The future of aerosol remote sensing will be propelled by advances in satellite payload technology. Instead of imaging multi-wavelength radiometers, like MODIS, we will expect the additional angular information pioneered by MISR plus the advantages of polarization. Imaging multi-wavelength multi-angle polarimeters were introduced to the community with POLarization and Directionality of the Earth's Reflectances (POLDER)/PARASOL (Deuzé et al., 2000, 2001; Chen et al., 2020). Future imaging polarimeters will improve upon the POLDER/PARASOL technology with better polarization accuracy, hyperspectral and hyper-angle capabilities, decreases in pixel size, and increases in the number of wavelengths with polarization and still allow for imaging. Super polarimeters with all of this technology in a single instrument are not necessary. Individual missions will tailor their instrument characteristics to meet particular mission needs, but these types of instrument advancement will be available for mission design. The Multi-viewing, Multi-channel, Multi-polarisation imaging mission (3MI) is the European follow-on to POLDER/PARASOL that will fly on EUMETSAT's second-generation polar-orbiting Meteorological Observing satellite series (MetOp SG) (Fougnie et al., 2018). The first of this series is planned for launch in 2024. The HyperAngle Rainbow Polarimeter 2 (HARP2) and the Spectro-polarimeter for Planetary Exploration One (SPEXone) will both fly on NASA's Plankton, Aerosol, Clouds, ocean Ecosystem (PACE) mission to be launched in 2024 (Werdell et al., 2019), and the Multi Angle Imager for Aerosols (MAIA: Diner et al., 2018) is scheduled for launch soon after. China has the Directional Polarized Camera (DPC) already in orbit aboard the Daqi-1 (DQ-1) satellite, with similar sensors repeated to launch until 2033. Finally, India is planning an aerosol mission, Nanosatellite for Earth Monitoring and Observation-Aerosol Monitoring (NEMO-AM), which will carry a multiangle polarimeter, the Multi Angle Dual Polarisation Imaging sensor (MADPI).

Similar instruments, with extended wavelength ranges, are planned for NASA's Atmospheric Observing System (AOS, 2023) launching later this decade. Beyond the AOS, we expect more hyperspectral capability, broader wavelength ranges including the thermal infrared, finer spatial resolution, broader swaths, and more angles. Twenty years from now these imaging multi-angle and hyperspectral polarimeters will be the norm for passive aerosol-characterizing sensors. Simple single-view multi-wavelength radiometers will still be flying through the 2030s, as they offer products across multiple disciplines, i.e., the Visible InfraRed Suite (VIIRS), Ocean and Land Colour Instrument (OLCI), MultiSpectral Instrument (MSI), and Medium Resolution Spectral Imager (MERSI), but the cutting edge aerosol characterization will have moved to the polarimeters by the end of the 2030s. The exception for radiometers will be the geostationary fleet of multispectral radiometers that we discuss below.

In the thermal infrared, single-view instruments will still have a role (DeSouza-Machado, 2010; Klüser et al., 2011). Hyperspectral thermal infrared spectrometers such as the Atmospheric Infrared Sounder (AIRS) have been used to detect and quantify dust aerosol optical depth (AOD) and layer height beginning with the Aqua mission (DeSouza-Machado et al., 2006) and advancing to other hyperspectral sounders in this spectral range, e.g., the Infrared Atmospheric Sounding Interferometer (IASI) (Clarisse et al., 2019). The community has recently directed greater focus to dust's contribution to Earth's energy balance through its perturbation of the radiation field at infrared wavelengths (Zheng et al., 2022; Song et al., 2022). New technology for hyperspectral thermal infrared spectrometers will improve signals and reduce spatial footprints and are our best hope for achieving mineral speciation of airborne dust from space.

