Predicting the Summer Rains of the Southwest United States Using Machine Learning

The summer rains in the Southwest United States are a vital source of water for the people. These seasonal rains are not well forecasted and with machine learning comes the question if it can help predict these summer rains. Machine learning is becoming a bigger topic for various fields. However, that does not mean it can always help to predict outcomes. Machine learning is used here to statistically predict the summer rains in the Southwest. In this paper, single variable linear regression and statistical trees are used to predict the summer rains. There is still more research to be done on this; however, linear regression and statistical trees still give some insight into the predictive ability. From the linear regression and statistical trees for two locations in Arizona and two locations in New Mexico, the east Arizona location is the most predictable. Statistical trees proved to have better performance compared to the single variable linear regressions. Despite the slight differences, both tools agree that the water vapor mixing ratio at 850 mb (Q850) is, of the 14 tested variables, the most important variable to consider when predicting the rain. Moreover, both tools agree that the atmospheric variables are significantly more relevant than the oceanic variables that are considered.