Role of the strengthened El Niño teleconnection in the May 2015 floods over the southern Great Plains
Abstract
The climate anomalies leading to the May 2015 floods in Texas and Oklahoma were analyzed in the context of El Niño teleconnection in a warmer climate. A developing El Niño tends to increase late‐spring precipitation in the southern Great Plains, and this effect has intensified since 1980. Anthropogenic global warming contributed to the physical processes that caused the persistent precipitation in May 2015: Warming in the tropical Pacific acted to strengthen the teleconnection toward North America, modification of zonal wave 5 circulation that deepened the stationary trough west of Texas, and enhanced Great Plains low‐level southerlies increasing moisture supply from the Gulf of Mexico. Attribution analysis using the Coupled Model Intercomparison Project Phase 5 single‐forcing experiments and the Community Earth System Model Large Ensemble Project indicated a significant increase in the El Niño‐induced precipitation anomalies over Texas and Oklahoma when increases in the anthropogenic greenhouse gases were taken into account.
1 Introduction
In May 2015, an El Niño had developed (Figure 1a) and as a consequence—at least in part—precipitation anomalies in Texas and Oklahoma were off the scale reaching over 200 mm above normal (Figure 1b); this was accompanied by dry anomalies in Kentucky and Tennessee. As the Texas news media echoed “enough rain fell in May to cover the entire state 8 inches deep” (CNN 1 June 2015) and in Houston alone, the flood damage was estimated to “top $45 million” (Associate Press 31 May 2015). While seasonal predictions, issued as early as March, had indicated increased May precipitation for the southern Great Plains (http://www.cpc.ncep.noaa.gov/products/NMME/), the extreme magnitude of the rainfall was not indicated nor anticipated—a challenge that is yet to be realized.
It is known that either the onset or a persistent El Niño can increase spring precipitation in the southern Great Plains at the same time reducing precipitation in the southeast U.S. [e.g., Ropelewski and Halpert, 1986, 1987; S.‐K. Lee et al., 2014]. Previous studies [e.g., Meehl and Teng, 2007; Stevenson et al., 2012; Wang et al., 2014] have found that in a warmer climate, the teleconnection that underlies the El Niño–Southern Oscillation (ENSO) and its associated impact on North America would change in terms of intensification and/or a positional shift of the resultant climate anomalies; this is regardless of the direction of future change in frequency and intensity that ENSO might take. Climate modifications that involve ENSO have forecast implications at both seasonal [Mo, 2010] and decadal [Meehl et al., 2014] timescales, and it is reasonable to question to what extent, if any, the rather extreme May 2015 precipitation event that occurred during an El Niño was induced through a warming climate. It is therefore worthwhile to conduct a climate diagnostics and an attribution analysis of the May 2015 high‐precipitation event that occurred over Texas and Oklahoma.
2 Data
We adopted 17 models from the Coupled Model Intercomparison Project Phase 5's (CMIP5) historical single‐forcing experiments that were driven by (1) natural‐only forcing including solar and volcano (NAT), (2) greenhouse gas (GHG)‐only forcing, and (3) all these historical forcings (ALL) including anthropogenic aerosols [Taylor et al., 2011]. Each experiment produced multiple members initialized from a long‐stable preindustrial (1850) control run up to 2005. Table S1 in the supporting information provides the full name, institute, ensemble size, and spatial resolution of these models. (To date, these are the only models that provide outputs for both GHG and NAT experiments.) In addition, we utilized 30 members produced by the Community Earth System Model version 1 (CESM1) through the Large Ensemble Project (LEP) [Kay et al., 2014]. The CESM1 ensemble simulations covered two periods: 1920–2005 with “ALL” forcing and 2006–2080 with RCP8.5 forcing (the total increase of radiative forcing at 8.5 W/m2/yr since preindustrial values). The CESM1 was used here because it simulates well the ENSO cycle and associated teleconnection in North America [Wang et al., 2014, 2015]. All model outputs were regridded to 2.5° longitude × 2.5° latitude resolution before averaging, to be comparable with the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research Reanalysis (R1) data [Kalnay et al., 1996]. Although the R1 is an older‐generation reanalysis, it is the only data set that covers the presatellite era (before 1979) and yet is still updated operationally. Meanwhile, we utilized four satellite era reanalyses to establish a consensus for the trend analysis, including MERRA [Rienecker et al., 2011], CFSR [Saha et al., 2010], ERA‐Interim [Dee et al., 2011], and the JRA‐25 [Onogi et al., 2007], detailed in Table S2. These reanalyses were averaged with equal weighting to form an ensemble, following Wang et al. [2013]. Other data sets included were the Extended Reconstructed Sea Surface Temperature (v3b) derived from the International Comprehensive Ocean‐Atmosphere Data Set [Smith and Reynolds, 2003] and global precipitation produced by NOAA's Precipitation Reconstruction over Land (PREC/L) and historical observations over ocean (PREC/O); these commence from 1948 [Chen et al., 2002]. The definition of ENSO was determined by the monthly mean Niño3.4 index provided by the Climate Prediction Center (CPC).
