• ARM, 1994: Balloon-borne sounding system (SONDEWNPN). Updated daily, ARM Climate Research Facility, accessed 2 June 2016, doi:10.5439/1021460.

    • Crossref
    • Export Citation
  • ARM, 1996: Soil Water and Temperature Profiling System (SWATS). Updated daily, ARM Climate Research Facility, accessed 16 June 2016, doi:10.5439/1150274.

    • Crossref
    • Export Citation
  • Betts, A. K., 1992: FIFE atmospheric boundary layer budget methods. J. Geophys. Res., 97, 18 52318 531, doi:10.1029/91JD03172.

  • Dirmeyer, P. A., Z. Wang, M. J. Mbuh, and H. E. Norton, 2014: Intensified land surface control on boundary layer growth in a changing climate. Geophys. Res. Lett., 41, 12901294, doi:10.1002/2013GL058826.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, doi:10.1029/2002JD003296.

    • Search Google Scholar
    • Export Citation
  • Ferguson, C. R., and E. F. Wood, 2011: Observed land–atmosphere coupling from satellite remote sensing and reanalysis. J. Hydrometeor., 12, 12211254, doi:10.1175/2011JHM1380.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Findell, K. L., and E. A. B. Eltahir, 2003a: Atmospheric controls on soil moisture–boundary layer interactions. Part I: Framework development. J. Hydrometeor., 4, 552569, doi:10.1175/1525-7541(2003)004<0552:ACOSML>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Findell, K. L., and E. A. B. Eltahir, 2003b: Atmospheric controls on soil moisture–boundary layer interactions. Part II: Feedbacks within the continental United States. J. Hydrometeor., 4, 570583, doi:10.1175/1525-7541(2003)004<0570:ACOSML>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 434439, doi:10.1038/ngeo1174.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Koster, R. D., M. J. Suarez, A. Ducharne, M. Stieglitz, and P. Kumar, 2000: A catchment-based approach to modeling land surface processes in a general circulation model 1. Model structure. J. Geophys. Res., 105, 24 80924 822, doi:10.1029/2000JD900327.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Koster, R. D., and et al. , 2006: GLACE: The Global Land–Atmosphere Coupling Experiment. Part I: Overview. J. Hydrometeor., 7, 590610, doi:10.1175/JHM510.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lintner, B. R., P. Gentine, K. L. Findell, and G. D. Salvucci, 2015: The Budyko and complementary relationships in an idealized model of large-scale land–atmosphere coupling. Hydrol. Earth Syst. Sci., 19, 21192131, doi:10.5194/hess-19-2119-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Owe, M., R. de Jeu, and T. Holmes, 2008: Multisensor historical climatology of satellite-derived global land surface moisture. J. Geophys. Res., 113, F01002, doi:10.1029/2007JF000769.

    • Search Google Scholar
    • Export Citation
  • Rienecker, M. M., and et al. , 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 36243648, doi:10.1175/JCLI-D-11-00015.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roundy, J. K., and E. F. Wood, 2015: The attribution of land–atmosphere interactions on the seasonal predictability of drought. J. Hydrometeor., 16, 793810, doi:10.1175/JHM-D-14-0121.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roundy, J. K., C. R. Ferguson, and E. F. Wood, 2013: Temporal variability of land–atmosphere coupling and its implications for drought over the southeast United States. J. Hydrometeor., 14, 622635, doi:10.1175/JHM-D-12-090.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roundy, J. K., C. R. Ferguson, and E. F. Wood, 2014: Impact of land–atmospheric coupling in CFSv2 on drought prediction. Climate Dyn., 43, 421, doi:10.1007/s00382-013-1982-7.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saha, S., and et al. , 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 10151057, doi:10.1175/2010BAMS3001.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saha, S., and et al. , 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 21852208, doi:10.1175/JCLI-D-12-00823.1.

  • 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, 577599, doi:10.1175/2009JHM1066.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Santanello, J. A., and et al. , 2011: Local land–atmosphere coupling (LoCo) research: Status and results. GEWEX News, Vol. 21, No. 4, International GEWEX Project Office, Silver Spring, MD, 7–9. [Available online at http://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, 324, doi:10.1175/JHM-D-12-023.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 58135829, doi:10.1175/JCLI-D-14-00680.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 125161, doi:10.1016/j.earscirev.2010.02.004.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Song, H.-J., C. R. Ferguson, and J. K. Roundy, 2015: 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, 541556, doi:10.1175/JHM-D-15-0045.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stommel, H., 1947: Entrainment of air into a cumulus cloud. J. Meteor., 4, 9194, doi:10.1175/1520-0469(1947)004<0091:EOAIAC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Susskind, J., J. M. Blaisdell, L. Iredell, and F. Keita, 2011: Improved temperature sounding and quality control methodology using AIRS/AMSU data: The AIRS Science Team version 5 retrieval algorithm. IEEE Trans. Geosci. Remote Sens., 49, 883907, doi:10.1109/TGRS.2010.2070508.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 19291945, doi:10.1175/JHM-D-14-0117.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Taylor, C. M., R. A. M. de Jeu, F. Guichard, P. P. Harris, and W. A. Dorigo, 2012: Afternoon rain more likely over drier soils. Nature, 489, 423426, doi:10.1038/nature11377.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    The basis for the coupling regime classification method (Roundy et al. 2013) where (a) is an example of the three variables (CTP, HI, and SM) used in the coupling classification and the resulting classification of the CTP–HI space based on SM, with (b) an example of a dry and wet coupling event for a grid (36.5°N, 97.5°W) in the SGP in the United States based on data from MERRA.

  • View in gallery

    The atmospheric profile and corresponding CTP and HI at 0730 UTC for Aqua (peach) and 0530 UTC for in situ (red) on a day during (a) a dry year (7 Jun 2006) and (b) a wet year (3 Jun 2007) for the SGP location (36.5°N, 97.5°W).

  • View in gallery

    Comparison of the (a) CTP, (b) HI, and (c) SM from satellite remote sensing (Aqua) and reanalysis (MERRA and CFSR) with in situ observations for a point in the SGP (36.5°N, 97.5°W) for the available data from 2003 to 2015. The regression line (red), Pearson correlation rp, and Spearman correlation rs are also given.

  • View in gallery

    The classified CTP–HI space from in situ observations, satellite (Aqua), and reanalysis (MERRA and CFSR) for a point in the SGP (36.5°N, 97.5°W). The given percentage is the percent of the in situ classification that is consistent with the given dataset.

  • View in gallery

    Comparison of in situ CDI with reanalysis (MERRA and CFSR) and satellite remote sensing (Aqua) for a point in the SGP (36.5°N, 97.5°W) for (a) monthly time series from 2003 to 2015; (b) scatterplots of the monthly values for May–September, with the dark gray points from 2003 to 2011 and the light gray points from 2012 to 2015; and (c) the spatial variability of the CDI in June 2007. The regression line (red), rp, and rs are also given in (b).

