1. Introduction
The rainfall over most of the earth’s land areas has a strong impact on the environment and local economy. Summer precipitation over the southeastern United States is influenced by synoptic-scale convective activity (Baigorria et al. 2007), landfalling tropical cyclones, and the North Atlantic subtropical high (Davis et al. 1997; Li et al. 2012, 2013). Despite the proximity to the Atlantic Ocean and numerous links between the sea surface temperature (SST) variability over the Pacific Ocean and the U.S. continent, summer precipitation in the southeastern United States is not strongly associated with the tropical Pacific or Atlantic SST anomalies but relies on southerly flow anomalies from the Gulf of Mexico into the region (Wang et al. 2010a,b). These anomalies are related to the internal atmospheric dynamics and thermodynamics (Seager et al. 2009).
Other than atmospheric processes, local land–atmosphere interactions, such as soil moisture content and latent heat flux, show a strong relationship with precipitation (Hohenegger et al. 2009; Sun and Wang 2012; Taylor et al. 2013; Froidevaux et al. 2014). Whether a wetter surface enhances (positive feedback) or reduces (negative feedback) the local rainfall depends on many factors. In numerical simulations, the soil moisture–precipitation feedback is sensitive to the spatial resolution and convective parameterization (Hohenegger et al. 2009; Taylor et al. 2013). Hohenegger et al. (2009) found that in the Alpine region the positive soil moisture–precipitation feedback simulated by a low-resolution, parameterized convection model became negative in the high-resolution, resolved convection version of the same model. For the Sahel region, Taylor et al. (2013) found the same positive feedback bias in the models with parameterized convection, and increasing the resolution of models with parameterized convection did not change the sign of the feedback. These studies attribute the successful simulation of the soil moisture–precipitation feedback in the convection-permitting models to the better simulation of the diurnal cycle. The superparameterization of cloud processes (Grabowski 2001; Khairoutdinov and Randall 2001) in a global climate model has also shown significant improvements in the representation of the diurnal cycle of precipitation over land in general (e.g., Khairoutdinov et al. 2005; DeMott et al. 2007; Pritchard and Somerville 2009) and the contiguous United States in particular (Dirmeyer et al. 2012). The superparameterization framework also belongs to the class of convection-permitting models. In this framework, convection and stratiform clouds are simulated by a cloud-process-resolving model (CRM) embedded in each grid column of the atmospheric GCM.
In a warmer climate, the models from phase 5 of the Coupled Model Intercomparison Project (CMIP5; Taylor et al. 2013) project an intensified variability of summer rainfall in the southeastern United States (Li et al. 2013). The projected increase of precipitation variability in the multimodel ensemble seems to be explained by the changes in the pattern and position of the North Atlantic subtropical high. While the regional response in the hydrological cycle is correlated with the dynamical interactions between the region and the large-scale circulation, local soil moisture–precipitation feedbacks can also alter these interactions. In all the CMIP5 models the formation and evolution of cloud processes are parameterized, and convective parameterizations can be biased toward the positive feedback (Hohenegger et al. 2009).
In this study, climate projections with a model with explicit representation of cloud processes are evaluated to explore the precipitation changes in the southeastern United States in the high-emissions RCP8.5 climate change scenario. The model is the superparameterized Community Climate System Model, version 4 (SP-CCSM4; Stan and Xu 2014). In SP-CCSM4, cloud processes are represented by a 2D CRM embedded in each grid column of the atmospheric model. The details of the CRM’s configuration are given in Khairoutdinov and Randall (2001, 2003). Climate change studies comparing simulations with the superparameterized and conventional CCSM4 show various regions where the two models produce different results (Zhu et al. 2014; Arnold et al. 2014). Therefore, the second objective of the paper is to compare the projection of precipitation in the southeastern United States between the superparameterized and conventional CCSM4 in the RCP8.5 experiments and explore possible reasons for any differences in the precipitation responses of the two models.
Hence, this paper is organized as follows. Datasets and methods are described in section 2. Section 3 evaluates the climatology of precipitation over the southeastern United States simulated in present-day conditions by the two models. Section 4 compares the projection of precipitation in SP-CCSM4 and CCSM4 and explores the causes responsible for the different response of precipitation to the climate change scenario in the two models. Two types of factors are considered: the influence of large-scale circulation on the southeastern U.S. precipitation and the impact of local land–atmosphere interactions such as soil moisture content and latent heat flux on precipitation. Section 5 discusses the relative importance of large-scale circulation versus local factors in influencing southeastern U.S. precipitation. Finally, conclusions are summarized in section 6.
2. Data and method
The observational dataset chosen for monthly precipitation is the Global Precipitation Climatology Project (GPCP) from January 1979 to December 2013 with a horizontal resolution of 2.5° latitude × 2.5° longitude on a global grid (Adler et al. 2003). The numerical models used in this study are the standard Community Climate System Model version 4.0 (Gent et al. 2011) and the superparameterized CCSM4 (Stan and Xu 2014). The horizontal resolution of both models is 0.9° latitude × 1.25° longitude for the atmospheric component and 1° latitude × 1° longitude for the ocean model. The main difference between the two models is the parameterization of moist convection. In CCSM4, convection is represented by the conventional parameterization schemes described in Neale et al. (2013), while SP-CCSM4 uses an explicit, 2D representation of cloud processes (e.g., Grabowski 2001; Khairoutdinov and Randall 2001; Grabowski 2004).
The control experiments for CCSM4 and SP-CCSM4 are initialized from a historical run of the conventionally parameterized CCSM4, and a 50-yr period is analyzed for each model. The atmospheric CO2 concentration in the control simulations is fixed at the same value (368.9 ppm). In the control experiment, the analyzed period is from model year 2001 to 2050 for CCSM4, whereas for SP-CCSM4, it is from model year 2006 to 2055.
The global warming experiments are carried out based on the protocol of the representative concentration pathway 8.5 (RCP8.5) scenario, in which the external forcing, such as the concentration of greenhouse gases, changes to yield an increase in radiative forcing of 8.5 W m−2 (relative to preindustrial values) in 2100 (Moss et al. 2010). Both SP-CCSM4 and CCSM4 have three ensemble members of 100 yr for the RCP8.5 experiments, with the same initial conditions and external forcing. The analysis in this paper is carried out on the ensemble mean of each model. In the RCP8.5 experiment, the analyzed period is between year 2006 and 2100 for both SP-CCSM4 and CCSM4.