3 Results
3.1 El Niño Factor
As shown in Figure 1c, the May 2015 anomaly of the 250 hPa stream function (as a departure from the 1981–2010 mean) depicts a standing trough over the southwest U.S.; this trough directed a series of short waves toward Texas and Oklahoma (not shown). The 850 hPa wind anomalies (vector) depict the intensified southerly flow that conveyed moisture from the Gulf of Mexico into the Great Plains; these upper and lower level features indicate a coupling of baroclinic (frontal) forcing and moisture supply and are of climatological importance in sustaining late‐spring rainfall in the southern Great Plains [Helfand and Schubert, 1995; Higgins et al., 1997; Santanello et al., 2012]. Additionally, as shown in Figure 1c, the anomalous trough over the southwest U.S. was accompanied by a subtropical anticyclone in the eastern Pacific; this formed a teleconnection pattern that resembles the atmospheric response to equatorial eastern Pacific sea surface temperature (SST) forcing [e.g., Mo, 2010]. Daily rainfall data over Texas and Oklahoma (not shown) indicated that the entire month of May, with the exception of 1–3 and 12 May, experienced consecutively large rainfall totals. The coexistence of an El Niño with large May rainfall echoes the observation of S.‐K. Lee et al. [2014] that a positive precipitation anomaly centered over Texas tends to occur in late spring during the onset of an El Niño; this is coincident with CPC's announcement of 2015 El Niño advisory in March (i.e., issued when El Niño conditions are observed and expected to build) so that by the time May came around, the warm‐tongue SST pattern (Figure 1a) had transitioned toward an El Nino onset.
To examine the long‐term change in the relationship between the ENSO teleconnection and precipitation anomalies in the southern U.S., we regressed the May Niño3.4 index with the monthly mean precipitation for two periods: 1948–1980 (Figure 2a) and 1981–2014 (Figure 2b, unit: mm/month/C)). Here all the data within either period were linearly detrended to minimize the effect of any interdecadal trends or cycles. While the east‐west contrast between a wetter Great Plains and drier southeast is apparent in both periods, the regressed precipitation over Texas and Oklahoma exhibits a distinctively stronger signal in the latter time period. Next, the role of increased GHG in the atmosphere in the ENSO‐precipitation relationship was analyzed by conducting the regression analysis for CMIP5's NAT‐only (Figure 2c) and GHG‐only (Figure 2d) experiments, covering the period 1970–2005. Despite differences in the general precipitation pattern as compared to observations (a result of inherent model biases), the GHG‐only forcing produces a significantly stronger precipitation response over Texas and Oklahoma. Moreover, a similar analysis performed using the CESM1 LEP data for the periods of 1940–1980 (Figure 2e) and 2010–2050 (Figure 2f; i.e., that reflects future precipitation outcomes) reveals the same tendency, i.e., stronger precipitation anomalies over Texas and Oklahoma in response to a transformational ENSO signature in a warmer climate. Noteworthy here is that since CESM1 is not included in the 17 CMIP5 models, this result also serves to validate the multimodel analysis. Since the regression is linear, the corollary is valid—that is, the La Niña‐induced precipitation deficit or drought would likewise result over Texas and Oklahoma and equally so, any overtones associated with a strong La Niña as was the case in 2011 [Peterson et al., 2012]. In the future, as can be inferred from Figure 2f, one may anticipate the tendency for an increased ENSO impact on the southern Great Plains precipitation regime.
