Characterization and predictability of droughts in the Midwest
Abstract
The characteristics and regionalization of droughts, especially in the Midwest, have not been investigated in depth. Drought regions for the Midwestern portion of the United States have also not been clearly defined. Principal component (PC) analysis has been used in the present study to investigate the drought data from Illinois, Indiana and Ohio. The stations surrounding these states were also included in the analysis. Monthly data used are Palmer's Drought Severity Index (PDSI), precipitation and temperature. The causal variables are precipitation and temperature and the effect variable is PDSI. The spatial variability of these data are analyzed and characterized first. The spatial-temporal relationship of data is studied by using unrotated and rotated principal components because rotated PCs can give the best representation of the spatial data in a region. Using the spatial principal components, droughts are classified into climatologically homogeneous regions. Homogeneous regions derived by using principal components of precipitation, temperature and PDSI are compared. The temporal characteristics are investigated by using spectral analysis and tests of significance to determine significant periodicities in the cause and effect variables. Principal components are used to estimate the predictability of droughts and to determine the causal connections between PDSI and precipitation. Finally, the physical relationships between rainfall and droughts are investigated to obtain a better understanding of the spatial causal relationship between precipitation and droughts. By using principal component analysis, the data are compressed by as much as 89%, while explaining at least 80% of the variance. The spatial rotated principal components, compared to unrotated components, accurately recover the input patterns of the original data and give more intuitively meaningful results. The spatial-temporal relationship between precipitation and PDSI has a much greater significance than the relationship between temperature and PDSI. Classification of droughts into homogeneous regions gives different results depending on the method used. It is possible to predict monthly PDSI more accurately by using past monthly PDSI than by using precipitation. The relationship between regional precipitation and drought is clearly brought out by the principal datum and estimated patterns.
Degree
Ph.D.
Advisors
Rao, Purdue University.
Subject Area
Hydrology|Civil engineering|Atmosphere
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.