1. Introduction
The southern plains of the United States are susceptible to extreme precipitation events in late spring, which pose substantial social and economic challenges. For example, prolonged droughts during the spring impact crop selection and the management of grazing lands (Raz-Yaseef et al. 2015; Baxter et al. 2021), while intensified spring rainstorms pose an increased risk of extreme rainfall and subsequent flooding (Wang et al. 2018). These precipitation variations directly affect agricultural sectors, influencing critical aspects such as crop yield and grazing conditions. In 2011, Texas witnessed an unprecedented drought following historically low March precipitation, resulting in substantial losses in livestock and hay production (Nielsen-Gammon 2012; Ziolkowska 2016; Fernando et al. 2016). Subsequently, the southern plains experienced a stark contrast in 2015 and 2016 when heavy spring rainfall led to catastrophic floods, causing extensive disruptions to socioeconomic conditions and loss of life (Wang et al. 2015). The consequences of severe flooding extend to vital infrastructures, including military facilities in Texas, Missouri, and Louisiana. Moreover, these events align with the amplified pattern of alternating water cycle extremes identified by Yoon et al. (2018). The adverse effects of these consecutive extreme events underscore the urgent demand for reliable late-spring precipitation forecasts. To enhance the predictability of late-spring precipitation in the southern plains, it is crucial to identify predictable sources on longer time scales, such as remote ocean forcing.
While El Niño–Southern Oscillation (ENSO) is widely recognized as the most prominent factor in predicting winter precipitation in the southern plains (National Integrated Drought Information System 2021), its teleconnection impact becomes elusive during the transition season from spring to summer (Timmermann et al. 2018). During La Niña events in the boreal winter, the atmospheric response to tropical Pacific forcing results in the meridional position of the jet stream and storm track in the western United States shifting poleward compared to El Niño events (Rasmusson and Mo 1993; Trenberth et al. 1998; Cook and Schaefer 2008; Lee et al. 2014; Zhang and Villarini 2018). These atmospheric conditions are prone to bring dryness in the southern plains during La Niña and wetness during El Niño (e.g., Ropelewski and Halpert 1986, 1987; Ting and Wang 1997; Dai et al. 1998; Schubert et al. 2009; Seager et al. 2014; Seager and Hoerling 2014). However, during late spring, the sea surface temperature (SST) anomalies associated with the ENSO transition become less systematic and can be chaotic (Fedorov and Philander 2000; McPhaden 2003; McPhaden and Zhang 2009; Timmermann et al. 2018), obscuring the remote impact of ENSO. It has been observed that the onset or persistence of El Niño may increase precipitation in the Southern Great Plains in the spring, but the increased precipitation may also occur during other ENSO phases (Ropelewski and Halpert 1987; Lee et al. 2014). Moreover, recent studies have pointed out that only half of the recorded La Niña events can explain summer droughts over the Great Plains (Pu et al. 2016; Fernando et al. 2016, 2019), implying that ENSO is an insufficient precursor for forecasting late-spring precipitation.
The remote impact from the Atlantic and Indian Oceans might complement the inadequate ENSO influence on spring precipitation in the southern plains (Schubert et al. 2004; Seager et al. 2014; Seager and Hoerling 2014; Tian et al. 2017). Variability in the tropical and North Atlantic climate can influence precipitation in the southern United States (Mo et al. 2009; Schubert et al. 2009; Hu and Feng 2012). Such variability can also trigger drought events, exemplified by the 1930s Dust Bowl, the 1950s drought, and the 2011 drought (Nigam et al. 2011; Seager et al. 2014). Land surface feedback can further intensify and prolong these droughts (Hong and Kalnay 2002; Koster et al. 2004; Mueller and Seneviratne 2012; Fernando et al. 2016; Williams et al. 2016). Aligning with prior research, the emerging idea of tropical interbasin interactions underscores the significant role of tropical Atlantic variability in global climate predictability via cross-basin atmospheric teleconnections (McGregor et al. 2014; Chikamoto et al. 2015; Kucharski et al. 2015; Li et al. 2016; Ruprich-Robert et al. 2017; Cai et al. 2019). Thus, the notion of tropical interbasin interaction offers the potential to enhance late-spring precipitation predictability in the southern plains.
Key inquiries of this notion revolve around the extent to which factors from the tropical Pacific and Atlantic enhance precipitation predictability in the southern plains, as well as the performance of current climate models in forecasting their oceanic origins. Addressing these challenges could advance our predictive skills and understanding of climate dynamics. To improve seasonal precipitation predictability, this study analyzes the relationship between April–June (AMJ) precipitation variability in the southern Great Plains and the large-scale atmospheric circulation influenced by ocean variability. Utilizing a combination of global and partial ocean assimilation approaches, as detailed in section 2, our aim is to identify the model biases and systematic errors in simulating precipitation (section 3) and the relative contributions from the tropical Pacific and the Atlantic to precipitation predictability in the southern Great Plains (section 4). In section 5, we focus on assessing the seasonal predictive skills of ocean precursors in two independent sets of hindcast experiments: the decadal climate prediction experiments based on the Community Earth System Model (CESM) and the North American Multi-Model Ensemble (NMME). CESM hindcast experiments span the extended period from 1960 to 2015, with the single model initialized annually on 1 January and a 10-yr lead time. The NMME, an ensemble prediction system based on seven models, covers a shorter period starting from 1982, with monthly initialization and a 9-month lead time. Therefore, the consistent results of these hindcast experiments enhance the robustness of our findings. Results are discussed in section 6, and conclusions are presented in section 7.
2. Data and model experiments
a. Observation datasets
We utilized gridded monthly precipitation datasets from the Global Precipitation Climatology Centre (GPCC) version 2022 with a resolution of 0.25° for the period of 1891–2020 (Schneider et al. 2015), geopotential height anomalies at 250 hPa (Z250) from the Japanese 55-year Reanalysis (JRA-55) with a resolution of 1.25° for 1958–2021 (Kobayashi et al. 2015), sea surface temperature from the Centennial In Situ Observation-Based Estimates version 2 (COBE2) SST with a resolution of 1° (Hirahara et al. 2014) for 1850–2021, and NOAA Optimum Interpolation (OI) SST V2 with a 1° latitude–longitude grid for 1982–2022. The historical time series of statewide monthly precipitation in Texas, spanning from 1895 to 2020, was retrieved from NOAA’s Climate at a Glance (NOAA 2023). This dataset consists of weather station-based daily precipitation observations, averaged statewide to provide a monthly measure reflective of the entire Texas region. We compiled these datasets for the analysis period of 1960–2015 to align with the CESM model experiments, and for 1982–2020 to match the NMME, as described below. We selected JRA-55 for its comprehensive data coverage aligning with the CESM model experiments, with similar results expected from the fifth major global reanalysis produced by ECMWF (ERA5) (Hersbach et al. 2020). Anomalies were calculated as deviations from the climatological mean for the respective analysis periods. All anomalies were linearly detrended at each grid point.
b. CESM decadal climate prediction system and ocean assimilation experimental design
The fully coupled climate model utilized in this study is a lower-resolution version of the CESM version 1.0.6 (Gent et al. 2011; Shields et al. 2012). The atmospheric and land models have a horizontal resolution of the T31 spectral grid, which corresponds to an approximate 3.75° horizontal resolution. They consist of 26 hybrid sigma/pressure coordinate levels in the atmosphere, 10 soil layers, and an aquifer water layer on land. The ocean and sea ice models use a curvature grid with a displaced North Pole, resulting in a 3° horizontal grid but a 1° latitude grid near the equator. The ocean model employs 60 vertical levels. This computationally efficient model resolution allows us to conduct numerous ensemble model integrations for the decadal climate prediction experiment using a fully coupled climate model and data assimilation techniques, covering an extended period of 55 years with a longer lead time of 10 years compared to seasonal forecasting systems (∼6000 years in total). No flux correction is applied when exchanging heat, water, and momentum fluxes between the atmosphere and the ocean.
In this study, we employed CESM to conduct a series of model experiments, as detailed by Chikamoto et al. (2019). The experiments comprised three sets, each utilizing a 10-member ensemble and consistent radiative forcings. These forcings included greenhouse gas and aerosol concentrations, solar cycle variations, and major volcanic eruptions, aligning with historical records up to 2005 and the IPCC RCP4.5 future emission scenario thereafter. The first set, referred to as the global (GLOB) run, involved a global ocean data assimilation experiment. For the period 1958–2015, we assimilated three-dimensional ocean temperature and salinity anomalies from the observation-based ocean reanalysis dataset (Balmaseda et al. 2013) into the ocean component of CESM. This assimilation employed the incremental analysis update (IAU) scheme (Bloom et al. 1996; Huang et al. 2002), which distributes analysis increments into each model time step during a specified assimilation window. This method enhances assimilation accuracy by using time-evolved forcing to incrementally correct the model, thereby improving the accuracy of the analysis field and significantly reducing initialization shock. The IAU scheme’s ability to maintain model continuity has led to its adoption in various data assimilation systems (Mochizuki et al. 2012; Tatebe et al. 2012; Zhang et al. 2015; Ham et al. 2016; Lei and Whitaker 2016; Chikamoto et al. 2019).
