Lord Kelvin would certainly agree that it is vital to measure the state of the atmosphere and its surface boundary in order to understand weather and climate. Physical quantities such as temperature, water vapour, wind speed and direction, pressure, precipitation, cloudiness, radiation, aerosols, atmospheric composition, land and ocean surface properties and many others need to be observed to achieve this understanding.
In this review paper, we summarize what is measured and how these measurements are made, thus describing current observational capabilities to define climate, its variability, and its weather extremes.
While the early motivation for meteorological observations was to describe and understand the atmosphere, weather prediction and climate projections are now driving forces for sustaining and increasing our observational capabilities. Prediction of local weather such as severe thunderstorms requires both observations and model grids at sub-kilometer resolution, while climate models benefit from detailed knowledge of atmospheric radiation, aerosols and surface properties. An additional benefit of atmospheric and related observations could be described as situational awareness.
Real-time knowledge of potential adverse effects of severe storms, lightning, flooding, fog, smoke, storm surge, etc. are beneficial to forecasters, broadcasters, emergency managers and first responders, utilities, transportation sector, outdoor venues and the general public. Knowledge of weather conditions is important for efficient production of renewable energy, while the changing climate affects energy security. Public health is sensitive to current temperature, humidity, chemistry and aerosol conditions, as well as to a changing climate. Food production and water resource management are also strongly affected by weather and climate variability.
Organizing a review paper on observations is challenging, as it could be done by variable, instrument, purpose, atmospheric location, scientific challenge or operational application. We have chosen a hybrid approach. First, an overview of the basic physical principles governing the primary remote sensing systems is presented. There are separate sections focusing on global observations for climate and synoptic-scale phenomena and their spatiotemporal variability, and on regional observations for mesoscale and convective-scale weather systems. Within these sections, we sequence though the quantities required and the observing systems that provide those measurements. The last major section covers observation-related issues such as coordination, data assimilation, curation and a look forward. We note that since the nature of observing systems is constantly changing and the specific systems operated by different countries and the availability of data vary, this paper will be as generic as possible. However, U. S. systems will often be used as examples. An extensive reference list is provided for those who wish to learn more about each observing system. The reader is also encouraged to search online for information on how to access data from the different observing systems in each country.
In Section 2 the fundamentals of remote sensors are discussed. The reader is referred to Fig. 1 for the frequency ranges of electromagnetic radiation used in many of the sensors; the shortest, visible wavelengths are employed by optical sensors and lidars, while the longest, microwave wavelengths are used by radiometers and radars. The field of satellite meteorology recently marked the 60th anniversary of the April 1, 1960 launch of the first satellite dedicated to satellite meteorology, the Television and Infrared Observational Satellite (TIROS-1). This event was celebrated in conjunction with the American Geophysical Union (AGU) and American Meteorological Society (AMS) Centennial meetings. Satellite meteorology traces its roots back to the launch of Sputnik on October 4, 1957, soon followed by Explorer 1 launched on January 31, 1958. The history of satellite meteorology with key milestones and instrument advancements is well documented in Smith et al. (1986), Kidder and Vonder Haar (1995), Lewis et 2016) is particularly noteworthy as it chronicles the life and times of Professor Verner Suomi, the acknowledged father of satellite meteorology and the co-inventor of the first cloud camera in geosynchronous earth orbit. The camera was carried on the Advanced Technology Satellite (ATS-1) launched in1966 as a research pathfinder. The routine sequence of cloud images possible from high above Earth provided the first depictions of synoptic scale motions from space. GOES-1, the first NOAA Geostationary Operational Environmental Satellite launched on October 16, 1975, initiated a series of operational satellites in geostationary orbit that followed to this day. Of additional historical note are the anthology sessions recorded at the 2019 Joint AMS-EUMETSAT-NOAA Satellite Conference (Session 10) held October 2, 2019 "Celebrating the 60th Anniversary of the First Weather Satellite, its Evolution, and International Partnership" (https://ams.confex.com/ams/JOINTSATMET/m eetingapp.cgi/Session/52323) and the 16th Annual Symposium on New Generation Operational Environmental Satellite Systems (Session 3) held with the AMS Annual Meeting on January 13, 2020 (https://ams.confex.com/ams/2020Annual/meetingapp.cgi/Session/ 53293). These presentations cover the evolution and improvement of the satellite capabilities, instruments, and measurements since the inception of satellite meteorology.
