Doctor of Philosophy (PhD)
Geography and Anthropology
Over the last century, scientists have raised questions about climate change and one of the more important variables that may undergo change is precipitation. Precipitation can affect many sectors including agriculture, socio-economic activities and hazard management. This dissertation addresses the temporal aspect of precipitation using 167 first order stations in the contiguous United States from 1951-2015. The three objectives of this dissertation are to perform (1) an annual analysis of the frequency of rain days in the United States and changing magnitudes of daily rainfall, (2) seasonal rain day frequency in the United States, and (3) use an artificial neural network (ANN) to address predictive accuracy between teleconnections and precipitation days throughout the United States. Similar methods will be used throughout these studies including mainly non-parametric trend analyses. Methods include Mann Kendall trend testing, sliding window correlations analyses, autoregressive forecasting models, and an artificial neural network. The annual Mann-Kendal test found that the majority of the Northeast and Midwestern states show upward trends in precipitation days, while negative trends are located in the Southeast and in clusters throughout the Northwest. The seasonal Mann-Kendal test found clusters of positive and negative trends. In the winter, the northwest and northern Rockies had significant negative trends and the upper Midwest and central Ohio valley had significant positive trends. Spring and summer also had many positive trends within regions. There were no significant negative trends throughout the United States in Fall. The annual sliding window correlation analysis revealed that the Northeastern United States had more significant changes during the earlier decades whereas the center part of the country had more significant changes in the later decades. The seasonal sliding window correlation showed decadal trends at a regional scale throughout the 1990s and 2000s especially. The autoregressive forecast model showed that precipitation days are expected to increase for most of the United States into the future. The ANN predicted approximately 66% of the month/region combinations above the no information rate of 51%. Even through teleconnections are beneficial for precipitation day prediction throughout the United States, other climate variables may help increase accuracy in predictions.
Bartels, Rudy J., "A Climatology of Precipitation Days Throughout the Continguous United States" (2018). LSU Doctoral Dissertations. 4587.