In the second set, we conducted partial ocean assimilation experiments, named equatorial Pacific (eqPAC) and Atlantic (ATL) runs, to investigate the potential impacts of specific ocean basins. These runs focused on assimilating observed ocean information exclusively in the equatorial Pacific (10°S–10°N) and the Atlantic (30°S–70°N), respectively, while maintaining the same model configuration as the GLOB run. Notably, our approach shares similarities with the AMIP-type experiments, which prescribe tropical SST forcing to the atmospheric general circulation model (e.g., Alexander et al. 2004). However, our partial assimilation method in the eqPAC and ATL runs allows for atmosphere–ocean interactions within both assimilated and nonassimilated regions, as well as dynamically coupled atmosphere and ocean responses to remote ocean forcings. For instance, in the ATL run, we assess the fully coupled atmosphere–ocean responses to Atlantic Ocean forcings, which includes evaluating the tropical Pacific SST variability. Thus, partial assimilation experiments, such as ours, are more adept at evaluating interbasin interactions compared to AMIP-type experiments that depend solely on atmospheric models.
In the last set, a series of 10-yr-long hindcast experiments were conducted, each initialized on 1 January annually from 1960 to 2015. The initial conditions for these experiments were derived from the GLOB run. To minimize model drift during prediction, we implemented bias correction during the assimilation of ocean temperature and salinity within the GLOB run. This correction aimed to preserve the model-simulated climatology and externally forced variability, while integrating observed internal variability (Chikamoto et al. 2019). Additionally, we extended the assimilation to include subsurface and deep ocean temperatures and salinities, which facilitated rapid model equilibration and allowed for a more efficient spinup process. Given that mean state biases are a primary cause of initialization shocks and artificial drift during forecasting, our methodology strategically addresses these issues to enhance the robustness of our hindcast experiments. The performance and outcomes of these experiments have been extensively documented in previous studies (Chikamoto et al. 2015, 2016, 2019, 2020b; Purich et al. 2016; Ham et al. 2017; Johnson et al. 2018, 2020; Stuivenvolt-Allen et al. 2021).
To mitigate the influence of model spinup, our analysis focuses on the period from 1960 to 2015 for both the global and partial ocean data assimilation runs. Anomalies were computed as deviations from the climatological mean over the 56-yr period (1960–2015) in each model experiment. Subsequently, all anomalies underwent linear detrending at each grid point to remove long-term trends. In the hindcast experiment, we specifically extracted a lead time of 4–6 months, corresponding to the targeted season of AMJ for the years 1960–2015. Climatology within the hindcast experiments was determined by calculating the ensemble mean for each respective lead time. Similar to the other model experiments and observations, anomalies in the hindcast experiment were calculated by deviations from the climatological mean for 1960–2015 and were linearly detrended. The detrending process is crucial to isolate natural variability from the climate response to radiative forcing, especially as SST over a more extended period demonstrates a prominent warming trend due to climate change.
c. NMME
To assess seasonal predictability from various forecasting systems, we compiled predicted SST and precipitation datasets from the NMME (Kirtman et al. 2014). These datasets comprise seasonal ensemble hindcast/forecast experiments conducted by multiple forecasting centers, each utilizing coupled climate models and distinct initialization techniques. Each model experiment exhibits differences in data coverage, lead time, output variables, and ensemble size. For consistency in our analysis, we aggregated 9-month-long hindcasted/forecasted products initiated every month from January 1982 to December 2020 from seven forecasting systems (refer to Table 1). All datasets are based on a 1° latitude–longitude grid on monthly time scales. Within each system, we initially calculate the ensemble mean and subsequently derive climatology and anomalies. The climatology is determined by averaging over the 30-yr period from January 1991 to December 2020, for each lead time and system. Anomalies are defined as deviations from this climatology. Finally, the multimodel ensemble is computed by averaging these products, with equal weight assigned to the seven forecasting systems, promoting a balanced representation of the collective forecasting capabilities. As described in the observation datasets above, we use OI SST to evaluate the NMME results, a common practice in NMME validation. Given the shorter validation period, linear trends are not removed from the anomalies to highlight the combined effect of internally generated variability and the climate response to radiative forcing. However, it should be noted that these linear trend components are indirectly removed when calculating the interbasin contrast, as discussed below.
Summary of NMME forecast systems, ensemble sizes, and references used in this study.
d. SVD analysis
To unravel coherent variability in precipitation between observational records and model simulations, we employed singular value decomposition (SVD) analysis on seasonally averaged precipitation anomalies from both sources. This statistical technique identifies the principal modes of covariability between pairs of variables (Bretherton et al. 1992). It decomposes the correlation (or covariance, as utilized in maximum covariance analysis) matrix into spatial and temporal patterns represented by left and right singular vectors and their corresponding expansion coefficients (ECs), respectively. The significance of each mode is quantified by the squared covariance fraction (SCF), with the primary mode designated as SVD1, accompanied by its ECs, labeled EC1. Homogeneous correlation maps are generated by aligning an EC with its corresponding field (e.g., the left EC with the left vector field), while heterogeneous correlation maps associate ECs with contrasting fields (e.g., the right EC with the left vector field).
SVD analysis is a widely recognized method for detecting air–sea coupling processes, notably between SST and 500-hPa geopotential height (Wallace et al. 1992). It is also used to calibrate forecast models and evaluate their potential predictability (Rukhovets et al. 1998; Feddersen et al. 1999; Chikamoto et al. 2013). In our study, observational data are represented by the left singular vectors/ECs, while model simulations correspond to the right. We employ homogeneous correlation maps to distinguish between signals in the observed and simulated data.
Our CESM ocean data assimilation experiments leverage SVD analysis to dissect the primary mode of seasonal precipitation anomalies, aiming to estimate potential predictability driven by ocean forcing. Traditional predictive skill assessments, typically reliant on direct grid-to-grid comparisons between observed and modeled datasets, are prone to distortion by inherent model biases and systematic errors. By focusing on the first SVD mode comparison between observational data and ocean data assimilation outputs, we effectively circumvent these inaccuracies. This method enables a precise evaluation of the impact of ocean forcing on predictability, thus enhancing our understanding of forecast accuracy despite notable model shortcomings. While other statistical methods, such as canonical correlation analysis (CCA), extended logistic regression (ELR), and average predictability time (APT), are available for estimating potential predictability, the foundational principle of SVD analysis remains central in addressing biases and errors in dynamical models (Ward and Navarra 1997; DelSole and Tippett 2009; Jia and DelSole 2011; Barnston and Tippett 2017).
3. Model performance in simulating Texas precipitation
To assess model performance in simulating precipitation, we analyzed the climatology and standard deviations of precipitation during AMJ in the GPCC, CESM GLOB run, and the multimodel ensemble mean of NMME initialized on 1 April (Fig. 1). Observed precipitation climatology shows a pronounced zonal contrast, with drier conditions in the southwest and wetter conditions in the southeast, leading to a substantial zonal gradient across Texas, ranging from 0.5 to 4.0 mm day−1 (Fig. 1a). While both CESM and NMME capture these regional distinctions (Figs. 1b,c), CESM’s coarse resolution hampers its ability to simulate localized patterns accurately. Similarly, while observed standard deviations align closely with climatological patterns, being lower in the southwest and higher in the southeast with a peak around eastern Texas and Louisiana, both models tend to underestimate these peak values, particularly in Texas and Louisiana (Figs. 1d–f). Despite these limitations, the models simulate the broader observed climatic features reasonably well.
(a)–(c) Climatology and (d)–(f) corresponding standard deviations of precipitation during AMJ derived from the (left) GPCC for 1960–2015, (center) CESM GLOB run for 1960–2015, and (right) NMME for 1991–2020. Model climatology is based on the ensemble mean of 10 members for the CESM GLOB run and the multimodel ensemble mean for NMME, where individual model ensemble means are averaged. Standard deviations are computed initially for individual members before averaging across the ensemble. NMME forecasts are initialized annually on 1 Apr, featuring a 1–3-month lead time.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
Precipitation correlation maps with statewide Texas precipitation, obtained from NOAA’s Climate at a Glance, highlight the model’s systematic errors (Fig. 2). During AMJ, the spatial pattern of GPCC exhibits higher positive correlation coefficients exceeding 0.7 across Texas (Figs. 2a,b), indicating consistency between two different precipitation data sources (i.e., GPCC and NOAA’s Climate at a Glance). A slightly weak negative correlation is also observed in California and the Pacific Northwest. These patterns persist through two different analysis periods (Figs. 2a,b), suggesting that AMJ Texas precipitation is localized and stationary. The NMME forecasts at 1–3-month lead time capture this positive correlation (Fig. 2d), although the region of positive correlation extends further westward. However, they fail to capture the weak negative correlation coefficients around California and northern Idaho, pointing to the model’s deficiencies in simulating localized precipitation patterns. The correlation coefficients in NMME forecasts are slightly smaller than those observed and decay more rapidly after 4–6-month lead time (Figs. 2d,f), posing a challenge for seasonal forecasts at longer lead times but a reasonable performance for short lead times (Infanti and Kirtman 2014; Khajehei et al. 2018). In contrast, the CESM GLOB run completely misses this positive correlation over Texas (Fig. 2c), instead showing a local maximum around Utah. Since the CESM GLOB run assimilates only ocean information and not atmosphere–land observations, the lower correlation compared to NMME highlights the importance of initializing atmosphere–land variables. Due to this low correlation, the CESM hindcast runs at 4–6-month lead time also show lower correlations with statewide Texas precipitation (Fig. 2e).
