A precipitation dipole in eastern North America



[1] We have identified a dipole in annual precipitation across eastern North America (ENA) east of 100°W between 30°N and 60°N. This dipole appears to create spatially coherent opposing variations in precipitation with a separation of the two regions around 45°N. Annual average precipitation over ENA appears to be stable and unimodal, suggesting that the amount of overall precipitation variability is a small fraction of the mean and is largely determined by similar large scale processes. Analysis of regional average time series at interannual (3–7 year) and decadal (10–16 year) scales indicates that the dipole over the ENA region is most clearly discernible at the decadal scale. Linear regression analysis between global sea surface temperatures (SSTs) and precipitation over the two subregions in ENA suggests that SST variations in several areas of the oceans tend to be associated with opposite precipitation anomalies in the two subregions of ENA.

1. Introduction

[2] Many recent observational data analyses suggest that precipitation has changed significantly during the 20th century in different parts of the globe [e.g., Dai et al., 1997; New et al., 2001]. There is growing evidence that precipitation and streamflow has increased all across the United States over the last several decades [Karl et al., 1995; Karl and Knight, 1998; Groisman et al., 2001; Kunkel et al., 2002; Lins and Slack, 1999; Small et al., 2006; Lettenmaier et al., 1994]. For example, Karl and Knight [1998] reported a 10% increase in annual precipitation across United States between 1910 and 1996. Our recent study [Small et al., 2006] attributed the widely reported increases in the annual 7-day low flow [Lins and Slack, 1999; Douglas et al., 2000] to large increases in precipitation across the eastern United States. The consistency of the reported trends in precipitation and streamflow and their spatial coherence suggests that precipitation over eastern United States has increased during last several decades.

[3] Precipitation and streamflow also appears to be changing across Canada. For example, Zhang et al. [2001] found that annual precipitation totals have decreased in southeastern Canada and increased in northern high latitudes. Gan [1998] analyzed monthly precipitation data for 1949–1989 and reported significant negative trends in winter precipitation. Recently, Dery and Wood [2005] found a 10% decrease in the total annual river discharge to the Arctic and North Atlantic Ocean for the period 1964 to 2003 and suggest that it is consistent with a corresponding decline in precipitation. The bulk of evidence suggests that annual precipitation and streamflow has generally decreased across southeastern Canada over the last 50 years.

[4] It appears that annual precipitation is increasing in the eastern United States and decreasing in southeastern Canada. This apparent dipole like variation in precipitation over eastern North America (ENA) is intriguing. The question of whether the underlying physical processes that tend to produce increases in precipitation over the eastern United States also cause decreases in southeastern Canada has important implications on how, and to what extent, long-term variations in precipitation can be predicted and managed at regional scales. Many previous studies have identified spatially coherent seasonal or longer variations in precipitation that tend to produce relatively dry conditions over some areas and wet conditions over others. For example, dipole-like variations in precipitation have been previously observed over the western United States [Dettinger et al., 1998; Cayan et al., 1998] and northern Eurasia [Fukutomi et al., 2003, 2004]. Isolating a dipole like precipitation variation over ENA, identifying the dominant timescale and spatial extent of such variations and linking the variations to a remote boundary forcing that may produce and maintain such a pattern is important for understanding and predicting regional climatic and hydrologic variations over ENA. The relationship among precipitation anomalies, boundary forcing, atmospheric circulations, and surface climate over eastern North America has been the subject of many observational and modeling studies. Most of these studies, however, have focused on either United States or Canada. This is perhaps one of the first studies to investigate this apparent seesaw pattern of precipitation variations over ENA, defined as the region between eastern United States and southeastern Canada east of Rockies between 30°N to 60°N.

[5] The present study is an attempt to characterize the nature of the apparent dipole-like precipitation variations over ENA region by addressing three related questions: 1) Is there a precipitation dipole over the ENA region?; 2) What is the characteristic time scale for such a dipole?; and (3) Can we identify associational linkages between the dipole in ENA precipitation and sea surface temperature anomalies?

2. Data

[6] We have used the global monthly gridded (2.5° × 2.5°) precipitation data (PREC/L) from 1948–2003 [Chen et al., 2002] for this study. The data set is derived from gauge observations collected in the Global Historical Climatology Network (GHCN), version 2, and the Climate Anomaly Monitoring System (CAMS) data sets. Global monthly gridded (5° × 5°) sea surface temperature (SST) anomalies [Kaplan et al., 1998], provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado, USA, from their web site at http://www.cdc.noaa.gov/, is also used in this analysis. This data set is produced using MOHSST5 version of the GOSTA data set [Kaplan et al., 1998]. Both the data sets were standardized by subtracting monthly means and dividing it by the monthly standard deviation on all the grid points and then taking the annual average to create an anomaly data set for SST and precipitation. The domain of the analysis using SST data extends from 20°S to 70°N. The domain of the precipitation analysis is eastern North America (ENA), east of 100°W between 30°N and 60°N.

