The airborne in situ water vapor measurements reported here are from the Jet Propulsion Laboratory Laser Hygrometer Mark2 (JLH Mark2), a tunable laser spectrometer with an open-path cell external to the aircraft fuselage (May, 1998). Water vapor is reported at 1 Hz (10 % accuracy), although the time response of the open-path cell is much faster than this because the instrument samples the free-stream airflow. This instrument has a redesigned optomechanical structure for greater optical stability and was first flown in this configuration on the NASA ER-2 high-altitude aircraft during the SEAC4RS field mission. Pressure and temperature, provided by the Meteorological Measurement System (MMS; Scott et al., 1990), are used in the data processing to calculate water vapor mixing ratios from spectra, as described in May (1998).

During SEAC4RS, nine aircraft flights targeted air parcels with recent convective influence (see Table 3 of Toon et al., 2016). Figure 1 shows the combined vertical profiles of JLH Mark2 water vapor from all 23 SEAC4RS flights. Outliers with high water vapor mixing ratios are the focus of this study. Enhanced water vapor was measured on 11 flights (Table 1). Here we define “enhanced water vapor” as mixing ratios greater than 2 standard deviations above the mean in situ measurement. For the overworld stratosphere in all 23 SEAC4RS flights, the mean H2O is 6.7 ± 1.5 ppmv at 380–400 K and 5.0 ± 0.8 ppmv at 400–420 K (Fig. 2). Thus, the threshold for enhanced water vapor is 9.7 ppmv at 380–400 K and 6.6 ppmv at 400–420 K. The majority of the measurements have background water mixing ratios characteristic of the overworld stratosphere, 4 to 6 ppmv. For the overworld stratosphere (potential temperature greater than 380 K), Fig. 1 shows enhanced water vapor at potential temperatures up to approximately 410 K (17.5 km of altitude). We define the “enhanced water region” as the layer of the overworld stratosphere where these events have been observed at 380–410 K of potential temperature corresponding to 16–17.5 km of altitude. Enhanced water vapor measured in situ by both the JLH Mark2 instrument (Fig. 1) and the Harvard Water Vapor instrument (Smith et al., 2015) on the NASA ER-2 aircraft indicated that the aircraft intercepted convectively influenced air. Other tracers measured on the aircraft did not change significantly in these plumes. For the SEAC4RS flights, the agreement between these two water vapor instruments is within ±10 % for stratospheric water. This is consistent with the AquaVIT laboratory intercomparison (Fahey et al., 2014) and other aircraft field missions (e.g., Rollins et al., 2014). The largest enhancements were observed on three flights that are described in detail in Sect. 4.

Aura MLS measures ∼ 3500 profiles each day of water vapor and other atmospheric species (Livesey et al., 2016). While the aircraft samples in situ water in a thin trajectory through the atmosphere, Aura MLS provides a larger-scale context. Expanding on the analysis of Schwartz et al. (2013), Aura MLS observations of stratospheric water vapor are presented here for the SEAC4RS time period in summer 2013. Aura MLS H2O has 0.4 ppmv of precision at 100 hPa for individual profile measurements with a spatial representativeness of 200 km along the line of sight (Schwartz et al., 2013). The results shown here use MLS version 4.2 data, but are not significantly different from the previous version 3.3. MLS observations over CONUS are at ∼ 14:10 local time (ascending orbit) and ∼ 01:20 local time (descending orbit) with successive swaths separated by ∼ 1650 km. The vertical resolution of the water vapor product is ∼ 3 km in the lower stratosphere (Livesey et al., 2016).

Aura MLS shows a seasonal maximum in water vapor over CONUS in July and August. The histogram of Aura MLS water vapor in Fig. 3 indicates that the July–August 2013 CONUS lower stratosphere was drier than the previous nine-summer MLS record (2004 to 2012). Nevertheless, enhanced lower stratospheric water vapor was observed by MLS in 2013 as a rare but detectable event. From the MLS histogram, the frequency of 100 hPa H2O > 8 ppmv was 0.9 % of the observations in July–August 2013 in the blue shaded box. Figure 4 shows that, out of all MLS 100 hPa water vapor retrievals over the 2-month period from July to August 2013, water greater than 8 ppmv was measured only nine times over North America (in the blue shaded box), three times near the west coast of Mexico and once over the Caribbean Sea.