  • View in gallery

    The monthly standardized anomaly rs (2003–15) of the CDI with P, TSM, DAT, VPD, EF, BLH, CTP, and HI [(a) provides the legend (colored shapes) and (b) provides a map of the six climate regions] from (c) Aqua CDI and MERRA, (d) Aqua CDI and CFSR, and (e) MERRA CDI and MERRA. The horizontal red dashed lines in (c)–(e) indicate statistical significance at p = 0.05 for the monthly values from May to September (r = 0.24) and all the monthly values’ significance (r = 0.16).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 117 117 15
PDF Downloads 110 110 12

Utility of Satellite Remote Sensing for Land–Atmosphere Coupling and Drought Metrics

View More View Less
  • 1 Department of Civil, Environmental, and Architectural Engineering, University of Kansas, Lawrence, Kansas
  • | 2 Hydrological Sciences Laboratory, Goddard Space Flight Center, Greenbelt, Maryland
© Get Permissions
Full access

Abstract

Feedbacks between the land and the atmosphere can play an important role in the water cycle, and a number of studies have quantified land–atmosphere (LA) interactions and feedbacks through observations and prediction models. Because of the complex nature of LA interactions, the observed variables are not always available at the needed temporal and spatial scales. This work derives the Coupling Drought Index (CDI) solely from satellite data and evaluates the input variables and the resultant CDI against in situ data and reanalysis products. NASA’s Aqua satellite and retrievals of soil moisture and lower-tropospheric temperature and humidity properties are used as input. Overall, the Aqua-based CDI and its inputs perform well at a point, spatially, and in time (trends) compared to in situ and reanalysis products. In addition, this work represents the first time that in situ observations were utilized for the coupling classification and CDI. The combination of in situ and satellite remote sensing CDI is unique and provides an observational tool for evaluating models at local and large scales. Overall, results indicate that there is sufficient information in the signal from simultaneous measurements of the land and atmosphere from satellite remote sensing to provide useful information for applications of drought monitoring and coupling metrics.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author e-mail: Joshua K. Roundy, jkroundy@ku.edu

Abstract

Feedbacks between the land and the atmosphere can play an important role in the water cycle, and a number of studies have quantified land–atmosphere (LA) interactions and feedbacks through observations and prediction models. Because of the complex nature of LA interactions, the observed variables are not always available at the needed temporal and spatial scales. This work derives the Coupling Drought Index (CDI) solely from satellite data and evaluates the input variables and the resultant CDI against in situ data and reanalysis products. NASA’s Aqua satellite and retrievals of soil moisture and lower-tropospheric temperature and humidity properties are used as input. Overall, the Aqua-based CDI and its inputs perform well at a point, spatially, and in time (trends) compared to in situ and reanalysis products. In addition, this work represents the first time that in situ observations were utilized for the coupling classification and CDI. The combination of in situ and satellite remote sensing CDI is unique and provides an observational tool for evaluating models at local and large scales. Overall, results indicate that there is sufficient information in the signal from simultaneous measurements of the land and atmosphere from satellite remote sensing to provide useful information for applications of drought monitoring and coupling metrics.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author e-mail: Joshua K. Roundy, jkroundy@ku.edu

1. Introduction

In the absence of strong advective influences, land–atmosphere (LA) coupling (Seneviratne et al. 2010) drives the diurnal cycle of clouds and precipitation that can greatly impact the water cycle. As a result, there has been a great deal of work to quantify LA interactions and feedbacks through observations and prediction models. Much of this work has been carried out by the Global Energy and Water Exchanges project (GEWEX) Global Land/Atmosphere System Study (GLASS) local land–atmosphere coupling (LoCo; Santanello et al. 2011) working group. As part of this work a suite of diagnostics has been developed, ranging in applicability from observations to models and spanning a broad range of spatiotemporal scales (Ferguson and Wood 2011; Lintner et al. 2015; Dirmeyer et al. 2014; Tawfik et al. 2015). For example, mixing diagrams (Betts 1992; Santanello et al. 2009; Stommel 1947) are recommended to analyze entrainment into clouds and boundary layer processes at a point scale. In contrast, the rainfall triggering feedback strength (TFS) of Findell et al. (2011) quantifies how rainfall frequency changes with surface evaporative fraction and requires model data over a period of 90 days or longer. Perhaps most well known is the model-based coupling strength of the Global Land–Atmosphere Coupling Experiment (GLACE; Koster et al. 2006): coherence among members Ω is computed for two model ensembles—one with prescribed soil moisture and the other with freely evolving soil moisture—and the difference ΔΩ is deemed the coupling strength. The overall applicability of these respective LA coupling metrics is inherently limited by the ability to observe the variables required by each, which for most remains only at the point scale or during short-term field experiments because of the simultaneous soil moisture, surface flux, boundary layer, and precipitation measurement requirements.

Satellite data offer the ability to obtain some of these variables globally and routinely (and thus have the most promise for GCM and model development applications), but have been limited to date (Ferguson and Wood 2011; Roundy et al. 2013; Taylor et al. 2012). To make satellite observations useful for informing and improving the LA interactions within the models requires further development of satellite-based metrics. The Coupling Drought Index (CDI) developed by Roundy et al. (2013) is such a metric since it has application to LA interactions and drought and can be calculated entirely from satellite remote sensing. The CDI is based on a classification of LA interactions into regimes built off of the work of Findell and Eltahir (2003a,b), who demonstrated the preferential tendency for convective rainfall over wet (i.e., wet advantage) versus dry soils (i.e., dry advantage), depending on low-level atmospheric humidity (HI) and instability [i.e., convective triggering potential (CTP)]. The CTP is a measure of atmospheric stability defined as the area between the temperature profile and a moist adiabat from 100 to 300 hPa above the surface. The HI is a measure of low-level boundary layer moisture given by the sum of the dewpoint depression at 50 and 150 hPa above the surface. Thus, the regimes are strictly a function of lower-troposphere temperature profiles and moisture condition.

The two-dimensional space that comprises the CTP and HI relationship can then be classified into regimes based on the ability of the soil moisture (SM) state to initiate convection (Findell and Eltahir 2003b). Later work by Ferguson and Wood (2011) applied this classification approach to different datasets and regions and showed that the classified space presented by Findell and Eltahir was too stringent. Roundy et al. (2013) developed a method of using local statistics of top-layer soil moisture to classify the wet-advantage and dry-advantage subspaces within the CTP–HI space regionally. This approach separates the CTP–HI space into bins and uses the two-sample Kolmogorov–Smirnov (KS) test to compare the distribution of SM in each bin against the climatological SM. Bins of the CTP–HI space with predominantly wetter soils are considered wet coupling and bins that are predominantly drier are dry coupling. Bins that are neither dry nor wet (in a climatological sense) are classified as transitional, and bins with few samples are considered atmospherically controlled. The rationale for this approach is that there is an inherent connection between the soil moisture and heat flux partitioning that causes a persistence into the dry and wet coupling regimes that is driven by the feedback between the land and the atmosphere. Because of the sensitivity of bin size and the significance level of the KS test, the classification uses an ensemble approach where each ensemble member utilizes a different bin size and significance level. This accounts for the sensitivity of these classification parameters and provides a means to quantify the uncertainty. The final discrete classification is determined based on the uncertainty in each bin [see Roundy et al. (2013) for more details]. Although there is similarity between the Findell and Eltahir (2003b) and Roundy et al. (2013) classification, the latter is based on soil moisture and includes days with and without convective precipitation. To denote this difference, the regime names in Roundy et al. are referred to simply as “dry coupling” and “wet coupling” to indicate the persistent nature of the overall dry and wet events, respectively.

