1. Introduction
A “climatology” of radiosonde-derived indices provides a valuable historical context for the forecasting of severe weather. Relating this climatology to near-real-time observations from meteorological sensors on satellites could provide a valuable tool in assessing the risk of severe weather (Doswell 2004; Breznitz 1984; Barnes et al. 2007; Golden and Adams 2000; Rothfusz et al. 2014; Cintineo et al. 2014). For example, convective available potential energy (CAPE) is strongly related to the largest hail sizes in convection over the southern Great Plains (Johns and Doswell 1992). In the case of severe weather, CAPE is one of the key indices that are considered when evaluating the potential for thunderstorm development (McNulty 1995). CAPE is a measure of energy available for convection (Moncrieff and Green 1972) and is proportional to the area of a thermodynamic diagram where the moist adiabat of an ascending saturated parcel is greater than the environmental temperature sounding bounded below by the level of free convection and bounded above by the equilibrium level (Johns and Doswell 1992; Blanchard 1998), including the area between the environmental temperature and the moist adiabat of the ascending parcel. Investigation of reanalysis fields has shown that the U.S. Great Plains region has some of the largest CAPE values in the world (Riemann-Campe et al. 2009).
Radiosondes have traditionally been launched two times per day from National Weather Service (NWS) sites at 0000 and 1200 UTC and have provided the operational vertical soundings used to compute convective instability indices. In the U.S. southern Great Plains, the typical spacing between NWS radiosonde launch sites is, on average, 315 km, and the launch times correspond to 0600 and 1800 central standard time (CST). Since diurnal convection tends to occur between these standard NWS launch times and often between the launch sites, the operational NWS soundings provide important but limited time and space information for short-term forecasting (Bevis et al. 1992). Numerical weather prediction (NWP) models typically assimilate these radiosonde observations at 0000 and 1200 UTC along with more frequent surface observations (Dee et al. 2011). The lack of upper-air observations between synoptic observation times may limit the ability of current NWP forecasts to represent the changing atmospheric conditions in the United States. The use of timely satellite observations combined with empirical forecasting methods can help to address this deficiency (Cintineo et al. 2014).
High-spectral-resolution infrared sensors on sun-synchronous polar-orbiting weather satellites can help to fill in the temporal sampling gap in atmospheric vertical profiles of temperature and water vapor over the United States (Smith et al. 2009; Chahine et al. 2006; Davis 2007). The sun-synchronous orbit precesses at a rate that is matched to the rotation of Earth such that the satellite nadir track crosses the equator at approximately the same local time during each orbit. The future operational weather satellite system is an international joint effort, with the late-morning orbit being sampled by the EUMETSAT MetOp series of satellites and the early-afternoon orbit being sampled by the NOAA Joint Polar Satellite System series (Hilton et al. 2012; Goldberg et al. 2013). In preparation for these operational satellite systems, NASA launched the Aqua satellite in 2002 as part of the NASA Earth Observing System program (Parkinson 2003). In particular, the Atmospheric Infrared Sounder (AIRS) sensor has been used to provide quantitative information about the troposphere for Earth science applications (Tian et al. 2013; Fetzer et al. 2006; Chahine et al. 2006). Algorithms to process NASA AIRS data in near–real time have been developed for direct broadcast applications (Smith et al. 2012; Weisz et al. 2013). Retrievals of temperature and moisture vertical profiles have the potential to provide a positive impact on forecasts of convective watches and warnings (Stensrud et al. 2009). The development of operational software packages for timely and automated analysis of hyperspectral infrared data on future polar-orbiting satellites continues to be an active area of research (Goldberg et al. 2013). Note that NWP centers currently assimilate AIRS data for medium range forecasting, but the infrared channels sensitive to the lower troposphere are excluded over land regions to avoid cloud contamination (McNally et al. 2006).
The objective of the work presented here is to make a quantitative assessment of the accuracy of CAPE derived from high-spectral-resolution infrared sounder observations. Ten years of satellite overpasses of the NASA Aqua satellite over the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) program Southern Great Plains (SGP) site provide the time/space-coincident dataset needed to make a statistical assessment (Stokes and Schwartz 1994). The DOE ARM SGP site was chosen as the assessment site because radiosondes are launched daily at 0600 and 1800 UTC, about 1.5 h prior to the Aqua satellite overpass. For comparison purposes, a reanalysis of NWP temperature and water vapor fields was also collected for the same time period.
This paper is organized as follows: Section 2 describes the data used, section 3 outlines the method, section 4 discusses the results, section 5 provides an uncertainty analysis, and section 6 gives the conclusions for this study. An appendix is included to illustrate the AIRS CAPE spatial coverage for two cases of severe weather in the SGP region.
2. Data
In situ vertical profiles used in this study are from Vaisala, Inc., RS92 radiosondes launched at the DOE ARM SGP central facility near Lamont, Oklahoma (36°36′18.0″N, 97°29′6.0″W; altitude 320 m) in the 10-yr time period from January 2005 through December 2014. The data were obtained online from the ARM archive (http://www.archive.arm.gov). The files used in this analysis are from an ARM value-added product (“sgp10rlprofmr1turn”), with observed pressure, temperature, and water vapor vertical profiles interpolated in time to 10-min intervals and interpolated in the vertical direction to fixed height levels. The interpolated profiles closest in time to each Aqua satellite overpass were extracted to represent the ARM-radiosonde vertical profiles. Several assessments have been made of the ARM-radiosonde accuracy by comparison with reference instruments at the SGP ARM site (Goldsmith et al. 1998; Turner and Goldsmith 1999; Revercomb et al. 2003; Miloshevich et al. 2009). The high time sampling of the in situ radiosonde data provides high-resolution vertical profiles of pressure, temperature, and relative humidity. In this study, the vertical profiles have been uniformly interpolated to 202 fixed vertical height levels from the surface to 15 km at a constant 75-m resolution with the exception of the lowest two levels, which are at a nominal height above ground of 2 and 30 m. The interpolation to the 75-m height bins preserves most of the high-resolution features of the original radiosonde; if one assumes an average ascent rate of 5 m s−1, the temporal sampling between layers is about 15 s.