Correlation maps of precipitation anomalies with statewide Texas precipitation derived from (a) GPCC for 1960–2015, (b) GPCC for 1982–2020, (c) CESM GLOB run for 1960–2015, (d) NMME at 1–3-month lead time for 1982–2020, (e) CESM HCST runs at 4–6-month lead time for 1960–2015, and (f) NMME at 4–6-month lead time for 1982–2020. Statewide Texas precipitation data are obtained from NOAA’s Climate at a Glance. CESM simulations are based on the ensemble mean of 10 members, while the NMME utilizes the multimodel ensemble mean. The 1–3- and 4–6-month lead times correspond to predictions initialized on 1 Apr and 1 Jan, respectively. Correlation coefficients of 0.26 and 0.31 achieve 95% statistical significance with 54 and 37 degrees of freedom, respectively, using a two-tailed Student’s t test.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
To evaluate the model performance in simulating moisture transports, we examine precipitable water and horizontal winds at 700 hPa in the JRA-55 reanalysis and CESM GLOB run (Fig. 3). The observed climatological conditions during AMJ are characterized by dominant westerly winds in the lower atmosphere across the United States (Fig. 3a). Near the West Coast of the United States, this jet stream splits meridionally, forming subpolar and subtropical jets. The southerly winds, as part of the subtropical jet, transport moist air toward the southern plains, marking the transition from a dry winter to a wet summer in the region. The CESM GLOB run reasonably simulates the overall structure of these lower-atmospheric circulations and precipitable water, despite its coarse resolution, although the meandering pattern of the subtropical jet (Fig. 3b) appears weaker than in the reanalysis dataset (Fig. 3a).
(a),(b) Seasonal climatology of precipitable water (shaded) and horizontal winds at 700 hPa (vector) for AMJ and (c),(d) their regressions with statewide Texas precipitation for 1960–2015, derived from the (left) JRA-55 reanalysis and (right) CESM GLOB run. The CESM GLOB run is based on the ensemble mean of 10 members. Statewide Texas precipitation data are obtained from NOAA’s Climate at a Glance. The reference vector is placed in the upper-right corner of each plot. Vectors with a magnitude less than 20% of the reference vector are omitted.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
In contrast to the observed climatological conditions, the CESM GLOB run exhibits systematic errors in representing the lower-atmospheric circulation and precipitable water patterns associated with Texas precipitation (Figs. 3c,d). The observed wind anomalies display cyclonic circulations centered in Mexico, accompanied by strong southerly wind anomalies in the southern plains, known as low-level jets (Fig. 3c). These southerly anomalies transport moist air from the ocean to Texas, contributing to the enhanced Texas precipitation through moisture advection by anomalous winds (i.e.,
4. Ocean impacts on late-spring precipitation variability in the Great Plains
a. Potential ocean impacts on Texas precipitation: Global ocean data assimilation run
To mitigate systematic errors in simulating precipitation variability in the southern plains, we applied SVD analysis to AMJ precipitation anomalies between the GPCC and CESM GLOB runs (left panels of Fig. 4). The first SVD mode reveals that the observed precipitation pattern exhibits higher correlations in Texas and Louisiana compared to surrounding states (Fig. 4a), which closely mirrors the pattern seen in the correlation map with statewide Texas precipitation (Fig. 2a). Since the CESM GLOB run assimilates only ocean observations, this first SVD mode suggests that ocean-induced precipitation variability significantly influences these regions. Indeed, the temporal variability of the first SVD mode captured extreme periods of precipitation, such as the wet season in 2015 and the dry season in 2011, correlating strongly with statewide Texas precipitation (correlation coefficient R = 0.84). Conversely, the model-simulated precipitation pattern is heavily displaced toward Utah rather than Texas (Fig. 4d), indicating a systematic error as previously described. This first SVD mode explains 36.9% of the total variance, with the correlation coefficient of their EC1 reaching 0.62. The second mode, explaining 22.3% of the total variance, demonstrates effective separation from the first mode, based on the sampling size of 56 years (North et al. 1982). These results imply that despite the model’s bias in simulating spatial precipitation patterns, the first SVD mode successfully captures the temporal evolutions of ocean-induced precipitation variability in the southern plains.
Homogeneous correlation maps of (a)–(c) GPCC, (d)–(f) CESM simulations, and (g)–(i) their corresponding ECs (EC1) associated with the first SVD mode of seasonal mean precipitation anomalies during AMJ between the GPCC and CESM simulations. The first SVD mode is derived from the correlation matrix for precipitation anomalies during AMJ in North America (20°–50°N, 125°–80°W) between the observations (GPCC) and the CESM simulations, including (left) GLOB, (center) eqPAC, and (right) ATL runs. The SCF of the first SVD mode accounts for 36.9% in GLOB, 31.6% in eqPAC, and 30.0% in ATL runs of the total variance. Black and red lines in (g)–(i) represent EC1 from GPCC and CESM simulations (red solid: ensemble mean; red shaded: maximum-to-minimum range of the 10 ensemble members), respectively. The correlation coefficient of EC1 between the GPCC and CESM simulations is noted in the upper-right corner of the bottom panels.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
The correlation maps of Z250 and SST anomalies associated with the EC1 further highlight the impact of large-scale climate factors on precipitation variability in the southern plains (Fig. 5). We removed the zonal mean from Z250 anomalies to emphasize the standing wave pattern (referred to as Z250*). The observed dataset revealed tropical interbasin contrasts, with positive Z250* and warmer SST anomalies concurring in the central tropical Pacific and negative Z250* and colder SST anomalies coexisting in the tropical Atlantic (Figs. 5a,b). These contrasts can act as a heat source for atmospheric teleconnections, generating a Pacific–North America (PNA)-like pattern through Rossby wave energy propagation (Fig. 5a). The ensemble mean of the GLOB run, correlated with the EC1, exhibited similar but amplified Z250* and SST patterns (Figs. 5c,d). Chikamoto et al. (2017) demonstrated that the interbasin interaction between the Pacific and Atlantic influences drought predictability in the western United States on decadal time scales. The findings presented here suggest that such interbasin interaction also impacts late-spring precipitation in the southern plains.
Correlation maps of (left) Z250 and (right) SST anomalies associated with the observed and model-simulated ECs of the first SVD mode. Observed correlation maps are derived from (a) JRA-55 and (b) COBE2 SST, corresponding to the observed EC1 (i.e., the black line in Fig. 4g). Model-simulated correlation maps are presented for (c),(d) the GLOB run, (e),(f) the eqPAC run, and (g),(h) the ATL run, calculated against the model-simulated EC1 in the GLOB run (i.e., the red line in Fig. 4g). Geopotential height anomalies (Z250*) are defined by the AMJ mean and the deviation from the zonal average at each latitude. Correlation coefficients of 0.26 achieve 95% statistical significance with 54 degrees of freedom using a two-tailed Student’s t test. The black box represents the area-averaged domain in the tropical Pacific (15°S–15°N, 180°–150°W) and ATL (15°S–15°N, 60°W–10°E).
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
We note some discrepancies in the Z250* anomaly pattern between the observations and the GLOB run. Particularly, the positive Z250* anomalies in the tropical central Pacific are shifted eastward in the model simulation compared to the observations, leading to a displacement of PNA-like teleconnection patterns (Figs. 5a,c). As a result, Fig. 4 reveals a significant bias in the model-simulated precipitation pattern. Nevertheless, it is clear that the low-resolution CESM effectively captures the large-scale climate dynamics, especially for the tropical interbasin contrast between the Pacific and the Atlantic. In fact, the temporal variability of the tropical interbasin contrast, calculated by Z250* anomalies in the CESM GLOB run, agrees well with the reanalysis (R = 0.93 in Fig. 6a). Given the high correlation between the tropical interbasin contrasts of Z250* and SST anomalies (R = 0.83 in observation and 0.91 in the CESM GLOB run), these contrasts can serve as an alternative metric for assessing late-spring precipitation predictability in the southern plains.
Standardized time series of the tropical Pacific–ATL contrast in (left) Z250* and (right) SST for observations (black) with the CESM simulations in (a),(b) the GLOB run (red) and the partial assimilation runs (purple) of (c),(d) the eqPAC run and (e),(f) the ATL run. Lines and shades indicate the ensemble average and the maximum-to-minimum range of 10 ensemble members, respectively. The tropical Pacific (15°S–15°N, 180°–150°W) and ATL (15°S–15°N, 60°W–10°E) domains are highlighted by the black boxes in Fig. 5. Correlation coefficients of CESM simulations with the observation are denoted in the upper-right corner (the number in the bracket indicates against the GLOB run instead of the observation).