A schematic of the three variables used in the classification (CTP, HI, and SM) and the classified CTP–HI space for a grid cell in the southern Great Plains (SGP) using the Modern-Era Retrospective Analysis for Research and Applications (MERRA) is given in Fig. 1a. As illustrated in Fig. 1a, the CTP is calculated by integrating the area between the moist adiabat and the temperature profile. The CTP in Fig. 1a is positive and indicates an unstable atmosphere. If the moist adiabat is cooler than the temperature profile, then the CTP is negative and indicates a stable atmosphere. The HI is a measure of the atmospheric humidity and is calculated as the sum of the dewpoint depression at 50 and 150 hPa above the surface. A large value of HI, as shown in Fig. 1a, is indicative of a dry atmosphere. As the dewpoint temperature approaches the temperature profile, the atmosphere moves closer to saturation and the HI decreases. A climatological sample of daily CTP, HI, and SM are then used to create the classification of the CTP–HI space. To do this, the CTP–HI space is broken up into bins and each is classified based on the soil moisture values that fall into that bin by the method described above. Once the CTP–HI space is classified, it is used to generate a daily coupling classification based on the location of the CTP and HI for that day. For example, given the classified CTP–HI space in Fig. 1, a day with CTP of 400 J kg−1 and an HI of 30°C would be classified as dry coupling.

Fig. 1.
Fig. 1.

The basis for the coupling regime classification method (Roundy et al. 2013) where (a) is an example of the three variables (CTP, HI, and SM) used in the coupling classification and the resulting classification of the CTP–HI space based on SM, with (b) an example of a dry and wet coupling event for a grid (36.5°N, 97.5°W) in the SGP in the United States based on data from MERRA.

Citation: Journal of Hydrometeorology 18, 3; 10.1175/JHM-D-16-0171.1

Multiple days with the same coupling classification are considered to be an event and are called dry or wet coupling events. These events can persist from days to weeks. An example of a persistent dry and wet coupling event that occurred in the same year (2000) for a grid cell in the SGP is given in Fig. 1b based on MERRA. Vertical dashed lines denote the beginning and ending of the event as determined by the daily classification, where the start of the event is the first day with a daily classification of dry or wet coupling and the end of the event is the last day of the consistent daily classification of dry or wet coupling. Persistent events, such as those depicted in Fig. 1b, can have large impacts on the local water and energy cycles. To demonstrate this, time series of daily average SM, evaporative fraction (EF; ratio of latent heat flux to available energy), boundary layer height (BLH), lifting condensation level (LCL; the level to which a parcel of air can be lifted adiabatically before it becomes saturated) deficit (difference between the LCL and BLH), and the nighttime and daytime precipitation are also included. The dry coupling event is typified by low soil moisture, a small surface EF, a large boundary layer height, and a large LCL deficit. Toward the end of the dry coupling event the BLH increases and the LCL deficit decreases because of an increase in BLH. In contrast, the wet coupling event has high soil moisture, a large EF, small BLH, and a small LCL deficit. The wet coupling event also shows a decrease in the LCL toward the end of the event; however, BLH also decreases, which indicates that the decrease in the LCL deficit is due to a decrease in the LCL because of the large latent heat flux. Daytime precipitation occurs during both dry and wet coupling events, but the precipitation is less frequent and smaller in magnitude during the dry coupling event. Although the persistence in these coupling regimes can be explained by LA feedbacks, it is important to note that advected moisture into the region also plays a key role (Song et al. 2015), and any dry or wet coupling event is ultimately a combination of local feedback mechanism and large-scale circulation patterns.

The cumulative negative (dry coupling promotes drying) and positive feedback (wet coupling promotes wetting) of these events is the foundation of the CDI, which is simply the number of dry coupling days minus the wet coupling days, divided by the total number of days over a period of time. CDI has a range from −1 (all wet coupling) to +1 (all dry coupling) and gives an average measure of coupling over the chosen time window. The CDI has been successfully applied in the evaluation of reanalysis and seasonal forecasts (Roundy et al. 2013, 2014; Roundy and Wood 2015).

One of the unique characteristics of metrics based on CTP, HI, and SM is that these variables can be derived from simultaneous measurements from instruments on board NASA’s Aqua satellite. Specifically, the Atmospheric Infrared Sounder (AIRS) provides temperature and moisture profiles that can be used to estimate the CTP and HI while measurements from the Advanced Microwave Scanning Radiometer for EOS (AMSR-E) instrument can be used to derive soil moisture. The simultaneous measurement of both land and atmospheric variables from the Aqua satellite provides a unique large-scale and observationally based dataset for developing coupling metrics suited for evaluating weather and climate models. This work aims to assess the utility and uncertainty of the satellite data for application to the coupling classification. This is done by first introducing the in situ, satellite, and reanalysis datasets and methods utilized in this study (section 2). Next, a comparison of the measurements and the derived variables (CTP, HI, and SM) across datasets is made (section 3a) followed by an evaluation of utilizing these variables to the coupling classification and CDI in section 3b. The CDI from remote sensing is then compared to other common surface and boundary layer variables from reanalysis in section 3c and is followed by discussion and conclusions in section 4.

2. Datasets and methods

a. Datasets

In this work, four different datasets are used to calculate the CTP, HI, and SM needed for the CDI classification, and they include satellite remote sensing, reanalysis, and in situ data. Table 1 provides a summary of the datasets used, the type of data, and the temporal range of the dataset that was utilized in this study. The satellite remote sensing data are from the NASA Aqua satellite, which includes AIRS as well as AMSR-E. The AIRS data used in this study are from the level 3, version 6, data product and provide 12 vertical levels of consistent measurements of temperature and humidity (Susskind et al. 2011). These AIRS observations are provided twice daily at 0130 and 1330 local time (LT) on a 1° × 1° global grid from August 2002 to present. Only the 0130 LT (descending overpass) data are used in this study, as they provide a better measure of the atmosphere in early morning before the impact of the daytime surface heat fluxes. These observations of atmospheric temperature and humidity enable the calculation of the CTP and HI. The measurements from the AMSR-E aboard Aqua are used to derive soil moisture from the Land Parameter Retrieval Model (LPRM; Owe et al. 2008) and are representative of the top-2-cm soil layer. Unfortunately, the AMSR-E instrument failed in 2011 and limits the availability of soil moisture data from 2002 to 2011.

Table 1.

A summary of the relative characteristics from each dataset used to derive the CDI.

Table 1.

Reanalysis products are also used in this study, as they provide global, continuous, and long-term records of the climate system constructed by combining observations and models. Reanalysis datasets also provide a means for initializing forecast models with the best temporally and spatially continuous estimates of Earth system variables for weather and climate forecasts. It is important to remember that although reanalysis assimilates observations, there is still a large component that is based on the parameterizations and assumptions inherent in the model. Therefore, while reanalyses may assimilate a similar set of observations, they may provide different representations of the climate because of the differences in the assimilation technique and modeling.

