Rainfall is a meteorological phenomena and its variations are link ed with climate con ditions in different region which change the local rainfall pattern. Th e understanding of long term (Seasonal or annual , inter - annual) rainfall variations is important for many sectors like agriculture, industry, health, irrigation, education so on in a region. Ping basin is one of major basin s of upper Chao Phraya in Thailan d. Thailand is under influence of Monsoon and tropical cyclone throughout the year. Present study developed the four seasonal rainfall forecasting models for Ping river basin using Artificial Neural Network (ANN) technique which is a data driven model wide ly used in non - linear analysis and forecast the future seasonal rainfall under climate change scenario . The large scale atmospheric variables (LSAVs) called as predictors for seasonal rainfall (i.e. F eb - Mar - A pr , M ay - J un - J ul , A ug - S ep - O ct and N ov - D ec - J an ) w ere identified using correlation maps. Using correlation analysis and ANN model ing , the most significant variables and their best combinations were determined . It was found best ANN model for each season to forecast the rainfall according to the performan ce indices. Quantile mapping bias correction applied for identified predictors (1971 - 2100) from tw o General Circulation Models (GC Ms); C SIRO Mk3.6 and MPI - ESM - MR, for RCP4.5 scenario. Then forecast the seasonal rainfall till 2100 using developed models. Th e extreme events; wet and dry for next 30 year period (till 2040) was defined according to thresholds at 90 th and 10 th percentile of observed climatology (1971 - 2000). Sea level pressure, surface air temperature, zonal wind, meridian wind, relative humidit y, sea surface temperature and precipitable wa ter were the identified predictors for different lead times (4 - 15 months) with 85 - 95% confident level. The forecast lead time was one month for MJJ a nd FMA seasonal rainfall whereas two month for A SO and NDJ se asonal rainfall. Accordingly, MJJ and NDJ rainfall over Ping basin is forecasted to decrease by 0.5mm and 0.14mm per year respectively from 2011 to 2100. On the other hand ASO and NDJ rainfall increases by 0.45mm and 0.03mm annually during the same period. The Gaussian distribution results shows an increase of extreme wet (above 90 th percentile) by 20% compared to observed period (10%) during monsoon season (ASO) under RCP4.5 scenario from MPI - ESM - MR model. Similarly, approximately 31% and 1 4% probability o f occurrences was forecasted for extreme dry condition during NDJ and FMA seasons respectively under RCP4.5 from CSIRO Mk3.6 model. Results obtained from the study provide understanding of significant variables on rainfall occurrence and fut ure rainfall variations over the Ping river basin. Effective planning and management of resources and implementation strategy can be developed accordingly.

A thesis submitted in partial fulfillment of the requirements for the Degree of Mas ter of Engineering in Water Engineering & Management

Thesis (M. Eng.) - Asian Institute of Technology, 2013