1. Introduction
Accurate representations of the land–atmosphere (LA) coupling processes are critical for weather forecasts and climate predictions (Seneviratne et al. 2006, 2010; Santanello et al. 2018). A lack of quantitative understanding of the nature and characteristics of LA coupling remains (e.g., Betts 2004; Ek and Holtslag 2004; Guillod et al. 2014; Santanello et al. 2018), owing to the multivariate and multiscale interactive processes between the land surface, planetary boundary layer (PBL), clouds and precipitation, and the limited observations over highly heterogeneous land surfaces with varying vegetation cover, land use, terrain, and soil texture.
Previous studies on LA coupling focus on the soil moisture–precipitation (SM–P) feedback and show discrepancies in the coupling strength between models and observations. Based on multiple weather and climate models, the Global Land–Atmosphere Coupling Experiments (GLACE) provided an estimate of the global distribution of LA coupling strength (Koster et al. 2004, 2006) and identified several “hot spots” where the soil moisture anomalies are strongly correlated with the summertime precipitation. Although the spatiotemporal scales may complicate the analysis (Guillod et al. 2015; Ferguson et al. 2012), there is often a lack of evidence for a strong LA coupling in observations over the model-identified hot spots, e.g., at the U.S. Great Plains (Lamb et al. 2012; Taylor et al. 2012; Findell et al. 2011; Ruiz-Barradas and Nigam 2013; Phillips and Klein 2014; Wei et al. 2016; Song et al. 2016; Tang et al. 2018). Using the global flux tower data, Dirmeyer et al. (2018) showed that models generally underestimate the atmospheric linkage between evaporative fraction (EF) and precipitation while overestimate the terrestrial linkage between soil moisture and evaporative fraction (SM–EF). This is consistent with Phillips et al. (2017) in which the SM–EF coupling is much stronger in both free-running simulations and constrained climate-model hindcasts than the Atmospheric Radiation Measurement (ARM) observations at the Southern Great Plains (SGP) site. Williams et al. (2016) found that the EF biases in models are attributed to the overestimation in the bare soil evaporation and can be improved by better representations of vegetation, supported by the strong correlation between the observed leaf area index and EF (Williams and Torn 2015). The stronger positive SM–P feedback in models is usually associated with a warm and dry bias over the central United States (Klein et al. 2006), which results from both overestimated surface shortwave radiation and underestimated EF (Zhang et al. 2018). Ma et al. (2018) further diagnosed that biases in EF are more important than biases in surface radiation in explaining the large warm bias in models and these biases are mainly associated with significant underestimation of precipitation.
At the ARM SGP site, long-term observations show two peaks in the diurnal cycle of surface precipitation during warm seasons (Zhang and Klein 2010). The primary nighttime peak is linked with the eastward propagating mesoscale convective systems (MCSs) into the SGP site that originate at the ridge of the Rocky Mountains in the afternoon (e.g., Jiang et al. 2006). The secondary afternoon peak is strongly associated with the local PBL development driven by the diurnal variation in surface heating. Using 10-yr observations at the ARM SGP site, Tao et al. (2019, hereafter TZ19) separated “locally generated” from “nonlocal” convective events and showed that the LA coupling strength amplifies on the afternoon deep convection days, in which surface evaporation is the dominant moisture source for precipitation. This suggests that the overestimated LA coupling at SGP in climate models may result from models’ incapability of simulating the propagating MCSs. Therefore, in the model evaluation against observations and the reconciliation of discrepancies, it is very important to carefully distinguish local from nonlocal convection regimes in disentangling the connections between model deficiencies and biases.
To comprehensively diagnose the model performance on the representation of LA coupling, integrative metrics and process-oriented analyses are highly desirable. To facilitate this purpose, the Local Land–Atmosphere Coupling (LoCo) metrics (Santanello et al. 2011b, 2018) were developed to quantify the complex SM–P relationship and feedbacks (e.g., Findell and Eltahir 2003a; Santanello et al. 2009, 2011a; Dirmeyer 2011; Tawfik and Dirmeyer 2014; Tawfik et al. 2015). LoCo metrics have been widely used and proven very useful in addressing model deficiencies. For example, using LoCo metrics and the ARM data, the SM–P feedbacks were diagnosed during extremely dry and wet conditions (Santanello et al. 2013) and in several prevailing reanalysis data products (Santanello et al. 2015).
In the model diagnosis, it is important to isolate the problems of physical parameterizations from the biased atmospheric or land surface states. With this in mind, we evaluate two models in their special configurations: 1) the Community Atmosphere Model (CAM version 5.1) short-range hindcast runs initiated every day with reanalysis data by the DOE Cloud-Associated Parameterization Testbed (CAPT) (CAM5-CAPT; Ma et al. 2020); and 2) the DOE Energy Exascale Earth System Model (E3SM) Atmosphere Model (EAM version 1) Regionally Refined Model (EAMv1-RRM; Q. Tang et al. 2019, hereafter TK19) runs with a higher resolution (0.25°) in the contiguous United States (CONUS) domain and the large-scale winds nudged toward analysis data at a coarser resolution (1°) outside the CONUS domain. Using these model configurations, we hope that biased model behaviors in LA coupling might be attributed to different factors, such as parameterizations, initial land–atmosphere conditions, and large-scale circulations.
In this study, we focus on locally generated convection regimes in which the LA coupling is the strongest and use LoCo metrics and the long-term ARM SGP data to evaluate the North American Regional Reanalysis (NARR; Mesinger et al. 2006) and two aforementioned climate model simulations. We aim to advance process-oriented diagnoses of local LA coupling and provide insights into model deficiencies. Particularly we intend to answer the following questions:
How well are the budgets of atmospheric moisture and surface energy and the diurnal cycle of clouds and precipitation represented in reanalysis data and climate models, especially in the locally generated convection regimes?
For each locally generated convection regime, how well is the local LA coupling represented in reanalysis and climate model simulations in comparison with ground-based observations?
The remaining parts of the paper are organized as follows. Section 2 presents the observational datasets, reanalysis, and climate model simulations. The analysis of moisture and energy budgets and the diurnal cycle of clouds and precipitation are presented in section 3, diagnosis of LA coupling in local convection regimes in section 4, and conclusions in section 5.
2. Datasets
a. ARM observations
We use ground-based data at the ARM SGP site during the warm seasons (May–August) from 2004 to 2012. Figure 1 shows the analysis domain with an area of 150 × 150 km2 centered around the SGP central facility. This is the same region as the subdomain E in TZ19 with the land cover dominated by winter wheat, where the domain-averaged leaf area index decreases from late spring to midsummer due to harvest. In the following, the hourly-mean domain averages are used.
(a) The variational analysis domain of this study (the irregular domain enclosed by the black line) and of the ARM continuous forcing dataset (green circle). The red dot represents the location of the ARM SGP central facility. The gray line at 37°N indicates the state boundary line between Oklahoma (below) and Kansas (above). (b) The analysis domain with grid points from NARR (green dots), CAM5-CAPT hindcasts (red x marks), and EAMv1-RRM nudged runs (blue crosses).
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
Detailed information on the ARM data is listed below:
The components in the atmospheric moisture budget and surface energy budget, as well as the near-surface temperature and humidity are obtained using the constrained variational analysis (Zhang and Lin 1997; Zhang et al. 2001) applied to the analysis domain (TZ19; https://portal.nersc.gov/project/capt/ARMForcingData/tang32/VARANAL_subdomains/forcing_data/domainE/), the same method as in the ARM continuous forcing dataset (Xie et al. 2004; S. Tang et al. 2019).
The ARM Best Estimate data products (ARMBE) (Xie et al. 2010; https://doi.org/10.5439/1095313) provide the hourly cloud fraction profiles that are derived from the Active Remotely-Sensed Cloud (ARSCL; https://doi.org/10.5439/1052058) data, a combination of cloud radar, micropulse lidar, and ceilometer observations (Clothiaux et al. 2000).
The balloon-borne sounding system samples the atmospheric temperature and moisture profiles four times a day at 0530, 1130, 1730, and 2330 local standard time (LST). The moisture profile is scaled by the total precipitable water vapor from the microwave radiometer measurement (LSSONDE; https://doi.org/10.5439/1027294).
The convective mixed-layer top height Zi, or the so-called daytime PBL height, is derived from the range-corrected signal-to-noise ratio (SNR) data collected by the 915-MHz Radar Wind Profiler (RWP; https://doi.org/10.5439/1025136) at the SGP central facility. The sharp humidity and temperature gradient at Zi results in a large change in the index of refraction
, which is proportional to the RWP SNR. Therefore, the height level with the maximum SNR often provides a good estimate of Zi (Bianco and Wilczak 2002; Bianco et al. 2008). The final estimation of Zi is a weighted average of three guesses: the height of maximum SNR, the height of maximum SNR gradient and the maximum height of SNR greater than the critical threshold. Figure 2 shows an example of the daytime evolution of the hourly-mean RWP-derived Zi. The PBL heights derived from radiosonde measurements are also shown for comparison. In Fig. 2, the RWP-derived Zi successfully captures the height level with the maximum increase (decrease) of potential temperature (water vapor mixing ratio). The RWP-derived Zi is in general very comparable to the radiosonde-derived PBL heights except the significantly underestimated one using the bulk Richardson number (Rib) equal to 0.25.