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
b. Identifying the sources of ocean forcing: Partial ocean data assimilation run
For diagnostic purposes, we further conducted partial ocean assimilation runs by assimilating observed ocean data either in the eqPAC run or in the ATL run. We then evaluated the relative impact of each basin on the interbasin climate contrasts by calculating correlation maps of Z250* and SST anomalies in each partial assimilation run with the EC1 of the CESM GLOB run obtained through the SVD analysis mentioned earlier. Consistent results between the partial assimilation run and the GLOB run would thus indicate contributions from the remote forcing of the assimilated ocean.
Our eqPAC run highlighted the consistent results with previous studies that have demonstrated the significant influence of the equatorial Pacific Ocean on global climate variability, including interbasin climate contrasts (McGregor et al. 2014; Chikamoto et al. 2015; Li et al. 2016; Cai et al. 2019). The correlation map of Z250* anomalies with the EC1 reveals the zonal Z250* contrast in the tropics between the central Pacific and the Atlantic, as well as the presence of a PNA-like teleconnection pattern (Fig. 5e). These atmospheric responses exhibit a similar pattern in the GLOB run (Fig. 5c), emphasizing the importance of equatorial Pacific forcing in determining tropical interbasin contrast and resultant precipitation variability in the southern plains. The SST correlation map with EC1 shows a consistent pattern over the tropical and North Pacific between the eqPAC and GLOB runs but with opposite signs over the tropical Atlantic (Figs. 5d,f). This result lends more direct support to the hypothesis that tropical Atlantic variability also contributes to the tropical interbasin contrast and precipitation variability in the southern plains.
Our ATL run of the partial ocean assimilation highlighted the additional role of the Atlantic alongside the equatorial Pacific. Cooler SST anomalies in the tropical Atlantic, compared to those in the central Pacific (see Fig. 5h), contributed to creating tropical interbasin contrast. This was evident in the lower Z250* anomalies in the Atlantic and higher Z250* anomalies in the central Pacific (Fig. 5g). Although the cooling in the tropical Atlantic did not generate positive SST anomalies in the central tropical Pacific (Fig. 5h), we still observed negative SST anomalies in the North Pacific and higher but weaker Z250* anomalies around Alaska (Fig. 5g). Therefore, the tropical interbasin interaction between the Pacific and the Atlantic, rather than the tropical Pacific alone, played a more important role in controlling the interbasin contrasts of SST and Z250* anomalies, thereby serving as an effective metric for assessing the late-spring precipitation variability over the southern plains.
To estimate the relative contributions of the equatorial Pacific and the Atlantic, we further conducted a time-series analysis of the interbasin contrasts of SST and Z250* (Fig. 6). The eqPAC run exhibited high correlation coefficients with the observations and the GLOB run for both interbasin contrasts of SST and Z250* anomalies (R = 0.82–0.88 in Figs. 6c,d). While relatively weaker, the ATL run did reveal significant correlation coefficients (R ranging from 0.49 to 0.62 in Figs. 6e,f). Based on these correlation coefficients, we deduced that the equatorial Pacific contributes about 70% (range: 67%–77%) to the interbasin contrasts, while the Atlantic accounts for the remaining 30% (range: 24%–38%). Our results imply that including ocean conditions from both the tropical Pacific and the Atlantic may boost seasonal precipitation predictability. This underscores the need to further examine the predictive abilities of interbasin climate contrasts in existing climate models, as elaborated in the next sections.
The Atlantic contributions to modulating late-spring precipitation in Texas are also highlighted by the SVD analysis between the observations and the partial assimilation experiments. Figure 4 also shows the first SVD modes between the observations and the partial ocean assimilation experiments. The first SVD modes for both model simulations explain about 30% of the total variance but demonstrate different spatial patterns. The observed precipitation pattern of the first SVD mode in the eqPAC run is characterized as a monopole pattern around the western United States along the 40°N (Fig. 4b). This first SVD mode also exhibits a minor contribution from the equatorial Pacific to the Texas precipitation variability, which makes ENSO remote impact less clear, as described in section 1. By contrast, the observed precipitation of the first SVD mode in the ATL run shows a dipole pattern between the Pacific Northwest and the southern plains (Fig. 4c), which resembles the SVD result between the GPCC and GLOB runs (Fig. 4a). This result echoes Johnson et al. (2020) in that the Atlantic remote impacts could modulate the atmospheric teleconnection pattern forced by the tropical Pacific.
5. Seasonal predictability
a. Hindcast experiments in CESM
To evaluate the seasonal predictability, we next examined hindcast experiments conducted using CESM. These experiments were initiated on 1 January annually from 1960 to 2015, based on the initial conditions derived from the CESM GLOB run. Following this, a 10-yr model free run commenced from these conditions. The hindcast experiments emphasized the AMJ season, corresponding to a 4–6-month forecast lead time. Anomalies were determined by computing the deviation from the long-term average between 1960 and 2015 during the 4–6-month lead time within the experiment. Each hindcast run incorporated 10 ensemble members, offering a robust dataset for examination.
The hindcast run adeptly replicates a large-scale pattern akin to that observed in the first SVD mode (Fig. 7). The correlation map of Z250* anomalies in the hindcast, linked with the PC1 in the GLOB run, showcases a pronounced PNA-like teleconnection pattern in the North Pacific (Fig. 7a). Moreover, zonal interbasin disparities of Z250* and SST anomalies between the tropical Pacific and Atlantic are noticeable, barring the equatorial zone (Figs. 7a,b). The diminished correlation at the equator might signify the ENSO spring predictable barrier’s existence. However, the hindcast run displays commendable predictive abilities for these interbasin contrasts of Z250* and SST anomalies, even with a 4–6-month lead time (Figs. 7c,d). It appears that the SST contrast displays superior predictive capabilities compared to the Z250* contrast, highlighting the prolonged ocean memory and interbasin influence on Texas precipitation beyond ENSO’s predictability bounds. Although the CESM hindcast run demonstrates poor performance in predicting Texas precipitation, as shown in Fig. 2e, it can still yield useful predictions by translating interbasin contrasts into meaningful forecasts.
Correlation maps of (a) Z250* and (b) SST anomalies in the hindcast experiment for the 4–6-month lead time associated with the first SVD mode in the ensemble mean of the GLOB run. The time series of the tropical Pacific–ATL contrast in (c) Z250* and (d) SST anomalies are shown for observation-based products (black), the ensemble mean of the GLOB run (red), and the ensemble mean of the hindcast experiment at the 4–6-month lead time (blue). The blue lines represent the ensemble mean, and the shaded area indicates the maximum-to-minimum range of 10 ensemble members in the hindcast experiment. The correlation coefficients between the hindcast run and the observation-based products, as well as the ensemble mean of the GLOB run (within the parentheses), are denoted in the upper-right corners of (c) and (d). A correlation coefficient of 0.235 corresponds to a statistically significant level of 95% with 50 degrees of freedom using a one-sided Student’s t test.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
b. Hindcast experiments in NMME
The zonal interbasin contrast of SST anomalies plays a pivotal role in forecasting seasonal mean precipitation in the southern plains. While current climate models have limitations in simulating precise precipitation due to inherent deficiencies, the SST anomalies’ contrast may prove to be more predictable. We analyzed the seasonal predictive capabilities of this contrast using the NMME, gathering data from seven climate forecasting systems spanning monthly initializations from 1982 to 2021. Given that the NMME offers operational SST forecasts, our insights might pave the way for real-time forecasting products for AMJ precipitation in the region.
Figure 8 presents the time series of this SST contrast alongside its predictive capabilities, with the multimodel ensembles crafted by giving equal importance to each forecasting system. Consistent with the CESM hindcast run, seasonal prediction systems aptly captured the zonal interbasin SST contrast, as depicted in Fig. 8. Most forecasting systems, especially those from NASA and Canada (NASA-GEOSS2S, CanCM4i-IC3, and GEM5-NEMO), demonstrate commendable predictive skills even at a 7–9-month lead time [anomaly correlation coefficient (ACC) > 0.6 and root-mean-square error (RMSE) < 0.9], whereas others like NCEP-CFSv2 and GFDL-CM2p5-FLOR series experience quicker declines in their predictive capabilities. Basically, the multimodel ensemble consistently outperforms individual systems from a 1–3- to 7–9-month lead time, underscoring its utility in offering skillful forecasts for the AMJ season in the southern plains. Due to the extended predictive skill of the interbasin SST contrast, the multimodel ensemble can maintain a relatively high level of skill in predicting Texas precipitation variability over a longer lead time (Fig. S1 in the online supplemental material).