Weather satellites are the omnipresent backbone of the global observing system (Fig. 2).
The principal satellite orbits are Low Earth Orbit (LEO) and Geostationary Earth Orbit (GEO), which together provide different perspectives of the atmosphere and earth below (World Meteorological Organization Space Programme, 2020). Schumann 2020), whereas the GEO satellites are the primary source of near real-time imagery used for nowcasting and the detection of rapidly evolving high impact environmental phenomena (Goodman et al., 2018(Goodman et al., , 2019;;Holmlund et al., 2021;Schmit et al., 2017Schmit et al., , 2018)). The major U.S. operational contributions to the baseline satellite observing system depicted in Fig. 2 are the Geostationary Operational Environmental Satellite (GOES) -R Series in GEO (Fig. 3) and Joint Polar Satellite System (JPSS) in LEO (Fig. 4). A "constellation" of satellites can also fly in formation to produce synchronized data from a collection of different instruments (e.g. the "A/C-train" constellation in Fig. 2 comprised of OCO-2, G-COMW, Aqua, Calipso, Cloudsat; https://atrain.nasa.gov, Stephens et al., 2018).
Fig. 3 The Geostationary Operational Environmental Satellite (GOES) R-Series satellite and instruments. Graphic courtesy of Lockheed Martin and the GOES-R Program.
Fig. 4 The Joint Polar Satellite System (JPSS) satellite and instruments. Graphic courtesy of Joseph Smith and the JPSS Program.
The GEO and LEO earth viewing instruments are described below, with further information on the missions, satellites, instruments, data products, and additional resources available at http://goes-r.gov and https://jpss.noaa.gov.
The sun-synchronous LEO orbit has 14 polar-orbiting passes per day with equatorial crossing times at the same time in the morning and afternoon. These orbits provide global imagery, atmospheric data (temperature, moisture, trace gases), and surface (sea and lake ice, ocean salinity and heat content, harmful algae blooms, vegetation health, flood inundation) observations from an altitude of ∼800 km. These measurements are also assimilated into the regional and global NWP models (e.g. European Meteorological
Operational satellite (METOP), NOAA's JPSS, CMA/NSMC (China Meteorological Administration/National Satellite Meteorological Center) Fēngyún-3 (FY-3) in Fig. 2). LEO orbits can also be precessing where the ground track varies day to day allowing observations throughout the diurnal cycle rather than the twice per day sunsynchronous coverage in polar orbits. The non-sun-synchronous, inclined orbits in LEO have more frequent lower latitude coverage typically from an orbital altitude of about 400 km, but with a longer revisit time to view the same exact spot on the Earth.
GEO satellites are located at 35,786 km (22,236 miles) above the earth's equator with the orbit matching the Earth's rotation. This allows the satellite to view the earth, atmosphere, and high impact weather at the same satellite sub-point continuously day and night. Together the LEO and GEO orbits and their instruments provide a broad spectrum of atmospheric, land, ocean, and ice measurements used in weather forecasting and analysis ( and ). The new generation international "GEO-Ring" satellite constellation provides full disk earth and atmosphere imagery and derived products (e.g. cloud mask, cloud height, cloud phase, precipitable water, stability indices, winds) over the full earth disk every ten minutes and at a high refresh cadence of 1-2.5 min over limited areas.
Current and planned constellations or CubeSat swarms of visible/infrared imagers and microwave radiometers may greatly augment the capability of the global observing system and increase the revisit time from twice per day to perhaps hourly or better (http s://www.nasa.gov/mission_pages/cubesats/), making these data of potentially great interest and value for nowcasting and regional to global-scale NWP. A CubeSat is a designation used to classify a tiny (nanosatellite) satellite made up of 10 × 10 × 11.35 cm units, designed to provide 10 × 10 × 10 cm or 1 L of useful volume while weighing no more than 1.33 kg (2.9 lb) per unit. The smallest standard size is 1U, although in recent years larger CubeSat platforms have been developed, most commonly 6U (10 × 20 × 30 cm) and 12U (20 × 20 × 30 cm) to extend the capabilities of CubeSats beyond academic and technology validation applications and into more complex science missions. Most CubeSats carry one or two scientific instruments as their primary mission payload.