(left) Time–height range-corrected SNR on 18 Jun 2012 at the ARM SGP central facility. The white dot-line denotes the RWP-derived mixed-layer top Zi, an average of three retrievals: number 1 as the height of maximum SNR; number 2 as the height of maximum SNR gradient; number 3 as the maximum height of SNR greater than the critical threshold (10 dB in this case). Vertical profiles of (center) potential temperature and (right) water vapor mixing ratio from the balloon sounding data at 1130 LST. The horizontal black solid line represents Zi from RWP. The horizontal dashed lines denote the derived Zi based on sounding profiles using four different algorithms (green: Heffter; red: Liu and Liang; bright blue: bulk Richardson number with 0.25 threshold; dark blue: bulk Richardson number with 0.5 threshold).
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
b. NARR
The NARR data have been widely used in the LA coupling studies (e.g., Findell et al. 2011). It is developed with the 2003 version of the National Centers for Environmental Prediction (NCEP) Eta model and a three-dimensional variational data assimilation (3DVAR) technique (Mesinger et al. 2006). The Eta model is coupled to the Noah land surface model (Ek et al. 2003) and uses the Yonsei University PBL scheme, which is a first-order nonlocal scheme with a countergradient term in the eddy-diffusion equation (Hong et al. 2006). One of the most important features of NARR is the direct assimilation of precipitation and radiances, as NARR is primarily designed to increase the reliability of land surface water and energy budgets. Variables from NARR are 3-h temporal averages with 29 vertical layers at 32-km horizontal resolution (Fig. 1b).
c. Climate model simulations
1) CAM5-CAPT hindcast runs
The multiyear hindcast experiment (Phillips et al. 2004; Ma et al. 2020) is conducted with the CAM5 (version cesm1_0_5, FC5 comp set, Neale et al. 2012) using the finite volume dynamical core at a horizontal resolution of 0.9° latitude × 1.25° longitude and with 30 vertical levels. Table 1 lists the physical parameterizations used in CAM5. The land model is the Community Land Model version 4.0 (CLM4) with the same horizontal resolution. The experiment contains a suite of 72-h global hindcasts starting at 0000 UTC every day in the years of 1997–2012. The initial atmospheric states including horizontal winds, temperature, specific humidity and surface pressure are directly taken from the ERA-Interim reanalysis (Dee et al. 2011). A long-term continuous nudging simulation was also performed to generate other necessary initial state variables (e.g., cloud and aerosol fields), which are not available from the ERA-Interim. Land initial conditions are obtained from a fully spun-up offline land model simulation forced by reanalysis and observational data including precipitation, surface winds, and surface radiative fluxes (Ma et al. 2015). In the following analysis, we concatenated each hindcast from 24- to 48-h lead time to form a day-2 time series of 9 years from 2004 to 2012. Day-1 data are not used to minimize the impact of model spinup (Ma et al. 2013, 2014).
The temporal/spatial resolution of the model outputs from the CAM5-CAPT hindcasts and the EAMv1-RRM nudged runs and the associated parameterizations used.
2) EAMv1-RRM nudged runs
The newly released DOE E3SMv1 is a fully coupled physical model designed to address DOE mission-relevant science questions (Golaz et al. 2019; Caldwell et al. 2019). The EAMv1 (Rasch et al. 2019) is developed based on the CAM version 5.3 (Neale et al. 2012; Terai et al. 2017) but includes substantial changes such as the enhanced vertical resolution (72 levels versus 30), a higher model top (0.1 versus 2 hPa) and newly adopted physical parameterizations (Table 1). The land component of E3SMv1 is developed from the CLM version 4.5 with new options for representing soil hydrology and biogeochemistry. Details of E3SMv1 simulated clouds and convective processes are documented in Xie et al. (2018) and Zhang et al. (2019). Considering the computational limitation for running global high-resolution models, an RRM capability is adopted by EAMv1 (TK19), which simulates the CONUS domain at 0.25° horizontal resolution while keeps the remaining outside area at 1° globally. The RRM has been proven useful for evaluating physical parameterizations and understanding atmosphere model behaviors at high resolution. We performed 9-yr (2004–12) EAMv1-RRM simulations nudged to the ERA-Interim analysis fields of horizontal velocities (Dee et al. 2011) with a 6-h relaxation time scale outside the CONUS. Here, the nudging coefficient map (same as Fig. 17 in TK19) has been carefully designed to 1) keep the CONUS domain run freely and 2) reduce the nudging noise due to the inconsistency between the model and analysis data over the free-running region. Details of EAMv1-RRM nudging runs are documented in TK19.
In the following, spatial averages of NARR, CAM5-CAPT, and EAMv1-RRM are compared with the ARM observations within the analysis domain in Fig. 1. In addition, we emphasize models’ performance against observations and do not intend to compare between the two climate models due to their different configurations, resolutions and model physics.
3. Overall performance of reanalysis and climate models
In this section, we first examine the daily-mean regional moisture and energy budgets and then investigate the diurnal evolution of clouds and precipitation during warm seasons at the analysis domain of SGP (Fig. 1).
The representation of LA coupling in climate models is a combined result of the large-scale circulations, the initial atmospheric/land surface conditions, physical parameterizations and the coupling between them. To isolate processes contributing to model biases, we take advantages of the model configurations in CAM5-CAPT hindcasts and EAMv1-RRM nudged runs and make day-to-day comparisons with observations within each of the three locally generated convection regimes: clear-sky, fair-weather shallow cumulus (ShCu) and late-afternoon deep convection days. Through this convection-regime classification, we are able to examine the strength of local LA coupling under different atmospheric conditions.
Details of the observational selection criteria of different convection regimes are summarized in Table 2. Based on the 9-yr ARM observations, we identify 66 clear-sky days, 48 fair-weather ShCu days, and 48 days with late-afternoon deep convection. In these three locally generated convection regimes, convective thermals, clouds and precipitation develop locally and are tightly coupled with the boundary layer development driven by the diurnally varying surface fluxes (Zhang and Klein 2010, 2013; Lareau et al. 2018; TZ19) so that the LA coupling is expected to be the strongest. In addition to the definition criteria listed in Table 2, Geostationary Operational Environmental Satellite (GOES) images by P. Minnis’s group at the National Aeronautics and Space Administration (NASA) Langley Center (available online at https://cloudsway2.larc.nasa.gov/) are also scrutinized to ensure that these regimes are least subject to the influences from synoptic or mesoscale weathers (Zhang and Klein 2010, 2013).
Definition criteria and sample size of the three locally generated convection regimes based on the ARM observations.
a. Atmospheric moisture budget
Comparison of the atmospheric moisture budget terms (daily means) among (a) all warm season (May–August 2004–12), and three local convection regimes: (b) clear-sky, (c) fair-weather shallow cumulus, and (d) late-afternoon deep convection. Different colors represent results from each of the four datasets (light pink: ARM observations; green: NARR; red: CAM5-CAPT hindcasts; blue: EAMv1-RRM nudged runs). The length of the vertical black lines denotes two standard errors; P, E, MFC, and dPW represent precipitation, evaporation, moisture flux convergence, and the atmospheric storage change, respectively.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
Overall, NARR agrees well with the ARM data on the warm-season moisture budget (Fig. 3a) yet it slightly underestimates the surface evaporation for all local convection regimes (Figs. 3b–d). This is consistent with Nigam and Ruiz-Barradas (2006), which indicated that the estimates of evaporation in NARR are still subject to errors as the precipitation assimilation does not directly modify evapotranspiration. Such negative bias in evaporation is also noted for both CAM5-CAPT and EAMv1-RRM. For example, on the observed clear-sky days, the daily-mean evaporation of EAMv1-RRM is only about half of that in the ARM data (Fig. 3b). Precipitation is in general insufficient in CAM5-CAPT but it is comparable with that observed on late-afternoon deep convection days (Fig. 3d). In addition, it rains in CAM5-CAPT on the observed fair-weather ShCu days (Fig. 3c). EAMv1-RRM largely underestimates the precipitation in total amount (Fig. 3a) and on late-afternoon deep convection days (Fig. 3d), however, it rains on both of the observed fair-weather clear-sky and ShCu days (Figs. 3b,c). This suggests potential model deficiencies in its convective trigger in the deep convection parameterization (Xie et al. 2019). Moreover, a large discrepancy exists between EAMv1-RRM and ARM observations in the daily-mean MFCs. EAMv1-RRM shows a larger water vapor divergence in total (Fig. 3a) while for local convection regimes, it significantly underestimates the associated convergence (Fig. 3d) or divergence (Figs. 3b,c). This hints at a considerable contribution of the large-scale atmospheric advective forcings to the bias in the LA coupling at SGP in EAMv1-RRM.