Along with advances in passive remote sensing imagers, we expect new capabilities in active lidars. The straightforward backscatter lidar in space (CALIOP) has demonstrated the unique aerosol information that only a lidar can provide. By 2043 standard aerosol missions will include lidars, and these lidars will surpass CALIOP with additional wavelengths, high-spectral-resolution lidar (HSRL) and/or Raman capability, as well as multiple beams to expand coverage. Ground-based and airborne lidars with these enhanced capabilities are demonstrating that, like polarimetry, the information content from enhanced lidar design allows for aerosol characterization in ways previously unattainable (Müller et al., 2014). Post-CALIOP lidars flying now or expected in the near future include the Chinese Aerosol and Carbon Detection Lidar (ACDL; Ke et al., 2022) and the European ATmospheric backscatter LIDar (ATLID; Groß et al., 2015). Both have HSRL capability. These are small steps forward, and we understand the expense of launching high-capability lidars in space, as the HSRL planned for NASA's AOS mission has recently been descoped because of cost. However, we remain optimistic. By 2043, these modest steps into lidar capability expansion could grow to achieve the ideal for aerosol characterization (Müller et al., 2014). This means that there could be at least one space payload offering three wavelengths of backscattering, two wavelengths of extinction profiles using HSRL technology, and one wavelength providing depolarization. Aerosol scientists have been advocating for this array of lidar payload capability for more than a decade, and despite the next set of planned aerosol missions falling short, we foresee the dream becoming reality in 20 years.

The lessons we learned from the past 20 years in terms of the advantages of formation flying will guide mission planning far into the future. Like the A-Train of the 2000s and 2010s, missions will be constructed to purposely match passive polarimeters with active lidars in ways that promote synthesized data processing, algorithms, science, and applications. The workhorse Sun-synchronous polar-orbiting mission that has been the mainstay of aerosol characterization will be supplemented with inclined orbits that have been more typical of precipitation-centered missions but that have much to offer for aerosols. Inclined orbits allow for a multitude of geometries and time-of-day samplings of the aerosol system over the course of weeks as well as multiple coincidences a day, with other satellites looking simultaneously at the same target at the ground. This provides a look into aerosol situations never encountered by MODIS, MISR, or their ilk. Part of this expanding constellation of aerosol observations from satellites will be geostationary (GEO) imagers. Capability from GEO platforms has already evolved from the two or three broad uncalibrated wavelengths that were ubiquitous at the time of Terra launch into multi-wavelength moderate-resolution sensors that mimic most of MODIS' capabilities. By 2043, these modern-day Advanced Baseline Imagers (ABIs) and the Advanced Himawari Imager (AHI), FengYu (FY), and Meteosat Third Generation (MTG) will be upgraded with spectrometers as per the NASA–NOAA joint Geostationary Extended Observations (GEO-XO) program (Frost et al., 2020). At that time we expect studies in progress to adapt polarimeters to GEO orbits. Other GEO missions include the Geostationary Extended Observations Imager (GXI), Flexible Combined Imager (FCI), Geostationary High speed Imager (GHI), and Advanced Geostationary Radiation Imager (AGRI). Lidar at that altitude may still be pending. With the Equator crowded with observatories in GEO orbit, overlap between sensors provides both a means to cross-calibrate and also multi-angle looks at each scene to increase the information content for aerosol retrieval (Bian et al., 2021).

Table 1 presents a partial list of planned space missions from the national agencies of the US, Europe, and China that have identified an aerosol objective. The list is not comprehensive by any measure, as many missions not specifically designed to characterize aerosol find a means to do so anyway. For example, the Earth surface Mineral dust source Investigation (EMIT) mission will advance characterization of dust aerosol by identifying the mineral content of dust on the ground before the particles are airborne. EMIT will help with aerosol characterization but is not included here because it is not designed to retrieve airborne aerosol parameters. Table 1 is also missing a comprehensive list of missions because many additional countries are finding resources to launch space observatories as space becomes more democratized, as we will discuss below. India, Japan, and Korea are already planning for future aerosol-capable missions in the next decades.

Table 1Partial list of planned missions from major space agencies in the US, China, and Europe and one each from India and South Korea being aerosol-objective or aerosol-capable without a specific objective. Compiled from WMO OSCAR (2011–2023).

^∗ Already launched and in orbit.