Two different reanalysis datasets are considered, NASA’s MERRA (Rienecker et al. 2011) and the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR; Saha et al. 2010). These datasets were chosen because of their global coverage and availability. MERRA is based on NASA’s Goddard Earth Observing System, version 5 (GEOS-5; Rienecker et al. 2011), which utilizes the Catchment LSM (Koster et al. 2000). The top soil layer in Catchment represents the uppermost 0–2-cm layer. MERRA has a 0.5° × 0.667° horizontal resolution over the globe with 72-layer vertical resolution, and the assimilated data are provided at 6-hourly increments from 1979 to present. CFSR includes the Global Forecast System, version 2, atmospheric component with 64 layers in the vertical with a horizontal resolution of T382 (0.313°); the coupled Modular Ocean Model, version 4, with 40 vertical layers; and the Noah land surface model (Ek et al. 2003), which has four soil layers that cover 0–10, 10–40, 40–100, and 100–200 cm, respectively. Although CFSR has a T382 horizontal resolution, the atmospheric data are archived at a 0.5° resolution, while the land surface data are archived at the T382 resolution. The original CFSR has a record length from January 1979 through March 2011. In April 2011, the updated Climate Forecast System, version 2, was put into operation to produce real-time CFSR data through the present (Saha et al. 2014). These combined datasets make up the whole of the CFSR data used in this study that provide 6-hourly analysis data from 1979 to present.

The last type of data utilized in this study are in situ data, and they provide direct measurements of the atmosphere and the land surface as part of the U.S. Department of Energy’s (DOE) continuous record of observational data from the ARM Southern Great Plains site (covering a large part of Oklahoma and Kansas). Because of this unique dataset that includes atmospheric and surface variables, the SGP has been the test site for a number of studies (Santanello et al. 2013, 2015). Specifically, radiosonde profiles (ARM 1994) and top-layer soil moisture from the Soil Water and Temperature Profiling System (SWATS; ARM 1996) from the ARM central facility (36.610°N, 97.4899°W) near Lamont, Oklahoma, were utilized. SWATS provides six levels of soil moisture measurements for two soil profiles (east and west) that are separated by a distance of 1 m. Only the measurements at 5 cm are utilized and are calculated as the average of the two profile measurements. The radiosonde data provide high-vertical-resolution measurements of atmospheric temperature and humidity that can be utilized for calculating the CTP and HI. The radiosonde data are routinely collected four times a day at approximately 0530, 1130, 1730, and 2330 UTC, and the soil moisture is collected hourly.

b. Methods

One of the challenges of comparing all the datasets is the different spatial resolution, domains, and temporal ranges. To make consistent comparisons, all the datasets are upscaled to the 1° × 1° global grid of the AIRS data through bin averaging. To make a comparison to the SGP site (36.610°N, 97.4899°W), the containing grid cell (36.5°N, 97.5°W) from the 1° × 1° global grid of the AIRS is used. In addition to spatial differences, there is also a temporal inconsistency between the datasets. The Aqua satellite data are acquired around 0130 LT (0730 UTC), where the reanalysis data are provided every 6 h (0000, 0600, 1200, and 1800 UTC) and the in situ data are also available approximately every 6 h (0530, 1130, 1730, and 2330 UTC), which is a 1.5- and 2-h difference for the reanalysis and in situ measurements, respectively. To account for this difference in time, the SM and atmospheric profile data from in situ and reanalysis are linearly interpolated to correspond with the satellite overpass. This temporal linear interpolation in time is done before calculating the CTP and HI. This temporal interpolation provides a reasonable estimate since the nighttime profiles of temperature and humidity are typically slowly varying in the early morning hours (e.g., 0000–0600 LT) in terms of their bulk structure in the lower troposphere, while SM evolves on much slower time scales overall.

These spatially and temporally consistent estimates of CTP, HI, and SM are used to classify the CTP–HI space and give a daily coupling classification following the procedures outlined in Roundy et al. (2013). Because of the spatial consistency of the coupling regimes, earlier work used all the grid cells in the entire southeastern United States for the classification (Roundy et al. 2013). While there is general consistency in the classification over regions with similar climate, a regional classification leads to abrupt spatial changes when moving across regional boundaries. To overcome this limitation, Roundy et al. (2014) included the local classification of each grid cell while maintaining regional consistency by utilizing the surrounding grids cells to provide a spatially consistent classification. As compared to utilizing the grid data only, incorporating the surrounding grid cells provides an increased sample size that leads to a robust ensemble that accounts for the uncertainty in the classification. This technique results in a classification with weakened spatial heterogeneity as compared to the raw atmospheric profiles and SM, but still represents the larger spatial patterns. This methodology of using the surrounding grid cells is used to provide the coupling classification for the reanalysis and remote sensing datasets.

As this is the first time that in situ observations have been used in the classification methodology, the classification of point data presents some challenges. One major challenge is producing a unique classification for the in situ data given the absence of surrounding grid cells to incorporate in the classification. One of the key aspects of the classification methodology is to quantify the uncertainty in the CTP–HI space by using an ensemble of bin sizes and significance levels. The ensemble parameters (i.e., number of bin sizes, significance levels, and uncertainty thresholds) were developed by Roundy et al. (2014) for gridded data that incorporate the nearest grid cells and have been used for a number of studies (Roundy et al. 2014; Roundy and Wood 2015; Santanello et al. 2015; Song et al. 2015). Applying the ensemble parameters from the gridded data to a single point drowns out the signal because of the impact of small bin sizes and strict significance levels for the smaller sample size. To account for this difference in the in situ data, a series of tests were performed with the Aqua data to adjust the ensemble parameters to achieve a consistent classification between utilizing a single grid cell only and the grid cell with the surrounding grid cells. This resulted in bin sizes ranging from 7 to 17 and significance levels from 10% to 15%, as compared to bins ranging from 10 to 35 and significance levels from 1% to 5%. The lower significance level indicates more uncertainty in the classification and that resulted in a point classification with a smaller regime classification. This is consistent with the results from Roundy et al. (2013) that showed that a smaller sample size resulted in a consistent yet smaller regime classification. Notwithstanding the smaller regime classification, the point specific classification parameters yield a consistent classification and are used for the in situ data.

To produce a unique classification that accounts for the characteristics of a dataset requires a training period that must be consistent across all the datasets because of the sensitivity of the training period on the classification. The maximum consistent training period across all datasets is a 9-yr period from 2003 to 2011. Although the classification of the CTP–HI space is only done for 2003–11, the daily coupling classification only requires daily values of CTP and HI once the CTP–HI space is classified. Therefore, the analysis will focus on the full period of data availability from 2003 to 2015 for all datasets (see Table 1). In this sense, the period from 2012 to 2015 acts as a cross validation period as the CDI is being applied for the period that is different from the training period.