The summer season is defined as June–August (JJA), and the domain of interest is the rectangular area (30°–40°N, 75°–85°W) in the southeastern United States; hereafter, the southeastern United States refers to this rectangle. Note that points over the ocean covered by the rectangle are not included in the analysis.
The projected changes of precipitation and other atmospheric variables as a result of the global warming scenario are represented as differences in the summer climatology between the first 20 yr and the last 20 yr. The relationship between precipitation and its influencing factors is investigated using the simultaneous correlation analysis and multivariate empirical orthogonal function (EOF) analysis (Kutzbach 1967; Wheeler and Hendon 2004). The monthly anomalies in the control runs are calculated by removing the annual cycle based on the climatological mean of each model. In the RCP8.5 experiment, the anomalies also have the linear trend of the whole analyzed period (2006–2100) removed. The multivariate EOF analysis is adopted to identify the primary modes of the combined variance among precipitation, local factors, and large-scale circulation factors. This analysis was further used to compare the relative influence of the local and large-scale factors onto the projected precipitation changes.
3. Simulation of precipitation in the control runs
Figure 1 shows the climatology of JJA rainfall simulated in the 50-yr control run by SP-CCSM4 (Fig. 1a), CCSM4 (Fig. 1b), and that in the observations (Fig. 1c). In the observations the maximum rainfall is located over Florida, with the seasonal mean reaching 6 mm day−1 (Fig. 1c). The spatial gradient of JJA precipitation shows decreasing rainfall from the coastal region to inland, where the seasonal mean is less than 6 mm day−1 [e.g., West Virginia, Virginia, and Kentucky (Fig. 1c)].
The climatology of JJA precipitation rate (mm day−1) in (a) SP-CCSM4, (b) CCSM4, and (c) observations (GPCP). (d) Time series of the regional average precipitation rate (mm day−1) over the rectangular region (30°–40°N, 75°–85°W) in (a)–(c). The black, red, and blue lines represent observations, the control simulation for SP-CCSM4, and the control simulation for CCSM4, respectively. In (a),(b) for the rectangular area, RMSE stands for the root-mean-square error between simulated precipitation and observation and R represents the spatial correlation between simulated precipitation and observation.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
Comparing to observations, SP-CCSM4 simulates drier summer precipitation climatology over the whole southeastern United States (Fig. 1a). The maximum value over West Virginia and Kentucky is about 4 mm day−1. In SP-CCSM4 the maximum over Florida is less than 3 mm day−1, which is about half of the observed climatology (Fig. 1a). Moreover, in SP-CCSM4, the JJA rainfall distribution pattern shows a small zonal gradient from the coast to the inland.
The climatology of JJA rainfall simulated in CCSM4 shows a different pattern from both observations (Fig. 1c) and SP-CCSM4 (Fig. 1b). The summer precipitation simulated in CCSM4 shows a maximum of 5.5 mm day−1 over South Carolina (Fig. 1b). The precipitation in Florida simulated in CCSM4 is also less than in observations but more than that in SP-CCSM4. In the CCSM4 model, regions like Georgia, Tennessee, and North Carolina also experience larger than observed rainfall (Fig. 1b).
Considering the climatology pattern, Florida is excluded from the analysis as both models show large biases of underestimated precipitation. Figure 1d shows the time series of precipitation averaged over the rectangular area. The 35-yr climatology of observed precipitation is about 4.2 mm day−1, shown as the thin black lines in Fig. 1d. Comparing to observations, over the rectangular area, SP-CCSM4 underestimates precipitation (red lines in Fig. 1d) and CCSM4 overestimates precipitation (blue lines in Fig. 1d). These differences between the observations and the control runs are statistically significant based on the two-sample t test with the significance level of 5%.
The comparison between simulated and observed precipitation climatology shows biases in both models. The root-mean-square error and anomaly correlation averaged over the rectangular area are 0.84 and 0.41 mm day−1 for CCSM4 and 1.01 and 0.37 mm day−1 for SP-CCSM4. These results suggest that both models have comparable skill in capturing the observed features of the summer precipitation over the southeastern United States. The anomaly correlation between the two models is 0.9.
4. Projection of precipitation in the RCP8.5 experiment
Earlier studies (e.g., Palmer et al. 2008; Davy and Esau 2014) showed that model biases represent a significant source of uncertainty and could impact the reliability of climate change projections. However, assuming that model bias is the background model drift, a climate change measured by the difference in climatology between the first 20 yr and the last 20 yr is free of the background overall mean bias. The caveat here is that model biases might not have a linear response to external forcing or to temperature-mediated feedbacks.
a. Projected changes in precipitation
In the global warming scenario, under the extreme RCP8.5 external forcing, the two models evolve in opposite directions (Fig. 2). The superparameterized model, SP-CCSM4, projects a homogeneous decrease (about 1 mm day−1) of precipitation across the southeastern United States to the central United States. (Fig. 2a). The decreased precipitation over the southeastern United States is statistically significant at the 5% significance level. The conventionally parameterized model, CCSM4, projects an increase of summer precipitation over the southeastern United States by the end of twenty-first century (Fig. 2b). The CCSM4 model projection shows a maximum increase of 1 mm day−1 in North Carolina and smaller values over the other parts of the rectangle (Fig. 2b).
The projected changes in the JJA precipitation rate climatology (mm day−1, average of the last 20 − the first 20 yr) in (a) SP-CCSM4 and (b) CCSM4. Color shading represents precipitation changes exceeding the 5% significance level based on the t test. Solid (dash dotted) lines represent positive (negative) changes. (c) The regional ensemble average precipitation over the region 30°–40°N, 75°–85°W in (a),(b) from year 2006 to 2100: RCP8.5 experiments in SP-CCSM4 (red) and in CCSM4 (blue). The horizontal straight line segments are climatology of the first and last 10 years.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
The spatial average over the rectangular area of the JJA rainfall shows a strong decreasing trend in SP-CCSM4 (red lines in Fig. 2c) but a relatively weaker increasing trend in CCSM4 (blue lines in Fig. 2c). While the changes of precipitation projected by CCSM4 are statistically significant only over a limited area of the rectangle (red shading in Fig. 2b), the changes in the regional averages are statistically significant.