The NWP reanalysis fields used in this study were from the ERA-Interim project (Dee et al. 2011) produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). DOE ARM radiosondes have been used to assess ECMWF model forecasts (Cheinet et al. 2005). ARM radiosondes have been available to the ERA-Interim assimilation system through the Global Telecommunications System, but it is not clear whether the 0600 and 1800 UTC radiosondes used in this study are actually assimilated (Dee et al. 2011). ERA-Interim has assimilated AIRS radiances into the model beginning in 2003, but no spectral channels are assimilated that are sensitive to the lower atmosphere over land areas such as the SGP region (McNally et al. 2006). The data were downloaded from an NCAR website that allows spatial and temporal subsetting of the ECMWF data files (the ERA-Interim project; ECMWF 2009). The original ECMWF grid representation contains spectral coefficients for a reduced n128 (nominally ~512 × 256) Gaussian grid with T255 truncation. This provides contiguous spatial coverage with spatial resolution of 0.7° in the SGP domain. The temperature and water vapor mixing ratio are specified in hybrid pressure coordinates at 60 levels (Berrisford et al. 2011). The model data were extracted over the SGP domain (109°–86°W and 25°–45°N) at 0000, 0600, 1200, and 1800 UTC for each day between 1 January 2005 and 31 December 2014. Model surface pressure was used to compute the pressure profile at each model level for each grid cell using the ECMWF formula for hybrid vertical coordinates (Berrisford et al. 2011). The requested model fields were in “netCDF” format (e.g., “ei.oper.an.ml.regn128sc.2013050100.nc”). The model fields used were “LNSP_GDS4_HYBL,” “T_GDS4_HYBL,” and “Q_GDS4_HYBL” for surface pressure, air temperature, and water vapor mixing ratio, respectively.
The satellite data used in this analysis were from the NASA AIRS sensor on the Aqua platform in a sun-synchronous orbit with local overpass times of about 0130 and 1330 (Chahine et al. 2006). As the satellite orbits from pole to pole, the sensor maps out a swath using a cross-track scan pattern. The AIRS measures the infrared wavelengths from 3.7 to 15.4 μm while the collocated AMSU-A sensor makes simultaneous measurements in the microwave region between 23 and 89 GHz. The AIRS sounding footprint is synchronized to the AMSU-A cross track with footprint diameter of 40 km at nadir growing to about 90 km at the edge of the scan (±49.5° cross-track scan from a 705-km orbit). The data product used was the NASA AIRS science team version-6 level-2 retrieval product that was obtained online from the NASA Goddard data archive (http://disc.sci.gsfc.nasa.gov/AIRS) (Susskind et al. 2003, 2011, 2014). Fields used include air temperature profiles (“TAirSup”) and water vapor profiles (“H2OCDSup”) at 101 fixed pressure levels from the support product. The AIRS surface variables used are the surface pressure (“PSurfStd”), near-surface air temperature (“TSurfAir”), and near-surface water vapor mixing ratio (“H2OMMRSurf”). The AIRS near-surface mixing ratio value along with the near-surface air temperature and surface pressure was used to compute a surface dewpoint temperature for each profile. According to the AIRS version-6-release level-2 product user guide (https://disc.gsfc.nasa.gov/AIRS/documentation/v6_docs/v6releasedocs-1/V6_L2_Product_User_Guide.pdf), PSurfStd is interpolated from the Global Forecast System forecast and corrected using the local topography from a digital elevation model of the retrieval field of view. (Additional details about the AIRS variables can be found at http://disc.sci.gsfc.nasa.gov/AIRS/documentation/v6_docs.)
In this analysis, the quality-control variables shown in Table 1 were used to create the comparison dataset. Following the quality-control approach of Bedka et al. (2010) and Roman et al. (2016), these include the use of a precipitable water vapor quality flag (“PWV_QC”) and a fractional-error estimate (“totH2OStdErr”). For quality control of the profile column, the quality variable “PGood” was used as a threshold to exclude invalid profiles, but only a very loose requirement (PGood > 500 hPa) was used in the analysis so as to include as many profiles as possible that sampled the lower troposphere. The AIRS estimated cloud fraction over the 45-km sounding footprint was used to exclude soundings with a cloud fraction of greater than 80%. An illustration of the quality control applied to the AIRS version-6 level-2 product is given in the appendix for two case-study days.
AIRS CAPE quality-control criteria used in this study.
3. Method
To obtain a statistically significant range of CAPE values in the U.S. southern Great Plains, vertical profiles of pressure, temperature, and water vapor were obtained for the period from 1 January 2005 to 31 December 2014 for a region centered on the ARM SGP central facility. The ARM site was chosen for this study because routine radiosonde launches at 0600 and 1800 UTC are within 1.5 h of the nominal satellite overpass times of the Aqua satellite. Only cases with radiosonde profiles having CAPE greater than 50 J kg−1 were included in the analysis. This threshold was used to eliminate the large number of zero (or small) CAPE values. To investigate spatial sampling issues, CAPE values from AIRS and ERA were selected within a radius of 50, 150, and 250 km of the ARM SGP central facility. The selected CAPE values for a time and space region are used to create histograms using a uniform bin size of 50 J kg−1. Normalizing by the sum of the histogram creates a probability distribution function (PDF) of CAPE. PDF percentiles at 25th, 50th, 67th, 75th, 95th, and 99th were tabulated to quantify the characteristics of each CAPE distribution.