Download Print Version | Download XLSX

The tag line for technology in 2043 will be more, more, more! Each sensor will be collecting orders of magnitude more data than MODIS does today. Polarized measurements require 3 times the data per wavelength than a simple radiometer to account for the three angles of polarization. Multiply that increase by the number of view angles beyond the single MODIS-like view, and we are likely now producing 15 to 60 times the number of data being produced today. The implementation of hyperspectral sensors or improvements in pixel size while maintaining broad swaths to view the entire globe further increases the data volume. Advancing lidars with additional wavelengths, HSRL, multiple beams, and quicker pulses also magnifies data volume quickly. A part of the technology evolution will be the need for spacecraft and especially spacecraft electronics, data systems, and transmission to keep up with data volumes. Optical transmission will replace radio frequencies, and infrastructure in space will serve as a way for stations to stage downlinking to overworked ground stations. For example, today the Ka band is the radio frequency of choice for satellite missions. NASA is promoting “new and improved” Ka-band technology available for licensing that achieves a data rate of 130 megabits per second (130 Mbps) (NASA Technology Solution Communications, 2023). In parallel, NASA recently announced the achievement of 200 gigabytes per second (Gbps) with a space-to-ground optical link (NASA Press Release, 2023), i.e., 3 orders of magnitude more data coming down with an optical system. On Earth, computing power will be fully cloud-based and ever expanding to hold the proliferation of collected data.

Today, the majority of aerosol retrieval algorithms interpret the signal measured by the satellite sensor in terms of radiative transfer through the atmosphere. The process is governed by this equation:

ρ_TOA is the solar radiance reflected by the Earth system measured at the top of the atmosphere by the satellite sensor, ρ_a is the contribution by the atmosphere alone, ρ_s is the reflectance of Earth's surface that transmits through the atmosphere with transmittance, T, and s is the spherical albedo with the (1−sρ_s) term representing multiple scattering between the atmosphere and Earth's surface. Each term is for specific wavelength, Sun, and sensor geometries, but that notation has been omitted for simplicity. The terms depend on surface properties (ρ_s and s) and the amount and optical properties of the gases and aerosols in the atmosphere (ρ_a and T). By making physically based assumptions about some surface, gas, and aerosol properties and calculating ρ_TOA, the calculated ρ_TOA can be compared with the measured ρ_TOA and assumptions adjusted until the calculated and measured ρ_TOA values are the same to within an acceptable error, consistent with measurement uncertainty.

The simplest physically based algorithms use one wavelength, one geometry at a time. These simple retrievals keep all assumptions constant except for the optical measure of aerosol loading, the AOD. Thus, the retrievals use one piece of information to retrieve one parameter, the AOD. By minimizing the difference between measured and calculated ρ_TOA simultaneously for more than one wavelength, more than one angle, or more than one polarization state, more information is introduced, and additional assumptions can become free parameters, leading to retrievals of aerosol parameters beyond AOD. For example, the MODIS dark target retrieval is based on the knowledge that independent, multispectral observations provide more information. However, not all wavelengths are completely independent, so that seven MODIS wavelengths from 0.47 to 2.1 µm yield, at most, three pieces of information (Tanré et al., 1996). Likewise, additional geometries and/or polarization states can be introduced, freeing even more parameters for retrieval. The simplest algorithms have employed radiative transfer models run in the forward direction with assumed aerosol, gas, and surface parameters, resulting in a look-up table (LUT) stored for future use. The retrieval can then move quickly through the LUT, interpolating between entries, and return the retrieved parameters in an operational setting.

Today, as information content increases and computer power grows to meet demand, LUTs are being supplemented by other methods to solve the radiative transfer equation. Techniques are being used to simultaneously retrieve multiple aerosol, gas, and surface parameters. These more flexible algorithms provide an alternative framework for retrieval, but they are still physically based on radiative transfer. As we march towards 2043, the LUT methods will be permanently jettisoned. The reason is that radiative transfer will become so sophisticated with better representation of aerosol optical properties that we would need an infinitely large LUT. Optimal estimation avoids LUT limitations but requires an initial guess and the risk of finding local minima. With parallel improvements (and greater sampling) of ground-based and airborne observations including in situ sampling, more properties can be constrained. The in situ measurements are especially important for characterizing irregularly shaped particles. Algorithms will be designed to ingest data from multiple collocated sensors, especially combinations of passive and active sensors (Liu et al., 2017) and combinations of spaceborne and ground-based sensor observations (Lopatin et al., 2013; Li et al., 2019). We expect joint retrievals of aerosol and trace gases, aerosol and land properties, aerosol and ocean color properties, and aerosol–gas–surface retrievals. Eventually there will be sufficient information by 2043 for retrieved properties to include vertical profiles of aerosol extinction and particle number concentration (Schlosser et al., 2022) as well as retrievals over clouds (Torres et al., 2012; Jethva et al., 2013) and over all land surface types, including snow and ice (Shi et al., 2019; Mei et al., 2020; Zhang et al., 2023). Characterization will include particle spectral absorption (Xu et al., 2021) and make progress towards constraining particle composition (Kacenelenbogen et al., 2022).

