Projection the long-term ungauged rainfall using integrated Statistical Downscaling Model and Geographic Information System (SDSM-GIS) model

N.N.A Tukimat,N.A Ahmad Syukri,M.A Malek

doi:10.1016/j.heliyon.2019.e02456

Abstract

An accuracy in the hydrological modelling will be affected when having limited data sources especially at ungauged areas. Due to this matter, it will not receiving any significant attention especially on the potential hydrologic extremes. Thus, the objective was to analyse the accuracy of the long-term projected rainfall at ungauged rainfall station using integrated Statistical Downscaling Model and Geographic Information System (SDSM-GIS) model. The SDSM was used as a climate agent to predict the changes of the climate trend in Δ2030s by gauged and ungauged stations. There were five predictors set have been selected to form the local climate at the region which provided by NCEP (validated) and CanESM2-RCP4.5 (projected). According to the statistical analyses, the SDSM was controlled to produce reliable validated results with lesser %MAE (<23%) and higher R. The projected rainfall was suspected to decrease 14% in Δ2030s. All the RCPs agreed the long term rainfall pattern was consistent to the historical with lower annual rainfall intensity. The RCP8.5 shows the least rainfall changes. These findings then used to compare the accuracy of monthly rainfall at control station (Stn 2). The GIS-Kriging method being as an interpolation agent was successfully to produce similar rainfall trend with the control station. The accuracy was estimated to reach 84%. Comparing between ungauged and gauged stations, the small %MAE in the projected monthly results between gauged and ungauged stations as a proved the integrated SDSM-GIS model can producing a reliable long-term rainfall generation at ungauged station.

Highlights

Uncontrolled emissions dispersed to the atmosphere as a main factor in increasing the greenhouse gasses (GHGs) level and contribute to the climate changes crisis
In the year of 2016, the majority part of east coast of Peninsular Malaysia had been flooded after couple weeks of continuous rain started on Nov, 2014
2018 proved the hybrid methods which combining the linear stochastic models with extreme learning machine (ELM) methods has good performance in improving the accuracy of rainfall projection

Summary

Introduction

Uncontrolled emissions dispersed to the atmosphere as a main factor in increasing the greenhouse gasses (GHGs) level and contribute to the climate changes crisis. Nor Aizam and Peter (2011) stated the most river in Malaysia were frequently affected by flood event due to the unpredictable rainfall variability. In the year of 2016, the majority part of east coast of Peninsular Malaysia had been flooded after couple weeks of continuous rain started on Nov, 2014. Most of the areas which near to the river were flooded because of abnormal rainfall intensity which 60% higher than the average monthly rainfall during normal condition. The downstream rivers became overflow caused by heavy rainfall at an upstream areas

Methods

Results

Conclusion