3. Results

a. Derived variable intercomparison

Observations from Aqua are first compared with in situ measurements of the three variables used in the CDI: CTP, HI, and SM. A comparison of the atmospheric profiles of temperature and humidity (given as dewpoint temperature) for the in situ observations and the satellite data are given in Fig. 2 for a day in a dry year (2006) and a wet year (2007) in the SGP. In comparing these datasets, the higher level of vertical detail in the radiosonde data is evident. Notwithstanding the low resolution in the vertical, the satellite profiles of atmospheric temperature show a good agreement with the in situ observations. In contrast, the lack of vertical resolution in the satellite observations is more damaging in terms of dewpoint temperature. These characteristics directly translate to the CTP and HI. For the CTP, there is good agreement between the in situ and satellite observations, with small relative differences. The HI, on the other hand, shows a larger disagreement between the in situ and satellite observations because of the lack of vertical detail in the dewpoint temperature from AIRS. These results are consistent for both the dry and wet years. The number of observations in the CTP–HI range (50–300 hPa above the surface) varies by day and location. For the examples shown in Fig. 2, the satellite observations have 3 and 2 measurements in the CTP–HI range compared to the 516 and 411 measurements from in situ. This represents a substantial difference in the vertical that is noticeable in Fig. 2 and is likely one of the main causes for the discrepancy between the CTP and HI.

Fig. 2.
Fig. 2.

The atmospheric profile and corresponding CTP and HI at 0730 UTC for Aqua (peach) and 0530 UTC for in situ (red) on a day during (a) a dry year (7 Jun 2006) and (b) a wet year (3 Jun 2007) for the SGP location (36.5°N, 97.5°W).

Citation: Journal of Hydrometeorology 18, 3; 10.1175/JHM-D-16-0171.1

The above analysis only considers two days chosen at random during a dry year and a wet year, but comparing the CTP and HI over a larger time period and extending the comparison to include the comparison of in situ observations with reanalysis can yield further insights. This comparison is given in Fig. 3 for the same location in the SGP but covering all available data from 2003 to 2015. For each variable, only days that have data from in situ, satellite, and reanalysis are shown in Fig. 3. This results in a CTP and HI comparison that includes data from 2003 to 2015, while the SM comparison only includes 2003–11 because of the short record of AMSR-E data. For the CTP, the satellite observations show the largest scatter with in situ observations with Pearson and Spearman correlations of 0.71 and 0.78, respectively, as compared to 0.94 and 0.96 for MERRA and 0.92 and 0.94 for CFSR. A similar relationship can be seen for the HI, with the reanalysis datasets showing a strong correlation with in situ, while the satellite data show much more scatter with Pearson and Spearman correlations of 0.74 and 0.73, respectively. Although the two days shown in Fig. 2 indicate that the HI is dry compared to the in situ observations, the regression line matches well with the one-to-one line with a slope of 0.97 and an x intercept of 1.26 that indicates that there is a wet bias (Aqua HI too low), particularly for the driest HI values from in situ. The larger scatter between satellite CTP and HI and in situ is likely due to the low resolution of the vertical levels from a satellite that fails to capture the fine details (see Fig. 2).

Fig. 3.
Fig. 3.

Comparison of the (a) CTP, (b) HI, and (c) SM from satellite remote sensing (Aqua) and reanalysis (MERRA and CFSR) with in situ observations for a point in the SGP (36.5°N, 97.5°W) for the available data from 2003 to 2015. The regression line (red), Pearson correlation rp, and Spearman correlation rs are also given.

Citation: Journal of Hydrometeorology 18, 3; 10.1175/JHM-D-16-0171.1

Figure 3c shows the in situ SM at the ARM site compared against the Aqua/AMSR-E SM retrieval and the reanalysis products. The SM for each dataset is normalized by the maximum and minimum value (essentially resulting in a moisture availability) in order to account for the difference in the dynamic ranges of SM in each product. CFSR shows the highest correlation with in situ SM with Pearson and Spearman correlations of 0.66 and 0.69, respectively. MERRA and Aqua/AMSR-E SM have slightly lower correlations of 0.58–0.63 and 0.56–0.6, respectively. Overall, the SM datasets show a greater spread and much lower correlations than the CTP or HI. There are three main reasons why the soil moisture data do not compare as well as the CTP and HI across the datasets. First, the inconsistency is likely partially due to the nature of soil moisture heterogeneity at a single site versus that of a large grid cell. While there is a similar difference in scale for the CTP and HI, the atmosphere is more homogeneous over the grid scale compared to the SM. Second, there are linear features present in the in situ data that show little sensitivity to changes in SM from the reanalyses and satellite. This is a known limitation of the SWATS instrument where it is insensitive to soil moisture variations at certain thresholds (and is being rectified by the installation of new SM instruments at the SGP sites). The third reason for the inconsistency is the difference in the depth of each of the measurements. The in situ observations are at 5 cm, while MERRA and Aqua cover the 0–2-cm layer and CFSR covers the 0–10-cm layer. This could be the reason that the CFSR matches better with the in situ since the average point of the top layer matches with the in situ measurement. Notwithstanding the spatial scale, measurement errors, and vertical difference in the measurements, there is still a reasonable amount of consistency that can capture larger regimes of SM that makes it a useful application with the CDI.

b. Coupling classification and CDI

This section extends the previous comparisons to the classification of the CTP–HI space and the CDI. As described above, the classification identifies areas in the two-dimensional space made up of the CTP and HI that have consistent statistics of soil moisture. Thus, the classified CTP–HI space is an integration of the three variables previously compared that identifies a connection or “coupling” of these variables. The classified CTP–HI space from in situ, satellite, and both reanalysis datasets is given in Fig. 4 for the SGP. All datasets show areas classified as dry and wet coupling and show relative consistency between wet and dry coupling locations within the CTP–HI space. The in situ classification has smaller regions of wet and dry coupling and a boxier shape because of the smaller sample size that necessitated an adjustment of the ensemble parameters as part of the classification algorithm.

Fig. 4.
Fig. 4.

The classified CTP–HI space from in situ observations, satellite (Aqua), and reanalysis (MERRA and CFSR) for a point in the SGP (36.5°N, 97.5°W). The given percentage is the percent of the in situ classification that is consistent with the given dataset.

Citation: Journal of Hydrometeorology 18, 3; 10.1175/JHM-D-16-0171.1

The overlap of dry and wet coupling regimes within the CTP–HI space with in situ classification is quantified as the number of bins in the CTP–HI space with the same coupling regime classification as in situ relative to the total number of bins defined as that coupling regime from the in situ and given as a percentage. MERRA shows a consistency of 100% and 96% for wet and dry coupling, respectively. CFSR has a consistency for both wet and dry coupling at 100%. The Aqua classification is 82% consistent with in situ classification for the wet coupling regime and 87% consistent for the dry coupling regime. Given this measure, there are two reasons that consistency could be less than 100%: 1) a difference in size of the regime space and 2) a difference in location. Given the small size of the in situ regimes, the lower consistency between in situ and satellite is due to the location, not the size. The Aqua classification shows a translation down the CTP dimension for both the wet and dry coupling regimes. Consistency in wet coupling is highest for both MERRA and CFSR, which also have the highest correlation with in situ data for CTP and HI. CFSR has the highest consistency with dry coupling and showed the highest correlation with SM. It is not surprising that the Aqua classification has a lower consistency with in situ compared to the reanalysis, given the differences in CTP, HI, and SM shown in Fig. 3. Even though the difference in the location of the regimes results in a lower consistency for the Aqua dataset, the overall patterns across the datasets are comparable. This difference in the location of the coupling regimes in the CTP–HI space across datasets was one of the main reasons that a local-dataset-specific classification of the CTP–HI space was developed by Roundy et al. (2013).