To better understand the large negative bias of precipitation in EAMv1-RRM, we further examine the precipitation in different months and diurnal periods (Fig. 4). Considering the different dominant moisture sources (TZ19), precipitation between 0000 and 0800 LST and between 1200 and 2000 LST are separated, which correspond to a primary nighttime precipitation peak and a secondary late-afternoon peak in the observed warm-season diurnal cycle of precipitation at SGP (Zhang and Klein 2010). In May, the total EAMv1-RRM precipitation is slightly lower than observations (Fig. 4a) with significant underestimation in nighttime precipitation (Fig. 4b) and some overestimation in afternoon rains (Fig. 4c). In June, however, EAMv1-RRM only produces 16% of nighttime precipitation and 60% of afternoon precipitation compared with the ARM data (Figs. 4b,c). Similar behavior persists in July and August for EAMv1-RRM. CAM5-CAPT also simulates less rain than the observed in general (Fig. 4a) but the greater afternoon rain compared with the ARM data tends to cancel out the insufficient nighttime rain in CAM5-CAPT. The underestimation of nighttime precipitation is largely due to the incapability of models to capture elevated convection above PBL that is associated with the propagation of MCSs (Xie et al. 2019). Such warm and dry bias in climate models may also be partly attributed to the impact of nocturnal low-level jet and irrigation (Qian et al. 2013).
(a) Composite mean surface precipitation rate (mm day−1) of ARM, NARR, CAM5-CAPT hindcasts, and EAMv1-RRM nudged runs in all warm seasons (May–August) and four months, respectively. (b) As in (a), but for nighttime precipitation between 0000 and 0800 LST. (c) As in (a), but for afternoon precipitation between 1200 and 2000 LST.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
b. Surface energy budget
The Taylor diagram (Taylor 2001) in Fig. 5a shows that the daily mean surface energy budget in warm seasons is overall not well represented in reanalysis nor in climate models, as the points are scattered far from the reference point (1, 0). NARR exhibits significant positive biases (percent bias > 30%) for surface upward shortwave radiation (SWUP) and SH flux, consistent with Kennedy et al. (2011). Using the 3-yr ARM SGP data, Kennedy et al. (2011) found that NARR significantly overestimates SWDN and SWUP because of too few clouds and the insufficient extinction by aerosols and water vapor. Both SWUP and surface albedo (not shown) are largely underestimated in CAM5-CAPT during warm seasons, with negative bias between 20% and 30% of observations (Fig. 5a). This is consistent with Van Weverberg et al. (2018), which indicated that albedo issues dominate over cloud issues in the large positive net surface shortwave radiation biases in CAM5. The surface energy budget components are in general poorly simulated in EAMv1-RRM. The correlations with observations are lower than 0.6 in EAMv1-RRM for all variables except for surface downward (LWDN) and upward longwave radiation (LWUP). Specifically, EAMv1-RRM shows the largest RMSE in EF [LH/(SH + LH)] and largest biases in SH flux. For EAMv1-RRM, errors in EF might be attributed to the large precipitation deficit and the misrepresentation of the land model in surface energy partition. The abovementioned issues in NARR, CAM5-CAPT, and EAMv1-RRM during warm seasons are robust in all three local convection regimes (Figs. 5b–d).
Taylor diagrams of daily mean surface energy budget components in (a) all warm seasons, and three local convection regimes: (b) clear-sky days, (c) fair-weather ShCu days, and (d) late-afternoon deep convection days. The ARM data are used as the reference point (1, 0). The numbers denote different variables (1: surface downward shortwave radiation, SWDN; 2: surface upward shortwave radiation, SWUP; 3: surface downward longwave radiation, LWDN; 4: surface upward longwave radiation, LWUP; 5: surface sensible heat flux, SHFLX; 6: surface latent heat flux, LHFLX; 7: evaporative fraction, EF). The colors denote different datasets (light pink: ARM; green: NARR; red: CAM5-CAPT hindcasts; blue: EAMv1-RRM nudged runs). Different symbols indicate the bias magnitude (in percent). Variables with standard deviations > 3.0 are noted in the text below the panel if any.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
To further explore the biases of EF in EAMv1-RRM, Fig. 6 shows the seasonal variation from May to August of daytime (0600–1800 LST) mean SH, LH, and EF. Consistent with the lack of precipitation in EAMv1-RRM (Fig. 4), Fig. 6c shows a significant negative bias in EF persisting since June with overestimated SH and underestimated LH fluxes. The EF is also underestimated in NARR, mainly attributed to a large positive bias in SH flux (Fig. 6a). This is consistent with Santanello et al. (2013), which found that the surface Bowen ratio (SH/LH) is overestimated by NARR. The daytime mean SH is well simulated in CAM5-CAPT, however, an underestimation in LH is shown since mid-June (Fig. 6b) corresponding to the insufficient precipitation in summer (Fig. 4). Overall, the SH and LH fluxes are fairly well captured in CAM5-CAPT, suggesting that the method of land model initialization in CAM5-CAPT is effective in reproducing a reasonable land state.
The seasonal variation of 2004–12 daytime mean (0600–1800 LST): (a) surface sensible heat flux, (b) surface latent heat flux, and (c) surface evaporative fraction from ARM observations (black), NARR (green), CAM5-CAPT hindcasts (red), and EAMv1-RRM nudged runs (blue). A moving average of 30 days is applied to smooth out short-term fluctuations and highlight longer-term trends.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
c. Diurnal cycle of surface precipitation and clouds
The warm-season diurnal cycle composites of surface precipitation and vertical cloud fraction are illustrated in Fig. 7. It is not surprising that NARR and ARM observations have excellent agreement on precipitation (Fig. 7a) as the precipitation is assimilated in NARR (Bukovsky and Karoly 2007; Ruane 2010). Both the 1° CAM5-CAPT and 0.25° EAMv1-RRM fail to capture the nocturnal precipitation peak at SGP. Instead, both models show a diurnal precipitation maximum during the middle of the day (Fig. 7a). This is a well-known problem of climate models with deep convection parameterizations (Dai and Trenberth 2004; Lee et al. 2007) based on the local convective available potential energy (CAPE), in which CAPE builds up after sunrise and reaches a maximum usually in phase with the surface heating (e.g., Zhang and McFarlane 1995; Dirmeyer et al. 2012). A recent work by Xie et al. (2019) demonstrated that the diurnal cycle of precipitation in E3SM can be dramatically improved with a revised convective trigger function that induces a dynamic constraint to relax the unrealistically strong coupling of convection trigger with surface heating, and an unrestricted air parcel launch level to capture elevated convection during nighttime at SGP. Different from the ARM data, the diurnal cycles of cloud fraction in both CAM5-CAPT (Fig. 7c) and EAMv1-RRM (Fig. 7d) are featured with persistent high clouds (CHGH, above 400 hPa). It is also noted that EAMv1-RRM significantly underestimates mid and low-level cloud fraction. These cloud errors contribute a fair amount to the radiation bias in Fig. 5.
The warm-season (May–August 2004–12) diurnal cycle composites of (a) surface precipitation (mm day−1) and vertical profile of cloud fraction from (b) ARM, (c) CAM5-CAPT hindcasts, and (d) EAMv1-RRM nudged runs.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
d. Statistics of “correct” and “wrong” cases in models
We now focus on the performance of climate models on individual local convective regime days. In particular, we ask, how many clear-sky, fair-weather ShCu and late-afternoon deep convection days identified from ARM observations are correctly simulated by climate models based on the diurnal cycle of clouds and precipitation? For both CAM5-CAPT and EAMv1-RRM, the distributions of “correct” and “wrong” cases are quantified for each locally generated convection regime in Table 3.
Statistics of the distributions of “correct” and “wrong” cases of CAM5-CAPT hindcasts and EAMv1-RRM nudged runs on the clear-sky days, fair-weather shallow cumulus days, and late-afternoon deep convection days identified from the ARM observations. For each regime, here list the four criteria defined based on the ARM observations. The criteria are ordered from I to IV according to the selection steps. “✓” denotes that model simulations are able to reproduce certain criterion while “x” hints at models’ incapability. CLOW, CMED, and CHGH represents the amount of low-level, midlevel, and high-level clouds, respectively.
In general, the observed clear-sky days are fairly well captured by models. About 74% of the ARM-observed clear-sky days are correctly simulated as clear-sky days in CAM5-CAPT (Table 3) while the other 20% of them have precipitation during the day. Twenty-three out of 48 of the observed ShCu days have no/little low-level clouds (CLOW, diurnal maximum < 5%) in CAM5-CAPT, with only 10 days correctly simulated as ShCu days. CAM5-CAPT misses about 85% of the observed late-afternoon deep convection days, primarily due to the problem with too-early diurnal precipitation peak (before 1500 LST).