3. Results

3.1. Identification of a Precipitation Dipole Over Eastern North America

[7] We performed principal component analysis (PCA) on unfiltered standardized precipitation time series to explore the presence of a dipole over the ENA region. Figure 1a shows the correlation between leading principal component of annual precipitation anomalies over the ENA region and the precipitation time series at each grid point. Figure 1a clearly shows two distinct regions with opposing precipitation variations with a regional separation around 41°N to 46°N. The associated first principal component (PC1) explains about 25% of the variance of the original standardized precipitation time series. We then prepared zonal averages of the annual precipitation anomalies at each of the 12 latitude bands in the domain and performed PCA using a correlation matrix. Figure 1b shows the leading EOF of zonally averaged precipitation anomalies. The EOF changes sign around 46°N with contrasting precipitation regimes on either side of this latitude. The leading PC of this zonally averaged precipitation captures 35% of the variance in the original standardized precipitation time series.

Figure 1.

EOF loadings for: (a) entire ENA region and, (b) zonally averaged precipitation anomaly time series from 1948–2003; and (c) 5-year running mean for two sub-regions in ENA.

[8] To address some of the limitations of PCA, such as domain dependence and sensitivity to sampling error, we analyzed precipitation anomalies averaged over the regions with contrasting precipitation variations. The two regional anomaly time series are defined as the average annual precipitation anomaly between 30°N and 45°N (ENA1 region) and between 46°N and 60°N (ENA2 region). The choice of regions is dictated by the EOF of the zonally averaged precipitation (Figure 1b). Figure 1c shows the 5- year running mean of the precipitation anomaly time series in ENA1 and ENA2 and demonstrates that there is a north-south contrast between the two regions, which became more pronounced after 1980s. Figure 1c suggests that when precipitation anomalies in one region are positive, those in the other region tend to be negative. The correlation between the two time series is low; hence it is an intriguing problem to establish the dominant time scale of the dipole in the region.

[9] To quantify how the overall variation of precipitation over ENA is associated with the north–south precipitation variations and identify the center latitude of the dipole, we estimated three spatial moments from the unfiltered zonally averaged precipitation time series following Dettinger et al. [1998]. The three moments are: (a) domain averaged precipitation for each year, (b) the central latitude of precipitation distribution for each year; and (c) the latitudinal spread of the precipitation distribution. The domain average precipitation (Figure 2a) is 71mm/month with a standard deviation of 4mm/month. The solid black line in Figure 2a represents the 5-year running mean for the entire ENA region. The precipitation distribution appears to be stable over the entire ENA region with no sharp trend in the time series. The domain average precipitation time series is highly correlated with the leading PC of Figure 1a (r = 0.83). Such an associational link provides further support of an existence of a north-south precipitation dipole over ENA.

Figure 2.

(a) Domain average precipitation, (b) central precipitation latitude, and (c) latitudinal spread.

[10] The central latitude of the precipitation dipole is defined as the center of mass in the north-south direction of the spatial precipitation distribution (Figure 2b). The average center of mass is estimated to be at 41.5°N with standard deviation of 0.9 degrees and is consistent with the leading EOF of zonally averaged precipitation anomalies that suggested a change of signs between 41°N to 46°N. Figure 2b along with Figure 1a and Figure 1b further support our choice of ENA1 (30°N–45°N) and ENA2 (46°N–60°N) as the two regions displaying characteristically different precipitation variations. The third moment, the latitudinal spread, is the weighted standard deviation of the zonal distribution of precipitation in each year across the latitudes and provides measure on unimodality of precipitation. The average spread in 56 years is about 8.0 degrees with standard deviation of 0.15 degrees, suggesting that the precipitation is unimodal over the entire region from its central latitude.