Although there are inconsistences among the datasets in terms of the coupling classification and the input, the coupling classification and resultant CDI are based on the temporal persistence in dry or wet coupling regimes, and it is arbitrary if the actual locations of the regimes (i.e., in Fig. 4) are consistent. Furthermore, once the CTP–HI space is classified using the soil moisture data, only the CTP and HI are needed to produce a daily classification and calculate the CDI. This is particularly fortunate for the Aqua satellite and allows the calculation of the CDI beyond 2011 even though soil moisture data are no longer available. A comparison of the time series of the monthly CDI from 2003 to 2015 is given in Fig. 5a and shows consistency in the temporal variability of the datasets. This is partially due to the CDI capturing consistent temporal relationships within coupling regimes that are not impacted by the inconsistencies previously discussed.

Fig. 5.
Fig. 5.

Comparison of in situ CDI with reanalysis (MERRA and CFSR) and satellite remote sensing (Aqua) for a point in the SGP (36.5°N, 97.5°W) for (a) monthly time series from 2003 to 2015; (b) scatterplots of the monthly values for May–September, with the dark gray points from 2003 to 2011 and the light gray points from 2012 to 2015; and (c) the spatial variability of the CDI in June 2007. The regression line (red), rp, and rs are also given in (b).

Citation: Journal of Hydrometeorology 18, 3; 10.1175/JHM-D-16-0171.1

Notwithstanding the consistency in the CDI among the datasets, there are some noticeable differences. First, the in situ CDI has a lower magnitude than the reanalysis. This is particularly noticeable for the extremely dry months (positive) and wet months (negative). This is likely due to the smaller area classified as dry and wet coupling in the CTP–HI space (Fig. 4). The satellite CDI magnitude is also smaller in amplitude as compared to reanalysis. This is consistent with Aqua not being able to capture the extremes of HI (as discussed earlier). However, the satellite and in situ CDI does capture the relative peaks of dry (2011 and 2012) and wet (2007) regimes well. In comparing the two reanalysis datasets, CFSR has a higher CDI than MERRA for most months. This is likely due to the larger boundary layer growth as a result of a persistent dry bias in the PBL (Santanello et al. 2015).

The consistency between the CDI of the datasets is primarily seen in the summer months (May–September), while the winter months generally have a low magnitude and there is more scatter across the datasets. This is not surprising given the dominant nature of the coupling regimes in the summertime. Since the summer months are more relevant to land–atmosphere interactions and the CDI, the monthly CDI is compared in Fig. 5b for the summer months. The dark gray points are for the training period (2003–11) while the light gray points are from 2012 to 2015. The overall correlations for reanalysis and satellite with in situ CDI are significant at a 99% confidence level across, with Pearson and Spearman correlations of 0.85 and 0.83 for MERRA, 0.8 and 0.7 for CFSR, and 0.68 and 0.68 for Aqua. There is also no noticeable degradation in the relationship with the in situ data outside of the training period. The relative rankings are consistent with the previous analyses that examined the variables and classification that go into the CDI (Figs. 3, 4). Specifically, MERRA is more consistent with in situ data at the SGP site, followed by CFSR and then Aqua.

Even though the MERRA is more consistent with the in situ data than the other datasets, it is important to remember that up to this point the analysis has only considered a single point and may or may not be representative of other locations. In fact, the ability to have observations over the globe is one of the major advantages of using satellite remote sensing to estimate the CDI. The CDI over the contiguous United States is shown in Fig. 5c for MERRA, CFSR, and Aqua for June 2007. There is overall consistency across the datasets, with the dominant spatial patterns evident in both reanalysis and satellite CDI and showing the wet conditions in the Northwest and the SGP and the drought in the Intermountain West and the Southeast. The spatial patterns are weaker for the satellite CDI, particularly for the magnitude and extent of the wet coupling area. Despite the weaker spatial patterns and limitations of the satellite data (vertical resolution, short record, and coarse spatial scale), they still capture the primary signals and have potential to yield useful information as a large-scale observation.

c. CDI relationship to other variables

The CDI captures the intensification, persistence, and recovery of drought through the persistence in the dry and wet coupling regime. While it is clear from Fig. 1 that there is a connection between variables typically associated with LA interactions and the coupling regimes, the relationships between CDI and these variables have never been quantified. Since these variables have different means and variance, each one is normalized to a standardized index by subtracting the mean and dividing by the standard deviation. In this manner, the CDI from satellite remote sensing is compared with other LA-associated variables from reanalysis. The monthly standardized anomalies of precipitation P, total soil moisture (TSM), daily average temperature (DAT), vapor pressure deficit (VPD), EF, BLH, CTP, and HI are compared to the monthly anomaly of CDI. The correlations of the Aqua CDI with the aforementioned variables from MERRA and CFSR are given in Figs. 6c and 6d, and the MERRA CDI compared to MERRA variables are given in Fig. 6e. Because Aqua itself is limited in terms of observing the majority of these individual variables, we compare Aqua CDI to the reanalysis products that are assumed to capture the bulk behavior of these coupling-related properties of the LA system. Each panel includes the correlation of the spatial average standard anomaly for six climate regions (colors) and the entire United States (gray and white boxes reflected around zero) for months in the May–September season (open shapes) and the full year (filled shapes).

Fig. 6.
Fig. 6.

The monthly standardized anomaly rs (2003–15) of the CDI with P, TSM, DAT, VPD, EF, BLH, CTP, and HI [(a) provides the legend (colored shapes) and (b) provides a map of the six climate regions] from (c) Aqua CDI and MERRA, (d) Aqua CDI and CFSR, and (e) MERRA CDI and MERRA. The horizontal red dashed lines in (c)–(e) indicate statistical significance at p = 0.05 for the monthly values from May to September (r = 0.24) and all the monthly values’ significance (r = 0.16).

Citation: Journal of Hydrometeorology 18, 3; 10.1175/JHM-D-16-0171.1

For all three comparisons, the CDI and precipitation show a higher correlation in the western portion of the United States that only shows a minor increase during the summer months. The spatial difference in the correlation between the CDI and precipitation is less pronounced in the Aqua–CFSR comparison as compared to the Aqua–MERRA or MERRA–MERRA comparison. This suggests that the MERRA precipitation (which is known to have major limitations in timing and intensity over much of CONUS) may be the cause of this spatial difference. There is also less of a seasonal difference in the MERRA–MERRA comparisons, suggesting a greater seasonal difference in the Aqua CDI compared to MERRA. Total soil moisture shows a similar relationship with CDI and precipitation in that it has a higher correlation in the west and a relatively small seasonal difference. In fact, for the Aqua–MERRA and MERRA–MERRA comparison the correlations are nearly the same. This suggests that there is a high correlation between precipitation and TSM in MERRA, as would be expected within the same reanalysis system. In contrast, the Aqua–CFSR comparison for TSM shows a much lower correlation and little spatial difference.

