ABSTRACT |
|
Interannual variations and intra-annual variation of regional-scale and
global-scale climate variables are characterized by principal component
analysis (PCA). Climate consistency is detected among the entire United States,
the North Central states, and the Ozark Highlands (OZHI). The regional-scale
modes of the OZHI climate are classified as the predictands of the statistical
climate model. Characteristic pattern and time coefficients are examined in
global-scale climate variables as the predictor of the models.
Relationships between regional-scale and global-scale climate variables are identified by the month lead cross-correlation analysis. The OZHI temperature in January and July are highly correlated to lead time global-scale climate variables in the tropical and subtropical Pacific and Atlantic and those of lead time in the eastern subtropical and midlatitude Pacific, respectively. The OZHI precipitation levels in January and May are highly correlated to lead time global-scale climate variables in the western tropical Pacific and in the western tropical Indian, and South Pacific Convergence Zone (SPCZ), respectively. From multiple linear regression (MLR) and principal components regression (PCR) analysis, the predictability of OZHI regional temperature and precipitation are discussed with model diagnostics and measurements of forecasting performance. This study suggests that PCR can clearly eliminate the multicollinearity among predictors. For the purpose of building the statistical climate model, the sensitivities of the main predictors (i.e., temperature and precipitation) are investigated, and relatively long-memory and short-memory predictors are uncovered. The sea surface temperatures have a relatively long-memory effect. |