Hierarchical linear model and its research on hierarchical characteristics of rainfall

Hierarchical linear model mainly analyze the effects in different levels and the variance changes of random effects to explore more information and results. In this paper, we do linear regression of monthly rainfall in Beijing area between 1996-2009 with hierarchical linear model. In this way it not only considers the factors affecting the rainfall, but also takes into account the cyclical and seasonal differences in rainfall. By means of a two-layer model, the accuracy of the rainfall regression is improved a lot, and the differences in monthly rainfall are made a good explanation. The results show that, season variability accounts for about 55% of the monthly rainfall variability; Variance of monthly rainfall decreased by 46% when adding the four factors (temperature, sunlight, humidity and air pressure) to level 1; Once fixed the season effects, only a little variance of slope 3(humidity and rainfall) and slope 4(air pressure and rainfall) are not explained.