In order to link the stratospheric water vapor encountered by the aircraft to the storm systems from which they may have originated, it is necessary to have a comprehensive continental-scale catalog of deep convection. Geostationary Operational Environmental Satellite (GOES) infrared imagery is used to assemble a catalog of OTs throughout the US and offshore waters. This catalog was acquired from the NASA LaRC Airborne Science Data for Atmospheric Composition data archive (http://www-air.larc.nasa.gov/cgi-bin/ArcView/seac4rs). Because OTs are correlated with storm intensity, the OT product was primarily developed to benefit the aviation community for more accurate turbulence prediction as well as the general public for earlier severe storm warnings. However, the product is also ideally suited for identifying storm systems that can moisten the stratosphere.

Infrared brightness temperatures are used to detect cloud top temperature anomalies within thunderstorm anvils. OT candidates are colder than the mean surrounding anvil, with the temperature difference indicative of both the strength of the convective updraft and the depth of penetration. For a description of the method, the reader is directed to Bedka et al. (2010). The horizontal spatial resolution of the OT product is dependent on the underlying satellite imagery resolution, i.e., the size of the GOES IR pixel, which is 7 km or less over the CONUS. Additional validation of OTs requires comparison with the Global Forecast System (GFS) Numerical Weather Prediction (NWP) model tropopause temperature. The maximum OT cloud height was derived based on knowledge of the (1) OT–anvil temperature difference, (2) the anvil cloud height based on a match of the anvil mean temperature near the OT and the GFS NWP temperature profile and (3) a temperature lapse rate within the UTLS region based on a GOES-derived OT–anvil temperature difference and NASA CloudSat OT–anvil height difference for a sample of direct CloudSat OT overpasses (Griffin et al., 2016). Griffin et al. (2016) find that 75 % of OT height retrievals are within 0.5 km of CloudSat OT height, so we conservatively estimate the accuracy of the OT altitude to 0.5 km. For SEAC4RS, every available GOES-East and GOES-West scan (typically 15 min resolution) was processed for the full duration of the mission, even for the non-flight days, yielding a detailed and comprehensive picture of the location, timing and depth of penetration of the convective storms over the entire CONUS. The output files include the OT coordinates, time, overshooting intensity in degrees K, which is related to the temperature difference between the OT and the anvil, and an estimate of maximum cloud height for OT pixels in meters.

The ability of GOES-East and GOES-West to observe an OT depends on its lifetime. OTs are transient events with lifetimes typically less than 30 min, but they can exceed 1 hour in well-organized storms such as mesoscale convective systems and supercell storms (Bedka et al., 2015; Solomon et al., 2016 and the references therein). Animations such as the following show the variability in OTs sampled by GOES at 1 min of resolution:

Infrared wavelength animation: http://cimss.ssec.wisc.edu/goes/srsor2015/

Visible wavelength animation:

http://cimss.ssec.wisc.edu/goes/blog/wp-content/uploads/

It is clear that some OTs are quite persistent and are both prominent and detectable in IR imagery, but the majority of OTs in these particular animations are short lived (< 10 min). Within these OTs, strong convective updrafts can transport ice to 16–18 km of altitude where turbulent processes, such as gravity wave breaking, mix tropospheric and stratospheric air (e.g., Mullendore et al., 2009, 2005; Wang, 2003; Homeyer et al., 2017), enabling the detrainment of ice and stratospheric hydration.

Bedka et al. (2010) showed that the OT detection algorithm has a false positive rate of 4.2 to 38.8 %, depending on the size of the overshooting and algorithm settings. As noted above, OTs are transient and can evolve quite rapidly. The storm top characteristics and evolution we see in the GOES data featured in this paper only capture a subset of the storm lifetimes, even if we were to have a 100 % OT detection rate, due to the 15 min resolution of the GOES imager. In addition, the relatively coarse GOES spatial resolution (up to 7 km over the northern latitudes of the US) can cause the Bedka et al. (2010) method to miss some small diameter and/or weak OT regions. We would be able to better map tracks of storm updrafts using data at 1 min of frequency like that shown by Bedka et al. (2015), but these data are not available over the broad geographic domains required for our analysis. Given uncertainties in back trajectories, GOES under-sampling and the fact that many OTs can be located in close proximity to one another, we are not able to make a direct connection between an individual OT and a stratospheric water vapor plume observed 1 day or more later. Rather, our analysis identifies a cluster of storms that are the best candidates for generating ice that sublimates into the enhanced water vapor plumes sampled by the ER-2.

Back trajectories were run from each flight profile for which enhanced water vapor was measured to determine whether the sampled air was convectively influenced. The trajectories were run with the FLEXPART model (Stohl et al., 2005) using NCEP Climate Forecast System version 2 (CFSv2) meteorology (Saha et al., 2014); the trajectory time step interval was 1 h. Trajectories were initialized every second along the flight track profiles and run backward for 7 days. A sampled air parcel was determined to be convectively influenced if the back trajectory from that parcel intercepted an OT region. The tolerances for a trajectory to be considered to have intercepted an OT cloud were ±0.25∘ latitude and longitude, ±3 h and ±0.5 km in altitude. These tolerances were chosen primarily due to the resolution of the NCEP meteorology used to run the trajectories (1∘ × 1∘) and based on personal communication with Leonard Pfister.