Only half of the ARM-observed clear-sky days are correctly simulated as clear-sky days in EAMv1-RRM, with the other 25% of them being days having too much daytime CHGH and another 25% with precipitation during the day. The performance of EAMv1-RRM degrades significantly in the other two local convection regimes. Only five ARM-observed ShCu days are correctly simulated as ShCu days in EAMv1-RRM while the majority (67%) of them are simulated as days with no/little CLOW. The ARM-observed late-afternoon deep convection days are either simulated as days with no precipitation/drizzling (44%) or with precipitation peaking too early (33%) in EAMv1-RRM.
4. LA coupling in local convection regimes
The day-to-day comparison above between observations and climate models indicates a low hit rate of the climate models (in hindcasts or nudged runs) to reproduce the ARM-observed ShCu and late-afternoon deep convection regimes. Particularly, such day-to-day comparison may not be the best way to evaluate EAMv1-RRM. Although the EAMv1-RRM runs are nudged toward analysis data every 6 h outside the CONUS domain, the model is free-running inside the CONUS domain. With this in mind, we reclassify the three local convection regimes in the 9-yr simulations of warm seasons from CAM5-CAPT and EAMv1-RRM separately, regardless of the day-to-day match with observations. In this model reclassification, we apply the same criteria as those of ARM observed local convective days (Table 2), except that we slightly relax the precipitation criterion for clear-sky and ShCu days in models as in Table 3.
During the 9-yr simulations, 136 (116) clear-sky days are reclassified from CAM5-CAPT (EAMv1-RRM), doubling the 66 clear-sky days identified from ARM observations. The occurrence frequency of ShCu days is lower in model simulations than that observed, e.g., 39 (12) ShCu days in CAM5-CAPT (EAMv1-RRM) versus 48 observed ShCu days in the ARM data. Ninety (42) late-afternoon deep convection days are simulated by CAM5-CAPT (EAMv1-RRM) while 48 days identified in ARM data. Notice that the correctly simulated days matching the day-to-day ARM observations in Table 3 are a subset of these reclassified convective regime days in models.
In the following, we focus on local convective regimes identified from ARM observations for NARR while for climate models, we use local convection regimes reclassified above from CAM5-CAPT and EAMv1-RRM separately. By doing so, the benefits are threefold: 1) it complements the composite day-to-day match comparison in section 3, which focuses on how well models can reproduce the ARM observed days of local convective regimes; 2) it features a statistical and climatological comparison on local convective regimes between the long-term ARM data and the corresponding climate model simulations with constrained large-scale winds; 3) it focuses on the local convective days that are correctly simulated and classified in model results using similar criteria as those defined from ARM observations. Specifically, in this section, we try to answer the questions as: how well the LA coupling processes are represented in climate models when models correctly capture the local convective regimes?
As shown in Fig. 8a, the diurnal precipitation maxima on model-based (both CAM5-CAPT and EAMv1-RRM) late-afternoon deep convection days are much lower than those on the observation-based late-afternoon deep convection days. The distinct diurnal cycle of the observed ShCu cloud fraction, which increases rapidly through the morning, peaks around early afternoon and diminishes through the late-afternoon, is generally reproduced in both models (Figs. 8b–d). But the simulated cloud fraction and cloud base are much lower than those on the observed ShCu days.
Diurnal cycle composites of (a) surface precipitation (mm day−1) on late-afternoon deep convection days and vertical profile of cloud fraction from (b) ARM, (c) CAM5-CAPT hindcasts, and (d) EAMv1-RRM nudged runs on fair-weather shallow cumulus days.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
In the following, we further evaluate the LA coupling in climate models and reanalysis for local convection regimes using three LoCo metrics: mixing diagrams (Santanello et al. 2009), the lifting condensation level (LCL) deficit (Santanello et al. 2011a), and the framework of convective triggering potential (CTP) and humidity index (HIlow) (Findell and Eltahir 2003a), which are calculated using the Coupling Metrics Toolkit (CoMeT; www.coupling-metrics.com). Through these metrics, we wish to discern the sources of model biases in the land–PBL–cloud–precipitation interaction.
a. Mixing diagrams
The mixing diagram approach relates the conservative variables, potential temperature θ and total water specific humidity q, to the water and energy budgets and the growth of PBL (Santanello et al. 2009, 2011a). By mixing diagram, we dissect the relative contributions of surface fluxes (sensible + latent) versus atmospheric fluxes (advection + entrainment) to the development of PBL in terms of the diurnal coevolution of θ and q. Previous studies often use θ and q at the surface to approximate mixed-layer averages due to the limited data of PBL heights. To improve this, we use the hourly RWP-derived Zi to separate the influences on the evolution of mixed-layer θ and q in the mixing diagrams, e.g., the contribution of advective (both horizontal and vertical) fluxes of heat and moisture within the PBL related to large-scale winds versus the contribution of entrainment fluxes from above the PBL related to local turbulent mixing and free troposphere conditions.
Figure 9 illustrates the composite clear-sky-day mixing diagrams from ARM observations, NARR, CAM5-CAPT, and EAMv1-RRM. The coevolution of Lυq and Cpθ (0730–1730 LST) is decomposed by vector components that represent the integrated fluxes of heat and moisture from the land surface (Vsfc), the advection (Vadv) and the entrainment at the PBL top (Vent as a residual). Six metrics are derived from these diagrams and summarized in Table 4, which include the Bowen ratio of the surface (βsfc) and the entrainment (βent), the entrainment ratio of heat (ESH) and moisture (ELH), and the advective flux ratio of heat (ASH) and moisture (ALH).
Clear-sky day mixing diagram of the PBL conservative variables, Lυq vs Cpθ, during the daytime evolution from ARM observations (black), NARR (green), CAM5-CAPT hindcasts (red), and EAMv1-RRM nudged runs (blue). Dots denote the composite hourly means from 0730 to 1730 LST. The text annotations depict the vector component contributions from surface (Vsfc), advection (Vadv), and entrainment fluxes (Vent) to the evolution. Note that we linearly interpolate the 3-hourly NARR data to hourly in this figure.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
The surface (βsfc) and entrainment (βent) Bowen ratios, the entrainment ratio of heat (ESH) and moisture (ELH), and the advective flux ratio of heat (ASH) and moisture (ALH) from the ARM observations, NARR, CAM5-CAPT hindcasts, and EAMv1-RRM nudged runs on clear-sky days. The flux values (W m−2) are derived using the mixing diagram theory and surface, advection, and entrainment flux vectors depicted in Fig. 9.
The observed temporal change in θ on clear-sky days is a direct result of PBL warming from the surface and entrainment with minimal warm advection (ASH = 0.04, Table 4). The large ESH (1.79) further indicates that the entrainment heat flux almost doubles the surface SH flux on the warming of the mixed layer. The overall diurnal change in q is relatively small compared with that in θ. The entrainment drying is comparable to but does not fully compensate the surface evaporation (ELH = −0.90), with the additional drying of the mixed layer from advection. The diurnal ranges of θ and q are in general similar between NARR and the ARM data but NARR tends to exhibit a slightly warm and dry bias near the surface, consistent with Santanello et al. (2015). Different from the observed slight warm advection, NARR presents a weak cold advection. The entrainment drying in NARR is similar to the observed while the entrainment warming is roughly 25% larger, which results in a slightly larger βent in NARR than in the ARM data.
In the early morning (0730 LST) of the clear-sky days from CAM5-CAPT, there already exist large warm and dry biases compared with the ARM observations and the simulated daytime change range is weaker in θ but stronger in q. The weaker increase in θ in CAM5-CAPT is mainly due to a cold advection almost balancing out the surface heating. In contrast to ARM observations, the q advection is moistening the PBL in CAM5-CAPT. The larger decrease of q in CAM5-CAPT than observations results from a much stronger dry air entrainment flux, which doubles the observed impact (ELH = −2.62) and is larger than the PBL moistening from the surface evaporation and advection. When the dry air entrainment ceases toward the late afternoon, the q increases rapidly from 1630 to 1730 LST in CAM5-CAPT.
