Plain Language Summary
Data from a laser-based instrument called a lidar (which can measure
clouds and aerosols) and an optically based lightning detection
instrument, both hosted on the International Space Station during
March-October 2017, were used to study global thunderstorms from space.
Because these types of instruments have only been rarely combined in the
past, this study focused on analyzing a small dataset in order to
determine how useful the instrument combination can be. The results
found that thunderstorm cloud-top heights slope downward from the
Equator toward the poles, similar to how the tropopause height also
slopes downward. Lidar measurements of cloud properties, like cloud-top
height and the amount of ice in the cloud, were quantitatively related
to lightning observations like flash rate. The lidar also was helpful in
finding instances where the lightning instrument accidentally detected
glint from the sun on water, snow, etc. instead of lightning, because
there were no lidar-detected clouds nearby. However, this rarely
occurred due to how the lightning instrument’s data are processed.
Additional fruitful scientific insights can be expected from other,
larger combined lidar/lightning datasets.
1 Introduction
1.1 Background
Lidar is a common tool for measuring clouds, aerosols, and atmospheric
state, and has been used for many decades from ground-based (e.g.,
Sassen, 1977), airborne (e.g., McGill et al., 2007), and spaceborne
platforms (e.g., Winker et al., 2006). A fundamental aspect of lidar is
its difficulty in penetrating deeply into optically thick clouds, such
as cumulonimbus (i.e., thunderstorms). Nevertheless, some studies have
successfully used lidar to study characteristics of deep convection,
thunderstorms, and even lightning. For example, Sassen (1977) took
advantage of relatively high-altitude cloud bases in Wyoming to use a
ground-based lidar to study the optical scattering characteristics of
melting precipitation in summertime thunderstorms. Airborne lidars have
been used to document many aspects of thunderstorms as well as their
surrounding environments. For example, Sassen et al. (2000) and Campbell
et al. (2005) used lidar to study the microphysical properties of deep
convective cloud tops and thunderstorm anvils. These studies often have
made use of polarization-diversity lidar measurements to infer
cloud-particle phase, habit, and orientation within these anvils,
qualitatively similar to how polarimetric microwave radar has been used
to study precipitation characteristics deep within thunderstorms (e.g.,
Kumjian & Ryzhkov, 2008). Outside of precipitation, airborne lidar has
been used to quantify wind profiles near deep convection (e.g., Cui et
al., 2020). Meanwhile, ground-based ozone lidars have been used to
document lightning-produced nitrogen oxides (NOx) in the vicinity of
thunderstorms (Wang et al., 2015). Lidar has even been proposed as a
method to remotely sense electromagnetic fields in order to estimate the
possibility of lightning strikes outside of clouds (Shiina et al.,
2006).
Combined observations of deep convection from radar, microwave
radiometers, infrared spectrometers, and lidar have also been made
(Heymsfield & Fulton, 1988; McGill et al., 2004; van Diedenhoven et
al., 2016). Indeed, these combined measurements with lidar and other
instruments form the scientific basis for the CloudSat and Cloud-Aerosol
Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) missions
(Mace et al., 2009), as well as the overall A-Train satellite
constellation (Delanoe & Hogan, 2010) and the future Atmosphere
Observing System (AOS, 2022). Such combined measurements can enable more
accurate retrievals of microphysical processes near cloud top.
However, detailed comparisons between lightning and lidar observations
are rare in the literature. One notable recent study was Allen et al.
(2021), which made use of airborne observations from the Fly’s Eye
Geostationary Lightning Mapper Simulator (FEGS; Quick et al., 2021), the
Cloud Physics Lidar (CPL; McGill et al., 2002), and an ultraviolet (UV)
visible spectrometer to estimate NOx production by lightning. The role
played by CPL was to measure cloud-top pressure, which played an
important role in the NOx calculations, while FEGS was used to document
the detection efficiencies of ground-based and spaceborne lightning
observations for the storms that were studied. However, the airborne
lidar and lightning observations were not directly compared to better
resolve thunderstorm structure or processes. Moreover, coupled
spaceborne lidar/lightning studies do not appear to exist in the
literature. Potential advantages and challenges of doing more direct
lidar/lightning comparisons are discussed below, particularly from the
perspective of spaceborne platforms.
