Abstract
Dominant spatiotemporal patterns of precipitation, modeled soil moisture, and vegetation are determined in North America within the recent observational record (late twentieth century onward). These data are from a gridded U.S.–Mexico precipitation product, retrospective long-term integrations of two land surface models, and satellite-derived vegetation greenness. The analysis procedure uses three statistical techniques. First, all the variables are normalized according to the standardized precipitation index procedure. Second, dominant patterns of spatiotemporal variability are determined using multitaper method–singular value decomposition for interannual and longer time scales. The dominant spatiotemporal patterns of precipitation generally conform to known and distinct Pacific SST forcing in the cool and warm seasons. Two specific time scales in precipitation at 9 and 6–7 yr correspond to significant variability in soil moisture and vegetation, respectively. The 9-yr signal is related to precipitation in late fall to early winter, whereas the 6–7-yr signal is related to earlysummer precipitation. Canonical correlation analysis is finally used to confirm that strong covariability between land surface variables and precipitation exists at these specific times of the year. Both signals are strongest in the central and western United States and are consistent with prior global modeling and paleoclimate studies that have investigated drought in North America.
Corresponding author address: Dr. Christopher L. Castro, Department of Atmospheric Sciences, The University of Arizona, Physics and Atmospheric Sciences Bldg., Rm. 520, 1118 East Fourth Street, Tucson, AZ 85721-0081. Email: castro@atmo.arizona.edu