Another noteworthy aspect of the precipitation projections is the change in the intensity, which can be seen in the histogram of precipitation intensity. The distributions of the southeastern United States precipitation are compared between the first 20 yr and the last 20 yr in SP-CCSM4 (Fig. 3a) and in CCSM4 (Fig. 3b). Over the rectangular region (30°–40°N, 75°–85°W), there are 90 grid points with 60 monthly values for the 20 summer seasons (JJA). Hence the total sample size for each 20-yr period is 5400. In SP-CCSM4, the distribution shifts toward lower precipitation rates in the last 20 yr of the RCP8.5 experiment (gray bars in Fig. 3a), indicating more frequent occurrence of low precipitation rates and less frequent occurrence of high precipitation rates. With a less frequent occurrence of intense precipitation, the JJA seasonal precipitation decreases in the last 20 yr of the projection by SP-CCSM4 (Fig. 2a). Conversely, in CCSM4, the shift of precipitation is toward more frequent occurrence of high precipitation rates at the end of the twenty-first century (gray bars in Fig. 3b), which is associated with the increased precipitation climatology over the rectangular area (Fig. 2b).
Histogram of summer monthly precipitation rate (mm day−1) over the rectangular region (30°–40°N, 75°–85°W) for the first 20 (black bars) and the last 20 yr (gray bars) in (a) SP-CCSM4 and (b) CCSM4.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
It is interesting to notice that in the superparameterized CAM, version 5, Kooperman et al. (2014) found that over the central and eastern United States the summer precipitation on daily time scales shifts toward high precipitation rates when the CO2 forcing is doubled. We can think of three possible explanations. First, there are differences between the radiation schemes of the host models and the aerosol–cloud interactions in the superparameterization of CAM, version 5; SP-CCSM4 does not include a treatment of the cloud–aerosol interactions. Second, the coupling between the atmosphere and ocean can also have an influence, and we will address this element in the next section. The other possibility is the time scale of analyzed variability. There can be enhanced but less frequent precipitation on daily time scales, which after averaging translates to less precipitation on monthly time scales projected in the superparameterized model.
b. The influence of large-scale circulation on precipitation
Li et al. (2012) showed that the western ridge of the North Atlantic subtropical high affects the regional precipitation over the southeastern United States by regulating the vertical motion and moisture transport in the area. Hence, as an effective factor for summer precipitation in the Southeast, the projection of the subtropical high is first analyzed in order to understand the different responses of the SP-CCSM4 and CCSM4 models to the global warming.
Figure 4 shows the projected change of sea level pressure (SLP), which is directly associated with the subtropical high. In SP-CCSM4, at the end of the twenty-first century, SLP increases significantly over the eastern United States, especially over the southeastern United States (Fig. 4a). From Georgia northeastward to West Virginia the magnitude of the change is about 80 Pa. This indicates that the subtropical high has a stronger intrusion into the southeastern United States. The high pressure systems prevailing over the southeastern United States in summer tend to inhibit vertical motions and maintain clear sky. The lack of clouds favors the warming of the surface by the downward solar radiation; the surface temperature increase in SP-CCSM4 is slightly larger than in CCSM4 (not shown). These conditions could explain the reduced precipitation projected in SP-CCSM4.
The projected changes in the JJA SLP climatology (Pa, climatology of the last 20 − the first 20 yr) in (a) SP-CCSM4 and (b) CCSM4. Color shading represents changes of SLP climatology exceeding the 5% significance level.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
CCSM4 projects an increase of SLP over the southern part of the southeastern United States and a decrease of SLP over the northern part of the southeastern United States (Fig. 4b). In the last 20 yr of the RCP8.5 experiment, SLP decreases over Virginia, West Virginia, North Carolina, and Kentucky but increases over Georgia and South Carolina in the projections of CCSM4. The decrease in the seasonal-mean SLP over Virginia and North Carolina indicates a weakening of the high pressure systems, which could further be associated with enhanced clouds and increased precipitation projected in CCSM4. Notice that changes of SLP in CCSM4 are not statistically significant over the southeastern United States. This indicates that the projected changes in the subtropical high have a much weaker impact on the precipitation over the southeastern United States in CCSM4 than in SP-CCSM4 where changes in SLP are statistically significant.
The changes in SLP mentioned above can impact the subtropical high in two ways: one is the effect on the strength of the subtropical high, and the other is the zonal and/or meridional expansion or intrusion of the subtropical high. Both models project an increase of 1–2 hPa in the strength of the North Atlantic subtropical high maximum (not shown). However, based on the 1020-hPa isobar of SLP, which is used as the boundary of the subtropical high (Davis et al. 1997), the position of the subtropical high is projected differently between the two models (Fig. 5).
The projected positions of the subtropical high in (a) SP-CCSM4 and (b) CCSM4 for the first- (blue lines) and the last-20-yr climatology (red lines). Thick lines represent the boundary of the subtropical high (the 1020-hPa isobar of 20-yr-averaged JJA SLP). Thin lines (blue and red) represent the ridge line of the subtropical high (the zero zonal wind line of the 20-yr-averaged zonal wind). Red diamonds represent the west ridge points (intersection of 1020-hPa isobar and the zero wind line) during the last-20-yr periods. Blue asterisks represent the west ridge points during the first-20-yr period. Vectors represent the projected changes in the horizontal wind (m s−1) at 925 hPa (the last-20-yr minus the first-20-yr climatology).
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
Figure 5a illustrates the boundary of the subtropical high based on the last 20 yr (thick red line) and the first 20 yr (thick blue line) climatological-mean SLP in SP-CCSM4. In response to global warming, the boundary of the subtropical high shifts northward into the southeastern United States in SP-CCSM4. The northward intrusion directly contributes to the increase of SLP in the rectangular area. Moreover, the ridge of the subtropical high (the zero zonal wind line of the 20-yr-averaged zonal wind, shown as thin lines in Fig. 5) indicates a northward displacement of 1° in the last 20 yr in SP-CCSM4. In the zonal direction, the western ridge points (intersections between the boundary line and ridge line), depicted as diamonds for each year in the last-20-yr period and stars for that in the first 20 yr, show a northwest displacement of the subtropical high.