By 2043, although radiative transfer and scattering processes will be better understood, completely physically based algorithms will be rare. Development of machine-learning (ML) retrieval algorithms will accelerate, such that there will be algorithms that produce a full array of aerosol characteristics by pumping satellite-measured quantities through ML-derived models. These models have no physical basis and do not rely on the radiative transfer equation (Eq. 1). Instead, the models are created by training on a formulation dataset that matches satellite-measured radiances with measured aerosol characteristics. For example, today we can train a ML model using satellite measurements collocated with AERONET observations (Lary et al., 2009; Di Noia and Hasekampf, 2018; Kang et al., 2022). In this way we can obtain aerosol characteristics such as AOD that match AERONET with great accuracy and precision (Lary et al., 2009), but we lose all physical understanding of how that AOD is retrieved. The challenge for exclusive ML methods like these cited will be the identification of accurate measurements of aerosol characteristics, other than AOD, in sufficient quantities to train the ML models. However, ML algorithms will not recognize a physical situation that falls outside of its training dataset, and they still need to accommodate measurement uncertainty.

There are other uses of ML in aerosol remote sensing besides directly obtaining aerosol parameters from satellite observations. Within a physically based algorithm there are many assumptions that can be aided by ML methods. For example, running a radiative transfer model is computationally expensive. A ML model derived from radiative transfer runs will be more accurate than current interpolations of LUTs and save time in an operational setting (Gao et al., 2021; Ukkonen, 2022). In other work, ML methods are used to classify specific aerosol types, such as dust by training on other satellite datasets (CALIPSO) (Lee et al., 2021). ML is used to bias-correct aerosol products derived using standard methods (Lipponen et al., 2022), and it can help with interpolation, constraining non-retrievable aerosol optical properties, cloud (Wei et al., 2020) or snow masking, and surface reflectance determination (Su et al., 2020). In 2043 aerosol retrieval algorithms will continue to be physically based, but when examined carefully, they will contain many elements of indispensable ML within.

The biggest noticeable change over the next 20 years will be the gradual extinction of “satellite aerosol products”. The entire purpose of what we know now as a satellite data product will be to feed major assimilation systems that produce a complete representation of the global aerosol system within the context of the complex Earth system (Zhang et al., 2008; Benedetti et al., 2009; Gelaro et al., 2017). We already see the widespread acceptance of assimilation products, as noted by the number of citations earned by key papers describing each system. For example, using Clarivate's Web of Science database (Clarivate, referenced January 2023) and searching “satellite AND aerosol”, we find the first returned reference to be Holben et al. (1998), the seminal AERONET paper with 5265 citations accumulated over 24 years. Next up is Gelaro et al. (2017), the fundamental reference for the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) assimilation products, with 3270 citations accumulated over only 9 years. We understand that not all citations to MERRA-2 will involve applications to MERRA-2 aerosol products and that having a single fundamental reference concentrates users' citations to a single paper, but the rate of acceptance of MERRA-2 and other systems' assimilation products over the past 5 years is astounding. See Fig. 1.

Assimilation systems are analogous to the science fiction concept of a “cyborg”, defined by the Oxford Dictionary of Science Fiction as “A hybrid being: half human, half machine (a contraction of `cybernetic organism')” (Prucher, 2006). In our context, the assimilation systems are a hybrid of half observations, half models. Satellites measure electromagnetic radiation that is processed through aerosol retrieval algorithms to determine aerosol parameters. Assimilation systems intercept parameters from the retrieval algorithms during processing to ingest those data into a global model that adjusts model representation of aerosol parameters to agree with the observations. Some assimilation systems draw from the aerosol retrieval algorithms close to the satellite observations, making use of radiance or reflectance directly (Randles et al., 2017). Some systems ingest the final retrieval products such as AOD (Zhang et al., 2008). In all the cases the global model is drawn closer to actual measurements at regular time steps and is expected to represent the actual aerosol system better than a model that is initialized and run forward from aerosol source inputs. The advantage of assimilation aerosol products is that users can access aerosol parameters for any time period, at any location, and at any altitude. The parameters will have the benefit of ALL available data from ALL satellites simultaneously. Users will not need to patch together products from different algorithms. The model does that for you and presents the results in an Earth system governed by physical restraints.