To analyze the degree of agreement between the AIRS profiles and the ARM radiosondes, scatterplots of the 10-yr matchup dataset were created and a correlation coefficient was computed. The results include a comparison in which the ARM-radiosonde 2-m temperature and dewpoint temperature are used in place of the AIRS surface estimates. Vertical profiles extracted at the SGP ARM site from the ERA-model fields were analyzed in a similar manner, including the substitution of ARM-radiosonde surface temperature and water vapor in place of the NWP-model surface estimates.
To analyze the dependence of CAPE on the vertical resolution of the temperature and water vapor profiles, the radiosonde profiles were smoothed with a vertical boxcar function at widths in multiples of 75 m (1, 3, 5, 7, 9, 15, 21, 27, 35, 41, 47, and 53). The CAPE computed from the smoothed temperature and water vapor profile was differenced from the original radiosonde CAPE, for each profile and linear regression was performed to determine the fractional change in CAPE.
4. Results
Figure 1 illustrates the seasonal variation of dewpoint-temperature vertical profiles and surface-based CAPE derived from the ARM radiosondes at the times of the NASA Aqua satellite overpasses. It shows the highest CAPE values in the spring and summer months, when the surface dewpoint is also relatively high.
Time series of (top) dewpoint temperature (K) from ARM SGP radiosondes vs height (km) and (bottom) the corresponding radiosonde-derived CAPE at NASA Aqua satellite overpass times for the 10-yr period shown. CAPE values greater than 3500 J kg−1 are not shown in this figure but are included in the analysis.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
a. CAPE climatology at the DOE ARM SGP site
A summary of the distribution of CAPE computed from ARM radiosondes at 0600 and 1800 UTC is shown in Table 2 for the period 2005–14. Values of CAPE that are less than 50 J kg−1 were excluded from the PDF analysis shown in Table 2. The PDF cumulative sum of radiosonde-derived CAPE exceeding 1000 J kg−1 is approximately 46% for this 10-yr period. The 50th percentile of SGP radiosonde CAPE values is 875 J kg−1, the 95th percentile is 2625 J kg−1, and the 99th percentile is 3925 J kg−1.
Climatological summary of CAPE in a region centered at the ARM SGP site central facility for the period of January 2005–December 2014. Fractional error relative to the ARM radiosonde is shown in parentheses. The last six rows in the first column show the percentiles of CAPE at which values are shown for each dataset.
The ERA-Interim analysis fields at 0600 and 1800 UTC were used to extract temperature and moisture profiles for the grid cells near the ARM SGP site. These are the analysis times that are closest to the Aqua satellite overpass times (about 1.5 h prior). Table 2 details the ERA-data percentile ranges for profiles within a radius of 50, 150, and 250 km of the ARM SGP central facility. Note that the PDF is computed for bins of width 50 J kg−1; therefore, the CAPE values ending in xx25 or xx75 represent the middle of a PDF bin. The PDF cumulative sum of ERA-derived CAPE exceeding 1000 J kg−1 is approximately 38% for this 10-yr period. The 50th percentile of ERA CAPE values is 675 J kg−1, the 95th percentile is 2675 J kg−1, and the 99th percentile is 3525 J kg−1 for the profiles within 50 km of the SGP ARM site. At the 50th percentile, the ERA CAPE values are 23% lower than the ARM-radiosonde CAPE values used as reference. At the 95th percentile, the ERA CAPE is within 5% of the ARM-radiosonde values. Thus, the ERA CAPE PDF is shifted to lower values for the bulk of the distribution (<75th percentile) as a percentage. The extreme CAPE values show an absolute shift but a smaller percentage error. There is little variation in the CAPE distribution with increasing distance from the ARM site despite the fact that the number of samples (gridcell profiles) greatly increases.
The AIRS level-2 granules at 0830 and 1930 UTC were used to extract retrieval profiles of temperature and dewpoint temperature that fall within a radius of the SGP ARM site. The PDF cumulative sum of CAPE exceeding 1000 J kg−1 is approximately 38% under the conditions that CAPE is greater than 50 J kg−1. Table 2 details the AIRS-data percentile ranges for profiles within a radius of 50, 150, and 250 km. At the 50th percentile, the AIRS CAPE values are 17% lower than the ARM-radiosonde CAPE values used as reference. At the 95th percentile, the AIRS CAPE is within 10% of the ARM-radiosonde values. The 50th percentile of AIRS CAPE values is 725 J kg−1, the 95th percentile is 2775 J kg−1, and the 99th percentile is 3775 J kg−1 for the profiles within 50 km of the SGP ARM site. Thus, the AIRS CAPE PDF is shifted to lower values for the bulk of the distribution as a percentage. The extreme AIRS CAPE values show an absolute shift relative to the radiosonde PDF but a smaller percentage error, similar to the ERA results. As shown in Table 2, the AIRS and ERA CAPE PDF distributions are remarkably similar to each other, with similar differences from the ARM-radiosonde CAPE distribution.
b. CAPE correlation at the DOE ARM SGP site
Using only the AIRS profiles closest to the DOE ARM SGP site, a time- and space-coincident comparison was made with the ARM radiosondes for the 10-yr period of 2005–14. A scatterplot of the CAPE derived from AIRS and radiosonde profiles is given in Fig. 2. A weak correlation of 0.355 is found between AIRS and radiosonde CAPE. As illustrated in Fig. 3, when the radiosonde surface temperature and dewpoint values are used in the AIRS CAPE calculation, the correlation increases to 0.94. This result suggests that the AIRS vertical profile of temperature and dewpoint is consistent with the radiosonde atmospheric profile but that the estimate of the surface temperature and dewpoint has a difference relative to the radiosonde surface values. This situation is further explored in the section on uncertainty analysis.
