56 / 2016-03-09 16:26:17

Spatial Interpolation of Weather Generator Parameters in Peninsular Malaysia using Locally Weighted Regression

spatial interpolation,hourly rainfall,weather generator,locally weighted regression,rainfall modelling

Having insufficient climate data is a critical problem in hydrological studies. Spatial interpolation methods were widely used to overcome the missing data problem. Previously, Advanced Weather Generator (AWE-GEN) parameters are only fitted for the specific locations at which meteorological observations exist. However, the spatial density of such locations may not be adequate for some hydrological purposes which require weather time series at remote locations. Thus, the spatial variability of AWE-GEN parameters are examined to overcome the problem of inadequate resolution of the observing stations. The rainfall and temperature parameters estimated in AWE-GEN are interpolated using Locally Weighted Regression (LWR) method. This model was validated by comparing the observed and the interpolated parameters produced monthly. The monthly statistical properties produced at one hour period is also compared. Results show that all rainfall parameters are well produced except for θ and μ_c. The monthly statistics at different aggregation periods (i.e. 1, 24 and 48 hours) are tested. The mean and variance are reproduced very closely to the observed mean and variance with the exception of the month of November. The lag-1 autocorrelation and the skewness seem to be well reproduced at 1 hour aggregation period except for January and June where the skewness is quite high. The simulated shows higher probability of no rainfall and lower transition probability from a wet spells at this period. Generally, LWR did not produce commendable result on rainfall simulation for ungauged sites in Peninsular Malaysia.