Hindcasting 2000 Years of Climate Data Across States in the South Central and Southwestern US

The effective planning of water resources requires accurate information about climate variability. The short time period for which instrument records exist, however, limits our knowledge of long-term temperature and precipitation variability. To overcome this limitation, a simultaneous reconstruction of annual temperature and precipitation values was conducted for eight states across a gradient of modern climate zones: Arizona (Desert), California (Mediterranean), Colorado (Semiarid to Alpine), Kansas (Semiarid to Humid Continental), Nevada (Semiarid to Arid), New Mexico (Semiarid), Texas (Semiarid to Humid Subtropical), and Utah (Semiarid). The reconstruction involved imputation of values based on the self-organized nonlinear data vector relations among 2,000 years (0-2000 AD) of reconstructed warm-season (June-August) Palmer drought severity index data (PDSI; [11]), and 114 years (1895-2009) of annual state precipitation and temperature data [13]. The reconstruction was verified against independent precipitation and temperature data for the years: 1896, 1900, 1911, 1919, 1923, 135, 1940, 1952, 1960, 1966, 1968 (La Niña), 1986, 1993, 1998 (La Niña), and 2005 (El Niño) using split- and cross-validation (leave one out) approaches. The Spearman Rho correlation among observed and imputed values was greater than 95% with a p-value of 0.001.

Quantile modeling [4] of the reconstructed temperature change data (annual temperature minus 2,000 year median) revealed that the long-term global climate was interrupted by short-term changes. For example, the so-called Medieval Warm Period (~900 to ~1250) and Little Ice Age (~1400 to ~1850) were two changes over the last two millennia that appeared independently in the northern hemisphere [15], and our Arizona, Colorado but not Kansas (Fig. 3.1) reconstructions. The muted peaks and increased uncertainty (0.05 and 0.95 quantile models) in these reconstructions corroborates previous findings that PDSI data are best suited to spatial rather than temporal reproduction of peaks [3]. Regionally, our reconstructions revealed 2,000 year trends with increasing temperature for Arizona, constant temperatures for Colorado, and decreasing temperatures for Kansas. These findings are attributed to a strong ENSO teleconnection with Arizona, mixed ENSO signals in Colorado because it is a region between El Niño and La Niña latitudes, and Kansas

Fig. 3.1 Temperature change reconstruction: (a) Northern hemisphere (After Loehle [15]); (b) Arizona (desert); (c) Colorado (alpine to semiarid) and (d) Kansas (semiarid to humid). Colored lines are quantile regression models with b-spline smoothing (upper 0.05, middle 0.5, and lower 0.95)

because it is in continental interior and influenced by the Gulf of Mexico. These findings suggest that the natural multicentennial and regional climate variability may be larger than commonly believed.

0 0

Post a comment