(left) AIRS and (right) ERA CAPE vs ARM-radiosonde CAPE. The correlation is 0.355 for AIRS and 0.544 for ERA when they are compared with ARM radiosondes.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
As in Fig. 2, but substituting the ARM-radiosonde surface temperature and dewpoint temperature for the surface-parcel values in the (left) AIRS and (right) ERA CAPE calculation. The correlation is now 0.95 for AIRS and 0.99 for ERA when they are compared with ARM radiosondes.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
A similar analysis of the ERA-model correlation with DOE ARM radiosondes is shown in Figs. 2 and 3. The correlation of ERA CAPE with ARM-radiosonde CAPE is 0.544 using the model surface estimates of temperature and dewpoint but increases to 0.99 when the radiosonde surface values are used. This result suggests that differences in the model surface values of temperature and dewpoint relative to the ARM SGP 2-m surface observations cause the scatter seen in Fig. 2 for ERA versus ARM radiosondes.
5. Uncertainty analysis
An analysis was performed to understand better the ARM-radiosonde, ERA, and AIRS CAPE sensitivity to spatial, temporal, vertical, and measurement errors.
a. Spatial sampling error
For the satellite-product comparison with the ARM SGP site, a spatial sampling error could exist when a valid AIRS profile is not coincident with the SGP site location and the closest valid profile is some distance away. To quantify this issue, an analysis was created for AIRS CAPE values within a given radius of the ARM SGP site. The NWP ERA data were analyzed at the same circular region for consistency with the AIRS analysis. Table 2 compares the PDF of the 10-yr climatology of AIRS CAPE located within 50, 150, and 250 km of the ARM SGP site. Note that this is not the area average; instead, it is a distribution of individual profiles for each region. As the number of sample points in the region increases, the PDF distribution does not change as the radius increases. Therefore, we conclude that there is no significant spatial sampling error in the AIRS profiles used in this study. The ERA PDF shows a very similar result, suggesting that the ARM SGP site represents well the climatology of the region out to at least 250 km from the central facility.
A common problem in the use of radiosonde profiles for the validation of satellite retrievals and NWP-model profiles is the mismatch between the in situ point-scale sampling of the radiosonde and the area average represented by the sounding footprint. This issue has been addressed in detail in Tobin et al. (2006) for AIRS overpasses of the ARM SGP site. The spatial assessment in that paper was performed in a relative sense in which higher-spatial-resolution geostationary sounder profiles are used to compute statistics within the AIRS–AMSU sounding footprint. Tobin et al. (2006) found that at the SGP ARM site the variability of the atmospheric state within the AIRS–AMSU footprint sampled by the radiosonde had zero bias, with a relatively small standard deviation of only ~0.2 K in temperature and 20% in water vapor for the profile levels above the surface. In contrast, this paper has found that there is a much larger problem in the estimate of the 2-m air temperature and dewpoint temperature that may be caused by the issue of point measurements versus area averages. A detailed examination of the surface parameters is included in the section on measurement error.
b. Temporal sampling error
A temporal sampling error could be important in CAPE estimation given the rapid boundary layer changes that occur as a result of surface heating during the day and cooling after sunset. Operational radiosondes launched at 0000 and 1200 UTC (0600 and 1800 CST) are not ideal for assessment of CAPE during midday in the SGP region. The ARM SGP site was chosen for this study of AIRS CAPE because radiosondes are launched at 0600 and 1800 UTC (about midnight and noon local time). The afternoon satellite overpass is at roughly 1330 CST (1930 UTC), with some variation daily. The difference between the radiosonde launch time and Aqua satellite overpass time difference is typically less than 2 h. To further minimize this relatively small temporal sampling error, the radiosonde data were interpolated to each Aqua satellite overpass time. This works well in a steady-state or slowly evolving environment but may fail in rapidly evolving situations such as the passage of a front or dryline or the passage of convective outflow. The ERA-Interim analyses are available only at 0000, 0600, 1200, and 1800 UTC. For this study, the 0600 and 1800 UTC ERA fields are used without time interpolation and thus represent the atmospheric state ~1–2 h prior to the satellite overpass.
The issue of temporal variation of the atmospheric state during ascent of the radiosonde has been addressed in detail in Tobin et al. (2006). In that study, pairs of Vaisala radiosondes were launched from the ARM SGP site with a separation of ~45–90 min prior to and at the Aqua overpass time. The comparison of the change in the profile between the first and second radiosonde ascent provides a measure of the variability at the point scale of the radiosonde. Tobin et al. (2006) found a mean zero bias in the temperature and water vapor between the radiosonde pairs but a standard deviation of ~0.5 K in temperature and ~20% in relative humidity. The relevance to the current study is that the point sampling from the radiosonde would be expected to introduce a random-variation scatter in the derived CAPE but not a systematic bias. The magnitude of the point-sampling variability is believed to be small, however, as evidenced by the relatively small residual scatter shown in Fig. 3 after removing the effect of the surface-parcel difference.
c. Vertical-resolution error
The high-vertical-resolution ARM radiosondes were used to investigate the dependence of derived CAPE on vertical resolution of the temperature and moisture sounding. A boxcar smoother was applied to each vertical profile for a range of boxcar full widths, and CAPE was recomputed for each smoothed profile. A least squares linear fit of the smoothed versus original CAPE values was computed for the 10-yr radiosonde dataset. Figure 4 plots the linear slope fit versus smoothing resolution. The error bars represent the 95% confidence interval (2 standard deviations) in the uncertainty of the linear slope fit. There is a 17%–19% reduction in CAPE for a full-width vertical smoothing between 2 and 3 km. Table 2 shows that AIRS has a CAPE reduction of 17% at the 50th percentile relative to the higher-resolution radiosondes. The AIRS–AMSU instrument suite is designed to provide a tropospheric retrieval uncertainty of 1.0 K over a 1-km layer for temperature and 10% over 2-km layers for water vapor (Fetzer et al. 2003). The actual vertical resolution for the hyperspectral infrared is ~2.5 km for temperature and ~3 km for moisture in the lower troposphere (Maddy and Barnet 2008). The inherent vertical smoothing of the AIRS infrared channels could explain the 17% systematic bias found between AIRS and ARM-radiosonde CAPE distributions. The ERA-model 50th percentile is biased by −23% relative to radiosonde profiles, which may be due to a vertical smoothing inherent in the NWP model, particularly with respect to the vertical layering of water vapor. The lower troposphere near the top of the boundary layer is particularly sensitive to vertical resolution.
