|
Seasonal rainfall forecast for cropping pattern planning using a modified k-nearestneighbor model |
|---|---|
| รหัสดีโอไอ | |
| Creator | 1. Uruya Weesakul 2. Nkrintra Singhratta 3. Phailin Yodpongpiput |
| Title | Seasonal rainfall forecast for cropping pattern planning using a modified k-nearestneighbor model |
| Publisher | Faculty of Engineering, Khon Kaen University |
| Publication Year | 2559 |
| Journal Title | KKU Engineering Journal |
| Journal Vol. | 43 |
| Journal No. | 3 |
| Page no. | 156-161 |
| Keyword | K-Nearest neighbor,Large-Scale atmospheric variables |
| ISSN | 0125-8273 |
| Abstract | Drought phenomena recently occurred in Thailand caused severe damage to agricultural sector particularly in Central, Northern and Northeastern regions. A reliable seasonal rainfall forecast model is needed to provide useful information for effective cropping pattern planning and water resource management. The study aims to develop a seasonal rainfall forecast model using a stochastic model K-Nearest Neighbor technique to forecast seasonal rainfall with a leading time of one year ahead. A Mun river basin, located in the Northeastern region, was selected as a case study. Monthly rainfall data from 152 rainfall stations distributed over the river basin, were collected over a period of 37 years during 1975 to 2011 Analysis of correlation between Large Scale Atmosphere Variables (LAV) around the study basin and seasonal rainfall over the river basin reveals that Surface Air Temperature (SAT), Sea Level Pressure (SLP), Surface Zonal Wind (U) and Surface Meridional Wind (V) over China Sea and Pacific Ocean have influence on variation of seasonal rainfall over the basin. These 4 LAV variables are used as predictors in the modified k-nn model to forecast seasonal rainfall. The likelihood Skill Score (LLH) was adopted as a technique to evaluate model performance. Test of performance of model, using seasonal rainfall for period of 37 years during 1975 to 2011, reveals that the model is able to predict seasonal rainfall with the reliability of around 60%, providing sufficient information for appropriate cropping pattern planning in the area. |