Similarly, the clear-sky days identified from EAMv1-RRM also start with too warm and too dry conditions in the initial early-morning time. The diurnal change in θ is very close to that observed. However, the surface moistening via evaporation is too weak in EAMv1-RRM, consistent with the above moisture and energy budget analysis (Figs. 3 and 5). The high βsfc in EAMv1-RRM, about four times of the observed, confirms that more energy at the surface goes to heating. On the other hand, the quadrupled ELH indicates that the entrainment heating and drying dominate the surface fluxes in EAMv1-RRM, which supports rapid and deep PBL growth. The advection tends to cool and dry the mixed layer, but the impact is much smaller compared with those from the surface and entrainment.
b. LCL deficit
The LCL deficit (Santanello et al. 2011a,b), defined as PBL top height minus LCL, serves as a complementary metric for “mixing diagram” to further diagnose the competing effects of SH and LH fluxes on the PBL growth and cloud formation. The cloud onset occurs when the PBL top touches the LCL, i.e., LCL deficit reaches zero. This can be achieved via high SH flux accompanied by rapid PBL growth or large LH flux accompanied by LCL fall. Figure 10 shows the daytime evolution composites of PBL, LCL, and LCL deficit on clear-sky and ShCu days. It should be noted that the PBL height from observations is estimated based on the 915-MHz RWP SNR (section 2a) while the one simulated by climate models is based on the Rib. Specifically, the PBL height refers to the height of Rib > 0.19 (0.3) in CAM5-CAPT (EAMv1-RRM). With this in mind, the PBL heights estimated using the SONDE-derived profile of Rib are also plotted in Fig. 10 for comparison but with a critical threshold of 0.25 and 0.50 (Fig. 2). As shown, the composite mean PBL height estimated from RWP is in good agreement with that derived from sounding based on the Rib. The LCL is calculated as a function of surface pressure, air temperature and relative humidity as in Romps (2017).
Composite daytime evolution of (a) planetary boundary layer height (PBL), (b) lifting condensation level (LCL), and (c) LCL deficit (PBL minus LCL) from ARM observations (black), NARR (green), CAM5-CAPT hindcasts (red), and EAMv1-RRM nudged runs (blue) on clear-sky days. (d)–(f) As (a)–(c) but on fair-weather shallow cumulus days. In (a) and (d), the pink triangles (dark green upside-down triangles) denote the PBL heights derived from the balloon sounding data by a method using the bulk Richardson number with a critical threshold of 0.25 (0.50). The length of the vertical lines denotes two standard errors.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
The diurnal variations of PBL, LCL, and LCL deficit in NARR are overall consistent with those observed on clear-sky days. The PBL height never reaches the LCL throughout the clear-sky days and results in a negative LCL deficit. The PBL on clear-sky days identified in CAM5-CAPT grows rapidly after the sunrise as a result of high SH flux and is significantly higher than that on the ARM-observed clear-sky days (Fig. 10a). The vigorous PBL development in CAM5-CAPT corresponds to the large warm and dry air entrainment dominating the PBL budget (Fig. 9). On the other hand, the corresponding simulated LCL is even higher, leading to a negative LCL deficit (Fig. 10c). In other words, the PBL height is far from reaching the LCL and thus supports clear skies. Similarly, the clear-sky days identified in EAMv1-RRM is mainly a result of the much higher LCL while the PBL evolution is generally comparable to that on the observed clear-sky days. The much higher LCL in climate models compared with ARM observations is a result of the too warm and too dry surface conditions in the early morning, consistent with the findings from the mixing diagram analysis in Fig. 9. To summarize, on clear-sky days, the initial biases in the surface conditions are relatively more important than the entrainment in explaining the errors on the LA coupling processes in climate models.
The observed LCL on fair-weather ShCu days (Fig. 10e) is similar to that on clear-sky days but with a stronger diurnal evolution of PBL (Fig. 10d). As shown in Fig. 10f, the growth of the observed PBL on fair-weather ShCu days is deep enough to touch the LCL for cloud formation around noon (indicated by positive LCL deficit in this case). Overall, NARR captures the diurnal variations of PBL, LCL, and LCL deficit on fair-weather ShCu days, except for an overestimation of PBL in the afternoon. The deeper PBL growth in NARR may result from its higher sensitivity to surface evaporation, as suggested by Santanello et al. (2015). Different from the ARM observations, the daytime evolution of PBL is much weaker on ShCu days than that on clear-sky days in CAM5-CAPT. But the decrease in LCL from clear-sky days to ShCu days is even greater. As a result, the growth of PBL is just high enough to touch the LCL for cloud formation (Fig. 10f). Similar results are also noted for ShCu days in EAMv1-RRM, where both PBL and LCL decrease significantly compared with those on clear-sky days. The much lower LCL and PBL on ShCu days in climate models compared with those in the ARM observations correspond to the lower cloud fraction and cloud base shown in Fig. 8.
c. CTP-HIlow framework
We use the CTP-HIlow framework to examine the contribution of initial conditions to model biases on the LA coupling processes for clear-sky, ShCu, and late-afternoon deep convection days. The CTP-HIlow framework provides information on whether land surfaces or atmospheric conditions are more likely to influence afternoon convection using the early-morning temperature and humidity profiles (Findell and Eltahir 2003a,b). The CTP is determined by integrating the departure of the temperature profile from the moist adiabatic between 100 and 300 hPa above the ground. A negative (positive) CTP indicates a stable (unstable) lower troposphere hard (easy) for convection initiation. The HIlow is a low-level humidity index, defined as the sum of the dewpoint depressions at 925, 825, and 625 hPa. The higher the HIlow, the drier the atmosphere. The height levels to calculate HIlow here are slightly different from the ones in Findell and Eltahir (2003a,b) and roughly correspond to the levels in the middle of the mixed layer, immediately above the mixed-layer top and in the free atmosphere, respectively, based on observations at SGP. The dewpoint depression is linearly correlated with relative humidity (Lawrence 2005), one of the identified dominant environmental factors in determining the vertical extent of ShCu and the shallow-to-deep convection transition at SGP (Zhang and Klein 2010, 2013).
Figure 11 shows the early-morning (0530 LST) CTP–HIlow distributions of ARM observations, NARR, CAM5-CAPT, and EAMv1-RRM. Here, for each local convection regime, the value within each bin box of 100 J kg−1 × 5°C represents the percentage of days that fall in the certain range of CTP and HIlow. As shown in ARM observations, different local convection regimes generally have no preferential tendencies over certain CTP values, although clear-sky days tend to have a larger percentage toward negative CTP compared with the other two convection regimes (Figs. 11a–c). In contrast, these locally generated regimes can be clearly distinguished by HIlow. Specifically, there are about 42% of the observed clear-sky days with HIlow > 40°C, 45% of the fair-weather ShCu days with 25°C < HIlow ≤ 40°C and 46% of the late-afternoon deep convection days with HIlow ≤ 25°C, respectively. This suggests that a transition between local convection regimes from clear-sky to ShCu to afternoon deep convection is closely associated with lower-troposphere humidity, consistent with our previous studies (Zhang and Klein 2010, 2013).
Distributions of the CTP and HIlow at 0530 LST for (a) clear-sky days, (b) fair-weather shallow cumulus days, and (c) late-afternoon deep convection days from the ARM observations. (d)–(f),(g)–(i),(j)–(l) As in (a)–(c), but for NARR, CAM5-CAPT hindcasts, and EAMv1-RRM nudged runs, respectively. Color contour denotes the percentage of days within each bin box (100 J kg−1 × 5°C) among the total number of sample-days in each convection regime. Four characteristic regions are identified: 1) CTP < 0, 2) CTP > 0 and HIlow > 40°C, 3) CTP > 0 and 25°C < HIlow < 40°C; 4) CTP > 0 and HIlow < 25°C. The sum percentages in each of the four regions are annotated by numbers.
Citation: Journal of Hydrometeorology 22, 2; 10.1175/JHM-D-20-0078.1
The CTP-HIlow distributions of NARR are in general consistent with ARM observations except a few slight shifts of the distribution. On clear-sky days, there is a 10% shift from cases with negative CTP to cases with 25°C < HIlow ≤ 40°C and positive CTP (Fig. 11d). This suggests that NARR tends to be more convectively unstable and moister on clear-sky days. Similar shift is also found on ShCu days but from cases with HIlow > 40°C (Fig. 11e), suggesting that the lower troposphere in NARR is moister than the observed early-morning conditions on ShCu days.
The above-noted trend of clear-sky days toward higher HIlow and late-afternoon deep convection days toward lower HIlow in observations is generally captured by CAM5-CAPT, but the corresponding range of CTP is much narrower than that observed. The early-morning CTP can reach as high as 500 J kg−1 for the observed local convection regimes compared to 200 J kg−1 in CAM5-CAPT. Moreover, about 80% of the ShCu days in CAM5-CAPT fall in the region with HIlow ≤ 25°C. This suggests that the lower atmospheric conditions on ShCu days in CAM5-CAPT tend to be much moister than those on the observed ShCu days. Similar results are noted in EAMv1-RRM, where almost all of the classified ShCu days have HIlow ≤ 40°C. The moister lower atmosphere in models compared with that observed may help explain the underestimated LCL on ShCu days identified in both CAM5-CAPT and EAMv1-RRM (Fig. 10e). This implies that on the ARM-observed ShCu days, cloud forms more often with high SH flux and a rapid PBL growth touching LCL, while on the model-simulated ShCu days, cloud forms more often with a lowered LCL due to a more humid lower troposphere.