The disadvantages of assimilation are that the resulting aerosol field can be variable, changing abruptly with model time step or with the availability of different aerosol input fields. For example, there will be much more data available during the day than at night, which may create day–night discontinuities. Furthermore, by drawing the model nearer to the observed variables, there can be arbitrary introduction or subtraction of aerosol mass. This can create biases in mass budget parameters, such as mass deposition. Furthermore, model parameterizations such as chemical transformation and interaction with clouds remain uncertain and likely will remain uncertain in 2043. Nevertheless, assimilation datasets will become the norm for the majority of aerosol data product consumers. The assimilation systems will run in the cloud and will be called on to produce the cyborg data at a user's will.

We note that satellite-retrieved aerosol products can interact with models without being assimilated. They can serve as validation and identify and characterize aerosol point sources such as smoke and volcanic aerosols. Using satellite-derived datasets as separate, independent data can improve parameterization of aerosol processes within models. Such processes include chemical transformation and aerosol–cloud interactions. Such independent constraints on models have important implications for climate prediction and air quality applications (Kahn et al., 2022).

Figure 1Number of citations received by Gelaro et al. (2017), the fundamental reference for the NASA MERRA-2 assimilation system products (green) and the combination of citations received by Levy et al. (2013), Hsu et al. (2013), and Lyapustin et al. (2018), representing the Dark Target, Deep Blue, and MAIAC aerosol algorithm products for MODIS (blue). Source: Clarviate Web of Science database. The point is to note the rapid acceleration of MERRA-2 citations as an indicator of the speed of the research community's acceptance of assimilation products, while traditional retrieval products such as MODIS aerosol products maintain a consistent presence in the research community. We anticipate that, over the next 20 years, assimilation product use will continue to grow while use of standard products will plateau and decrease.

Download

A gold rush to space has begun and will intensify over the next 20 years as space becomes crowded with a variety of sensors in various Earth orbits. During this period we will realize a true democracy in space as non-traditional players such as the space agencies from the developing world and from a wide range of commercial sector companies join traditional national agencies in blanketing the skies with satellites and sensors. While only a fraction of these satellite sensors will be Earth-viewing and aerosol-capable, the result will yield a tremendous volume of data useful for aerosol retrieval.

Satellites will come in all sizes. Traditional agencies will still launch large, expensive, and robust flagship missions with multiple platforms and multiple sensors on each platform. The flagship missions will continue demonstrating lower risk, higher accuracy, greater reliability, and longevity. The flagship missions will be supplemented by and integrated with constellations of smaller satellites and nano-sized cubesats. The smaller satellites will be orders of magnitude less expensive and thus riskier and with more limited capability than the larger observatories. However, the multitude of the smaller units renders the loss of any particular unit insignificant to the greater objectives. With so many satellites flying, observing Earth, and returning measurements for aerosol retrieval, each satellite will be seeking a particular niche. All orbit types will be covered, including polar-orbiting, inclined orbit, and geosynchronous, and near-real-time observations will be achieved.

The challenge of so many data sources is the maintenance of data quality. There will be a lot of data, and there will be a lot of bad data. Eventually the scientific and user communities will organize to set standards and protocols to regulate data quality. Such standards will rely on the cohesion of the data communities for enforcement. An example of community establishing standards that are enforced through peer review are the protocols published by the International Ocean Colour Coordinating Group (IOCCG) for satellite ocean color sensor validation. Not all data will be free and open access. Commercialization of data products will aid in enforcement of standards, as different entities will be competing to sell their products based on many characteristics, including quality. Evaluation of that quality may move into the hands of independent companies or coordinating groups. Commercial entities that launch satellites intend to produce specialized data products for specific customers. Sometimes these customers will be national agencies looking for input into their assimilation systems. At other times commercial entities will produce their own assimilation system cyborg products and sell those results to their customers.