In contrast to SP-CCSM4, there is weaker northward intrusion of the subtropical high into the southeastern United States projected in CCSM4 (Fig. 5b). In the first 20 yr, the boundary of the subtropical high reaches only the south part of Florida (thick blue line in Fig. 5b). Under the strong external forcing, the poleward boundary of the subtropical high projected in CCSM4 (thick red line in Fig. 5b) is still located in Florida and does not show a strong northward displacement into the southeastern United States. Compared to SP-CCSM4, the changes of the ridge line of the subtropical high are smaller in CCSM4 (thin lines in Fig. 5b). However, considering the location of the western ridge points of the subtropical high (diamonds and stars in Fig. 5b), there is a stronger westward movement of the subtropical high projected in CCSM4 than that in SP-CCSM4.
The wind vectors in Fig. 5 represent changes in the wind climatology at 925 hPa (the last 20 yr minus the first 20 yr). Consistent with the northward movement of the subtropical high, SP-CCSM4 projects enhanced southerly wind (Fig. 5a), which provides a vigorous moisture transport into the Great Plains but a relatively weaker moisture transport into the southeastern United States. However, in CCSM4, as a result of the weaker northward intrusion of the subtropical high into the rectangular area, the change in the wind vector is dominated by enhanced eastward wind over the rectangular area (Fig. 5b). As a result, this anomalous eastward wind in CCSM4 could transport more moisture from the Gulf of Mexico into the southeastern United States, which would provide favorable conditions for precipitation in the conventionally parameterized model.
Considering the influence of large-scale circulation on summer precipitation over the southeastern United States, the northward intrusion of the subtropical high projected in SP-CCSM4 is stronger than the CCSM4’s projection. The northward intrusion of the subtropical high in SP-CCSM4 is associated with a significant increase of SLP, which inhibits upward motion over the southeastern United States, and weaker moisture transport into the rectangular area. These conditions further contribute to the reduced precipitation projected in SP-CCSM4.
Before we analyze the impact of local effects on the precipitation patterns projected by CCSM4 and SP-CCSM4, we would like to revisit the contrasting results between the uncoupled superparameterized CAM (SP-CAM), version 5, and SP-CCSM4 in the context of the North Atlantic subtropical high. The existence of the subtropical high is related to the SST pattern and the thermal land–sea contrast (Miyasaka and Nakamura 2005). Therefore, in an uncoupled simulation in which only the atmospheric CO2 concentration is increased but the SST is fixed, the response of the subtropical high to global warming can be different than in a simulation with all the feedbacks active. For instance, in an uncoupled simulation the land surface temperature responds to the changes in the atmosphere and the land–sea thermal contrast will be different than in a coupled model.
c. The influence of land–atmosphere interaction on the precipitation changes
The large-scale circulation plays an important role in modulating precipitation in the free troposphere. However, the initiation and development of convection are closely related to the planetary boundary layer (PBL). Hence, in addition to large-scale factors, local effects, such as soil moisture content and latent heat flux, are further analyzed in this section. The effect of soil moisture on rainfall is complicated because on one hand the availability of soil moisture determines the evaporation at the surface and hence the latent heat, and on the other hand moisture precipitating over land can have an influence on the rainfall if it evaporates back into the atmosphere. The projection of the soil moisture–rainfall feedbacks in CCSM4 and SP-CCSM4 will be evaluated in this section. The analysis of the hydrological cycle includes changes in the soil moisture content and the relationship between the evaporation at the surface and rainfall variability.
Figures 6a and 6b show the correlation of precipitation with the soil moisture content in the RCP8.5 experiment for SP-CCSM4 and CCSM4, respectively. The linear trend of precipitation and soil moisture content induced by the transient external forcing is not included in the correlation analysis. Both SP-CCSM4 (Fig. 6a) and CCSM4 (Fig. 6b) show significant positive soil moisture–precipitation feedback in their projections. In other words, a wet (dry) surface corresponds to a positive (negative) precipitation anomaly in both models.
The correlation coefficients (colored shading) between projected JJA soil moisture content (kg m−2) and projected JJA precipitation rate (mm day−1) in the RCP8.5 experiments in (a) SP-CCSM4 and in (b) CCSM4. Dotted areas represents statistically significant correlations exceeding the 5% significance level. (c),(d) Projected changes in soil moisture content (kg m−2; climatology of the last 20 yr minus first 20 yr) in SP-CCSM4 and CCSM4, respectively. Color shading in (c) and (d) represents changes statistically significant at the 5% level.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
Although both models project a positive soil moisture–precipitation feedback, the projected changes in the soil moisture content are different between SP-CCSM4 (Fig. 6c) and CCSM4 (Fig. 6d). SP-CCSM4 projects a significant decrease of soil moisture content over the rectangle, except along the east coast of Virginia, North Carolina, and South Carolina where the soil moisture increases in the last 20 yr of the RCP8.5 experiment (Fig. 6c). Conversely, CCSM4 projects a significant increase of soil moisture content over the southeastern United States (Fig. 6d). Hence, with the positive soil moisture–precipitation feedback in both models, the decreased soil moisture content is associated with decreased precipitation in SP-CCSM4, while the increased soil moisture content is accompanied by increased precipitation in CCSM4.
The soil moisture–precipitation feedback is maintained through the dynamics and thermodynamics of the surface fluxes. The soil moisture affects the partition between the sensible and latent heat fluxes in the boundary layer (Nemias 1952). The ratio between the two fluxes affects the stability of the boundary layer, which controls the convective development. Hence, changes in latent heat flux and sensible heat flux are further analyzed in Fig. 7. Both models project a decrease in the latent heat flux in response to the external forcing. SP-CCSM4 shows on average a significant decrease on the order of 10 W m−2 in the last 20 yr (Fig. 7a). The decrease in latent heat flux projected by CCSM4 is weaker than that in SP-CCSM4 and is not statistically significant over a large portion of the southeastern United States (Fig. 7c).
