Author(s): Sutat Weesakul; Nkrintra Singhrattna; Panuwat Pinthong; Supote Thammasithirong
Linked Author(s):
Keywords: Climate change; Rainfall forecasting; Real time reservoir operation
Abstract: At present the change of future climate causes variability on hydrological condition. Improved capability of reservoir for effective real time operation is required in order to cope with such changes, especially when the reservoir has a comparable size with the inflow. Study area is Ubolratana river basin in the northeast of Thailand having catchment area12,089 square kilometer. Reservoir capacity is 2,263 million cubic meter. The objective of study is to propose the real time reservoir system with meteo-hydrological forecasting using large scale climate variables. There are 3 parts for the system; the first one is hydrological forecasting using large scale atmospheric variables for seasonal rainfall. The sea surface temperature, sea level pressure, surface zonal and meridian winds over different regions which are Pacific and Indian Oceans are used to statistically related with historical seasonal rainfall varying from 4 to 15 months prior to the start of the season. Based on 95%confidence levels, the identified predictors from 1980 to 2007 are obtained from a general circulation model (GCM) called GFDL-R30. The statistical model can forecast and determine future climate on seasonal rainfall. The multisite daily rainfall generator based on a Markov chain and Monte Carlo approach is used to generate daily rainfall. The second model is modeling of rainfall-runoff process which NAM is used in the present study. The last part is reservoir operation model. Genetic-Neurofuzzy model has been developed and used to identify membership function. The input variables are inflow to dam, water demand, current storage and local downstream flow condition. The additional accumulated ten days inflow is introduced as a new variable which can improved model performance which is due to the characteristic of river basin and reservoir size. Reliability parameter for both flood and water demand function is selected as optimized variable. The model show improved flood reliability from 90%to 98%.
Year: 2014