Surprisingly, the early-morning atmospheric conditions in the local convective regimes classified from EAMv1-RRM are too stable even on the model-simulated late-afternoon deep convection days (Fig. 11i). The stable lower troposphere maintains in the CTP-HIlow distributions at 1130 LST right before the transition or precipitation onset. To further explore the model-simulated late-afternoon deep convection days in EAMv1-RRM, we separate the precipitation into convective and large-scale precipitation. It is found that large-scale precipitation contributes about 40% to the total precipitation on the late-afternoon deep convection days in EAMv1-RRM (not shown). This is in contrast to the precipitation on the late-afternoon deep convection days in CAM5-CAPT, which is primarily the convective precipitation from the deep convection scheme (Zhang and McFarlane 1995). For the other 60% of the late-afternoon deep convection days in EAMv1-RRM dominated by convective precipitation and with a stable lower troposphere, the deep convection is triggered by positive CAPEs coming from middle-to-upper atmosphere.
5. Summary
We use the 9-yr warm-season (May–August 2004–12) observations at the ARM SGP site to assess the performance of NARR and climate model simulations (CAM5-CAPT hindcasts and EAMv1-RRM nudged runs) in representing the land–PBL–cloud–precipitation coupling processes. To isolate biases from models’ incapability of simulating the propagating MCSs, we focus on local convection regimes including clear-sky, fair-weather ShCu and late-afternoon deep convection days, in which the strongest local LA coupling is expected.
Overall, NARR agrees pretty well with the ARM observations on the daily atmospheric moisture budget and surface energy budget, presumably because of the assimilated precipitation and radiances in NARR (Ruane 2010). However, NARR still significantly overestimates the SWDN, SWUP and surface SH flux (Fig. 5a) which may result from too few clouds and insufficient extinction by aerosols and water vapor suggested by Kennedy et al. (2011). The warm-season precipitation is insufficient in CAM5-CAPT, which is mainly attributable to the inadequate representations of propagating MCSs (Figs. 3a and 7a). The daily atmospheric moisture budget is overall poorly simulated in EAMv1-RRM, which shows a large negative bias in both surface precipitation and evaporation (Fig. 3). Moreover, there is a discrepancy between EAMv1-RRM and the ARM observations in the daily MFCs for local convection regimes. It suggests that large-scale atmospheric conditions also contribute to the biases in EAMv1-RRM at SGP. Additionally, EAMv1-RRM overestimates (underestimates) the surface SH (LH) flux, resulting in a large RMSE in EF (Fig. 5). This could be a result of the combination of a large deficit in precipitation or a misrepresentation of surface evaporation or energy partition and the erroneous feedback between them (Fig. 6).
About 74% and 48% of the ARM-observed clear-sky days are correctly simulated as clear-sky days in CAM5-CAPT and EAMv1-RRM, respectively. But the corresponding hit rate for ShCu days are low, where the majority of the ARM-observed ShCu days are simulated as days with no/little low-level clouds in either model. Only a few of the ARM-observed late-afternoon deep convection days are captured by CAM5-CAPT while about half of these days have diurnal maximum precipitation at hours outside of 1500–2100 LST in CAM5-CAPT (Table 3). The ARM-observed late-afternoon deep convection days in EAMv1-RRM are either simulated as days with no precipitation/drizzling (44%) or days with precipitation peaking too early (33%).
We reclassify the local convection regimes in both CAM5-CAPT and EAMv1-RRM using similar criteria as we classify the days using the ARM data and further evaluate the LA coupling in reanalysis and climate models using three LoCo metrics. This kind of analysis allows us to diagnose the model performance from a different perspective and in a statistical and climatological manner and complements the assessment of local convective regimes regardless of the day-to-day match with observations.
The LA coupling processes in NARR are in general comparable to those observed except that NARR tends to be slightly warmer and drier on the observed clear-sky days (Fig. 9), consistent with Santanello et al. (2015). The clear-sky days reclassified in both CAM5-CAPT and EAMv1-RRM are initially offset as too warm and too dry in the early morning (Fig. 9), which result in an overestimated LCL throughout the day (Fig. 10). In contrast to the ARM observations, the diurnal evolution obtained from the simulations (both CAM5-CAPT and EAMv1-RRM) shows a lower PBL and LCL on ShCu days than the one on clear-sky days. The growth of PBL is just high enough to touch the LCL for cloud formation but both the cloud fraction and the cloud base are much lower than those on the ARM-observed ShCu days (Fig. 8). In observations, ShCu forms as a result of strong SH flux that drives the rapid development of PBL; while in models, ShCu days start with a relatively more humid lower-troposphere that leads to a lowered LCL (Figs. 10 and 11). The total precipitation simulated on the model-based late-afternoon deep convection days is much weaker in intensity compared with that observed (Fig. 8). Moreover, days with late-afternoon deep convection in models tend to present a stable early-morning lower atmosphere more frequently, suggesting that deep convection is triggered more often by elevated instabilities in models.
It should be noted that CAM5-CAPT hindcasts and EAMv1-RRM nudged runs adopt different physical parameterizations as well as different resolutions over the central United States (Table 1). Therefore, in this study, we do not intend to compare the performance between the two models in representing the local LA coupling. However, the results here do imply that it still remains a challenge for EAMv1-RRM, with 0.25° resolution and nudging-to-observations techniques applied outside the CONUS domain, to capture the regional scale and diurnal variability in LA coupling over the central United States. The method of model initialization used in CAM5-CAPT hindcasts is more effective than EAMv1-RRM nudged runs in reproducing more realistic large-scale environments and land surface conditions, thus providing a better perspective to dissect model biases in the LA coupling during the local diurnal convective events. To better understand the role of physical parameterizations in the representation of the local LA coupling processes in climate models, hindcasts with constrained initial and large-scale conditions are desirable for a hierarchy of model development versions at various resolutions.
Acknowledgments
We sincerely thank three anonymous reviewers for their constructive comments, which helped to improve the manuscript. Many thanks go to Dr. Ahmed Tawfik for providing the codes of the LoCo metrics. Data from the U.S. Department of Energy (DOE) as part of the Atmospheric Radiation Measurement (ARM) Climate Research Facility Southern Great Plains site were used. C. Tao and Y. Zhang were mainly supported by the DOE Office of Science Early Career Research Program (ECRP) awarded to Y. Zhang. C. Tao also acknowledges support by the DOE ARM program. Y. Zhang also acknowledges support by the DOE Atmospheric System Research (ASR) program. Q. Tang was supported by the DOE Energy Exascale Earth System Model (E3SM) project. H-Y. Ma and V. Ghate were supported by the DOE ASR program. S. Tang and S. Xie were supported by the DOE ARM program. J. Santanello was supported by NASA’s Science Utilization of SMAP (SUSMAP) program. Lawrence Livermore National Laboratory is operated by Lawrence Livermore National Security, LLC, for the U.S. DOE under contract DE-AC52-07NA27344. We have provided the accesses to the observational datasets used in this work. All the other datasets, including modeling data and the list of local convective event days, will be made available upon request.
REFERENCES
Betts, A. K., 2004: Understanding hydrometeorology using global models. Bull. Amer. Meteor. Soc., 85, 1673–1688, https://doi.org/10.1175/BAMS-85-11-1673.
Bianco, L., and J. M. Wilczak, 2002: Convective boundary layer depth: Improved measurements by Doppler radar wind profiler using fuzzy logic methods. J. Atmos. Oceanic Technol., 19, 1745–1758, https://doi.org/10.1175/1520-0426(2002)019<1745:CBLDIM>2.0.CO;2.
Bianco, L., J. M. Wilczak, and A. B. White, 2008: Convective boundary layer depth estimation from wind profilers: Statistical comparison between an automated algorithm and expert estimations. J. Atmos. Oceanic Technol., 25, 1397–1413, https://doi.org/10.1175/2008JTECHA981.1.
Bogenschutz, P. A., A. Gettelman, H. Morrison, V. E. Larson, C. Craig, and D. P. Schanen, 2013: Higher-order turbulence closure and its impact on climate simulations in the community atmosphere model. J. Climate, 26, 9655–9676, https://doi.org/10.1175/JCLI-D-13-00075.1.
Bretherton, C. S., and S. Park, 2009: A new moist turbulence parameterization in the community atmosphere model. J. Climate, 22, 3422–3448, https://doi.org/10.1175/2008JCLI2556.1.
Bukovsky, M. S., and D. J. Karoly, 2007: A brief evaluation of precipitation from the North American regional reanalysis. J. Hydrometeor., 8, 837–846, https://doi.org/10.1175/JHM595.1.
