The response of cloud processes to an aerosol perturbation is one of the largest uncertainties in the anthropogenic forcing of the climate. It occurs at a variety of timescales, from the near-instantaneous Twomey effect to the longer timescales required for cloud adjustments. Understanding the temporal evolution of cloud properties following an aerosol perturbation is necessary to interpret the results of so-called “natural experiments” from a known aerosol source such as a ship or industrial site. This work uses reanalysis wind fields and ship emission information matched to observations of ship tracks to measure the timescales of cloud responses to aerosol in instantaneous (or“snapshot”) images taken by polar-orbiting satellites.
As in previous studies, the local meteorological environment is shown to have a strong impact on the occurrence and properties of ship tracks, but there is a strong time dependence in their properties. The largest droplet number concentration (
The liquid water path (LWP) enhancement and the
The response of a cloud to an aerosol perturbation is fundamentally time sensitive. Increasing the number of cloud condensation nuclei (CCN) increases the number of cloud droplets at cloud base
Changes in droplet size can also impact precipitation processes, leading to further changes in liquid cloud properties, notably the liquid water path (LWP) and cloud fraction
One of the largest uncertainties when using observations to constrain aerosol–cloud interactions is the impact of meteorological covariations, where aerosol and cloud properties are both correlated to the same meteorological factors (such as relative humidity). Variations in this factor will then generate relationships between aerosol and cloud properties, even without a causal impact of aerosol on cloud
The majority of satellite observations provide a static picture of the Earth, limiting their ability to characterise liquid cloud temporal development. Previous studies have addressed this by using multiple observations to build a composite diurnal cycle.
These studies have focused on existing variability in aerosol. This allows cloud temporal development to be investigated at a global scale, but limits the ability to measure timescales directly. Exogenous aerosol perturbations (such as those from ships) are emitted independently of meteorological factors. With a limited spatial extent, they have clearly identifiable polluted and control regions, allowing the impact of the aerosol on the cloud field to be inferred
In this work, the temporal development of ship tracks is used to quantify the timescales of aerosol–cloud interactions in marine boundary layer clouds and how these are affected by meteorology. Although ship tracks appear in satellite images as linear cloud formations, they have no ability to transmit information along their length. This means that they can be considered as a chain of independently perturbed clouds with a similar initial aerosol perturbation
This work uses the ship track and ship locations from
For each ship, the estimated trajectory of the emitted
Ship location data have to be interpolated between sparse AIS observations, which can lead to significant error in the locations. A comparison with ship meteorological reports suggests that this interpolation error can often be as large as 100 km in the ship position, compounding further in the estimated emission trajectories. To avoid this interpolation uncertainty, only cases where the normalised Fréchet distance (the maximum distance between points on the two trajectories when they are traversed to minimise this distance) between the reconstructed plume and the identified ship track is less than 0.5 are included to ensure an unambiguous match between the ship and ship track
The identification of polluted regions is based on the method from
Example track chunks and the identification of polluted pixels along the track shown in Fig.
With a buffer region of two pixels (
An example ship track.
The classifications are then aggregated into 15 min (10 km) segments. This is a short enough time period to allow the initial development of the track to be resolved and these two measures are approximately equal for relative wind speed of 40
Cloud properties are recorded along the length of the ship track for both polluted pixels and the corresponding clean/control region outside the track (Fig.
The retrieved cloud properties in this work are from the MODIS Aqua collection 6.1 cloud product
The uncertainties throughout this work are calculated using a bootstrap method
To compare the radiative effect of ship tracks in different conditions, a potential radiative forcing (PRF) is calculated following Eq. (
The PRF is calculated for each segment individually, with
The results from this work are split into two sections; the first deals with the macrophysical properties of the ship track (length, width, detectability, CF), whereas the second focuses on the microphysical properties and the liquid water path.
The median ship track in this study is last observed at a distance of about 200 km from the source ship, with longer ship tracks more commonly observed behind ships with higher
There is a large variation in the time to initial observation (time to obs.) and the time to last observation for the ship tracks in this work (Fig.
Many of these ship tracks have gaps (segments where the track is not detected). The maximum detected fraction is around 1–2 h after emission, where each 15 min segment has an approximately 60 % chance of containing a detected ship track (Fig.
The majority of segments do not contain a detected track; what limits ship track formation in these segments? For each segment lacking a detected track, four tests are applied in order. T1: is the segment outside a MODIS image (granule)? T2: is there no cloud in the segment? T3: is the segment ice cloud? T4: is the
Reasons for track non-detection in segments as a function of time since emission. The tests are applied in the order T1–4.
Although many ship tracks are located close to the edge of a MODIS image (“granule”), only a small number of ship tracks disappear because they reach the edge of the MODIS granule. Around 30 % of ship tracks reach a granule edge 20 h after emission. In around 25 % of segments, a lack of cloud prevents the detection of a ship track. This is expected given the CF in this region and is approximately constant with time since emission. Similarly, the small impact of overlying ice cloud on the detection of ship tracks is primarily due to the low ice CF in subtropical subsidence regions.
The impact of a small
Ship track width (defined as the maximum cross-track distance between polluted pixels within a segment) increases gradually with time since emission (Fig.