The projected changes in the JJA latent heat flux climatology (W m−2) in (a) SP-CCSM4 and (c) CCSM4. (b),(d) As in (a),(c), but for sensible heat. Color shading represents changes statistically significant at the 5% level.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
In the case of sensible heat flux, both models project an increase in the last 20 yr: about 20 W m−2 in SP-CCSM4 (Fig. 7b) and about 12 W m−2 in CCSM4 (Fig. 7d). Considering the increase of sensible heat flux in both models, the stronger (weaker) reduction of latent heat flux in SP-CCSM4 (CCSM4) contributes to the stronger (weaker) increase of the Bowen ratio, indicating a deeper (shallower) boundary layer and a reduced (increased) potential for convective development, which further contribute to reduced (enhanced) precipitation.
These results suggest that changes in latent heat flux play an important role in explaining the projection of summer precipitation. The relationship between the latent heat flux and precipitation is further evaluated between SP-CCSM4 (Fig. 8a) and CCSM4 (Fig. 8b). In SP-CCSM4, there is a significant positive correlation between the latent heat flux and precipitation dominating the southeastern United States (Fig. 8a). This indicates that the enhanced latent heat flux or evaporation at the surface is positively correlated with the rainfall, and vice versa in SP-CCSM4. In contrast to SP-CCSM4, the latent heat flux and precipitation are negatively correlated over the southeastern United States in CCSM4, although the correlation does not pass the significance test (Fig. 8b). In CCSM4, the increase in soil moisture in response to surface warming does not enhance the surface evaporation, and the increase in precipitation cannot be explained as a local response to balance the enhanced precipitation.
As in Figs. 6a and 6b, but for latent heat flux (W m−2). (c),(d) As in (a),(b), but for the simulated latent heat in the control. Dotted area represents the statistically significant correlation passing the 5% significance level.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
One can easily argue that the correlation between the latent heat flux and precipitation depends on the parameterization of cloud processes and that global warming does not cause the different response we see here. The correlations between the latent heat flux and precipitation anomalies for the control run of SP-CCSM4 (Fig. 8c) and CCSM4 (Fig. 8d) show patterns consistent with the climate change runs, although the northwest part of the rectangular area shows a stronger correlation.
The analysis of the hydrological cycle over the southeastern United States suggests that in SP-CCSM4 the warming of the surface reduces the relative humidity of the atmosphere and reduces the rainfall, whereas in CCSM4 the convective parameterization responds directly to the surface warming regardless of the energy available at the surface for evaporation. It is also possible that in CCSM4 the amount of water vapor in the atmosphere is supplied by the large-scale transport of moisture into the region. The relative importance of remote versus local effects is the subject of the next section.
5. Discussion
To compare the relative impact of large-scale and local factors, a multivariate EOF analysis is first carried out on a combined dataset composed of precipitation, latent heat flux (which represents the local influence on precipitation), and SLP (which represents the influence from the subtropical high).
First, multi-EOF analyses are performed for the 50-yr control simulations in SP-CCSM4 and CCSM4. Based on North’s rule of thumb (North et al. 1982), the first leading EOF modes in both SP-CCSM4 and CCSM4 are uniquely separated from the other modes (figure not shown). In SP-CCSM4, the correlation coefficients among the first leading principal components (PC1s) of these patterns are 0.71 between precipitation and latent heat flux and 0.35 between precipitation and SLP (Table 1), which are both statistically significant. These values of the correlation coefficients indicate that, for the control simulation of SP-CCSM4, the influence of local effects on precipitation are relatively stronger than that of the large-scale effects. The correlation coefficients among PC1s of the control run in CCSM4 are also significant, but the local effect plays a similar role in modulating precipitation as the large-scale factor. The correlations are 0.62 for latent heat flux and 0.59 for SLP in the control run of CCSM4 (Table 1).
The correlation coefficients between the PC1s of latent heat flux (LH) and SLP with the PC1 of precipitation from the multivariate EOF analysis in SP-CCSM4 and CCSM4 for the control and RCP8.5 experiments. Numbers in parentheses represent the p values of the correlation coefficient.
In both SP-CCSM4 (Fig. 9a) and CCSM4 (Fig. 9d), the pattern of the first EOF mode for precipitation shows increased precipitation over the west side but decreased precipitation over the east side of the rectangular domain. This precipitation pattern is accompanied by increased SLP in both SP-CCSM4 (Fig. 9c) and CCSM4 (Fig. 9f). This indicates that when the subtropical high extends westward over the continent, the precipitation is suppressed over the southeastern United States. A large difference between the two models stands out in the pattern of the latent heat flux. SP-CCSM4 shows a negative latent heat flux anomaly pattern (Fig. 9b), whereas CCSM4 is dominated by a positive latent heat flux anomaly (Fig. 9a). These results combined with the correlations between precipitation and latent heat flux (Figs. 8c,d) explain the precipitation pattern for each model.
The first leading EOF mode for the simulated (a),(d) precipitation rate (mm day−1), (b),(e) latent heat flux (W m−2), and (c),(f) SLP (Pa) in SP-CCSM4 and CCSM4, respectively, in the control experiment.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
In previous sections, we found that the two models project a different response of the subtropical high and soil moisture–latent heat flux for the climate change scenario. Here, we will compare the relative importance of the impact from the local and large-scale effects on summer precipitation in the global warming scenario, based on the multi-EOF analysis of the 95-yr RCP8.5 experiment. Because the RCP8.5 experiment is a transient run, the long-term linear trend is excluded from the multivariate EOF analysis. In SP-CCSM4, the pattern of the first EOF mode of precipitation reveals increased precipitation in Kentucky, West Virginia, and western North Carolina but decreased precipitation in Georgia, South Carolina, and along the U.S. East Coast (Fig. 10a). This pattern of precipitation in SP-CCSM4 is accompanied by enhanced (positive anomaly) latent heat flux (Fig. 10b) and increased (positive anomaly) SLP over the southeastern United States (Fig. 10c). In CCSM4, the pattern of precipitation is similar to that in SP-CCSM4, with enhanced precipitation to the west and decreased precipitation along the eastern coast (Fig. 10d). However, the pattern of precipitation in CCSM4 is accompanied by reduced (negative anomaly) latent heat flux (Fig. 10e) and increased (positive anomaly) SLP in CCSM4 (Fig. 10f).
As in Fig. 9, but for the projected precipitation rate, latent heat flux SLP in the RCP8.5 experiment.