Caldwell, P. M., and Coauthors, 2019: The DOE E3SM coupled model version 1: Description and results at high resolution. J. Adv. Model. Earth Syst., 11, 4095–4146, https://doi.org/10.1029/2019MS001870.
Clothiaux, E. E., T. P. Ackerman, G. G. Mace, K. P. Moran, R. T. Marchand, M. A. Miller, and B. E. Martner, 2000: Objective determination of cloud heights and radar reflectivities using a combination of active remote sensors at the ARM CART sites. J. Appl. Meteor., 39, 645–665, https://doi.org/10.1175/1520-0450(2000)039<0645:ODOCHA>2.0.CO;2.
Dai, A., and K. E. Trenberth, 2004: The diurnal cycle and its depiction in the community climate system model. J. Climate, 17, 930–951, https://doi.org/10.1175/1520-0442(2004)017<0930:TDCAID>2.0.CO;2.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–828, https://doi.org/10.1002/qj.828.
Dirmeyer, P. A., 2011: The terrestrial segment of soil moisture–climate coupling. Geophys. Res. Lett., 38, L16702, https://doi.org/10.1029/2011GL048268.
Dirmeyer, P. A., and Coauthors, 2012: Simulating the diurnal cycle of rainfall in global climate models: Resolution versus parameterization. Climate Dyn., 39, 399–418, https://doi.org/10.1007/s00382-011-1127-9.
Dirmeyer, P. A., and Coauthors, 2018: Verification of land–atmosphere coupling in forecast models, reanalyses, and land surface models using flux site observations. J. Hydrometeor., 19, 375–392, https://doi.org/10.1175/JHM-D-17-0152.1.
Ek, M. B., and A. A. Holtslag, 2004: Influence of soil moisture on boundary layer cloud development. J. Hydrometeor., 5, 86–99, https://doi.org/10.1175/1525-7541(2004)005<0086:IOSMOB>2.0.CO;2.
Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, https://doi.org/10.1029/2002JD003296.
Ferguson, C. R., E. F. Wood, and R. K. Vinukollu, 2012: A global intercomparison of modeled and observed land–atmosphere coupling. J. Hydrometeor., 13, 749–784, https://doi.org/10.1175/JHM-D-11-0119.1.
Findell, K. L., and E. A. Eltahir, 2003a: Atmospheric controls on soil moisture–boundary layer interactions. Part I: Framework development. J. Hydrometeor., 4, 552–569, https://doi.org/10.1175/1525-7541(2003)004<0552:ACOSML>2.0.CO;2.
Findell, K. L., and E. A. Eltahir, 2003b: Atmospheric controls on soil moisture–boundary layer interactions. Part II: Feedbacks within the continental United States. J. Hydrometeor., 4, 570–583, https://doi.org/10.1175/1525-7541(2003)004<0570:ACOSML>2.0.CO;2.
Findell, K. L., P. Gentine, B. R. Lintner, and C. Kerr, 2011: Probability of afternoon precipitation in eastern United States and Mexico enhanced by high evaporation. Nat. Geosci., 4, 434–439, https://doi.org/10.1038/ngeo1174.
Gettelman, A., and H. Morrison, 2015: Advanced two-moment bulk microphysics for global models. Part I: Off-line tests and comparison with other schemes. J. Climate, 28, 1268–1287, https://doi.org/10.1175/JCLI-D-14-00102.1.
Golaz, J.-C., V. E. Larson, and W. R. Cotton, 2002: A PDF-based model for boundary layer clouds. Part I: Method and model description. J. Atmos. Sci., 59, 3540–3551, https://doi.org/10.1175/1520-0469(2002)059<3540:APBMFB>2.0.CO;2.
Golaz, J.-C., and Coauthors, 2019: The DOE E3SM coupled model version 1: Overview and evaluation at standard resolution. J. Adv. Model. Earth Syst., 11, 2089–2129, https://doi.org/10.1029/2018MS001603.
Guillod, B. P., and Coauthors, 2014: Land-surface controls on afternoon precipitation diagnosed from observational data: Uncertainties and confounding factors. Atmos. Chem. Phys., 14, 8343–8367, https://doi.org/10.5194/acp-14-8343-2014.
Guillod, B. P., B. Orlowsky, D. Miralles, A. J. Teuling, and S. I. Seneviratne, 2015: Reconciling spatial and temporal soil moisture effects on afternoon rainfall. Nat. Commun., 6, 6443, https://doi.org/10.1038/ncomms7443.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Hsu, J., and M. J. Prather, 2009: Stratospheric variability and tropospheric ozone. J. Geophys. Res., 114, D06102, https://doi.org/10.1029/2008JD010942.
Iacono, M., J. Delamere, E. Mlawer, M. Shephard, S. Clough, and W. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.
Jiang, X., N.-C. Lau, and S. A. Klein, 2006: Role of eastward propagating convection systems in the diurnal cycle and seasonal mean of summertime rainfall over the U.S. Great Plains. Geophys. Res. Lett., 33, L19809, https://doi.org/10.1029/2006GL027022.
Kennedy, A. D., X. Dong, B. Xi, S. Xie, Y. Zhang, and J. Chen, 2011: A comparison of MERRA and NARR reanalyses with the DOE ARM SGP data. J. Climate, 24, 4541–4557, https://doi.org/10.1175/2011JCLI3978.1.
Klein, S. A., X. Jiang, J. Boyle, S. Malyshev, and S. Xie, 2006: Diagnosis of the summertime warm and dry bias over the U.S. Southern Great Plains in the GFDL climate model using a weather forecasting approach. Geophys. Res. Lett., 33, L18805, https://doi.org/10.1029/2006GL027567.
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.
Koster, R. D., and Coauthors, 2006: GLACE: The Global Land–Atmosphere Coupling Experiment. Part I: Overview. J. Hydrometeor., 7, 590–610, https://doi.org/10.1175/JHM510.1.
Lamb, P. J., D. H. Portis, and A. Zangvil, 2012: Investigation of large-scale atmospheric moisture budget and land surface interactions over U.S. Southern Great Plains including for CLASIC (June 2007). J. Hydrometeor., 13, 1719–1738, https://doi.org/10.1175/JHM-D-12-01.1.
Lareau, N. P., Y. Zhang, and S. A. Klein, 2018: Observed boundary layer controls on shallow cumulus at the ARM southern Great Plains site. J. Atmos. Sci., 75, 2235–2255, https://doi.org/10.1175/JAS-D-17-0244.1.
Lawrence, M. G., 2005: The relationship between relative humidity and the dewpoint temperature in moist air: A simple conversion and applications. Bull. Amer. Meteor. Soc., 86, 225–234, https://doi.org/10.1175/BAMS-86-2-225.
Lee, M., and Coauthors, 2007: An analysis of the warm-season diurnal cycle over the continental United States and Northern Mexico in general circulation models. J. Hydrometeor., 8, 344–366, https://doi.org/10.1175/JHM581.1.
Liu, X., and Coauthors, 2012: Toward a minimal representation of aerosols in climate models: Description and evaluation in the community atmosphere model CAM5. Geosci. Model Dev., 5, 709–739, https://doi.org/10.5194/gmd-5-709-2012.
Liu, X., P.-L. Ma, H. Wang, S. Tilmes, B. Singh, R. C. Easter, S. J. Ghan, and P. J. Rasch, 2016: Description and evaluation of a new four-mode version of the Modal Aerosol Module (MAM4) within version 5.3 of the Community Atmosphere Model. Geosci. Model Dev., 9, 505–522, https://doi.org/10.5194/gmd-9-505-2016.
Ma, H.-Y., S. Xie, J. S. Boyle, S. A. Klein, and Y. Zhang, 2013: Metrics and diagnostics for precipitation-related processes in climate model short-range hindcasts. J. Climate, 26, 1516–1534, https://doi.org/10.1175/JCLI-D-12-00235.1.
Ma, H.-Y., and Coauthors, 2014: On the correspondence between mean forecast errors and climate errors in CMIP5 models. J. Climate, 27, 1781–1798, https://doi.org/10.1175/JCLI-D-13-00474.1.
Ma, H.-Y., and Coauthors, 2015: An improved hindcast approach for evaluation and diagnosis of physical processes in global climate models. J. Adv. Model. Earth Syst., 7, 1810–1827, https://doi.org/10.1002/2015MS000490.
Ma, H.-Y., and Coauthors, 2018: CAUSES: On the role of surface energy budget errors to the warm surface air temperature error over the central U.S. J. Geophys. Res. Atmos., 123, 2888–2909, https://doi.org/10.1002/2017JD027194.
Ma, H.-Y., and Coauthors, 2020: A multi-year short-range hindcast experiment for evaluating climate model moist processes from diurnal to interannual time scales. Geosci. Model Dev. Discuss., https://doi.org/10.5194/gmd-2020-39.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, https://doi.org/10.1175/BAMS-87-3-343.
