Abstract
The intraseasonal evolution of the North American monsoon in southeast Arizona during the 1980–93 period is investigated using a neural network–based nonlinear classification technique known as the self-organizing map (SOM). The goal of the SOM algorithm is to discover meaningful low-dimensional structures hidden in the high-dimensional observations. Various daily lagged atmospheric fields (850-hPa meridional winds, 700-hPa specific humidity, 500-hPa geopotential heights, and 850–500-hPa thickness) for the summer season (June–July–August–September) of the 1980–93 period are used in the nonlinear classification of monsoon modes. Special emphasis is given to the wettest monsoon modes. The neural network classification successfully captures the multidimensional interaction of the atmospheric variables during the monsoon evolution, and shows monsoon “bursts” and “breaks” in a given year. Spectral analysis of daily summer rainfall in the study area reveals a significant peak in the 12–18-day band; a secondary and significant peak is also found near 40 days. Thus, monsoon bursts and breaks seem to be modulated by low-frequency variability.
The SOM nonlinear classification shows that the mature phase of the monsoon is associated with two distinct intraseasonal (>10 days) wet monsoon modes. The signature of the wettest monsoon mode is a zonal three-cell anomalous midtropospheric height pattern over the North Pacific–North American sector, suggesting a large-scale dynamical mechanism, possibly linked to sea surface temperature (SST) anomalies in the North Pacific. This zonal mode, which is most frequent in July and August, is characterized by an enhanced and northeastward-displaced monsoon ridge, large amounts of midtropospheric moisture over the study area, and an out of phase relationship between precipitation in the southwest United States and precipitation in the Great Plains. The zonal mode has been recognized in longer datasets and it is the most typical mode that characterizes the mature phase of the monsoon in the southwest United States. In contrast, the second wettest intraseasonal monsoon mode does not show a monsoon ridge, but a meridional three-cell anomalous midtropospheric height pattern along the west coast of North America, weak height anomalies over the rest of North America, and large amounts of moisture over the study area. Importantly, this meridional mode, which is most frequent in August and September, does not show out of phase links to Great Plains precipitation. The meridional wet mode also shows an anomalous low-level cyclonic circulation off the west coast of central-south Mexico suggesting that convective activity off the southern Mexican coast—possibly associated with the intertropical convergence zone—may cross over the Isthmus of Tehuantepec toward the Gulf of Mexico and the southern United States. This would explain the weak link between precipitation in the Southwest and precipitation in the Great Plains during August and September of the 1980–93 period.
At more regional scales, the zonal wet mode is also characterized by a latitudinal gradient of SST anomalies between Baja California and southern Mexico and reversed low-level flow over the Gulf of California. Looking at extreme wet monsoons outside of the study period (e.g., 1955, 1959, 1999) it is shown that the positive SST anomaly pattern along the Pacific coast of Baja California, which characterized wet events during 1980–93, can be completely reversed during other extreme wet events. These contrasting results suggest that interaction between local and remote forcing mechanisms over the study area are complex during extreme events and needs further investigation.
Corresponding author address: Dr. Tereza Cavazos, Departamento de Oceanografía Física, CICESE, P.O. Box 434844, San Diego, CA 92143. Email: tcavazos@cicese.mx