Citation: Journal of Climate 28, 20; 10.1175/JCLI-D-14-00765.1
The difference between the patterns of latent heat flux in the RCP8.5 experiments indicates that mechanisms used to convert the evaporation at the surface into precipitation have a different response to the extreme external forcing in SP-CCSM4 than in CCSM4. The opposite signs of the latent heat flux patterns also indicate that the relationship between latent heat flux and precipitation in the RCP8.5 experiment is different in SP-CCSM4 than in CCSM4. The negative latent heat flux pattern in CCSM4 has the potential of generating a warm surface temperature anomaly, which is likely to trigger the convective parameterization to produce precipitation. In SP-CCSM4, the positive latent heat flux anomaly leads to cooling of the surface and an increase in surface pressure, conditions favorable to subsidence rather than precipitation.
Compared to the control simulation, in the RCP8.5 experiment, the latent heat flux plays a relatively larger role in modulating the precipitation than the subtropical high in both SP-CCSM4 and CCSM4 (Table 1). In the RCP8.5 experiment, the correlation coefficients between PC1 of latent heat flux and PC1 of precipitation are 0.80 in SP-CCSM4 and 0.75 in CCSM4, respectively. But the correlation coefficient between PC1 of SLP and PC1 of precipitation is 0.47 in SP-CCSM4 and 0.40 in CCSM4 (Table 1). Based on the correlation analysis of the principal component time series, we speculate that the influence of the surface latent heat flux on precipitation could be twice the influence of the subtropical high on precipitation in SP-CCSM4 and CCSM4.
In addition to the local influence of the latent heat flux on precipitation, the large-scale moisture transport and moisture convergence can influence the southern U.S. precipitation during the summer (Higgins et al. 1997; Small et al. 2007; Hu et al. 2011; Hu and Feng 2012; Thibeault and Seth 2014). Thibeault and Seth (2014) analyzed the projection of summer precipitation over the northeastern North America in the CMIP5 models and found that changes in moisture convergence were critical for explaining the response of the summer precipitation to the RCP8.5 scenario. In our analysis of the moisture budget and moisture convergence over the southeastern United States, both SP-CCSM4 and CCSM4 project an increase of the moisture convergence into the rectangular area (figures not shown). This suggests that sensitivity of cloud parameterizations to local influences may contribute more to the southeastern U.S. precipitation than the large-scale moisture budget does in the SP-CCSM4 and CCSM4 models.
6. Conclusions
In this study, the simulation and projection of the hydrological cycle in summer over the southeastern United States is evaluated in SP-CCSM4 and further compared to simulation and projections in CCSM4. In the control experiment, both SP-CCSM4 and CCSM4 show biases in simulating the climatology of summer precipitation. SP-CCSM4 underestimates while CCSM4 overestimates summer precipitation over the southeastern United States.
Under the extreme forcing of the RCP8.5 scenario, SP-CCSM4 projects a significant decrease of the southeastern U.S. precipitation, whereas CCSM4 projects a significant increase of precipitation by the end of the twenty-first century. Furthermore, the distribution of summer precipitation over the 30°–40°N, 75°–85°W region shows a shift toward more frequent occurrences of precipitation with weak intensity in the last 20 yr of the RCP8.5 experiment with SP-CCSM4. This shift of the distribution could explain the decrease in the mean precipitation projected by this model. The distribution of precipitation in CCSM4 shifts toward more frequent occurrences of intense precipitation, which could explain the increase in precipitation over the southeastern United States.
To understand what causes the difference between the hydrological cycles projected in SP-CCSM4 and CCSM4, first, the impact of the North Atlantic subtropical high on precipitation is compared between the two models. The analysis of changes in SLP shows that SP-CCSM4 projects a significant enhancement of the high pressure system prevailing over the southeastern United States during the summer. The increase of SLP projected in SP-CCSM4 is accompanied by an inhibition of convection and rainfall. The pattern of projected SLP changes is different in CCSM4 compared to SP-CCSM4. The CCSM4 model projects an increase of SLP in the south part of the southeastern United States and a decrease of SLP over the north part of the southeastern United States, which are not statistically significant, indicating a weaker intrusion of the subtropical high into the southeastern United States in CCSM4.
Further analysis of the position of the subtropical high shows a larger response of the subtropical high in the global warming scenario simulated by SP-CCSM4 than that in CCSM4. In SP-CCSM4, at the end of the twenty-first century the boundary of the subtropical high is displaced northward relative to the mean position at the beginning of the simulation. In contrast to SP-CCSM4, the position of the subtropical high boundary projected in CCSM4 shows mostly a westward movement under strong external forcing and less northward displacement. Moreover, the ridge line of the subtropical high projected in SP-CCSM4 has a larger northward movement than that projected in CCSM4. Both the boundary and ridge line of the subtropical high indicate a stronger intrusion of the subtropical high into the southeastern United States projected in SP-CCSM4 than that in CCSM4. The stronger intrusion of the subtropical high in SP-CCSM4 is associated with subsidence and relatively weaker moisture transport into the southeastern United States. These conditions do not favor convection and precipitation and explain the decreased precipitation projected in SP-CCSM4.
In addition to the large-scale factors, local influences such as soil moisture content and latent heat flux were also investigated. The correlation between soil moisture content and precipitation in the RCP8.5 experiments indicates a positive soil moisture–precipitation feedback in both models. However, the changes in the soil moisture content are different between SP-CCSM4 and CCSM4. A significant decrease of soil moisture content is projected over the southeastern United States in SP-CCSM4, while significantly increased soil moisture dominates in the projection by CCSM4. As a result, under global warming a drier land surface corresponds to the reduced precipitation projected in SP-CCSM4, whereas CCSM4 projects that a wetter land surface is associated with increased precipitation.