Morrison, H., and A. Gettelman, 2008: A new two-moment bulk stratiform cloud microphysics scheme in the NCAR Community Atmosphere Model (CAM3), Part I: Description and numerical tests. J. Climate, 21, 3642–3659, https://doi.org/10.1175/2008JCLI2105.1.
Neale, R. B., and Coauthors, 2012: Description of the NCAR Community Atmosphere Model (CAM 5.0). NCAR Tech. Note NCAR/TN-486+STR, 274 pp., www.cesm.ucar.edu/models/cesm1.0/cam/docs/description/cam5_desc.pdf.
Nigam, S., and A. Ruiz-Barradas, 2006: Seasonal hydroclimate variability over North America in global and regional reanalyses and AMIP simulations: Varied representation. J. Climate, 19, 815–837, https://doi.org/10.1175/JCLI3635.1.
Park, S., and C. S. Bretherton, 2009: The university of Washington shallow convection and moist turbulence schemes and their impact on climate simulations with the community atmosphere model. J. Climate, 22, 3449–3469, https://doi.org/10.1175/2008JCLI2557.1.
Phillips, T. J., and S. A. Klein, 2014: Land-atmosphere coupling manifested in warm-season observations on the U.S. Southern Great Plains. J. Geophys. Res. Atmos., 119, 509–528, https://doi.org/10.1002/2013JD020492.
Phillips, T. J., and Coauthors, 2004: Evaluating parameterizations in general circulation models: Climate simulation meets weather prediction. Bull. Amer. Meteor. Soc., 85, 1903–1916, https://doi.org/10.1175/BAMS-85-12-1903.
Phillips, T. J., and Coauthors, 2017: Using ARM observations to evaluate climate model simulations of land-atmosphere coupling on the U.S. southern Great Plains. J. Geophys. Res. Atmos., 122, 11 524–11 548, https://doi.org/10.1002/2017JD027141.
Qian, Y., M. Huang, B. Yang, and L. K. Berg, 2013: A modeling study of irrigation effects on surface fluxes and land-air-cloud interactions in the southern Great Plains. J. Hydrometeor., 14, 700–721, https://doi.org/10.1175/JHM-D-12-0134.1.
Rasch, P. J., and Coauthors, 2019: An overview of the atmospheric component of the energy Exascale Earth System Model. J. Adv. Model. Earth Syst., 11, 2377–2411, https://doi.org/10.1029/2019MS001629.
Romps, D. M., 2017: Exact expression for the lifting condensation level. J. Atmos. Sci., 74, 3891–3900, https://doi.org/10.1175/JAS-D-17-0102.1.
Ruane, A. C., 2010: NARR’s atmospheric water cycle components. Part II: Summertime mean and diurnal interactions. J. Hydrometeor., 11, 1220–1233, https://doi.org/10.1175/2010JHM1279.1.
Ruiz-Barradas, A., and S. Nigam, 2013: Atmosphere-land surface interactions over the Southern Great Plains: Characterization from pentad analysis of DOE ARM field observations and NARR. J. Climate, 26, 875–886, https://doi.org/10.1175/JCLI-D-11-00380.1.
Santanello, J. A., C. D. Peters-Lidard, S. V. Kumar, C. Alonge, and W.-K. Tao, 2009: A modeling and observational framework for diagnosing local land–atmosphere coupling on diurnal time scales. J. Hydrometeor., 10, 577–599, https://doi.org/10.1175/2009JHM1066.1.
Santanello, J. A., C. D. Peters-Lidard, and S. V. Kumar, 2011a: Diagnosing the sensitivity of local land–atmosphere coupling via the soil moisture–boundary layer interaction. J. Hydrometeor., 12, 766–786, https://doi.org/10.1175/JHM-D-10-05014.1.
Santanello, J. A., and Coauthors, 2011b: Results from Local Land-Atmosphere Coupling (LoCo) Project. GEWEX News, Vol. 21, No. 4, International GEWEX Project Office, Silver Spring, MD, 7–9, www.gewex.org/gewex-content/files_mf/1432209597Nov2011.pdf.
Santanello, J. A., C. D. Peters-Lidard, A. Kennedy, and S. V. Kumar, 2013: Diagnosing the nature of land–atmosphere coupling: A case study of dry/wet extremes in the U.S. Southern Great Plains. J. Hydrometeor., 14, 3–24, https://doi.org/10.1175/JHM-D-12-023.1.
Santanello, J. A., J. Roundy, and P. A. Dirmeyer, 2015: Quantifying the land–atmosphere coupling behavior in modern reanalysis products over the U.S. Southern Great Plains. J. Climate, 28, 5813–5829, https://doi.org/10.1175/JCLI-D-14-00680.1.
Santanello, J. A., and Coauthors, 2018: Land–atmosphere interactions: The LoCo perspective. Bull. Amer. Meteor. Soc., 99, 1253–1272, https://doi.org/10.1175/BAMS-D-17-0001.1.
Seneviratne, S. I., D. Luthi, M. Litschi, and C. Schar, 2006: Land-atmosphere coupling and climate change in Europe. Nature, 443, 205–209, https://doi.org/10.1038/nature05095.
Seneviratne, S. I., T. Corti, E. L. Davin, M. Hirschi, E. B. Jaeger, I. Lehner, B. Orlowsky, and A. J. Teuling, 2010: Investigating soil moisture–climate interactions in a changing climate: A review. Earth-Sci. Rev., 99, 125–161, https://doi.org/10.1016/j.earscirev.2010.02.004.
Song, H.-J., C. R. Ferguson, and J. K. Roundy, 2016: Land–atmosphere coupling at the Southern Great Plains Atmospheric Radiation Measurement (ARM) field site and its role in anomalous afternoon peak precipitation. J. Hydrometeor., 17, 541–556, https://doi.org/10.1175/JHM-D-15-0045.1.
Tang, Q., S. Xie, Y. Zhang, T. J. Phillips, J. A. Santanello, D. R. Cook, L. D. Riihimaki, and K. L. Gaustad, 2018: Heterogeneity in warm-season land-atmosphere coupling over the U.S. Southern Great Plains. J. Geophys. Res. Atmos., 123, 7867–7882, https://doi.org/10.1029/2018JD028463.
Tang, Q., and Coauthors, 2019: Regionally refined test bed in E3SM Atmosphere Model version 1 (EAMv1) and applications for high-resolution modeling. Geosci. Model Dev., 12, 2679–2706, https://doi.org/10.5194/gmd-12-2679-2019.
Tang, S., S. Xie, M. Zhang, Q. Tang, Y. Zhang, S. Klein, and D. R. Cook, 2019: Differences in eddy-correlation and energy-balance surface turbulent heat flux measurements and their impacts on the large-scale forcing fields at the ARM SGP site. J. Geophys. Res. Atmos., 124, 3301–3318, https://doi.org/10.1029/2018JD029689.
Tao, C., Y. Zhang, S. Tang, Q. Tang, H.-Y. Ma, S. Xie, and M. Zhang, 2019: Regional moisture budget and land-atmosphere coupling over the US Southern Great Plains inferred from the ARM long-term observations. J. Geophys. Res. Atmos., 124, 10 091–10 108, https://doi.org/10.1029/2019JD030585.
Tawfik, A. B., and P. A. Dirmeyer, 2014: A process-based framework for quantifying the atmospheric preconditioning of surface triggered convection. Geophys. Res. Lett., 41, 173–178, https://doi.org/10.1002/2013GL057984.
Tawfik, A. B., P. A. Dirmeyer, and J. A. Santanello, 2015: The heated condensation framework. Part I: Description and Southern Great Plains case study. J. Hydrometeor., 16, 1929–1945, https://doi.org/10.1175/JHM-D-14-0117.1.
Taylor, C. M., R. De Jeu, F. Guichard, P. P. Harris, and W. A. Dorigo, 2012: Afternoon rain more likely over drier soils. Nature, 489, 423–426, https://doi.org/10.1038/nature11377.
Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106, 7183–7192, https://doi.org/10.1029/2000JD900719.
Terai, C. R., P. M. Caldwell, S. A. Klein, Q. Tang, and M. L. Branstetter, 2018: The atmospheric hydrologic cycle in the ACME v0.3 model. Climate Dyn., 50, 3251–3279, https://doi.org/10.1007/s00382-017-3803-x.
Van Weverberg, K., and Coauthors, 2018: CAUSES: Attribution of surface radiation biases in NWP and climate models near the U.S. Southern Great Plains. J. Geophys. Res. Atmos., 123, 3612–3644, https://doi.org/10.1002/2017JD027188.
Wei, J., H. Su, and Z.-L. Yang, 2016: Impact of moisture flux convergence and soil moisture on precipitation: A case study for the southern United States with implications for the globe. Climate Dyn., 46, 467–481, https://doi.org/10.1007/s00382-015-2593-2.