3.2. Time Scale of the Dipole

[11] Precipitation is primarily episodic, yet, there are low frequency variations that modulate it on longer time scales. The observation that the precipitation variations are contrasting over such a large region suggests that there are spatial structures in precipitation that recur from episode to episode. To further explore the nature of this apparent north-south dipole in precipitation variations, we created time-latitude plots of zonally averaged annual precipitation anomalies from 1948–2003 (Figure 3). We present zonally averaged unfiltered precipitation anomaly time series (Figure 3a), 3–7 year band-passed anomalies (Figure 3b) and 10–16 year band-passed anomalies (Figure 3c). The 3–7 year timescale typically represents El Nino–Southern Oscillation (ENSO) [Diaz and Markgraf, 1992] while the 10–16 year scale captures variability associated with several other atmospheric teleconnections and sea surface temperature anomalies that exhibit decadal [e.g., Latif and Barnett, 1994; Robertson, 2001] scale variability. The variations of precipitation in the time-latitude plots of the unfiltered and 3–7 year time series (Figures 3a and 3b) do not reveal any distinct spatial contrast between ENA1 and ENA2. Figure 3c, however, suggests that a dipole like variation in precipitation between these two regions may arise from decadal scale (10–16 year) fluctuations. For example, between 1983–1986 ENA1 had positive anomalies (wet years) while ENA2 had negative anomalies (dry years) while during 1989–1992 anomalies reversed signs for these two regions. Similar contrasts are also evident during 1963–1968 (ENA1 negative and ENA2 positive) while 1969–1974 (ENA1 negative and ENA2 positive). These contrasts are clearly visible at the decadal scale time series (Figure 3c). Because a free atmosphere without a coupling to an external boundary forcing cannot sustain perturbations over such long timescales [Borges and Sardeshmukh, 1995], the observed temporal persistence of the precipitation anomalies suggests that a large scale remote forcing may contribute to produce these patterns. Another indication of decadal scale origin of these contrasts come from the strong correlation (−0.54) between ENA1 and ENA2 decadal time series as opposed to much weaker correlation for unfiltered and interannual time series. It is also worthwhile to note that decadal scale time series show 39 years of contrasting variations in precipitation anomalies in ENA1 and ENA2 as opposed to 20 years at the interannual time scale with a correlation of 0.14. The spatial coherence and temporal persistence of a dipole like precipitation variations in ENA1 and ENA2 at the decadal or longer time scales suggest possible influence of remote boundary forcing. The associated 5-year running mean for unfiltered, interannual and decadal time scales is shown in Figures 1c, 3d and 3e, respectively. Figure 1c shows the opposing dipole type precipitation variation in ENA1 and ENA2. On interannual scale, the contrast is weak but the seesaw type precipitation variation is clearly evident at the decadal scale (Figure 3e). Such a pattern of dipolar variations suggests that the north-south contrast is more easily observed at decadal scales.

Figure 3.

Time-latitude diagram of zonal annual precipitation anomalies for 1948–2003: (a) Unfiltered, (b) filtered for 3–7 years (interannual); (c) filtered for 10–16 years (decadal); (d) domain average 5-year running mean over ENA1 & ENA2 region at the interannual scale and (e) similar to Figure 3d but for the decadal scale.

3.3. Relationships of Precipitation Variations With SST Anomalies

[12] To explore the possible influence of remote oceanic forcing from SST variations, we regressed the annual precipitation over ENA1 and ENA2 regions against global sea surface temperature (SST) anomalies from 20°S to 70°N (Figure 4). Figures 4a and 4b suggest that a strong linear relationship exists between precipitation and SST anomalies over several regions of the ocean. We have highlighted four such regions where the linear relationship between precipitation and SST is strong and exhibits correlations with precipitation in ENA1 that are of the opposite sign as those with precipitation in ENA2. Region 1 (North Atlantic) shows a strong negative correlation with ENA1 precipitation but a positive correlation with ENA2 suggesting that the SST anomalies from this region may contribute to the formation of contrasting precipitation variations. The linear analysis also suggests that SST variations in regions 2 and 4 in the North Pacific may also produce strongly contrasting patterns of precipitation in ENA1 and ENA2. SSTs in the Tropical Indian Ocean (region 3) also indicate a moderate opposing contrast in correlation. The presence of spatially coherent regions of correlations in the linear regression maps of ENA1 and ENA2 precipitation with SSTs suggest that spatially coherent SSTs may act as boundary forcing to produce a dipole like precipitation pattern over ENA. We plan to explore the physical mechanisms associated with these regions of spatially coherent SST variability and precipitation dipole over the ENA region in a future study.

Figure 4.

Linear correlation between precipitation anomalies over ENA1 & ENA2 with SST.

4. Concluding Remarks

[13] The objective of the study was to identify and characterize a dipole in annual precipitation in the ENA region, determine the characteristic timescale of the process and identify regions in the global oceans that may be forcing the opposing variations in precipitation. Our results suggest that a seesaw variation of wet and dry conditions exists across eastern North America (ENA) between the eastern United States and southeastern Canada. This apparent seesaw variation over 30°N to 60°N appears to create spatially coherent opposing trends in precipitation at decadal time scales after 1950s. We have found that between latitudes 41°N–46°N, precipitation time series in one region becomes anticorrelated with the other. Our analysis shows that this observed dipole like variations in precipitation is most discernible at the decadal scale (10–16 years). Our results show that the precipitation variation is unimodal over the entire region, which may suggest that precipitation over this region is influenced by the same regions of the ocean. Linear regression analysis with global northern hemisphere SSTs suggests that the precipitation in the ENA region, when separated into two distinct regions, is being influenced by SST anomalies from the same regions in the ocean. We have indicated four such regions where we have observed moderate to strong opposing correlations between SSTs and precipitation over ENA1 and ENA2. The SST anomalies in the North Atlantic, North Pacific and Tropical Indo-Pacific oceans exhibit linear relationships with precipitation over ENA1 that are opposite to those in ENA2. Further investigation is needed to identify physical mechanism(s) that may link SST variations and ENA precipitation for different seasons. Understanding of underlying mechanism of association with stable SST regions would aid in prediction of precipitation over the ENA region.


[14] This work was supported, in part, by a grant from the National Aeronautics and Space Administration (NAG5-11684 and NNG05GM16G) and Tufts School of Engineering.