Standardized zonal IBCs during the AMJ season for OI SST (dotted line) and NMME products at (a) 1–3-, (c) 4–6-, and (e) 7–9-month lead time, along with their respective predictive skills based on (b) ACC and (d) RMSE. The ensemble mean of individual forecasting systems is represented by colored asterisks and lines, while the black line represents the multimodel ensemble for those products. The correlation coefficient between the OI SST (dotted line) and the multimodel ensemble (black line) is indicated in the upper-right corner of (a), (c), and (e). In (b) and (d), the x axis denotes the center month for the lead time, with “month 2” representing a 1–3-month lead time, for example.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
Further analysis on predictability (Fig. 9) reveals that while SST anomalies in the central tropical Pacific and the tropical Atlantic exhibit interannual-to-decadal fluctuations, the former generally boasts higher predictability. For example, at 4–6-month lead time, the multimodel ensemble highlighted a higher anomaly correlation coefficient for central tropical Pacific SST variability (R = 0.88) compared to that in the tropical Atlantic (R = 0.79), as displayed in Figs. 9a and 9c. Consistent results are obtained in the detrended time series (R = 0.89 and R = 0.72, respectively). These features also appear in individual models, with notable exceptions such as NCEP-CFSv2 (Figs. 9b,d). In the tropical Atlantic, a wider intermodel spread is observed during the 2013–20 period (Fig. 9c), which contributes to the considerable multimodel spread in interbasin SST contrasts as illustrated in Fig. 8c. This spread may be potentially relevant to the significant SST warming in the tropical Atlantic attributed to climate change (Cai et al. 2019). To enhance predictive capabilities in this region, it may be necessary to improve model performance in capturing both the internally generated variability and the response to external forcings. Nevertheless, throughout the entire lead time, the multimodel ensemble generally demonstrates superior predictive skills in both the central tropical Pacific and Atlantic, with substantial contributions from NASA and Canadian systems. These results underscore the effectiveness of the multimodel ensemble approach for operational SST forecasts.
Standardized SST anomalies during the AMJ season, averaged for (a) the central tropical Pacific and (c) the tropical ATL, in the OI SST (dotted line) and NMME products at a 4–6-month lead time. Additionally, their respective predictive skills, based on the ACC, are denoted in the upper-right corners of (b) the central tropical Pacific and (d) the tropical ATL SST anomalies. Here, the R values without parentheses represent anomalies including trends, while those within parentheses indicate linearly detrended anomalies. The ensemble mean of individual forecasting systems is represented by colored asterisks and lines, while the black line represents the multimodel ensemble for those products. The correlation coefficient between the OI SST (dotted line) and the multimodel ensemble (black line) is indicated in the upper-right corner of (a) and (c). In (b) and (d), the x axis denotes the center month for the lead time, with month 2 representing a 1–3-month lead time, for example.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
c. Developing IBC index for precipitation anomalies
Up to this point, we have demonstrated the baseline performance of models in predicting the direct oceanic influences on precipitation variability in the southern plains. As previously shown, climate models exhibit reasonable accuracy in predicting the interbasin contrasts of SST and Z250*. By introducing the interbasin SST contrast (IBC) index, we can translate its predictions into precipitation anomalies in the southern plains. Here, we defined the IBC index as the difference in standardized SST anomalies between the central tropical Pacific (15°S–15°N, 180°–150°E) and the tropical Atlantic (15°S–15°N, 60°E–10°W) as highlighted in the boxes of Fig. 5. The IBC index becomes positive when tropical Pacific SST anomalies are warmer than Atlantic ones. Figure 10 illustrates coherent variability between the observed IBC index during AMJ and the statewide Texas precipitation anomalies during AMJ (correlation coefficient R = 0.50). Notably, the correlation with the Niño-3.4 index during the prior DJF is significantly reduced (Fig. 10a, R = 0.18). As the NMME demonstrates high predictive skills for interbasin contrasts, our results highlight the effectiveness of the IBC index as an alternative metric to predict late-spring precipitation anomalies in the southern plains.
Standardized time series of (a) Niño-3.4 index during DJF and (b) IBC index during AMJ, with Texas precipitation during AMJ (black line) obtained from the COBE2 SST and the NOAA Climate at a Glance. Correlation coefficients between these time series are denoted in the upper-right corner. Red and blue bars indicate positive and negative anomalies, respectively. The amplitudes larger than one standard deviation are highlighted by deeper colors. (c) A scatterplot between the Niño-3.4 index during DJF and the IBC index during AMJ. The colored circles in (c) represent the standardized statewide Texas precipitation during AMJ.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
The composite analysis of precipitation anomalies further underscores the effectiveness of the IBC index compared to ENSO. Positive and negative phases for these phenomena, defined using criteria of ±1 standard deviation for the Niño-3.4 index during the prior DJF and the IBC index during AMJ, reveal various combinations of ENSO and IBC phases. Due to the positively and weakly correlated nature of the IBC and Niño-3.4 indices (R = 0.24), we isolated the same phases of the IBC and Niño-3.4 indices from the combinations. Examination of precipitation anomalies during these IBC phases without ENSO reveals a mirror image with substantial amplitudes in the southern plains and opposite but slightly weaker anomalies in the Pacific Northwest (Figs. 11c,f). While precipitation anomalies over Texas are influenced by the remote impact of ENSO (Figs. 11b,e), considering ENSO forcing alone makes southern plains precipitation anomalies less distinct (Figs. 11a,d). These results suggest that precipitation variability in the southern plains during AMJ is more directly influenced by concurrent oceanic forcing rather than by wintertime ENSO teleconnections and the resulting local atmosphere–land feedback mechanisms.
Composite maps of precipitation anomalies during AMJ for (a) the El Niño years without the positive interbasin contrast (IBC+) phases (1966, 1973, 1983, 1987, 1998, 2003, 2010, and 2016), (b) concurrent years of El Niño and positive IBC (1992), (c) the positive IBC years without El Niño (1972, 1976, 1982, 1991, 1994, 1997, 2002, and 2015), (d) the La Niña years without the negative IBC (1971, 1974, and 1976), (e) concurrent years of La Niña and negative IBC (1989, 1999, 2000, 2008, and 2011), and (f) the negative IBC years without La Niña (1973, 1988, 1996, 1998, and 2010). The number of composite years is denoted within the parentheses of each subplot title. The ENSO and interbasin contrast years are detected by criteria of one standard deviation in the Niño-3.4 index during DJF and the IBC index during AMJ. Black dots indicate statistical significance above the 90% level using two-sided Student’s t tests for composite anomalies, except in (b) where the sample size is too small.
Citation: Journal of Climate 37, 22; 10.1175/JCLI-D-24-0013.1
The contrasted features between ENSO and IBC are evident in the scatterplot of Fig. 10c: Distributions of green and brown circles (representing positive and negative Texas precipitation anomalies, respectively) are more distinctly separated along the vertical axis than the horizontal axis, suggesting that the IBC index is more effective in assessing the phase of precipitation anomalies. For example, among nine El Niño and nine positive IBC years, there are five and six positive Texas precipitation anomalies, respectively. Notably, these nine El Niño years include one large negative anomaly below −1.5 standard deviations, while the positive IBC years feature two large positive anomalies above 1.5 standard deviations. Similarly, among eight La Niña and 10 negative IBC years, there are four and six negative Texas precipitation anomalies, respectively, with one and two large negative anomalies below −1.5 standard deviations, respectively. This result aligns with the findings of Seager et al. (2014), which highlight the weaker precipitation anomalies during MAM due to the weak La Niña amplitude in this season, although their AMIP-type model experiments overestimate this forcing. Since the primary distinction between the IBC and ENSO indices involves the consideration of the Atlantic Ocean, our results underscore the significance of the Atlantic’s contribution to late-spring precipitation variability.
6. Discussion
We should discuss the seasonal dependencies that influence our decision to utilize the Niño-3.4 indices during DJF rather than AMJ. ENSO predominantly affects boreal winter teleconnections. By contrast, AMJ is characterized by the phase transition and weaker amplitude of ENSO, which leads to diverse impacts on Texas precipitation (Kumar and Hoerling 1998; Lee et al. 2014; Seager et al. 2014; Fernando et al. 2016). Nevertheless, using the Niño-3.4 indices during AMJ yields a statistically significant simultaneous correlation with Texas precipitation (R = 0.40 in Fig. S2a). This finding aligns with the observation that the IBC index shows a higher simultaneous correlation with the Niño-3.4 indices during AMJ (R = 0.67 in Fig. S2c), attributable to the inclusion of SST variability in the tropical Pacific. However, Texas precipitation exhibits a stronger simultaneous correlation with the IBC index (R = 0.50 in Fig. 10b) than with the Niño-3.4 indices (R = 0.40 during AMJ and R = 0.18 during DJF in Fig. 10a). Rather than focusing solely on the tropical Pacific, our findings emphasize the importance of interbasin interactions between the Pacific and Atlantic.