NOAA, NASA, U.S. Space Force, Los Alamos National Lab (LANL), and other agencies as well as commercial companies are exploring smallsats and Cubesats for making and GeoOptics (CICERO series) provide real-time GPS radio occultation (RO) temperature and moisture data to NWP centres worldwide (see Section 3.a.1). Intercalibration of these measurements will be a challenge with each instrument providing a different view geometry and atmospheric path.
The Radiative Transfer Equation (RTE) is central to retrieving atmospheric temperature and moisture at infrared and microwave wavelengths, and models the propagation of terrestrial emitted energy through the atmosphere by absorption, scattering, emission and reflection of gases, clouds, suspended particles, and the surface (Deepak, 1977;Janssen, 1993). The observed radiances can be converted to brightness temperature and inverted to obtain atmospheric structure and properties such as profiles of temperature and water vapour, clouds (height, fraction, optical thickness, size), aerosol, dust, surface temperature, and surface types, including bare soil, desert, concrete, etc., due to different surface emissivity. The weighting function for radiation at a given wavelength represents contributions from various atmospheric layers to the radiance reaching the top of the atmosphere with each spectral band affected by an absorbing molecule (gas) such as water vapour, SO , O , and CO , having its own weighting function.
The spectral reflective and emissive properties (spectral signatures) of different targets makes it possible to derive a host of useful quantities. Planck's radiation law is central to understanding how the satellite senses energy and how that data is translated into different products, and is expressed in Eq. ( 1),
where λ is the wavelength (cm), h is the Planck constant (6.62607004 × 10 m kg s ), and c the speed of light (3 × 10 m s ). Brightness temperature is uniquely related to radiance for a given wavelength by the Planck function B(λ,T), with k the Boltzman constant (1.380649 × 10 J K ) that relates the average relative kinetic energy of particles in a gas with the thermodynamic temperature of the gas. The spectral radiance B describes the spectral emissive power per unit area, per unit solid angle for particular wavelengths or frequencies of radiation. Equation 30.1 shows that for increasing temperature, the total radiated energy increases and the peak of the emitted spectrum shifts to shorter wavelengths. Thus, given an observed radiance B, the Brightness Temperature T is the temperature, in Kelvin, of a blackbody that emits the observed radiance.
The upwelling flux, or energy per unit time, is sensed by a detector on the satellite. The energy detected from this area is called irradiance (E). Since it is referring to an area, irradiance is expressed in Watts per metre squared or Joules per second per metre squared. The satellite telescope collects the energy over a certain solid angle, which is expressed in steradians. Measuring the monochromatic irradiance over this solid angle constitutes the radiance. Hence the units for radiance are Watts per metre squared per wavelength interval per steradian.
The visible channels sense reflected solar radiation that has a wavelength of 0.4-0.75 micrometers. Since this is the wavelength interval over which the human eye is sensitive, the channel is called visible, frequently abbreviated as VIS. Since visible imagery is produced by reflected sunlight (radiation), it is only available during daylight hours.
Cloud top texture and features in the visible (overshooting cloud tops, above-anvil cirrus plumes) can indicate the potential for high impact and severe weather (Bedka et al., 2018). The earth's atmosphere, clouds and surface all absorb and reflect incoming solar radiation, with the earth's surface absorbing about half. The surface, clouds and atmosphere then re-emit part of this absorbed solar energy as heat or infrared (IR) energy.
The satellite senses IR energy at wavelengths of ∼4-14 micrometers. As some of the reemitted heat energy passes through the atmosphere, clouds and atmospheric gases absorb a portion of the energy. The energy can then be re-emitted in all directions at the same wavelength range. Thus, infrared channels sense radiation emitted by the earth's surface, earth's atmosphere, and cloud tops. A major advantage of the IR channel is that it can sense energy at night, making 24-hour imagery possible.