Unlike soil moisture content, the latent heat flux shows different relationships with precipitation in SP-CCSM4 and CCSM4 in the RCP8.5 experiment. In SP-CCSM4, a significant positive latent heat flux–precipitation relationship dominates the southeastern United States, whereas in CCSM4 a negative latent heat flux–precipitation relationship takes place over the southeastern United States. This indicates that southeastern U.S. precipitation in SP-CCSM4 and CCSM4 has a different response to the change of latent heat flux released from the surface. Moreover, changes in latent heat flux are different between SP-CCSM4 and CCSM4. Over the southeastern United States, latent heat flux is significantly reduced in SP-CCSM4’s projection but is barely changed in CCSM4’s projection. With similar increases in sensible heat flux in both models, weaker (stronger) latent heat flux in SP-CCSM4 (CCSM4) indicates a deeper (shallower) boundary layer and reduced (enhanced) precipitation, which contributes to drier (wetter) soil moisture and further leads to weaker (stronger) latent heat flux.
Further comparison between the relative importance of the local influence and large-scale factors is done using a multivariate EOF analysis of precipitation, latent heat flux, and SLP in both SP-CCSM4 and CCSM4. In the RCP8.5 experiment, we find different latent heat flux patterns associated with similar precipitation and SLP patterns in SP-CCSM4 and CCSM4. This further verifies that SP-CCSM4 and CCSM4 have a different latent heat flux–precipitation relationship in the RCP8.5 experiment. In the RCP8.5 experiment, the rainfall patterns in both models show increased precipitation to the west but decreased precipitation along the coast. In both models, increased SLP in the U.S. Southeast is associated with the positive precipitation pattern in the RCP8.5 experiment. However, accompanying the precipitation pattern, there is enhanced latent heat flux projected in SP-CCSM4 but decreased latent heat flux in CCSM4 over the southeastern United States in the RCP8.5 experiment. Despite the difference in the first EOF mode of the latent heat flux, both SP-CCSM4 and CCSM4 show a stronger local impact on precipitation than the remote impact from the subtropical high in the RCP8.5 experiment. The correlation coefficient between the time series of the dominant patterns of precipitation and latent heat flux in both models is larger than the correlation coefficient between the time series of the dominant patterns of precipitation and SLP.
The comparison of precipitation between SP-CCSM4 and CCSM4 sheds a new light on understanding the influence of soil moisture content, the latent heat flux, and the North Atlantic subtropical high on the southeastern U.S. hydrological cycle. In SP-CCSM4 and CCSM4, compared to the remote influence of the subtropical high, local factors such as the soil moisture content and latent heat flux may play a relatively more important role in modulating the precipitation over the southeastern United States. The conventional CCSM4 seems to be sensitive to thermal energy, while SP-CCSM4 has a significant positive relationship between precipitation and latent heat flux. This result indicates that projection in precipitation could depend on the cloud parameterization scheme, which can have different sensitivities to local surface changes.
The results in this study are limited to two models. Additional simulations with the superparameterization scheme implemented into different climate models could provide more thorough comparisons between conventional cloud parameterization schemes and convection-permitting schemes.
Acknowledgments
We thank all the scientists and software engineers who contributed to the development of the CCSM4 and the superparameterization. This work has been supported by the Regional and Global Climate Modeling Program funded by the U.S. Department of Energy, Office of Science under Award SC0006722. This research used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the U.S. Department of Energy under Contract DE-AC02-05CH11231. An earlier version of the paper was significantly improved by suggestions provided by three anonymous reviewers and Eric Altshuler.
REFERENCES
Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 1147–1167, doi:10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.
Arnold, N. P., M. Branson, M. A. Burt, D. S. Abbot, Z. Kuang, D. A. Randall, and E. Tziperman, 2014: Effects of explicit atmospheric convection at high CO2. Proc. Natl. Acad. Sci. USA, 111, 10 943–10 948, doi:10.1073/pnas.1407175111.
Baigorria, G. A., J. W. Jones, and J. J. O’Brien, 2007: Understanding rainfall spatial variability in Southeast USA at different timescales. Int. J. Climatol., 27, 749–760, doi:10.1002/joc.1435.
Davis, R. E., B. P. Hayden, D. A. Gay, W. L. Phillips, and G. V. Jones, 1997: The North Atlantic subtropical anticyclone. J. Climate, 10, 728–744, doi:10.1175/1520-0442(1997)010<0728:TNASA>2.0.CO;2.
Davy, R., and I. Esau, 2014: Global climate models’ bias in surface temperature trends and variability. Environ. Res. Lett., 9, 114024, doi:10.1088/1748-9326/9/11/114024.
DeMott, C. A., D. A. Randall, and M. Khairoutdinov, 2007: Convective precipitation variability as a tool for general circulation model analysis. J. Climate, 20, 91–112, doi:10.1175/JCLI3991.1.
Dirmeyer, P. A., and Coauthors, 2012: Simulating the diurnal cycle of rainfall in global climate models: Resolutions versus parameterization. Climate Dyn., 39, 399–418, doi:10.1007/s00382-011-1127-9.
Froidevaux, P., L. Schlemmer, J. Schmidli, W. Langhans, and C. Schär, 2014: Influence of the background wind on the local soil moisture–prediction feedback. J. Atmos. Sci., 71, 782–799, doi:10.1175/JAS-D-13-0180.1.
Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 4973–4991, doi:10.1175/2011JCLI4083.1.
Grabowski, W. W., 2001: Coupling cloud processes with the large-scale dynamics using the cloud-resolving convection parameterization (CRCP). J. Atmos. Sci., 58, 978–997, doi:10.1175/1520-0469(2001)058<0978:CCPWTL>2.0.CO;2.
Grabowski, W. W., 2004: An improved framework for superparameterization. J. Atmos. Sci., 61, 1940–1952, doi:10.1175/1520-0469(2004)061<1940:AIFFS>2.0.CO;2.
Higgins, R. W., Y. Yao, and X. L. Wang, 1997: Influence of the North American monsoon system on the U.S. summer precipitation regime. J. Climate, 10, 2600–2622, doi:10.1175/1520-0442(1997)010<2600:IOTNAM>2.0.CO;2.
Hohenegger, C., P. Brockhaus, C. S. Bretherton, and C. Schär, 2009: The soil moisture–precipitation feedback in simulations with explicit and parameterized convection. J. Climate, 22, 5003–5020, doi:10.1175/2009JCLI2604.1.
Hu, Q., and S. Feng, 2012: AMO- and ENSO-driven summertime circulation and precipitation variations in North America. J. Climate, 25, 6477–6495, doi:10.1175/JCLI-D-11-00520.1.