Calibration will be a continuing concern with the proliferation of small and nano-satellite data flooding the market. As stated above, there will be a lot of bad data, much of them due to poor calibration of each individual sensor, as evidenced by today's fleet of mostly poorly calibrated small satellite sensors. This will change. Calibration takes many forms and includes radiometric, polarimetric, geometric, and spectral components, and each component requires its own intense scrutiny that not all satellite-launching entities are ready to provide. To mitigate some of the chaos, a commercial pre- and post-launch calibration industry will spring up. These companies will offer pre-launch calibration facilities and take responsibility for a sensor's calibration, even verifying that calibration using vicarious methods once the sensor is in orbit. Eventually, either governmental or space industry bodies will move to a certification process similar to the Underwriters Laboratories (UL) or Conformite Europeenne (CE) marks common today to denote adherence to a specified set of standards for electronic systems. There will also be a need for in-space calibration options, such as was proposed for the Climate Absolute Radiance and Refractivity Observatory (CLARREO) mission. CLARREO would maintain highly accurate, traceable reflectance measurements from Earth and the Moon through its lifetime, which could be transferred to other instruments in orbit.

Validation of aerosol products is another issue. With the democratization of space, the leading national space agencies will no longer hold a monopoly on space access. Ironically, instead, the government agencies that pioneered aerosol characterization from space will take the lead on suborbital measurements to support the validation side of producing high-quality aerosol products. Government-sponsored field experiments and ground networks will provide the basis for the entire international community, including the commercial sector, to evaluate their products and improve their algorithms.

Today's basis for validation of aerosol datasets is the ground-based AERONET network (Holben et al., 1998; Giles et al., 2019). The AERONET concept will continue for decades into the future, but the technology itself will evolve. Instruments will move from finite wavelength bands to hyperspectral spectrometers and will continue AERONET's current trend into the shortwave infrared. Originally, AERONET instruments had invested in polarimetry, which turned out to be a capability ahead of its time. The community has caught up, and AERONET will expand the polarimetric capabilities of their networks and incorporate polarimetry into their retrievals. Data capacity, storage, and transmission will expand, as will the network itself. AERONET's subsidiary, the Marine Aerosol Network (MAN), will continue to collect data on ocean-going cruises and ships of opportunity. The MAN instruments will also upgrade their technology and data capabilities. The AERONET retrieval that produces reliable aerosol characterization in addition to AOD, including particle size distribution, single scattering albedo, complex refractive indices, and non-sphericity, will become a more important asset over time as aerosol satellite remote sensing advances in information content, requiring evaluation of satellite-retrieved products that include a wide range of aerosol parameters. However, we note that, while the uncertainty in AERONET AOD is sufficiently small to offer true validation to a collocated satellite AOD product, the uncertainty of AERONET inversion products (particle size distribution, single scattering albedo, etc.) is not. The AERONET inversion products are themselves subject to assumptions and caveats and may have greater error than certain satellite parameters. Still, the inversion products provide a vital service in the evaluation of satellite products, providing a standardized framework of high-quality, widely distributed aerosol characterization for nearly immediate comparisons with satellite products.

The AERONET concept of a widespread, high-quality, and aerosol-observing ground network will proliferate, as we are already witnessing with the Interagency Monitoring of Protected Visual Environments (IMPROVE: Malm and Hand, 2007), the European Aerosol Research Lidar Network (EARLINET: Pappalardo et al., 2014), the Aerosol, Cloud and Trace gases Research Initiative (ACTRIS, 2023), the Surface Particulate Matter Network (SPARTAN: Snider et al., 2015), and other networks. The different networks offer different types of aerosol characterization necessary to keep up with the expansion of retrieved aerosol parameters from space. In situ measurements of aerosol aerodynamic properties will continue to be essential for the assessment of the aerosol health impact and to link aerodynamic and optical measurements, which form the basis for satellite products. Ground lidar networks (Welton et al., 2001; Pappalardo et al., 2014) assisted by improved airport ceilometers will expand high temporal characterization of the aerosol vertical profile. As satellite retrieval algorithms upgrade their radiative transfer, refine their assumptions, incorporate ML, and become cyborgs, so will the ground network retrieval algorithms. More and more often validation datasets will include the resulting aerosol characterization obtained by simultaneous retrieval of several collocated instruments. Ideally this will include passive measurements with collocated lidar and ground-level or balloon-borne in situ measurements.