Recent research [Meehl and Teng, 2007; Stevenson et al., 2012; Wang et al., 2015] indicates that ENSO teleconnection and its regional impact would intensify in response to increasing SST. Under such a premise we regressed the 250 hPa stream function with the Niño3.4 index (Figure 3a) for the same two time periods as in Figure 2. The ENSO‐induced “great arch” teleconnection emanating from the central equatorial Pacific is visible in both periods. However, the post‐1980 time period features a noticeably stronger circulation amplitude; this includes a deepened, standing trough west of Texas as in May 2015. The deepened trough, together with an abnormally strong subtropical jet that extends into Baja California (not shown), is indicative of enhanced synoptic waves directed toward the southern Great Plains. In Figures 3c and 3d we show the tropical precipitation (shaded) and SST (contoured) regressions during the different time periods: After 1980, the El Niño‐induced precipitation and SST anomalies reveal marked increases over the equatorial central Pacific—in fact, at the center (outlined by a yellow circle), precipitation has increased 1.7 times in the latter period while the mean SST increased by about 1°C (Figures 3e and 3f).
Focusing on wintertime, Zhou et al. [2014] found that the tropical Pacific precipitation anomalies associated with ENSO would intensify in a warmer climate while extending eastward over the equatorial eastern basin. Such an increase in precipitation, which is similar to what Figure 3d shows, leads to substantial change in latent heating that can further intensify the teleconnection response [e.g., Branstator, 1985; Palmer, 2014; Wang et al., 2015]. To examine further, we conducted a comparable analysis to that of Figure 3 but this time with respect to the CMIP5 models in order to detect any change in the ENSO teleconnection between that of NAT and GHG; this is shown in the supporting information Figure S1 for the time period 1970–2005. The results are alike in that under CHG forcing, an intensified teleconnection wave train linked to a deepened trough in the southwestern U.S. is observed, in association with precipitation anomaly enhancement over the equatorial central Pacific. Altogether, the results presented here (i.e., Figures 2 and 3) as well as those in the current literature support the strengthening of the ENSO teleconnection due to a warmer climate.
3.2 Other Factors
The cause of widespread flooding is manifold and cannot be explained solely by any single climate process. Additional circulation features associated with the extreme rainfall of May 2015 do exist: By conducting a power spectral analysis for zonal wave numbers in the May 2015 stream function within the 30°–50°N latitudinal zone, a wave 2 regime and a wave 5 regime emerged (see Figure S2). While the wave 2 regime reflects the ENSO‐induced circulation anomaly that is inherently of longer wavelength [Wallace and Gutzler, 1981], the wave 5 regime echoes an increasingly influential mode of the so‐called circumglobal teleconnection [Branstator, 2002; Schubert et al., 2011; Teng et al., 2013]. Focusing on the latter, we performed a zonal harmonic analysis on the stream function anomaly following Wang et al. [2013]; this wave 5 component is shown in Figure 4a. A clear short‐wave train emerges encompassing the deepened trough west of the flooded region and the anomalous ridge to the east. We next computed the linear trend (slope) of this wave 5 stream function for the time period of 1980–2014 from the ensemble of modern‐era reanalyses; this is shown in Figure 4b. The trend reveals a distinct wave pattern in the wave 5 regime, and the phase of this intensified short‐wave train is remarkably coincident with the May 2015 anomaly. This result suggests an intensification of the short‐wave circulation and is resonant with the findings of Meehl and Teng [2007] and J.‐Y. Lee et al. [2014], i.e., that increased ENSO amplitude that ensues from a warmer climate produces a prominent wave 5 pattern within the teleconnection.