ARM SGP radiosonde CAPE smoothing error vs vertical resolution (2005–14).
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
d. Measurement error
The results shown in Figs. 2 and 3 revealed a sensitivity of CAPE to surface-parcel temperature and dewpoint temperature; that is, a difference in the surface-parcel estimate almost entirely explains the low correlation in the computed CAPE values of AIRS and ERA relative to ARM radiosondes. To quantify this effect, the ARM-radiosonde surface temperature and dewpoint temperature were used as a reference to compute the difference in the surface-parcel temperature and dewpoint temperature estimates of the AIRS and ERA profiles closest to the ARM SGP launch site. This complete set includes cases of zero and nonzero CAPE; that is, it represents all meteorological conditions for the 10-yr period that pass the quality control of Table 1. The mean differences over the entire dataset are less than 1°C, with a standard deviation of ~2°C. This very good statistical agreement in the mean disguises an error when CAPE is nonzero. Table 3 shows the result of analyzing the 10-yr matchup dataset for the subset of cases with CAPE greater than a minimum cutoff. The most notable feature in Table 3 is how the difference in surface dewpoint changes from near zero for all CAPE values to −2°C for the subset with CAPE values that are greater than a minimum value of 50 J kg−1. This difference is the same for both AIRS and ERA. As the CAPE minimum threshold increases to values of greater than 2500 J kg−1, the bias difference grows, with ERA exceeding −5°C and AIRS exceeding −7°C. Surface air temperature difference also grows with increasing CAPE, but the difference is less than one-half as large as the dewpoint temperature difference. Table 3 shows similar behavior between ERA and AIRS for CAPE up to 1500 J kg−1. For higher CAPE values, the AIRS bias difference exceeds that of ERA although both have equally large standard deviations. Note that the ARM site is a point measurement whereas AIRS and ERA represent the mean of a much larger area (45 × 45 km2). This spatial-scale mismatch could explain some of the apparent differences.
Error relative to ARM SGP radiosondes in surface-parcel temperature and dewpoint temperature estimates from ERA and from AIRS using the quality control shown in Table 1.
To characterize the extent to which differences in the surface-parcel estimates lead to differences in the derived CAPE estimates, a correlation coefficient was computed between AIRS and ARM radiosonde for nonzero CAPE values with a range of quality-control criteria. Decreasing the AIRS cloud-fraction cutoff increases the correlation with ARM radiosondes from 0.35 to 0.5, but the highest correlation (>0.8) is achieved only when the surface dewpoint of AIRS is within 1°C of the ARM-site radiosonde. This is illustrated in Fig. 5. The left-hand panels show the effect of changing the cloud-fraction cutoff from <0.8 to <0.1, and the respective right-hand panels show the additional effect of restricting the subset to dewpoint temperature with agreement better than 1°C. This result demonstrates that the scatter in the matchup between AIRS and ARM SGP radiosonde CAPE values is primarily driven by a difference in the estimation of the surface-parcel dewpoint temperature. This conclusion is consistent with that of Sanders (1986), who found that temperatures of air parcels lifted from the surface are sensitive mainly to surface dewpoint for very warm humid air. The comparison of the right-hand panels of Fig. 5 shows that a high correlation (>0.84) can be obtained even for AIRS cloud fractions up to 0.8 as long as the surface dewpoint estimate is within 1° of surface measurements. Note that Fig. 3 shows that the correlation increases to 0.95 when the AIRS surface-parcel values are replaced with the ARM-radiosonde values.
AIRS vs ARM-radiosonde data with AIRS cloud fraction of (top left) less than 0.8 and (bottom left) less than 0.1. (top right), (bottom right) The respective relationships when only a subset with surface dewpoint within 1° is used.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
Figure 6 shows a similar increase in the correlation coefficient of ERA with ARM radiosondes for the cases in which ERA surface dewpoint is in agreement with ARM surface dewpoint within 1°C. The correlation coefficient between ERA and ARM radiosondes increases from 0.5 to 0.78, but the number of available profiles decreases dramatically. Figure 3 shows that this correlation increases to 0.99 when the ARM-radiosonde surface values are used in the ERA CAPE calculation.
ERA vs ARM-radiosonde data (top) for all data and (bottom) for a subset with surface dewpoint within 1°, showing an increase in correlation coefficient from 0.54 to 0.78.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
6. Conclusions
Several authors have validated AIRS retrieved temperature and moisture vertical profiles (Tobin et al. 2006; Divakarla et al. 2006; Fetzer et al. 2006; Bedka et al. 2010, Roman et al. 2016), but only a limited study has been published previously on the accuracy of CAPE derived from AIRS profiles relative to radiosondes (Botes et al. 2012). That study commented on the lack of correlation between CAPE derived from AIRS and CAPE derived from the small number of radiosonde profiles considered, but no explanation was provided for the result. For the current study, a long time series of AIRS and radiosonde matchups was created to characterize the systematic biases and random-noise characteristics of the hyperspectral infrared satellite retrievals used to derive CAPE. As shown in Table 2, the AIRS and ERA CAPE distributions share similar characteristics, including a smaller median value relative to ARM radiosondes. A −17% bias in the 50th percentile of the AIRS CAPE distribution was found. This underestimate of the CAPE PDF is consistent with the lower vertical resolution of the satellite and NWP products relative to that of the radiosondes used as reference. The relatively poor correlation of AIRS and ERA CAPE with matched ARM SGP radiosondes (0.35 and 0.55, respectively) is explained by a difference in estimation of the surface-parcel temperature and dewpoint. Replacement of the AIRS (ERA) surface-parcel temperature and dewpoint with the ARM-radiosonde surface values while retaining the AIRS (ERA) profile above 2-m height causes the correlation to increase to 0.94 (0.99). Rather than resulting from a difference in the AIRS or ERA surface air temperature, this result may simply be a variation in the surface point measurement when compared with the much larger average satellite or ERA sounding footprint. In any case, it strongly suggests that the AIRS temperature and dewpoint vertical profiles above the surface are providing an accurate characterization of the atmosphere thermodynamic structure. This conclusion implies that the hyperspectral infrared sounding profiles can be used successfully to compute surface-based CAPE values when combined with surface observations. Future work in this area will focus on the incorporation of near-real-time Automated Surface Observing System surface observations with hyperspectral sounder profiles from direct broadcasts.