The challenge posed by climate model biases merits discussion. Present climate models exhibit mean-state biases in simulating the tropical Atlantic climate (Richter and Xie 2008; Luo et al. 2018; Kumar and Zhu 2018; Richter et al. 2018). This, consequently, impacts the model’s proficiency in portraying the tropical Pacific climate (McGregor et al. 2018; Kajtar et al. 2018; Chikamoto et al. 2020a). Additionally, the model shows pronounced systematic biases in replicating continental precipitation variability, attributed to shortcomings in representing the atmospheric response to distant oceanic forces (Wills et al. 2022). Our SVD analysis of observations and partial assimilation runs underscores disparities in precipitation patterns, notably in the ATL run (Fig. 4). Such model inaccuracies can compromise the predictive acumen of interbasin contrasts. Indeed, the NMME displays reduced predictive capabilities for SST variability in the tropical Atlantic compared to the central tropical Pacific (Fig. 9). The ensemble size’s influence on augmenting the predictability of these SST anomalies appears marginal; for instance, NASA-GEOSS2S (four members) outperforms CFS-v2 (24 members) as evidenced in Figs. 8 and 9. While numerous research endeavors have sought to amplify the predictability of the tropical Pacific climate (Timmermann et al. 2018), our findings underscore the need to refine climate model precision for the tropical Atlantic to further boost seasonal predictability.
In our analysis, we have primarily utilized CESM experiments to delineate the distinct contributions of the tropical Pacific and the Atlantic, yet further investigation with NMME is essential to enhance the predictability of AMJ precipitation. The statewide Texas precipitation correlation maps indicate a considerable reduction in predictive skill to around 0.5 at the 4–6-month lead time, as shown in Fig. 2f. This decline underscores the challenges of predicting drought onsets and terminations within the NMME framework (Seager et al. 2020). Conversely, the IBC index sustains higher predictive skills, surpassing 0.6 even at the 7–9-month lead time (Fig. 8b), thereby underscoring the feasibility of attaining reliable seasonal precipitation forecasts for the southern plains.
These findings lead to several follow-up research questions: How are precipitation predictive skills influenced by the phases of ENSO and the IBC, both independently and in combination? To what extent can models accurately forecast extreme rainfall events, such as those in 2011 and 2015? What mechanisms contribute to the IBC’s superior predictive skills? How do initialized atmosphere–land conditions affect precipitation predictability? Furthermore, what causes the model diversity in predicting interbasin interactions and the subsequent precipitation anomalies in the southern plains? An additional unresolved issue concerns the optimal method for evaluating predictability: Should assessments be based on the predictive skills of individual ensemble members, which reflect internal variability and SST forcing, or on the ensemble mean, which represents only the forced component? Resolving these questions could significantly advance our understanding of seasonal predictability and the factors governing precipitation extremes. From an operational standpoint, developing a simple statistical model that leverages the relationship between dynamically predicted SST variability and observed Texas precipitation would be beneficial when the climate model performance is not high enough to simulate precipitation variability.
7. Conclusions
This study investigated the oceanic influence on late-spring precipitation variability in the southern Great Plains. Through partial ocean assimilation experiments using CESM, conducted in both the equatorial Pacific (eqPAC) run and the Atlantic (ATL) run, we confirm their substantial impact on large-scale climate patterns and the resulting precipitation response in the United States. While the eqPAC run aligns with prior research emphasizing the equatorial Pacific Ocean’s significant role in global climate variability, the ATL run highlights the critical supplementary role of the Atlantic Ocean, underscoring the importance of Pacific–Atlantic interaction in enhancing predictability. Our time-series analysis estimates that the equatorial Pacific and the Atlantic contribute approximately 70% and 30%, respectively, to these interbasin contrasts. These findings suggest that considering oceanic conditions in both basins is essential to improving late-spring precipitation predictability in the southern Great Plains, a critical factor for effective water resource management and agricultural production. In line with the CESM model experiments, our evaluation of the NMME indicates that the zonal interbasin contrast of SST anomalies could exhibit higher predictability than precipitation itself. This insight, coupled with the substantial predictive skills exhibited by the NMME for the interbasin SST contrast, could pave the way for more accurate real-time forecasting products for late-spring precipitation in the southern plains. A recent study utilizing regional dynamical downscaling from the CESM GLOB run demonstrated that an enhanced horizontal resolution, coupled with a convective-permitting model, can significantly improve the simulation of May precipitation extremes in the southern plains (Chang et al. 2024). This is achieved through a more precise depiction of jet stream dynamics and convective events. Insights from this study have the potential to inform and refine future research endeavors aimed at advancing subseasonal to seasonal forecasts of extreme precipitation events, particularly through leveraging SST signals known to drive such extremes via interbasin contrasts.
Acknowledgments.
We express our deep appreciation to the three anonymous reviewers and the Editor Dr. Yuko Okumura for their constructive and insightful comments, which have greatly improved the manuscript. The CESM experiment in this paper was conducted by the University of Southern California Center for High-Performance Computing and Communications (https://hpcc.usc.edu), with high-performance computing support from Yellowstone (ark:/85065/d7wd3xhc) and Cheyenne (https://doi.org/10.5065/D6RX99HX) provided by NCAR’s Computational and Information Systems Laboratory, sponsored by the NSF. We gratefully acknowledge the support from the agencies that contribute to the NMME-Phase II system (NOAA, NSF, NASA, and DOE). Special thanks to the climate modeling groups—Environment Canada, NASA, NCAR, NOAA/GFDL, NOAA/NCEP, and University of Miami—for producing and sharing their valuable model output. NOAA/NCEP, NOAA/CTB, and NOAA/CPO provided coordinating support and led the development of the NMME-Phase II system. We also extend our appreciation to NCEP, IRI, and NCAR personnel for their assistance in creating, updating, and maintaining the NMME archive. This work is supported by the U.S. Department of Defense, Strategic Environmental Research and Development Program (SERDP, RC20-3056). Y. C. and S.-Y. W. were additionally supported by the U.S. Department of Interior, Bureau of Reclamation (R22AP00220 and R24AP00321), the U.S. Department of Energy Grant (DESC0016605), and the Utah Agricultural Experiment Station, Utah State University (approved as journal paper 9778). Y. C. is supported by the U.S. Geological Survey and Utah Center for Water Resources Research for the 104b Grant (G21AP10623-04).
Data availability statement.
GPCC and COBE2 SST sets are provided by the NOAA PSL, Boulder, Colorado, United States, from their website at https://psl.noaa.gov. JRA-55 data were provided by the Japan Meteorological Agency, and they are available through their website at https://jra.kishou.go.jp/JRA-55/index_en.html. The data from partial assimilation experiments are available from the Utah Climate Center website at https://climate.usu.edu/people/yoshi/data/2023-CPM_TX_data/data.html. Texas precipitation data were obtained from NOAA’s Climate at a Glance and can be accessed through their website at https://www.ncei.noaa.gov/access/monitoring/climate-at-a-glance/statewide/time-series. The NMME System Phase II data are available at https://www.earthsystemgrid.org/search.html?Project=NMME.
REFERENCES
Alexander, M. A., N.-C. Lau, and J. D. Scott, 2004: Broadening the atmospheric bridge paradigm: ENSO teleconnections to the tropical West Pacific-Indian Oceans over the seasonal cycle and to the North Pacific in summer. Earth Climate: Ocean-Atmosphere Interaction and Climate Variability, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 85–103, https://doi.org/10.1029/147GM05.
Balmaseda, M. A., K. Mogensen, and A. T. Weaver, 2013: Evaluation of the ECMWF ocean reanalysis system ORAS4. Quart. J. Roy. Meteor. Soc., 139, 1132–1161, https://doi.org/10.1002/qj.2063.
Barnston, A. G., and M. K. Tippett, 2017: Do statistical pattern corrections improve seasonal climate predictions in the North American Multimodel Ensemble models? J. Climate, 30, 8335–8355, https://doi.org/10.1175/JCLI-D-17-0054.1.
Baxter, L. L., C. P. West, C. P. Brown, and P. E. Green, 2021: Cover crop management on the Southern High Plains: Impacts on crop productivity and soil water depletion. Animals, 11, 212, https://doi.org/10.3390/ani11010212.
Bloom, S. C., L. L. Takacs, A. M. da Silva, and D. Ledvina, 1996: Data assimilation using incremental analysis updates. Mon. Wea. Rev., 124, 1256–1271, https://doi.org/10.1175/1520-0493(1996)124<1256:DAUIAU>2.0.CO;2.
Borovikov, A., R. Cullather, R. Kovach, J. Marshak, G. Vernieres, Y. Vikhliaev, B. Zhao, and Z. Li, 2019: GEOS-5 seasonal forecast system. Climate Dyn., 53, 7335–7361, https://doi.org/10.1007/s00382-017-3835-2.
Bretherton, C. S., C. Smith, and J. M. Wallace, 1992: An intercomparison of methods for finding coupled patterns in climate data. J. Climate, 5, 541–560, https://doi.org/10.1175/1520-0442(1992)005<0541:AIOMFF>2.0.CO;2.
Cai, W., and Coauthors, 2019: Pantropical climate interactions. Science, 363, eaav4236, https://doi.org/10.1126/science.aav4236.