This sensitivity of track width to
Differences in the cloud regimes indicated by
The high
Given the strong impact of cloud occurrence on the disappearance of ship tracks (Fig.
Ship track length as function of meteorological parameters.
Normalising by the total number of ship tracks remaining in the total sample highlights a much clearer role for meteorology, with ship tracks in low average EIS environments only half as likely to have a lifetime longer than 10 h compared those in a higher EIS environment due to the lower cloud fraction (Fig.
Despite a strong controlling influence on the formation of ship tracks
Wind speed has a stronger influence on ship track length, with higher wind speeds producing shorter ship tracks. Although ship track formation is more common at high wind speeds
As well as creating increases in
By estimating the ship aerosol trajectories, ship tracks are located even when there are no surrounding clouds (preventing the detection algorithm from operating). By selecting cases with a low
By selecting cases with a low
The previous section showed that the background meteorological state has a controlling influence on the macrophysical properties of ship tracks and their sensitivity to aerosol. These are not independent from the microphysical properties of the ship track, particularly the
Detected pixels (more than 2 standard deviations above the background
In segments where both the polluted and detected
Previous studies have shown a strong link between the ship emissions and the detected
A clearer signal is found using the polluted
Although noisy, some patterns can be observed in the LWP enhancement (
As well as being a clear function of time, the
As Fig.
The
Relative humidity also affects
The dependence of the
The impact of aerosol on the liquid water path is often quantified as the sensitivity of LWP to
The relationship between the mean LWP and
The LWP sensitivity to
To include this development, the average sensitivity (calculated individually for each segment using the
This bi-directional LWP sensitivity response is clear in (Fig.
A similar divergence in the sensitivity is observed as a function of cloud top humidity (Fig.
While these results support those found in previous studies, there are a number of important factors that must be taken into account when interpreting those studies. First, as demonstrated in model studies
By averaging the changes in LWP over hour-long periods, the temporal changes in LWP after the aerosol emission become clearer (Fig.
Development of LWP along ship tracks for
As seen in the sensitivity, when looking at cases with a clean background (Fig.
For ship tracks forming in polluted environments, there is a small decrease in LWP (Fig.
The combination of these increases in
The integrated potential forcing as a function of time since emission including (top row) and excluding (bottom row) CF changes (Eq.
The liquid cloud fraction has previously been shown to be the primary control on ship track formation
Although EIS has some control over the radiative effect of ship tracks, background
The
The aerosol impact through a modification of entrainment proceeds at a slower pace. In Fig.
While these timescales match previous work, the
An
One potential cause of an
If the initial value of the sensitivity of
This work has shown that many ship track properties vary significantly along the length of the track. As ship tracks do not transmit information along their length, they can be considered a collection of semi-independent segments, with the same initial aerosol perturbation, but at different times since emission. Using the ship location and the local wind field for reanalysis, the time since emission is inferred, allowing the timescales of the relevant cloud and aerosol processes to be measured. However, the results in this work come with some caveats.
The MODIS images are still only a snapshot of the cloud field, such that the time axis determined from the ship and cloud motion is not a real time axis. The unperturbed clouds will also develop over the time period
Although many ship tracks are intersected by CloudSat/CALIPSO, it is not enough to build up a picture of the precipitation development in these ship tracks. MODIS views every segment in each track; with 1209 tracks, almost 100 000 segments are used in this study. CloudSat will typically only view one segment per track (if any), so that resolving the development to the same detail requires 80 times as many ship tracks. Expanding this work to a global scale will not only allow the inclusion of other regions for ship track formation
This study uses
Cloud responses to aerosol perturbations are not instant but instead develop over characteristic timescales. This work uses ship position information and reanalysis wind fields to develop a time axis for satellite-observed ship tracks, providing a method for measuring the timescales of aerosol–cloud interactions from individual satellite images. The advected emission locations also provide an estimate for ship track locations in regions where they are too weak to be detected by existing methods (Fig.
While ships with higher
After an initial increase in track width (within the first 5 h), track width is relatively insensitive to the time since emission (Fig.
The reanalysis wind field is used to locate ship tracks in otherwise cloud-free environments (Fig.
The microphysical properties of the ship track, particularly the
While the
When combined, the radiative impact of the ship tracks investigated in this work is concentrated in the first 5–10 h after emission (Fig.
Although uncertainties remain, this study demonstrates how isolated aerosol perturbations can be used to measure the timescales of aerosol impacts on cloud properties, showing that the
Ship emissions were derived from location data from exactEarth. MODIS data used in this work were acquired from Level-1 and Atmosphere Archive and Distribution System (LAADS) Distributed Active Archive Center (DAAC), located in the Goddard Space Flight Center in Greenbelt, Maryland, USA (
The supplement related to this article is available online at:
EG designed the study and performed the analysis. TS and TG assisted with the interpretation of the results. All authors provided comments and suggestions on the manuscript.
The authors declare that they have no conflict of interest.
The authors would like to thank the reviewers for their helpful comments on the manuscript.
This research has been supported by a Royal Society University Research Fellowship (grant no. URF/R1/191602) and the European Research Council, H2020 grant no. 821205 (FORCeS).
This paper was edited by Zhanqing Li and reviewed by Kentaroh Suzuki and one anonymous referee.