Hu, Q., S. Feng, and R. J. Oglesby, 2011: Variations in North American summer precipitation driven by the Atlantic multidecadal oscillation. J. Climate, 24, 5555–5570, doi:10.1175/2011JCLI4060.1.
Khairoutdinov, M., and D. Randall, 2001: A cloud resolving model as a cloud parameterization in the NCAR Community Climate System Model: Preliminary results. Geophys. Res. Lett., 28, 3617–3620, doi:10.1029/2001GL013552.
Khairoutdinov, M., and D. Randall, 2003: Cloud resolving model of the ARM summer 1997 IOP: Model formulation, results, uncertainties, and sensitivities. J. Atmos. Sci., 60, 607–625, doi:10.1175/1520-0469(2003)060<0607:CRMOTA>2.0.CO;2.
Khairoutdinov, M., D. Randall, and C. DeMott, 2005: Simulations of the atmospheric general circulation using a cloud-resolving model as a superparameterization of physical processes. J. Atmos. Sci., 62, 2136–2154, doi:10.1175/JAS3453.1.
Kooperman, G. J., M. S. Pritchard, and R. C. J. Somerville, 2014: The response of US summer rainfall to quadrupled CO2 climate change in conventional and superparameterized versions of the NCAR Community Atmosphere Model. J. Adv. Model. Earth Syst., 6, 859–882, doi:10.1002/2014MS000306.
Kutzbach, J. E., 1967: Empirical eigenvectors of sea-level pressure, surface temperature and precipitation complexes over North America. J. Appl. Meteor., 6, 791–802, doi:10.1175/1520-0450(1967)006<0791:EEOSLP>2.0.CO;2.
Li, L., W. Li, and Y. Kushnir, 2012: Variation of the North Atlantic subtropical high western ridge and its implication to southeastern US summer precipitation. Climate Dyn., 39, 1401–1412, doi:10.1007/s00382-011-1214-y.
Li, L., W. Li, and Y. Deng, 2013: Summer rainfall variability over the Southeastern United States and its intensification in the 21st century as assessed by CMIP5 models. J. Geophys. Res. Atmos., 118, 340–354, doi:10.1002/jgrd.50136.
Miyasaka, T., and H. Nakamura, 2005: Structure and formation mechanisms of the Northern Hemisphere summertime subtropical highs. J. Climate, 18, 5046–5065, doi:10.1175/JCLI3599.1.
Moss, R. H., and Coauthors, 2010: The next generation of scenarios for climate change research and assessment. Nature, 463, 747–756, doi:10.1038/nature08823.
Neale, R. B., J. Richter, S. Park, P. H. Lauritzen, S. J. Vavrus, P. J. Rasch, and M. Zhang, 2013: The mean climate of the Community Atmosphere Model (CAM4) in forced SST and fully coupled experiments. J. Climate, 26, 5150–5168, doi:10.1175/JCLI-D-12-00236.1.
Nemias, J., 1952: The annual course of month-to-month persistence in climatic anomalies. Bull. Amer. Meteor. Soc., 33, 279–285.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moening, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699–706, doi:10.1175/1520-0493(1982)110<0699:SEITEO> 2.0.CO;2.
Palmer, T. N., F. Doblas-Reyes, A. Weisheimer, and M. Rodwell, 2008: Toward seamless prediction: Calibration of climate change projections using seasonal forecasts. Bull. Amer. Meteor. Soc., 89, 459–470, doi:10.1175/BAMS-89-4-459.
Pritchard, M. S., and R. C. J. Somerville, 2009: Assessing the diurnal cycle of precipitation in a multi-scale climate model. J. Adv. Model. Earth Syst., 1, 12, doi:10.3894/JAMES.2009.1.12.
Seager, R., A. Tzanova, and J. Nakamura, 2009: Drought in the southeastern United States: Causes, variability over the last millennium, and the potential for future hydroclimate change. J. Climate, 22, 5021–5045, doi:10.1175/2009JCLI2683.1.
Small, R. J. O., S. P. De Szoeke, and S.-P. Xie, 2007: The Central American midsummer drought: Regional aspects and large-scale forcing. J. Climate, 20, 4853–4873, doi:10.1175/JCLI4261.1.
Stan, C., and L. Xu, 2014: Climate simulations and projections with a super-parameterized climate model. Environ. Modell. Software, 60, 134–152, doi:10.1016/j.envsoft.2014.06.013.
Sun, S., and G. Wang, 2012: The complexity of using a feedback parameter to quantify the soil moisture–precipitation relationship. J. Geophys. Res., 117, D11113, doi:10.1029/2011JD017173.
Taylor, C. M., C. E. Birch, D. J. Parker, N. Dixon, F. Guichard, G. Nikulin, and G. S. Lister, 2013: Modeling soil moisture-precipitation feedback in the Sahel: Importance of spatial scale versus convective parameterization. Geophys. Res. Lett., 40, 6213–6218, doi:10.1002/2013GL058511.
Thibeault, J. M., and A. Seth, 2014: A framework for evaluating model credibility for warm-season precipitation in northeastern North America: A case study of CMIP5 simulations and projections. J. Climate, 27, 493–510, doi:10.1175/JCLI-D-12-00846.1.
Wang, H., R. Fu, A. Kumar, and W. Li, 2010a: Intensification of summer rainfall variability in the southeastern United States during recent decades. J. Hydrometeor., 11, 1007–1018, doi:10.1175/2010JHM1229.1.
Wang, H., S. Schubert, M. Suarez, and R. Koster, 2010b: The physical mechanisms by which the leading patterns of SST variability impact U.S. precipitation. J. Climate, 23, 1815–1836, doi:10.1175/2009JCLI3188.1.
Wheeler, M. C., and H. H. Hendon, 2004: An all-season real-time multivariate MJO index: Development of an index for monitoring and prediction. Mon. Wea. Rev., 132, 1917–1932, doi:10.1175/1520-0493(2004)132<1917:AARMMI>2.0.CO;2.
Zhu, X., L. Xu, and C. Stan, 2014: Projections of the tropical Atlantic vertical wind shear and its relationship with ENSO in SP-CCSM4. J. Climate, 27, 8342–8356, doi:10.1175/JCLI-D-14-00002.1.