Chang, H.-I., Y. Chikamoto, S. S.-Y. Wang, C. L. Castro, M. D. LaPlante, C. B. Risanto, X. Huang, and P. Bunn, 2024: Enhancing extreme precipitation predictions with dynamical downscaling: A convection-permitting modeling study in Texas and Oklahoma. J. Geophys. Res. Atmos., 129, e2023JD038765, https://doi.org/10.1029/2023JD038765.
Chikamoto, Y., and Coauthors, 2013: An overview of decadal climate predictability in a multi-model ensemble by climate model MIROC. Climate Dyn., 40, 1201–1222, https://doi.org/10.1007/s00382-012-1351-y.
Chikamoto, Y., and Coauthors, 2015: Skillful multi-year predictions of tropical trans-basin climate variability. Nat. Commun., 6, 6869, https://doi.org/10.1038/ncomms7869.
Chikamoto, Y., T. Mochizuki, A. Timmermann, M. Kimoto, and M. Watanabe, 2016: Potential tropical Atlantic impacts on Pacific decadal climate trends. Geophys. Res. Lett., 43, 7143–7151, https://doi.org/10.1002/2016GL069544.
Chikamoto, Y., A. Timmermann, M. J. Widlansky, M. A. Balmaseda, and L. Stott, 2017: Multi-year predictability of climate, drought, and wildfire in southwestern North America. Sci. Rep., 7, 6568, https://doi.org/10.1038/s41598-017-06869-7.
Chikamoto, Y., A. Timmermann, M. J. Widlansky, S. Zhang, and M. A. Balmaseda, 2019: A drift-free decadal climate prediction system for the Community Earth System Model. J. Climate, 32, 5967–5995, https://doi.org/10.1175/JCLI-D-18-0788.1.
Chikamoto, Y., Z. F. Johnson, S.-Y. S. Wang, M. J. McPhaden, and T. Mochizuki, 2020a: El Niño–Southern Oscillation evolution modulated by Atlantic forcing. J. Geophys. Res. Oceans, 125, e2020JC016318, https://doi.org/10.1029/2020JC016318.
Chikamoto, Y., S.-Y. S. Wang, M. Yost, L. Yocom, and R. R. Gillies, 2020b: Colorado River water supply is predictable on multi-year timescales owing to long-term ocean memory. Commun. Earth Environ., 1, 26, https://doi.org/10.1038/s43247-020-00027-0.
Cook, A. R., and J. T. Schaefer, 2008: The relation of El Niño–Southern Oscillation (ENSO) to winter tornado outbreaks. Mon. Wea. Rev., 136, 3121–3137, https://doi.org/10.1175/2007MWR2171.1.
Dai, A., K. E. Trenberth, and T. R. Karl, 1998: Global variations in droughts and wet spells: 1900–1995. Geophys. Res. Lett., 25, 3367–3370, https://doi.org/10.1029/98GL52511.
DelSole, T., and M. K. Tippett, 2009: Average predictability time. Part II: Seamless diagnoses of predictability on multiple time scales. J. Atmos. Sci., 66, 1188–1204, https://doi.org/10.1175/2008JAS2869.1.
Feddersen, H., A. Navarra, and M. N. Ward, 1999: Reduction of model systematic error by statistical correction for dynamical seasonal predictions. J. Climate, 12, 1974–1989, https://doi.org/10.1175/1520-0442(1999)012<1974:ROMSEB>2.0.CO;2.
Fedorov, A. V., and S. G. Philander, 2000: Is El Niño changing? Science, 288, 1997–2002, https://doi.org/10.1126/science.288.5473.1997.
Fernando, D. N., and Coauthors, 2016: What caused the spring intensification and winter demise of the 2011 drought over Texas? Climate Dyn., 47, 3077–3090, https://doi.org/10.1007/s00382-016-3014-x.
Fernando, D. N., S. Chakraborty, R. Fu, and R. E. Mace, 2019: A process-based statistical seasonal prediction of May–July rainfall anomalies over Texas and the Southern Great Plains of the United States. Climate Serv., 16, 100133, https://doi.org/10.1016/j.cliser.2019.100133.
Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 4973–4991, https://doi.org/10.1175/2011JCLI4083.1.
Ham, Y.-G., H.-J. Song, J. Jung, and G.-H. Lim, 2016: Development of the nonstationary incremental analysis update algorithm for sequential data assimilation system. Adv. Meteor., 2016, 4305204, https://doi.org/10.1155/2016/4305204.
Ham, Y.-G., Y. Chikamoto, J.-S. Kug, M. Kimoto, and T. Mochizuki, 2017: Tropical Atlantic-Korea teleconnection pattern during boreal summer season. Climate Dyn., 49, 2649–2664, https://doi.org/10.1007/s00382-016-3474-z.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hirahara, S., M. Ishii, and Y. Fukuda, 2014: Centennial-scale sea surface temperature analysis and its uncertainty. J. Climate, 27, 57–75, https://doi.org/10.1175/JCLI-D-12-00837.1.
Hong, S.-Y., and E. Kalnay, 2002: The 1998 Oklahoma–Texas drought: Mechanistic experiments with NCEP global and regional models. J. Climate, 15, 945–963, https://doi.org/10.1175/1520-0442(2002)015<0945:TOTDME>2.0.CO;2.
Hu, Q., and S. Feng, 2012: AMO- and ENSO-driven summertime circulation and precipitation variations in North America. J. Climate, 25, 6477–6495, https://doi.org/10.1175/JCLI-D-11-00520.1.
Huang, B., J. L. Kinter, and P. S. Schopf, 2002: Ocean data assimilation using intermittent analyses and continuous model error correction. Adv. Atmos. Sci., 19, 965–992, https://doi.org/10.1007/s00376-002-0059-z.
Infanti, J. M., and B. P. Kirtman, 2014: Southeastern U.S. rainfall prediction in the North American Multi-Model Ensemble. J. Hydrometeor., 15, 529–550, https://doi.org/10.1175/JHM-D-13-072.1.
Jia, L., and T. DelSole, 2011: Diagnosis of multiyear predictability on continental scales. J. Climate, 24, 5108–5124, https://doi.org/10.1175/2011JCLI4098.1.
Jia, L., and Coauthors, 2015: Improved seasonal prediction of temperature and precipitation over land in a high-resolution GFDL climate model. J. Climate, 28, 2044–2062, https://doi.org/10.1175/JCLI-D-14-00112.1.
Johnson, Z. F., Y. Chikamoto, J.-J. Luo, and T. Mochizuki, 2018: Ocean impacts on Australian interannual to decadal precipitation variability. Climate, 6, 61, https://doi.org/10.3390/cli6030061.
Johnson, Z. F., Y. Chikamoto, S.-Y. S. Wang, M. J. McPhaden, and T. Mochizuki, 2020: Pacific Decadal Oscillation remotely forced by the equatorial Pacific and the Atlantic Oceans. Climate Dyn., 55, 789–811, https://doi.org/10.1007/s00382-020-05295-2.
Kajtar, J. B., A. Santoso, S. McGregor, M. H. England, and Z. Baillie, 2018: Model under-representation of decadal Pacific trade wind trends and its link to tropical Atlantic bias. Climate Dyn., 50, 1471–1484, https://doi.org/10.1007/s00382-017-3699-5.
Khajehei, S., A. Ahmadalipour, and H. Moradkhani, 2018: An effective post-processing of the North American multi-model ensemble (NMME) precipitation forecasts over the continental US. Climate Dyn., 51, 457–472, https://doi.org/10.1007/s00382-017-3934-0.
Kirtman, B. P., and D. Min, 2009: Multimodel ensemble ENSO prediction with CCSM and CFS. Mon. Wea. Rev., 137, 2908–2930, https://doi.org/10.1175/2009MWR2672.1.
Kirtman, B. P., and Coauthors, 2014: The North American Multimodel Ensemble: Phase-1 seasonal-to-interannual prediction; phase-2 toward developing intraseasonal prediction. Bull. Amer. Meteor. Soc., 95, 585–601, https://doi.org/10.1175/BAMS-D-12-00050.1.
Kobayashi, S., and Coauthors, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–48, https://doi.org/10.2151/jmsj.2015-001.
Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305, 1138–1140, https://doi.org/10.1126/science.1100217.
Kucharski, F., F. S. Syed, A. Burhan, I. Farah, and A. Gohar, 2015: Tropical Atlantic influence on Pacific variability and mean state in the twentieth century in observations and CMIP5. Climate Dyn., 44, 881–896, https://doi.org/10.1007/s00382-014-2228-z.
Kumar, A., and M. P. Hoerling, 1998: Annual cycle of Pacific–North American seasonal predictability associated with different phases of ENSO. J. Climate, 11, 3295–3308, https://doi.org/10.1175/1520-0442(1998)011<3295:ACOPNA>2.0.CO;2.
Kumar, A., and J. Zhu, 2018: Spatial variability in seasonal prediction skill of SSTs: Inherent predictability or forecast errors? J. Climate, 31, 613–621, https://doi.org/10.1175/JCLI-D-17-0279.1.