Figure 5 shows the details of the 8 August 2013 ER-2 aircraft flight. This flight was the transit flight from Palmdale, California (34.6∘ N, 118.1∘ W), to Ellington Field in Houston, Texas (29.6∘ N, 95.2∘ W). In addition to sending the NASA ER-2 aircraft to the destination base, the scientific goal of this flight was to create five profiles for the North American monsoon region plus the aircraft ascent and final descent. This flight shows a dramatic transition (from west to east) of background stratospheric water to enhanced water. In the lowermost stratosphere (350 K < θ < 380 K), water can be highly variable, but at 90 hPa it is generally unusual to observe water vapor greater than 6 ppmv. As shown in Fig. 5c, there is a gradient in water vapor from west to east: 4.0 to 4.4 ppmv at 90 hPa (17 km) over the west coast of CONUS (black and blue points) and greater than 10 ppmv at 90 hPa over Texas (green points). Simultaneous Aura MLS retrievals also demonstrate a west-to-east water vapor gradient on this day (lines and filled circles in Fig. 5c). Both JLH Mark2 and Aura MLS water vapor exceed the thresholds for enhanced water vapor.

An analysis of the 8 August 2013 case is shown in Fig. 6. For clarity, only some example trajectories (a subset of our analysis) are shown. These are displayed as thin blue traces in panels (b) and (c). The intersections of the example trajectories with coincident OT are shown as red squares in panels (b) and (c). All overshooting convective tops within ±3 h of the red squares are shown by the green symbols in panels (b) and (c). Back trajectories from the flight track follow the anticyclonic NAM circulation over Western Mexico, the Great Plains and the Mississippi Valley (Fig. 6b). Every one of the example back trajectories intersects OT, as shown by the red symbols in Fig. 6b. For this flight, coincidences with overshooting convection are dominated by overshooting clouds over the Mississippi Valley and the Great Plains. All overshooting convections within the tolerances prescribed (see Sect. 3.2) for the back trajectories are shown by the green symbols in Figs. 6b and 5c. Figure 6c demonstrates the range of altitudes reached by the coincident overshooting convection and how many convective overshooting cells were coincident. The high resolution of the convective overshooting data meant that there could be multiple coincident convective overshooting cells for a single location on a back trajectory. It is significant that some of the green overshooting cells are higher altitude than the red coincident points, suggesting that overshooting air parcels descended slightly before mixing with the surrounding air. Figure 6d indicates that the source of the enhanced water was dominated by overshooting clouds within 7 days prior to intercept by the aircraft.

The NASA ER-2 flight on 16 August 2013 was designed to survey the North American monsoon in a triangular flight path from Houston, Texas to the Imperial Valley in Southern California, to southeastern Colorado and back to Texas. The NASA ER-2 aircraft performed six dives, encountering enhanced stratospheric water at 16 to 17 km of altitude (Fig. 7a). As shown in Fig. 7b, back trajectories intersect overshooting tops over the South Central US (Texas, Oklahoma and Arkansas) and also over the Sierra Madre Occidental mountain range on the west coast of Mexico. This case is an example of the classic North American monsoon circulation with a moisture source over the Sierra Madre Occidental (Adams and Comrie, 1997), in which air parcels are transported from OT in Mexico around the anticyclone to the CONUS (Fig. 7b). The altitude range of the convective overshoot is typically 16 to 17 km, as shown in Fig. 7c. The time between OT and intercept by the aircraft ranges from 2 to 7 days (Fig. 7d).

The 27 August 2013 flight performed six dives to sample the North American monsoon. Stratospheric water was enhanced to 15 to 20 ppmv at altitudes ranging from 16.0 to 17.5 km (Fig. 8a). The ER-2 aircraft intercepted highly enhanced stratospheric water from a mesoscale convective complex over the Upper Midwest, which had overshooting tops over northern Minnesota and northern Wisconsin (Toon et al., 2016), as shown in Fig. 8b. Figure 8c shows an abundance of OT above 17 km (green). Generally speaking, the OTs appear at higher altitudes in the northern CONUS and southern Canada than in the central CONUS. Figure 8d shows that the air masses were sampled in situ by the ER-2 aircraft over Illinois and Indiana 1 to 2 days after the intense storm. As is a common theme for all these experiment days, a portion of the back trajectories also trace back to overshooting tops over the Sierra Madre Occidental 1 week prior.