1.2 Potential advantages of using spaceborne lidar to study
thunderstorms
Some potential advantages of spaceborne lidar include its ability to
provide a more accurate measurement of cloud-top characteristics than,
for example, spaceborne radar. Cloud-top height has previously been
related to lightning flash rate (e.g., Price & Rind, 1992). In
addition, lightning within the overshooting tops of thunderstorms has
been shown to be of significance to thunderstorm electrification and
charge structure (MacGorman et al., 2017). However, echo-top height as
determined by radar strongly depends on the sensitivity of the radar
itself. For example, the Ku-band radar used in the Tropical Rainfall
Measuring Mission (TRMM), and the Ku- and Ka-band radars used in the
Global Precipitation Measurement (GPM) mission, have minimum
reflectivity sensitivities in excess of 10 dBZ (Kummerow et al., 1998;
Hou et al., 2014). This means that these radars are not sensitive to the
weaker echoes associated with true cloud tops (Hagihara et al., 2014).
Lidar has been used to help diagnose cloud-top height underestimates in
thermal imagery as well (e.g., Sherwood et al., 2004).
In addition, lidars are able to infer the dominant phase of hydrometeors
near cloud top (Yoshida et al., 2010), as well as other microphysically
related attributes like cloud optical depth and ice water content (IWC)
and path (IWP; e.g., Avery et al., 2012). That being said, optical
measurements do not penetrate deeply into thick clouds, particularly
deep convection. However, studies like Rutledge et al. (2020) have
demonstrated that near-cloud-top measurements of thunderstorms are
useful as their optical properties have (among other things)
implications for the detection efficiency of optically based lightning
mappers like the Lightning Imaging Sensor (LIS; Kummerow et al., 1998;
Blakeslee et al., 2020). In addition, lidars are capable of providing
vertical structure information within the portions of clouds they do
penetrate.
Lidars have been in space for longer than a decade (Winker, 2022),
providing climate-quality records of cloud properties. This time period
overlaps both LIS and Geostationary Lightning Mapper (GLM; Rudlosky et
al. 2019) observations for many years. This means that there may be
enough conjunctions between these two observations that useful analysis
of thunderstorm properties could be performed. Moreover, lidars are
planned to be part of the forthcoming AOS (2020), while GLMs and other
spaceborne observations of lightning are also planned into the near
future (e.g., Holmlund et al., 2021).
Though these data were not examined in this study, lidars are capable of
detecting and categorizing aerosol properties (Ceolato & Berg, 2021).
Since aerosols have been shown to be important for the strength and
properties of convection, including thunderstorms (Khain et al., 2005),
combining lidar observations of both cloud and aerosol properties with
lightning observations could provide potentially useful information
about thunderstorm-aerosol interactions. Moreover, it would be of
interest to perform composition studies like Allen et al. (2021) using
spaceborne platforms to provide a more global perspective on NOx
production by lightning.
1.3 Potential challenges of using spaceborne lidar to study
thunderstorms
All the above being said, there are significant challenges to using
lidar to study thunderstorm properties. In addition to lidar’s
well-known inability to penetrate thick clouds, the sampling
characteristics of spaceborne lidar and lightning observations are very
different. For example, lightning observations from LIS and GLM are
distributed horizontally, and include information about flash
two-dimensional (2D) location (Rudlosky et al., 2019; Blakeslee et al.,
2020). But they do not measure the vertical structure of lightning. On
the other hand, lidars typically measure along a nadir curtain (Yorks et
al., 2016; Winker, 2022), which means they provide vertically
distributed observations that lack horizontal context. Thus, care needs
to be taken when comparing spaceborne lidar and lightning datasets, but
no well-tread analysis pathways exist with comparing these two datasets
unlike, e.g., radar and lightning datasets (Rust & Doviak, 1982; Lopez
& Aubagnac, 1997; Wiens et al., 2005; Carey et al., 2019). There is
utility in exploring a small overlapping dataset of spaceborne lidar
observations of thunderstorms, to better understand the advantages and
challenges of doing this combined analysis, without committing
significant resources on a “wild goose chase” if the analysis does not
yield much scientific value.
1.4 Goals of this study
As it turns out, such a small overlapping dataset exists. For just short
of 8 months in 2017, the Cloud-Aerosol Transport System (CATS) lidar
(Yorks et al., 2016) overlapped with a LIS instrument on the
International Space Station (ISS). As will be shown in this study, these
co-located instruments provided a useful dataset for demonstrating the
value of using lidar to study thunderstorm characteristics. This paper
will discuss the advantages and challenges associated with this combined
analysis, and will provide a path forward for more detailed analysis
using larger overlapping datasets.
