Atmospheric moisture budget and its regulation of the summer precipitation variability over the Southeastern United States

Abstract

Atmospheric moisture budget and its regulation of the summer (June–July–August) precipitation over the Southeastern United State (SE U.S.) were examined during 1948–2007 using PRECipitation REConstruction over Land and multiple reanalysis datasets. The analysis shows that the interannual variation of SE U.S. summer precipitation can be largely explained by the leading Empirical Orthogonal Function mode showing a spatially homogenous sub-continental scale pattern. Consequently, areal-averaged precipitation was investigated to focus on the large-scale rainfall changes over the SE U.S. The wavelet analysis identifies an increased 2–4 year power spectrum in recent 30 years (1978–2007), suggesting an intensification of the interannual variability. Analysis of the atmospheric moisture budget indicates that the increase in precipitation variability is mainly caused by moisture transport, which exhibits a similar increase in the 2–4 year power spectrum for the same period. Moisture transport, in turn, is largely controlled by the seasonal mean component rather than the subseasonal-scale eddies. Furthermore, our results indicate that dynamic processes (atmospheric circulation) are more important than thermodynamic processes (specific humidity) in regulating the interannual variation of moisture transport. Specifically, the North Atlantic Subtropical High western ridge position is found to be a primary regulator, with the ridge in the northwest (southwest) corresponding to anomalous moisture divergence (convergence) over the SE U.S. Changes in moisture transport consistent with the increased frequency of these two ridge types in recent 30 years favor the intensification of summer precipitation variability.

Appendix: Spatial heterogeneity of SE U.S. summer precipitation

To assess the heterogeneity of SE U.S. summer precipitation, cluster analysis is applied to the 60-year summer precipitation records at each SE U.S. grid point. The clustering algorithm used in this study is K-means clustering method (Kalkstein et al. 1987). The algorithm assigns the SE U.S. grid point into n clusters so that the variance within each cluster is minimized and the variance between each cluster is maximized.

The within cluster variance is calculated as the square distance between grid point precipitation to their corresponding cluster centroid:

$$ \frac{1}{M - 1}\sum\limits_{i = 1}^{M} {\left( {P_{i} - P_{c} } \right)^{2} } , $$

(7)

where M is the number of grid point in the cluster.

The between cluster variance is calculated as the square distance between cluster centroids and the domain average over the SE U.S.:

$$ \frac{1}{n - 1}\sum\limits_{i = 1}^{n} {\left( {P_{c\_i} - \left[ P \right]} \right)^{2} } , $$

(8)

Where n is the number of clusters identified over the SE U.S. domain; and $\left[ P \right]$ is the areal-averaged summer precipitation over the SE U.S.

In the K-means algorithm, the determination of the number of clusters is subjective. In the analysis, we determine the number of clusters by tracing the within and between cluster variances, respectively, in response to the number of clusters assigned in K-means algorithm. The algorithm is repeated 30 times.

From our analysis, the within-cluster (between-cluster) variance rapidly decreases (increases) as the number of clusters increases; however, they saturate as the number of SE U.S. precipitation clusters increases to six. Thus, the six clusters are sufficient to generally reflect the spatial heterogeneity of SE U.S. summer precipitation. The six identified SE U.S. summer precipitation clusters are shown in Fig. 11.

Fig. 11

SE U.S. summer precipitation clusters identified using K-means algorithm (“Appendix”)

From our analysis, the interannual variation of SE U.S. summer precipitation shows certain spatial heterogeneity (inter-cluster spread), in addition to the dominance of the large-scale domain-wide signal (EOF mode 1, Fig. 1). However, the two major features concerning the temporal variation of SE U.S. summer precipitation are in general agreement among clusters. First, the summer precipitation experiences increased variability during the recent 30 years among all six clusters. Averaged over the six clusters, the standard deviation of the summer precipitation variability increases by 0.15 mm day⁻² in recent 30 years (Fig. 12), although the increase is relatively small in cluster 5 (northern Florida and the coastal regions of Georgia, Fig. 12). Second, moisture divergence largely explains the interannual variation of summer precipitation in each of the six clusters. Summer precipitation shows close linear relationship with the moisture divergence term ($\nabla \cdot \overline{{\int_{0}^{{p_{s} }} {q\vec{V}dp} }}$) (Fig. 13), and the correlation coefficient passes the $\alpha = 0.01$ significance level.

Fig. 12

The standard deviation of summer precipitation within each cluster as shown in Fig. 11. The blue slashed bars represent the 1948–1977 period, and the red slashed bars represent the 1978–2007 period

Fig. 13

The JJA mean moisture divergence anomaly (total) versus precipitation anomaly within each SE U.S. clusters (blue dots): a–f. The red solid lines are the best least squares fitting lines and the red dashed lines are the $y = - x$ line

The cluster analysis suggests that interannual variation of summer precipitation over the six SE U.S. clusters shares similar characteristics in terms of the recent precipitation variability change and the dominant moisture budget processes. That is: (1) the summer precipitation variability has increased domain wide in recent 30 years (Fig. 12); (2) moisture transport is the predominant process for the interannual variation of SE U.S. summer precipitation (Fig. 13). See Figs. 11, 12 and 13.