The authors speculate that the AIRS estimate of the surface 2-m humidity is an extrapolation of vertical gradients in water vapor mixing ratio from above and into the boundary layer. When the atmosphere is well mixed this approach appears to work well, but atmospheres with high CAPE are typically found to have a capping temperature inversion at the top of the atmosphere. In that case, the atmosphere is not well mixed and often has high temperature and specific humidity levels near the surface. This study suggests the value of improving the vertical resolution of the satellite retrievals in the boundary layer. The existing selection of hyperspectral infrared channels used in the AIRS algorithm is not able to sample the boundary layer with sufficient vertical resolution to detect low-altitude moisture gradients. Possible improvements could be made by using a more continuous set of spectral channels including on and off of weak water vapor lines, improving the treatment of infrared surface emissivity, improving the radiative transfer for the downwelling emission at the surface that reflects off the surface, and making use of an NWP 2-m surface analysis in the retrieval first guess.
Acknowledgments
The authors thank Genevieve Burgess for assisting in the acquisition of AIRS products and Greg Blumberg for discussions on the use of the SHARPpy software. We acknowledge the support of NOAA Grant NA10NES4400013. In addition, acknowledgment is made of the contribution of Dave Turner for the ARM data obtained from the U.S. DOE archive (at http://www.archive.arm.gov).
APPENDIX
AIRS CAPE Case-Study Examples
This study was motivated by two deadly tornados that struck the Oklahoma City, Oklahoma, area in May 2013: one in Moore, Oklahoma, on 20 May and one in El Reno, Oklahoma, on 31 May. Although the study results are primarily statistical, this appendix serves as an illustration of how the spatial coverage of soundings from the new generation of polar-orbiting weather satellites could be useful to operational forecasters. A snapshot of the cloud coverage in the larger SGP region is shown in Fig. A1 at the time of the afternoon Aqua satellite overpass. Comparison of the two satellite images suggests that the 20 May case had more deep convection earlier in the day than the 31 May case. In fact, the 20 May Moore tornado touched down at 1456 CST 15 km north of Norman, Oklahoma, whereas the 31 May El Reno tornado touchdown was not until 1803 CST about 75 km northwest of Norman. For the same time as the visible image shown in Fig. A1, Fig. A2 shows the CAPE derived from the AIRS observations at 1930 UTC in comparison with CAPE derived from the ERA-model field at 1800 UTC. The star symbol represents the location of the ARM site (36°36′18.0″N, 97°29′6.0″W), and the circle symbol is the Norman NWS site (35°13′21.0″N, 97°26′21.0″W) near Oklahoma City, separated by a distance of 153 km. The AIRS and ERA CAPE values have similar magnitudes, and both indicate severe CAPE. In the 20 May case, however, the AIRS estimate of highest CAPE covers a larger spatial extent to the southwest of the NWP-model estimate. In the 31 May case, the location of the highest AIRS CAPE is displaced westward of that shown in the ERA even though the satellite observations are nominally 1.5 h later than the analysis time. These two cases illustrate how satellite observations can help to refine the interpretation of NWP-model fields by either confirming the latest model forecasts or adjusting the timing, location, and intensity of potential convection hours before it becomes severe. The satellite-product quality control also must be considered, so as to remove low-quality or incomplete profiles. The cloud fraction for the AIRS sounding footprint was found to be a key variable for the quality control of AIRS CAPE. Figure A3 illustrates the AIRS cloud fraction within each sounding footprint and the corresponding CAPE quality mask. An AIRS cloud fraction of less than 0.8 is one of four criteria shown in Table 1 that was used to create a quality mask for AIRS CAPE that was used in this study. Figure A3 illustrates that the quality mask developed in this work retains useful AIRS CAPE estimates west, southwest, and south of Norman in the preconvective environment for both of these severe-weather cases.
Aqua MODIS illustrates the cloud development in the SGP region at (top) 1930 UTC 20 May 2013 ~1 h prior to the Moore tornado and (bottom) 1930 UTC 31 May 2013 ~4 h prior to the El Reno tornado near Oklahoma City.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
CAPE values (J kg−1) computed from (left) ERA-Interim at 1800 UTC and (right) AIRS satellite observation at 1930 UTC (top) 20 and (bottom) 31 May 2013.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
AIRS estimates of cloud fraction for (top left) 20 and (bottom left) 31 May 2013. (top right), (bottom right) The respective corresponding AIRS-derived CAPE quality masks, where blue indicates excluded soundings and red shows accepted soundings.
Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0267.1
REFERENCES
Barnes, L. R., E. C. Gruntfest, M. H. Hayden, D. M. Schultz, and C. Benight, 2007: False alarms and close calls: A conceptual model of warning accuracy. Wea. Forecasting, 22, 1140–1147, doi:10.1175/WAF1031.1.
Bedka, S., R. Knuteson, H. Revercomb, D. Tobin, and D. Turner, 2010: An assessment of the absolute accuracy of the Atmospheric Infrared Sounder v5 precipitable water vapor product at tropical, midlatitude, and Arctic ground-truth sites: September 2002 through August 2008. J. Geophys. Res., 115, D17310, doi:10.1029/2009JD013139.