The daily average temperature and vapor pressure deficit show a higher correlation with the CDI and a greater seasonal difference as compared to precipitation and soil moisture. The Aqua CDI correlation with DAT and VPD nearly doubles during the summertime compared to the full year and is more spatially homogeneous. The MERRA CDI and the MERRA VPD have less of a seasonal difference in correlation and there is consistency in the correlation across the different regions of the country, with a higher correlation in the west for both the DAT and VPD that is consistent with P and TSM. This same relationship is weaker for Aqua–CFSR as compared to Aqua–MERRA and MERRA–MERRA. This suggests a consistent spatial relationship between the CDI and MERRA variables that may be a unique attribute to MERRA and not CFSR.

The evaporative fraction has one of the lowest correlations with the CDI across all comparisons and also shows little difference in the seasonal correlation. The CDI is not well correlated with evaporative fraction for Midwest and especially the Northeast, while the highest correlations are generally seen in the south and the high plains. BLH shows similar spatial patterns, but CDI shows an overall higher correlation with BLH and an increase in the seasonal variability as compared to the evaporative fraction. This indicates that the CDI is more strongly correlated with the atmospheric side of LA coupling and shows the greatest strength in the areas that are considered hotspots (Koster et al. 2006). However, since the BLH is highly correlated with the sensible heat flux, it may be that the energy cycle side of the land surface plays an important role in the CDI evolution. The Aqua CDI is also more strongly correlated with the BLH from CFSR compared to MERRA BLH, which is consistent with Santanello et al. (2015), who found the MERRA BLH to be underestimated and lacking sensitivity to extremes.

The correlation of the CDI to the CTP and HI is among the strongest and is not surprising given that the CTP and HI are used to derive the CDI. The CTP has a larger seasonal difference in its correlation to the CDI as compared to the HI. In contrast the HI shows more spatial variability in its correlation with the CDI. Overall, the satellite-based CDI shows slightly lower correlations with other reanalysis variables then those seen internally within MERRA. This is not surprising given that reanalysis variables should be more consistent, while the satellite observations are more independent.

4. Discussion and conclusions

The aim of this study is to assess the utility of CDI-based variables and metrics derived from satellite remote sensing for global applications by comparing them with in situ observations and reanalysis datasets. Overall, the Aqua CDI performs well at a point, spatially, and in time (trends) compared to in situ and reanalysis products. This is especially promising given the inherent limitations in vertical profile resolution and soil moisture retrieval, as advances in satellite-based profiles (e.g., improved AIRS retrievals) and soil moisture retrievals (e.g., SMAP) will provide improved estimates of LA- and CDI-related quantities in the future. The satellite observations of atmospheric temperature and humidity profiles and the derived metrics compare well with in situ observations, although differences exist, mainly due to the limitation of vertical resolution of the satellite data (Fig. 2).

Although the lower vertical resolution of the atmospheric satellite data resulted in lower correlations of the CTP and HI from satellite with in situ data, the satellite data have sufficient correlation with in situ data to capture the main signal (Fig. 3). Both reanalysis datasets show an equally strong correlation with the in situ observations for CTP and HI, while the satellite data show a lower correlation with in situ HI as compared to the CTP. It should be noted that to date there has been very little focus or evaluation of AIRS, level 3, profile retrievals over land because of inherent difficulties in retrieving lower troposphere and PBL thermodynamics (due to factors such as limited weighting functions and surface emissivity). Moisture retrieval is inherently more difficult than temperature, and thus the results are not unexpected in that temperature (and CTP) performs better than moisture (and HI) against this small sample. The AIRS support product (level 2) has a finer vertical (100 levels) and spatial (45 km) resolution that may improve somewhat on the retrieval of lower-tropospheric humidity and temperature. In addition, the latest version of AIRS (version 6.28, to be released publically in version 7 in 2017) shows some improvements related to humidity retrieval that are due to improved IR channel sampling. However, any major improvements in space-based CTP–HI retrieval and vertical resolution must come with next-generation satellite missions dedicated to retrieving PBL profiles.

Given the large sample of days required by the CDI, it is likely that the bulk signal of the CTP and HI and its relative variability over the more than 9-yr period will still provide a self-consistent representation of dry and wet coupling regimes and variability. Figure 3 bears this out and suggests that despite the scatter, there are still decent correlations in CTP and HI that can be exploited to represent dry versus wet regimes. Likewise, the large scatter in SM should not prohibit the SM data from being representative of dry versus wet regimes and surface conditions. Combining the CTP, HI, and SM to identify areas of dry and wet coupling, the in situ classification has a high consistency with the reanalysis while satellite observations have the lowest consistency (Fig. 4). This is not surprising given that MERRA and CFSR showed the highest correlations with in situ data of CTP, HI, and SM (Fig. 3). Although there is a lack of consistency in the exact location of the dry and wet coupling regimes within the CTP–HI space across all the datasets, all the datasets, including satellite, indicate similar shapes and relative locations of the regimes. This indicates that all datasets show the existence of these regimes. This is particularly a novel finding of this study since this work represents the first time that in situ observations have been applied to the Roundy et al. coupling classification. The in situ and satellite remote sensing CDI provides a unique combination of observations that allows for an evaluation of model data at local and large scales that could be exploited in future studies. It is important to note that the in situ comparisons are only valid at a single point over the SGP. While the SGP is an ideal location to have such in situ observations, it would be ideal to compare in situ data from other areas with satellite remote sensing. It is expected that in mountainous, perpetually cloudy, and cold regions it is unlikely to retrieve profiles as well down to the surface. However, this is a promising start, and it indicates that satellite data (despite its limitations) can provide the information needed for such complex metrics as the CDI.

Applying the classification of the CTP–HI space to daily classification of the coupling state and the calculation of the CDI indicated similar results in that the in situ CDI showed the strongest consistency with MERRA; however, the monthly CDI from satellite still had a temporal correlation of 0.68 with the in situ observations (Figs. 5a,b). Furthermore, the spatial patterns of CDI for satellite remote sensing are consistent with the reanalysis for June 2007 over the United States. This indicates that both temporal and spatial patterns are largely captured by the CDI from satellite remote sensing and further demonstrates the potential of the Aqua dataset. There is, however, a smaller magnitude both in space and in time in the CDI compared to the reanalysis. The lower-magnitude CDI is especially noticeable during wet coupling, as indicated in Figs. 5a and 5c. This limitation could be partially due to missing values in the record, particularly during the wet coupling regime when there is more cloud cover that can limit the satellite observations. Missing values make the CDI move closer to zero, since it has the potential to reduce the numerator but not change the denominator in calculating the CDI. Future work will explore a revised CDI that would be less impacted by cloud cover and more relevant for satellite application. The limitation in the CDI magnitudes could also be partially due to the lack of resolution in the vertical from the atmospheric observations from satellite as shown in Fig. 2 along with difficulties in observing atmospheric humidity. It is hoped that through improvements in instruments and algorithms the quality of the satellite data will be increased and this limitation can be overcome.