The similarity between the two wave 5 circulations (Figures 4a and 4b) accompanied by a stronger low‐level jet (LLJ) in the Great Plains (Figure 1c) implies a coupling enhancement in the classic trough‐LLJ setting [Uccellini, 1980], and such coupling produces the majority of precipitation in the southern Great Plains during the late spring [Wang and Chen, 2009]. Readers are referred to Text S1 for further explanation of the relevant synoptic processes. The long‐term change in this trough‐LLJ coupling was further examined by computing the linear trend of the column water vapor fluxes
, integrated up to 300 hPa (Figure 4c). A distinct band of southerly
forms over the southern Great Plains signifying an intensified LLJ that is coupled with the deepened upper level trough to the west, as was noted in Barandiaran et al. [2013]. While Weaver et al. [2009] related the springtime LLJ intensification to interdecadal variation in the North Atlantic, Cook et al. [2008] linked the increased LLJ with the anthropogenic global warming.
Though the cause of flood is not the focus here, a final comment that should be realized is that ground conditions in Texas more than likely were “preconditioned” to initiate flooding conditions (J. Meng, NOAA/NCEP/EMC, personal communication, 2015): As Figure S3 shows, April 2015 was abnormally wet in southern Texas (though not exceptionally) whereupon soil moisture in the Huston area was already above normal for that time of year which carried forward into early May near saturation conditions.
4 Summary
The record precipitation that occurred over Texas and Oklahoma during the month of May 2015 was the result of a series of climate interactions and anomalies. Foremost is the role that ENSO played: A developing El Niño has a tendency to increase spring precipitation over the southern Great Plains, and this effect was found to have intensified since 1980; this intensification was concomitant with a warmer atmosphere due to anthropogenic GHG. Specifically, the intensified ENSO teleconnection appears to be triggered by enhanced latent heating in the equatorial central Pacific and is associated with broad SST warming in the tropics. In essence, there was a detectable effect of anthropogenic global warming on the teleconnection and moisture transport leading to May 2015's high precipitation. Previous studies as well as this one point to the following processes: (1) long‐term warming of the tropical Pacific acting to strengthen the atmospheric response to ENSO, (2) El Niño modulating the wave 5 circulation pattern in a warmer climate and its phase lock with the May 2015 anomaly, (3) enhancement of the Great Plains LLJ and associated moisture supply in late spring, and (4) the LLJ's coupling with the deepened spring trough at upper levels. All the aforementioned processes together with the attribution analyses of CMIP5 and CESM1 models analyzed here point toward the exacerbating effect of increasing GHG on the springtime precipitation over Texas and Oklahoma during a developing El Niño—this being so currently (i.e., 2015) and in the future. Furthermore, the diagnostic analyses detailed here, in which increased extreme events and a warmer climate were shown to be dynamically linked, are keys in the provision of seasonal predictions as a guide to future occurrences and intensities of extreme weather events.
Acknowledgments
Soil moisture information shared by Jesse Meng is appreciated. W.‐R. Huang was supported by the Ministry of Science and Technology of Taiwan under MOST 104‐2111‐M‐003‐001 and MOST 103‐2111‐M‐003‐001. H.‐H. Hsu is supported by NSC 100‐2119‐M‐001‐029‐MY5 supported by MOST, Taiwan. S.‐Y. Wang and R.R. Gillies are supported by the Bureau of Reclamation grants R11AC81456 and R13AC80039. ENSO indices are provided by the CPC at http://www.cpc.ncep.noaa.gov/data/indices/sstoi.indices. All data utilized in this study are freely available from the following sources: the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/, the CMIP5 Data Portal at http://cmip‐pcmdi.llnl.gov/cmip5/data_portal.html, and the CESM1 Large Ensemble Community Project at http://www.cesm.ucar.edu/experiments/cesm1.1/LE/.
The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.