Berrisford, P., and Coauthors, 2011: The ERA-Interim archive version 2.0. European Centre for Medium-Range Weather Forecasts ERA Rep. 1, 23 pp. [Available online at http://www.ecmwf.int/sites/default/files/elibrary/2011/8174-era-interim-archive-version-20.pdf.]
Bevis, M., S. Businger, T. A. Herring, C. Rocken, R. A. Anthes, and R. H. Ware, 1992: GPS meteorology: Remote sensing of atmospheric water vapor using the global positioning system. J. Geophys. Res., 97, 15 787–15 801, doi:10.1029/92JD01517.
Blanchard, D. O., 1998: Assessing the vertical distribution of convective available potential energy. Wea. Forecasting, 13, 870–877, doi:10.1175/1520-0434(1998)013<0870:ATVDOC>2.0.CO;2.
Botes, D., J. R. Mecikalski, and G. J. Jedlovec, 2012: Atmospheric Infrared Sounder (AIRS) sounding evaluation and analysis of the pre-convective environment. J. Geophys. Res., 117, D09205, doi:10.1029/2011JD016996.
Breznitz, S., 1984: The false alarm effect. Cry Wolf: The Psychology of False Alarms, Lawrence Erlbaum Associates, 9–16.
Chahine, M. T., and Coauthors, 2006: AIRS: Improving weather forecasting and providing new data on greenhouse gases. Bull. Amer. Meteor. Soc., 87, 911–926, doi:10.1175/BAMS-87-7-911.
Cheinet, S., A. Beljaars, M. Köhler, J.-J. Morcrette, and P. Viterbo, 2005: Assessing physical processes in the ECMWF model forecasts using the ARM SGP observations. European Centre for Medium-Range Weather Forecasts ARM Rep. 1, 25 pp. [Available online at http://www.ecmwf.int/sites/default/files/elibrary/2005/8658-assessing-physical-processes-ecmwf-model-forecasts-using-arm-sgp-observations.pdf.]
Cintineo, J. L., M. J. Pavolonis, J. M. Sieglaff, and D. T. Lindsey, 2014: An empirical model for assessing the severe weather potential of developing convection. Wea. Forecasting, 29, 639–653, doi:10.1175/WAF-D-13-00113.1.
Davis, G., 2007: History of the NOAA satellite program. J. Appl. Remote Sens., 1, 012504, doi:10.1117/1.2642347.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Divakarla, M. G., C. D. Barnet, M. D. Goldberg, L. M. McMillin, E. Maddy, W. Wolf, L. Zhou, and X. Liu, 2006: Validation of Atmospheric Infrared Sounder temperature and water vapor retrievals with matched radiosonde measurements and forecasts. J. Geophys. Res., 111, D09S15, doi:10.1029/2005JD006116.
Doswell, C. A., III, 2004: Weather forecasting by humans—Heuristics and decision making. Wea. Forecasting, 19, 1115–1126, doi:10.1175/WAF-821.1.
Doswell, C. A., III, and E. N. Rasmussen, 1994: The effect of neglecting the virtual temperature correction on CAPE calculations. Wea. Forecasting, 9, 625–629, doi:10.1175/1520-0434(1994)009<0625:TEONTV>2.0.CO;2.
ECMWF, 2009: ERA-Interim project. National Center for Atmospheric Research Computational and Information Systems Laboratory Research Data Archive, accessed 12 January 2015, doi:10.5065/D6CR5RD9.
Fetzer, E., and Coauthors, 2003: AIRS/AMSU/HSB validation. IEEE Trans. Geosci. Remote Sens., 41, 418–431, doi:10.1109/TGRS.2002.808293.
Fetzer, E., B. H. Lambrigtsen, A. Eldering, H. H. Aumann, and M. T. Chahine, 2006: Biases in total precipitable water vapor climatologies from Atmospheric Infrared Sounder and Advanced Microwave Scanning Radiometer. J. Geophys. Res., 111, D09S16, doi:10.1029/2005JD006598.
Goldberg, M. D., H. Kilcoyne, H. Cikanek, and A. Mehta, 2013: Joint Polar Satellite System: The United States next generation civilian polar-orbiting environmental satellite system. J. Geophys. Res. Atmos., 118, 13 463–13 475, doi:10.1002/2013JD020389.
Golden, J., and C. Adams, 2000: The tornado problem: Forecast, warnings, and response. Nat. Hazards Rev., 1, 107–118, doi:10.1061/(ASCE)1527-6988(2000)1:2(107).
Goldsmith, J. E. M., F. H. Blair, S. E. Bisson, and D. D. Turner, 1998: Turn-key Raman lidar for profiling atmospheric water vapor, clouds, and aerosols. Appl. Opt., 37, 4979–4990, doi:10.1364/AO.37.004979.
Halbert, K. T., W. G. Blumberg, and P. T. Marsh, 2015: SHARPpy: Fueling the Python cult. Fifth Symp. on Advances in Modeling and Analysis Using Python, Phoenix, AZ, Amer. Meteor. Soc., 402. [Available online at https://ams.confex.com/ams/95Annual/webprogram/Paper270233.html.]
Hart, J. A., J. Whistler, R. Lindsay, and M. Kay, 1999: NSHARP, version 3.10. National Centers for Environmental Prediction Storm Prediction Center Rep., 33 pp.
Hilton, F., and Coauthors, 2012: Hyperspectral Earth observation from IASI: Five years of accomplishments. Bull. Amer. Meteor. Soc., 93, 347–370, doi:10.1175/BAMS-D-11-00027.1.
Johns, R. H., and C. A. Doswell III, 1992: Severe local storms forecasting. Wea. Forecasting, 7, 588–612, doi:10.1175/1520-0434(1992)007<0588:SLSF>2.0.CO;2.