Field experiments featuring aircraft-borne measurements will be indispensable for validating the satellite observations at the radiance or polarization level (Level 1) and the retrieved aerosol particle properties in regions and situations of high scientific interest or societal relevance. Aircraft fly in three dimensions, sampling situations between network stations and providing vertical profiles of parameters where lidar is absent. Field experiments will also be essential for confirming ground-based validation sites, as now these sites will be producing a suite of retrieved parameters that require their own validation. We will find that field experiment measurements will be the only truly independent dataset for validation. With the abundance of suborbital data, quality again becomes an issue, and national agencies will be the only community players with the resources to maintain high-quality, long-term validation datasets.

At the opposite end of the spectrum will be the proliferation of low-cost citizen science networks, including the use of smartphone imagery. These commercial networks will expand across populated areas and will contribute to characterizing the global aerosol system simply by the abundance of available observations. However, there will not be the same oversight to maintain the quality of these datasets, and users will need to calibrate the data from citizen scientist measurements with government-maintained and quality-assured sources.

We have described a vibrant global measurement system focused on aerosol remote sensing as we move forward into the next 20 years. This picture assumes a continuing need for aerosol characterization, which will be driven by pressing science questions and societal priorities. What will be the aerosol science questions of the coming decades?

Quantifying changes to the aerosol system will be essential as we continue to struggle to catch up to climate change. By 2043 we will have 70 years of Total Ozone Measuring System (TOMS), 60 years of AVHRR and 40 years of MODIS and VIIRS aerosol time series. While most of these time series offer insight into global and regional changes to aerosol loading, inferring changes to aerosol types and particle properties will be especially important.

We will still be working towards understanding aerosol–cloud–precipitation processes, how these have changed in time, and how these processes interact with atmospheric dynamics.

The role of aerosols in transporting nutrients around the globe and depositing them into terrestrial and oceanic ecosystems will continue to be a pressing issue in the coming decades. Likewise, the role of aerosols in transporting pathogens will still be under investigation in the coming years.

Maintaining healthy air quality for all the world's populations will continue to call for understanding of global aerosol distributions and characterization in a way that will require satellite remote sensing.

Underlying these issues will be the goal of using the characterization of the global aerosol system to make our planet healthier and environmentally just.

No computer code was developed for this Opinion. The one figure was plotted using commercial software package Kaleidagraph 5.01 from Synergy Software.

Citation data for the one figure (Fig. 1) were accessed from Clarivate Web of Science. Clarivate is a commercial entity and is not open source. The Web of Science accessible through the subscriptions of many research and academic institutions provides the quickest and easiest method to attribute the number of citations of particular articles. In this Opinion, the quick access and analysis of the citation statistics made a general point not essential to the main conclusions of the Opinion. Any reader interested in comparing similar citation results could satisfy their curiosity by using any citation database accessible to them. The authors are confident that the same general point can be made about the rise in the citations of assimilation datasets versus the relatively stable number of citations of traditional satellite aerosol products over the same time period.

The manuscript resulted from discussions between the three authors over a time span of years. LR wrote the initial draft. RL and JVM edited this draft and added their own insights. LR took responsibility for addressing reviewers' comments, with RL and JVM editing her responses and subsequent versions of the Opinion.

LR and JVM hold ownership stakes in GRASP SAS in addition to their primary employment at the University of Maryland Baltimore County. GRASP SAS is involved in the commercial small satellite industry and in producing aerosol remote sensing products from its own satellites and from publicly available satellite measurements. GRASP SAS's subsidiary, Airphoton Inc., designs, builds, and sells ground-based instruments for aerosol measurements and payloads for space.

Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.

This article is part of the special issue “20 years of Atmospheric Chemistry and Physics”. It is not associated with a conference.

This paper was edited by Xiaohong Liu and James Allan and reviewed by Ralph Kahn and Zhanqing Li.