Notwithstanding the shortcomings of the satellite data, they still have the potential to yield useful information as a large-scale observational record. As compared to other variables, the CDI has the strongest correlation with the CTP and HI, from which it is derived, but it also has strong correlation with VPD and DAT. These correlations are the highest over the United States during the summertime when land–atmosphere feedbacks play a stronger role in the evolution of the daytime temperature and humidity. The CDI also has a reasonable correlation with BLH. The correlations between CDI and the various variables were also lower when comparing satellite CDI to reanalysis variables as compared to reanalysis CDI. This is not surprising, as there should be a level of consistency between the variables from the same reanalysis product. The results indicate that the CDI has the strongest relationship with atmospheric variables (DAT and VPD) that are greatly influenced by the land surface heat fluxes, for example, sensible and latent heat fluxes; however, it is not extensively correlated with any one variable and has its own unique characteristics. These unique characteristics could make it a useful drought-monitoring tool, as it has the potential to integrate multiple drivers and impacts of drought that may be missed by indices typically utilized for drought monitoring.

Overall, this work demonstrates that there is sufficient information in the simultaneous measurements of the land and atmosphere from satellite remote sensing to provide useful information to the applications of drought monitoring and coupling metrics that can be used to evaluate GCMs. While it is recognized that the variables and metrics currently available through satellite remote sensing are not always the optimal choice for LA coupling metrics, it is hoped that through further development, satellite-based CDI can be utilized to provide new insights and application relevant for drought monitoring and prediction.

Acknowledgments

This research was partially supported by an appointment to the NASA Postdoctoral Program at the Goddard Space Flight Center, administered by Oak Ridge Associated Universities. Data were obtained from the Atmospheric Radiation Measurement (ARM) Program sponsored by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division. This financial and data support is gratefully acknowledged.

REFERENCES

  • ARM, 1994: Balloon-borne sounding system (SONDEWNPN). Updated daily, ARM Climate Research Facility, accessed 2 June 2016, doi:10.5439/1021460.

    • Crossref
    • Export Citation
  • ARM, 1996: Soil Water and Temperature Profiling System (SWATS). Updated daily, ARM Climate Research Facility, accessed 16 June 2016, doi:10.5439/1150274.

    • Crossref
    • Export Citation
  • Betts, A. K., 1992: FIFE atmospheric boundary layer budget methods. J. Geophys. Res., 97, 18 52318 531, doi:10.1029/91JD03172.

  • Dirmeyer, P. A., Z. Wang, M. J. Mbuh, and H. E. Norton, 2014: Intensified land surface control on boundary layer growth in a changing climate. Geophys. Res. Lett., 41, 12901294, doi:10.1002/2013GL058826.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, doi:10.1029/2002JD003296.

    • Search Google Scholar
    • Export Citation
  • Ferguson, C. R., and E. F. Wood, 2011: Observed land–atmosphere coupling from satellite remote sensing and reanalysis. J. Hydrometeor., 12, 12211254, doi:10.1175/2011JHM1380.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Findell, K. L., and E. A. B. Eltahir, 2003a: Atmospheric controls on soil moisture–boundary layer interactions. Part I: Framework development. J. Hydrometeor., 4, 552569, doi:10.1175/1525-7541(2003)004<0552:ACOSML>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Findell, K. L., and E. A. B. Eltahir, 2003b: Atmospheric controls on soil moisture–boundary layer interactions. Part II: Feedbacks within the continental United States. J. Hydrometeor., 4, 570583, doi:10.1175/1525-7541(2003)004<0570:ACOSML>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 434439, doi:10.1038/ngeo1174.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Koster, R. D., M. J. Suarez, A. Ducharne, M. Stieglitz, and P. Kumar, 2000: A catchment-based approach to modeling land surface processes in a general circulation model 1. Model structure. J. Geophys. Res., 105, 24 80924 822, doi:10.1029/2000JD900327.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Koster, R. D., and et al. , 2006: GLACE: The Global Land–Atmosphere Coupling Experiment. Part I: Overview. J. Hydrometeor., 7, 590610, doi:10.1175/JHM510.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lintner, B. R., P. Gentine, K. L. Findell, and G. D. Salvucci, 2015: The Budyko and complementary relationships in an idealized model of large-scale land–atmosphere coupling. Hydrol. Earth Syst. Sci., 19, 21192131, doi:10.5194/hess-19-2119-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Owe, M., R. de Jeu, and T. Holmes, 2008: Multisensor historical climatology of satellite-derived global land surface moisture. J. Geophys. Res., 113, F01002, doi:10.1029/2007JF000769.

    • Search Google Scholar
    • Export Citation
  • Rienecker, M. M., and et al. , 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 36243648, doi:10.1175/JCLI-D-11-00015.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roundy, J. K., and E. F. Wood, 2015: The attribution of land–atmosphere interactions on the seasonal predictability of drought. J. Hydrometeor., 16, 793810, doi:10.1175/JHM-D-14-0121.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roundy, J. K., C. R. Ferguson, and E. F. Wood, 2013: Temporal variability of land–atmosphere coupling and its implications for drought over the southeast United States. J. Hydrometeor., 14, 622635, doi:10.1175/JHM-D-12-090.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roundy, J. K., C. R. Ferguson, and E. F. Wood, 2014: Impact of land–atmospheric coupling in CFSv2 on drought prediction. Climate Dyn., 43, 421, doi:10.1007/s00382-013-1982-7.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saha, S., and et al. , 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 10151057, doi:10.1175/2010BAMS3001.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saha, S., and et al. , 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 21852208, doi:10.1175/JCLI-D-12-00823.1.

  • 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, 577599, doi:10.1175/2009JHM1066.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Santanello, J. A., and et al. , 2011: Local land–atmosphere coupling (LoCo) research: Status and results. GEWEX News, Vol. 21, No. 4, International GEWEX Project Office, Silver Spring, MD, 7–9. [Available online at http://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, 324, doi:10.1175/JHM-D-12-023.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 58135829, doi:10.1175/JCLI-D-14-00680.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 125161, doi:10.1016/j.earscirev.2010.02.004.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Song, H.-J., C. R. Ferguson, and J. K. Roundy, 2015: 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, 541556, doi:10.1175/JHM-D-15-0045.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stommel, H., 1947: Entrainment of air into a cumulus cloud. J. Meteor., 4, 9194, doi:10.1175/1520-0469(1947)004<0091:EOAIAC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Susskind, J., J. M. Blaisdell, L. Iredell, and F. Keita, 2011: Improved temperature sounding and quality control methodology using AIRS/AMSU data: The AIRS Science Team version 5 retrieval algorithm. IEEE Trans. Geosci. Remote Sens., 49, 883907, doi:10.1109/TGRS.2010.2070508.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 19291945, doi:10.1175/JHM-D-14-0117.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Taylor, C. M., R. A. M. de Jeu, F. Guichard, P. P. Harris, and W. A. Dorigo, 2012: Afternoon rain more likely over drier soils. Nature, 489, 423426, doi:10.1038/nature11377.

    • Crossref
    • Search Google Scholar
    • Export Citation
Save