Maddy, E. S., and C. D. Barnet, 2008: Vertical resolution estimates in version 5 of AIRS operational retrievals. IEEE Trans. Geosci. Remote Sens., 46, 2375–2384, doi:10.1109/TGRS.2008.917498.
McNally, A. P., P. D. Watts, J. A. Smith, R. Engelen, G. A. Kelly, J. N. Thépaut, and M. Matricardi, 2006: The assimilation of AIRS radiance data at ECMWF. Quart. J. Roy. Meteor. Soc., 132, 935–958, doi:10.1256/qj.04.171.
McNulty, R. P., 1995: Severe and convective weather: A central region forecasting challenge. Wea. Forecasting, 10, 187–202, doi:10.1175/1520-0434(1995)010<0187:SACWAC>2.0.CO;2.
Miloshevich, L. M., H. Vömel, D. N. Whiteman, and T. Leblanc, 2009: Accuracy assessment and correction of Vaisala RS92 radiosonde water vapor measurements. J. Geophys. Res., 114, D11305, doi:10.1029/2008JD011565.
Moncrieff, M. W., and J. S. A. Green, 1972: The propagation and transfer properties of steady convective overturning in shear. Quart. J. Roy. Meteor. Soc., 98, 336–352, doi:10.1002/qj.49709841607.
Parkinson, C. L., 2003: Aqua: An Earth-observing satellite mission to examine water and other climate variables. IEEE Trans. Geosci. Remote Sens., 41, 173–183, doi:10.1109/TGRS.2002.808319.
Revercomb, H. E., and Coauthors, 2003: The ARM program’s water vapor intensive observation periods. Bull. Amer. Meteor. Soc., 84, 217–236, doi:10.1175/BAMS-84-2-217.
Riemann-Campe, K., K. Fraedrich, and F. Lunkeit, 2009: Global climatology of convective available potential energy (CAPE) and convective inhibition (CIN) in ERA-40 reanalysis. Atmos. Res., 93, 534–545, doi:10.1016/j.atmosres.2008.09.037.
Roman, J., R. Knuteson, T. August, T. Hultberg, S. Ackerman, and H. Revercomb, 2016: A global assessment of NASA AIRS v6 and EUMETSAT IASI v6 precipitable water vapor using ground-based GPS SuomiNet stations. J. Geophys. Res. Atmos., 121, 8925–8948, doi:10.1002/2016JD024806.
Rothfusz, L., C. Karstens, and D. Hilderband, 2014: Next-generation severe weather forecasting and communication. Eos, Trans. Amer. Geophys. Union, 95, 325–326, doi:10.1002/2014EO360001.
Sanders, F., 1986: Temperatures of air parcels lifted from the surface: Background, application and nomograms. Wea. Forecasting, 1, 190–205, doi:10.1175/1520-0434(1986)001<0190:TOAPLF>2.0.CO;2.
Smith, W. L., Sr., and Coauthors, 2009: Technical note: Evolution, current capabilities, and future advances in satellite nadir viewing ultra-spectral IR sounding of the lower atmosphere. Atmos. Chem. Phys., 9, 5563–5574, doi:10.5194/acp-9-5563-2009.
Smith, W. L., Sr., E. Weisz, S. V. Kirev, D. K. Zhou, Z. Li, and E. E. Borbas, 2012: Dual-regression retrieval algorithm for real-time processing of satellite ultraspectral radiances. J. Appl. Meteor. Climatol., 51, 1455–1476, doi:10.1175/JAMC-D-11-0173.1.
Stensrud, D. J., and Coauthors, 2009: Convective-scale warn-on-forecast system: A vision for 2020. Bull. Amer. Meteor. Soc., 90, 1487–1499, doi:10.1175/2009BAMS2795.1.
Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Radiation Measurement (ARM) program: Programmatic background and design of the cloud and radiation test bed. Bull. Amer. Meteor. Soc., 75, 1201–1221, doi:10.1175/1520-0477(1994)075<1201:TARMPP>2.0.CO;2.
Susskind, J., C. D. Barnet, and J. M. Blaisdell, 2003: Retrieval of atmospheric and surface parameters from AIRS/AMSU/HSB data in the presence of clouds. IEEE Trans. Geosci. Remote Sens., 41, 390–409, doi:10.1109/TGRS.2002.808236.
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, 883–907, doi:10.1109/TGRS.2010.2070508.
Susskind, J., J. M. Blaisdell, and L. Iredell, 2014: Improved methodology for surface and atmospheric soundings, error estimates, and quality control procedures: The Atmospheric Infrared Sounder science team version-6 retrieval algorithm. J. Appl. Remote Sens., 8, 084994, doi:10.1117/1.JRS.8.084994.
Tian, B., E. J. Fetzer, B. H. Kahn, J. Teixeira, E. Manning, and T. Hearty, 2013: Evaluating CMIP5 models using AIRS tropospheric air temperature and specific humidity climatology. J. Geophys. Res. Atmos., 118, 114–134, doi:10.1029/2012JD018607.
Tobin, D. C., and Coauthors, 2006: Atmospheric Radiation Measurement site atmospheric state best estimates for Atmospheric Infrared Sounder temperature and water vapor retrieval validation. J. Geophys. Res., 111, D09S14, doi:10.1029/2005JD006103.
Turner, D. D., and J. E. M. Goldsmith, 1999: Twenty-four-hour Raman lidar water vapor measurements during the Atmospheric Radiation Measurement program’s 1996 and 1997 water vapor intensive observation periods. J. Atmos. Oceanic Technol., 16, 1062–1076, doi:10.1175/1520-0426(1999)016<1062:TFHRLW>2.0.CO;2.
Weisz, E., W. L. Smith Sr., and N. Smith, 2013: Advances in simultaneous atmospheric profile and cloud parameter regression based retrieval from high-spectral resolution radiance measurements. J. Geophys. Res. Atmos., 118, 6433–6443, doi:10.1002/jgrd.50521.