Lee, S.-K., B. E. Mapes, C. Wang, D. B. Enfield, and S. J. Weaver, 2014: Springtime ENSO phase evolution and its relation to rainfall in the continental U.S. Geophys. Res. Lett., 41, 1673–1680, https://doi.org/10.1002/2013GL059137.
Lei, L., and J. S. Whitaker, 2016: A four-dimensional incremental analysis update for the ensemble Kalman filter. Mon. Wea. Rev., 144, 2605–2621, https://doi.org/10.1175/MWR-D-15-0246.1.
Li, X., S.-P. Xie, S. T. Gille, and C. Yoo, 2016: Atlantic-induced pan-tropical climate change over the past three decades. Nat. Climate Change, 6, 275–279, https://doi.org/10.1038/nclimate2840.
Lin, H., and Coauthors, 2020: The Canadian Seasonal to Interannual Prediction System version 2 (CanSIPSv2). Wea. Forecasting, 35, 1317–1343, https://doi.org/10.1175/WAF-D-19-0259.1.
Luo, J.-J., G. Wang, and D. Dommenget, 2018: May common model biases reduce CMIP5’s ability to simulate the recent Pacific La Niña-like cooling? Climate Dyn., 50, 1335–1351, https://doi.org/10.1007/s00382-017-3688-8.
McGregor, S., A. Timmermann, M. F. Stuecker, M. H. England, M. Merrifield, F.-F. Jin, and Y. Chikamoto, 2014: Recent Walker circulation strengthening and Pacific cooling amplified by Atlantic warming. Nat. Climate Change, 4, 888–892, https://doi.org/10.1038/nclimate2330.
McGregor, S., M. F. Stuecker, J. B. Kajtar, M. H. England, and M. Collins, 2018: Model tropical Atlantic biases underpin diminished Pacific decadal variability. Nat. Climate Change, 8, 493–498, https://doi.org/10.1038/s41558-018-0163-4.
McPhaden, M. J., 2003: Tropical Pacific Ocean heat content variations and ENSO persistence barriers. Geophys. Res. Lett., 30, 1480, https://doi.org/10.1029/2003GL016872.
McPhaden, M. J., and X. Zhang, 2009: Asymmetry in zonal phase propagation of ENSO sea surface temperature anomalies. Geophys. Res. Lett., 36, L13703, https://doi.org/10.1029/2009GL038774.
Mo, K. C., J.-K. E. Schemm, and S.-H. Yoo, 2009: Influence of ENSO and the Atlantic multidecadal oscillation on drought over the United States. J. Climate, 22, 5962–5982, https://doi.org/10.1175/2009JCLI2966.1.
Mochizuki, T., and Coauthors, 2012: Decadal prediction using a recent series of MIROC global climate models. J. Meteor. Soc. Japan, 90A, 373–383, https://doi.org/10.2151/jmsj.2012-A22.
Mueller, B., and S. I. Seneviratne, 2012: Hot days induced by precipitation deficits at the global scale. Proc. Natl. Acad. Sci. USA, 109, 12 398–12 403, https://doi.org/10.1073/pnas.1204330109.
National Integrated Drought Information System, 2021: Southern plains drought early warning system 2021–2025 strategic action plan. NIDIS Tech. Rep., 30 pp., https://www.drought.gov/documents/southern-plains-drought-early-warning-system-2021-2025-strategic-action-plan.
Nielsen-Gammon, J. W., 2012: The 2011 Texas drought. Tex. Water J., 3, 59–95, https://doi.org/10.21423/twj.v3i1.6463.
Nigam, S., B. Guan, and A. Ruiz-Barradas, 2011: Key role of the Atlantic Multidecadal Sscillation in 20th century drought and wet periods over the Great Plains. Geophys. Res. Lett., 38, L16713, https://doi.org/10.1029/2011GL048650.
NOAA, 2023: Climate at a glance: Global time series. National Centers for Environmental Information, https://www.ncei.noaa.gov/access/monitoring/climate-at-a-glance/global/time-series.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699–706, https://doi.org/10.1175/1520-0493(1982)110<0699:SEITEO>2.0.CO;2.
Pu, B., R. Fu, R. E. Dickinson, and D. N. Fernando, 2016: Why do summer droughts in the Southern Great Plains occur in some La Niña years but not others? J. Geophys. Res. Atmos., 121, 1120–1137, https://doi.org/10.1002/2015JD023508.
Purich, A., and Coauthors, 2016: Tropical Pacific SST drivers of recent Antarctic sea ice trends. J. Climate, 29, 8931–8948, https://doi.org/10.1175/JCLI-D-16-0440.1.
Rasmusson, E. M., and K. Mo, 1993: Linkages between 200-mb tropical and extratropical circulation anomalies during the 1986–1989 ENSO cycle. J. Climate, 6, 595–616, https://doi.org/10.1175/1520-0442(1993)006<0595:LBMTAE>2.0.CO;2.
Raz-Yaseef, N., D. P. Billesbach, M. L. Fischer, S. C. Biraud, S. A. Gunter, J. A. Bradford, and M. S. Torn, 2015: Vulnerability of crops and native grasses to summer drying in the U.S. Southern Great Plains. Agric. Ecosyst. Environ., 213, 209–218, https://doi.org/10.1016/j.agee.2015.07.021.
Richter, I., and S.-P. Xie, 2008: On the origin of equatorial Atlantic biases in coupled general circulation models. Climate Dyn., 31, 587–598, https://doi.org/10.1007/s00382-008-0364-z.
Richter, I., T. Doi, S. K. Behera, and N. Keenlyside, 2018: On the link between mean state biases and prediction skill in the tropics: An atmospheric perspective. Climate Dyn., 50, 3355–3374, https://doi.org/10.1007/s00382-017-3809-4.
Ropelewski, C. F., and M. S. Halpert, 1986: North America precipitation and temperature associated with the El Niño/Southern Oscillation (ENSO). Mon. Wea. Rev., 114, 2352–2362, https://doi.org/10.1175/1520-0493(1986)114<2352:NAPATP>2.0.CO;2.
Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115, 1606–1626, https://doi.org/10.1175/1520-0493(1987)115<1606:GARSPP>2.0.CO;2.
Rukhovets, L., H. M. van den dool, and A. G. Barnston, 1998: Forecast-observation pattern relationships in NCEP medium range forecasts of non-winter Northern Hemisphere 500-mb height fields: Research note. Atmos.–Ocean, 36, 55–70, https://doi.org/10.1080/07055900.1998.9649606.
Ruprich-Robert, Y., R. Msadek, F. Castruccio, S. Yeager, T. Delworth, and G. Danabasoglu, 2017: Assessing the climate impacts of the observed Atlantic multidecadal variability using the GFDL CM2.1 and NCAR CESM1 global coupled models. J. Climate, 30, 2785–2810, https://doi.org/10.1175/JCLI-D-16-0127.1.
Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 2185–2208, https://doi.org/10.1175/JCLI-D-12-00823.1.
Schneider, U., A. Becker, P. Finger, A. Meyer-Christoffer, B. Rudolf, and M. Ziese, 2015: GPCC full data monthly product version 7.0 (at 0.5°, 1.0°, 2.5°): Monthly land-surface precipitation from rain-gauges built on GTS-based and historic data. Global Precipitation Climatology Centre, accessed 9 October 2024, https://doi.org/10.5676/DWD_GPCC/FD_M_V7_050.
Schubert, S., and Coauthors, 2009: A U.S. CLIVAR project to assess and compare the responses of global climate models to drought-related SST forcing patterns: Overview and results. J. Climate, 22, 5251–5272, https://doi.org/10.1175/2009JCLI3060.1.
Schubert, S. D., M. J. Suarez, P. J. Pegion, R. D. Koster, and J. T. Bacmeister, 2004: Causes of long-term drought in the U.S. Great Plains. J. Climate, 17, 485–503, https://doi.org/10.1175/1520-0442(2004)017<0485:COLDIT>2.0.CO;2.
Seager, R., and M. Hoerling, 2014: Atmosphere and ocean origins of North American droughts. J. Climate, 27, 4581–4606, https://doi.org/10.1175/JCLI-D-13-00329.1.
Seager, R., L. Goddard, J. Nakamura, N. Henderson, and D. E. Lee, 2014: Dynamical causes of the 2010/11 Texas–Northern Mexico drought. J. Hydrometeor., 15, 39–68, https://doi.org/10.1175/JHM-D-13-024.1.
Seager, R., J. Nakamura, and M. Ting, 2020: Prediction of seasonal meteorological drought onset and termination over the southern Great Plains in the North American Multimodel Ensemble. J. Hydrometeor., 21, 2237–2255, https://doi.org/10.1175/JHM-D-20-0023.1.
Shields, C. A., D. A. Bailey, G. Danabasoglu, M. Jochum, J. T. Kiehl, S. Levis, and S. Park, 2012: The low-resolution CCSM4. J. Climate, 25, 3993–4014, https://doi.org/10.1175/JCLI-D-11-00260.1.