Linear mixed models with spatial correlation for an analysis of rice yields in each province of Thailand

Abstract

The objectives of this research are to propose a linear mixed model (LMM) with spatial correlation for an analysis of rice yields in Thailand, to estimate the rice yield in each month of all provinces in Thailand, to investigate factors influencing on the rice yields, and to construct the maps of rice yields. A linear mixed model (LMM) in which the spatial effects follow the conditional autoregressive model (CAR) is used. The estimated rice yields are used to construct the rice yield maps. The dependent variables are the rice yield in each month of each province. The data are secondary data at a provincial level. The factors considered are rainfall, averaged temperatures, and regions. The results show that the factors influencing on the rice yields are rainfall, averaged temperature, and region. The amount of regional effects, ordering from largest to smallest values, are northestern region, southern region, western region, central region, eastern region, and northern region, respectively. The top ten provinces and months with high yields, ranking from largest to smallest values, are Ubon Ratchathani in November (1,328,000 ton), Surin in Novermber (1,070,000 ton), Si Sa Ket in Novermber (1,023,000 ton), Buri Ram in November (939,400 ton), Nakhon Ratchasima in November (907,500 ton), Roi Et in Novermber (852,800 ton), Chiang Rai in Novermber (769,800 ton) Khon Kaen in Novermber (692,800 ton) Maha Sarakham in Novermber (676,200 ton), Udon Thani in Novermber (620,600 ton), respectively. The rice yeilds maps are easy for readers to identify which areas have high or low yields. They are a useful tool for planning, decision making and implementing plans to help rice famers to increase their rice production.