| Abstract | Hydroclimates like precipitation and streamflow are related to large-scale atmospheric variables via oceanic-atmospheric circulations. Unu sual patterns of circulation are connected to changes in climatic conditions due to factors like population and technology growth, urbanization and economic development. El N iño and La Niña, defined as anomalous sea surface temperatures over the tropica l Pacific Ocean, interrupt the Walker circulation and cause variability in hydroclimates and anomalous weather events. The case study (i.e. the Ping River Basin which is a sub-bas in of the Upper Chao Phraya River Basin) is located in northern Thailand. The Ping Ri ver Basin covers an area of 33,899 km 2 . The climate is classified as tropical monsoon with all average monthly temperatures greater than 18 ° C, and the highest temperature occurs in a period p rior to the monsoon season. Moreover, precipitation being less than 61 mm per month is found in one or more months. The Ping River Basin experiences dry season rainfall (from November to April) which is inversely related to air temperature in th e summer season (i.e. March-April-May or MAM). From 1951 to 2007, an increasing trend in dry season rainfall (by 16.3 mm over 57 years) is consistent with a decreasing trend of MAM temperature (drops by 0.6°C over 57 years). Furthermore, a higher MAM temperature in fluences the land-sea temperature gradient and strengthens the monsoon. The pre-monso on season rainfall (i.e. May-June- July or MJJ) is inversely correlated to MAM tempera ture, whereas the monsoon season rainfall (i.e. August-September-October or ASO) is positively related to MAM temperature. In El Niño years, the MJJ and ASO rain falls tend to decrease and vice versa in La Niña years Using correlation maps, seasonal rainfall (i.e. MJJ , ASO, NDJ and FMA) of the study basin can be statistically related to large-scale a tmospheric variables (sea surface temperature, sea level pressure, surface zonal and meridian winds) at long lead times, varying from 4 to 15 months prior to the start of t he season. Atmospheric predictors are identified over different regions (such as the Paci fic and Indian Oceans) based on significant relationships with rainfall at 95% conf idence levels. The gridded monthly data of identified predictors from 1961 to 2100 is obtai ned from a general circulation model (GCM) called GFDL-R30 and used in a statistical mod el to forecast and determine the effects of future climate on seasonal rainfall of t he study basin. A modified k-nearest neighbor (k-nn) model is used to downscale the rainfall of the study basin from large-scale atmospheric variables. The m odified k-nn model is a nonparametric approach, which locally fits a regression between d ependent (e.g. rainfall) and independent variables (e.g. atmospheric predictors) using a sma ll set of neighbors ( k ) at any given point. k and the order of polynomial ( p ) are selected using a generalized cross validation (GCV) method. In terms of the effects of future climate u nder two scenarios (A2 and B2), the 2011-2100 MJJ and ASO rainfall of the Ping River Ba sin is predicted to decrease by 0.11- 6.16 mm per year. Increasing trends of 0.02-5.91 mm per year are associated with the 2011-2100 dry season rainfall (i.e. NDJ and FMA). F urthermore, the monsoon season rainfall will have more chances of being dry and le ss chances of being wet. In contrast, future climate will affect more chances of wetness and less chances of dryness for the dry season rainfall. The wet season will tend to shift by two seasons, from ASO to FMA, under A2 and by one season, from ASO to NDJ, under B2. To compare two algorithms of rainfall-runoff models , the SIMHYD and HEC-HMS models have been studied. Both models are calibrate d from April 1999 to March 2003 and validated from April 2003 to March 2007. Four effic iency indexes (the deviation of volume (D v ), correlation coefficient ( r ), normalized root mean square error (NRMSE) and the Nash-Sutcliffe efficient index ( E )) are used to evaluate the model performance. The SIMHYD model shows poor performance, due to the hom ogeneity of basin characteristics, in capturing average monthly streamflow at the stat ions that cover a large drainage area. Comparing the performance of these models at six ga uging stations, the HEC-HMS model performs better than the SIMHYD model in capturing average monthly streamflow. Moreover, the HEC-HMS model can capture low flow be tter than the SIMHYD model. Although its performance is not consistent at all s tations, more efficiency in high flow simulation is associated with the HEC-HMS model as it shows less NRMSE and greater E . Therefore, the HEC-HMS model has been selected to s imulate the 2011-2100 daily streamflow using rainfall ensembles obtained from t he multisite daily rainfall generator. The effects of future climate under both scenarios present decreasing average discharges in the dry and wet seasons. The shift in peak discharg e from mid-September to the end of September or the beginning of October is expected t o be observed. With the exception of Station P75, P67 and 061302, the dry spells will be shorter in the future compared to historical records. Less severity of shortage durin g the dry spell is also predicted for all stations except P67 and 061302. Under A2, with the exception of Station P24A, 061302 and P14, the wet spells will be shorter. Under Scen ario B2, wet spells will be shorter at all stations. The intensity of abundance is less than t hat seen in historical records. Anomalous low flow in the wet season and anomalous high flow in the dry season have also been examined using thresholds of the observed Q 90 in the wet season (Q 90,wet ) and the observed Q 10 in the dry season (Q 10,dry ). Due to future climate alterations, the magnitude s of simulated Q 90,wet and Q 10,dry will likely be lower than those of observations. W ith the exception of Station P75 and P21, dry spells in the wet season will be shortened. Anomalous low flow in the wet season during 2011-21 00 is less severe than historical records. In terms of anomalous high flow in the dry season, a shorter duration of wet spells with lesser intensity of abundance will be found at four stations under A2 and at three stations under B2. |