode l s was performed by comparing empirical functions of the data not used in calibration process with their theoretical counterparts . Then comparison of three models at different levels of time aggregations was done with Cv- ratio function which is the ratio of coefficient of variation at an arbitrary time scale to that at a time scale equal to one day. Based on this criterion it has been found that no one model is superior to the other models. However, by recalibrating the models with all six sets of data grouped together, MRM became the only mod el that gave theoretical Cv- ratio functions that are consistent with the individual empirical cv-ratio functions of all six rainfall stations. Therefore, MRM is preferred over NSCM and RPM as the point process model for Rechna Doab.