2 Data and Methodology
2.1 ISS LIS
The International Space Station Lightning Imaging Sensor (ISS LIS) is a
high-speed camera (500 frames per second) affixed to a telescope that
detects lightning via monitoring transients at 777.4 nm. LIS is a
modified flight spare of the TRMM LIS (1997-2015) instrument, hosted
within the 5th Space Test Program – Houston (STP-H5) payload, launched
in 2017.
ISS LIS extends TRMM LIS time series observations, expands latitudinal
coverage, provides near-realtime data to operational users, and enables
cross-sensor calibrations (e.g., with GLM). A thorough review of the ISS
LIS sensor is provided in Blakeslee et al. (2020). The instrument’s
flash detection efficiency is approximately 60%, with sub-pixel
(< 4 km) location accuracy and sub-frame (< 2 ms)
timing accuracy. Quality-controlled, flash-level data and science
backgrounds from ISS LIS (Blakeslee 2000a, b) were used in this study.
These data are available starting 1 March 2017.
2.2 CATS
The CATS lidar made range-resolved measurements of clouds and aerosols
at 1064 and 532 nm during 2015-2017 (Yorks et al., 2016). CATS had a
vertical resolution of 60 m and a horizontal resolution of 5 km. The
Level 2 products used in this study included vertical feature mask
(e.g., liquid vs. frozen cloud, or aerosol type), as well as profiles of
cloud properties (e.g., IWC, cloud optical depth).
CATS overlapped on the ISS with LIS during 1 March through 29 October
2017. After this period the CATS instrument’s mission concluded.
Coincidentally, the original CATS ray-tracing code was adapted and used
within ISS LIS geolocation routines (Lang, 2019), which have been
demonstrated to provide the aforementioned sub-pixel (< 4-km)
location accuracy for LIS-detected lightning (Blakeslee et al., 2020).
The previous TRMM-based LIS geolocation code was found to be not easily
translated to the ISS.
2.3 Combining LIS with CATS
Combining the 2D horizontally distributed LIS dataset is not
straightforward to do with the vertically profiling CATS lidar
measurements. Since CATS provided a nadir-focused curtain, the first
step to combining the datasets was to threshold on LIS flash centroid
distance from the CATS ground track. During the nearly 8-month overlap
period, this study determined that 8246 ISS LIS flashes had centroids
within 25 km of the CATS ground track. The 25-km threshold was chosen to
balance obtaining a co-location dataset large enough to enable useful
statistical analysis, while still only comparing lightning that was
close enough to the ground track to be potentially physically related to
CATS-measured cloud properties. Sensitivity experiments were also
performed with 50- and 10-km ground-track distance thresholds. These
results (not shown), were found to be qualitatively similar to the 25-km
results discussed in this paper.
An example of a typical LIS/CATS matchup is shown in Fig. 1. Near 0903
UTC on 15 March 2017, CATS shows high-level cloud (close to 17 km MSL
maximum height), a few km below which the signal is attenuated (Fig.
1a). The cloud itself is identified as primarily ice by the CATS feature
mask (Fig. 1b). This is suggestive of the anvil region above and around
deep convection. Meanwhile, more than a dozen ISS LIS flashes are
identified whose centroids are within 25 km of the CATS ground track
around this same time. This figure (and hundreds like it, not shown)
demonstrate that these very different datasets can be combined to
provide meaningful qualitative information about thunderstorms.
In order to explore the ability to retrieve more quantitative
conclusions about thunderstorms using the combined dataset, an automated
statistical analysis of CATS-retrieved cloud properties near ISS LIS
measured lightning was performed. When lightning was observed, the
maximum values of cloud parameters (e.g., IWP, cloud-top height, optical
depth) within 50 km along the CATS track were determined. In addition,
clusters of lightning flashes were identified (e.g., like in Fig. 1)
using the Density-Based Spatial Clustering of Applications with Noise
(DBSCAN) algorithm (Ester et al., 1996) with a minimum of 1 flash to
create a cluster, and a maximum distance of 50 km between flashes in a
cluster. The total number of flashes in each of these clusters was
similarly compared to maximum values of CATS cloud properties within 50
km along the ground track. The use of the 50-km distance threshold along
the track was to allow for the possibility that CATS did not gain a
well-placed overpass of the relevant thunderstorm’s core.