The troposphere has collision-broadened molecular absorption features and nonresonant absorption by liquid water, making passive microwave remote sensing in the region from 3-300 GHz an important complement to visible and infrared radiation. A weakly absorbing water vapour line at 22 GHz, a strong oxygen band centred at 60 GHz, and a 183 GHz water vapour line are important features for determining atmospheric temperature and humidity structure (Janssen, 1993). Passive microwave sounders and imagers (Section 4) also provide information on precipitation rate, total precipitable water, land surface emissivity and snow cover from window channels at, for example, between 23 and 150 GHz. Over radiometrically cold ocean regions, changes in brightness temperature due to the absorption/emission by liquid hydrometeors at frequencies below the 50-60 GHz oxygen absorption band are directly related to rainfall.
Over radiometrically warm land surfaces, scattering by precipitation-sized ice particles at frequencies above the oxygen absorption band are used to indirectly estimate rainfall.
From microwave measurements at window channels, the retrieval of surface and atmospheric parameters is dominated by the effects of surface emissivity, which is a function of the surface type, view angle, and surface roughness. Over oceans where the emissivity is low and uniform, atmospheric emission and scattering are dominant and the retrieval of atmospheric constituents can be achieved with high accuracy. Over land, where the emissivity is generally high and in excess of 0.9, the absorption signal of the non-precipitating atmosphere is weak. The retrieval problem is compounded by the more highly variable emissivity due to land surface type, soil moisture and vegetation cover. Thus, the retrieval of atmospheric parameters over land is generally limited and restricted to empirical methods (Ferraro et al., 2005).
6 Atmospheric temperature and moisture sounding Weather radar has its roots in World War II, where it was used for military applications (Fletcher et al., 1990). The APQ-13 radar used during the war was adopted as a military weather radar after the war and followed by the CPS-9 and the civilian, WSR-57 in the summarizes the types of radars currently in use and some of their characteristics, along with their typical spatial resolutions and weather phenomena whose features they can resolve. See also Fig. 1 for a graphical description of the frequency bands used.
2 Basic operating principles i Radar reflectivity and precipitation estimation Radars are "active" devices in that they transmit (electromagnetic) radiation. Passive instruments, on the other hand, only receive (and detect) radiation. Most operational surveillance radars transmit pulses of radiation of very short time duration, so that the range to the target can be determined from the time difference between the transmitted signal and the backscattered signal.
The radiation transmitted by radars is absorbed in the atmosphere by hydrometeors and by airborne targets such as insects and birds. This radiation is then re-radiated and scattered in all directions. A small portion of this radiation is backscattered to the radar.
Most operational and research radars send out pulses periodically. The rate at which pulses are transmitted is known as the pulse repetition frequency (PRF) in pulses per second. The PRF must be low enough so that backscattered radiation from the targets reaches the receiver before the next in the series of pulses is transmitted. If the next in a series of pulses of signal is received back at the radar before the signal from actual target, there is "range folding." The maximum unambiguous range (in m) is
Range folding may be mitigated by sending out series of pulses at more than one PRF and comparing signals.
The power of the signal received back by the radar is given by the "radar equation"
(omitted here for brevity) for distributed targets that fill the radar volume (e.g. Doviak & Zrnić, 2006). The strength of the signal increases with increased pulse length, but the spatial resolution of the radar volume decreases with increased pulse length. The intensity of the backscattered signal received decreases as the square of the distance from the target (the range) and varies as the sizes, concentration, and electrical characteristics of the targets. For example, ice crystals have a lower reflectivity than liquid-water droplets and wet hail is more reflective than dry hail. The intensity of the backscattered signal after the beam of radiation has travelled through extensive precipitation and back may be attenuated, as some energy is lost through absorption, so that not all the incident power is re-radiated. Generally, attenuation increases with decreasing wavelength, so S-band radars suffer from the least attenuation, while C-band radars suffer from more attenuation, and X-band radars may suffer from so much attenuation that the radar signal goes to extinction before it even detects the outer fringes of the precipitation associated with a weather system. S-band radars have therefore been preferred for surveillance, but because their antennas must be large, they are chosen when the site is fixed. Airborne and ground-based mobile radars, most commonly operating at X band or C band, require smaller antennas, so that attenuation is usually a significant problem.
Because attenuation varies with wavelength, multi-wavelength radars have been used to correct for attenuation, since the characteristics for each wavelength are known and the difference in the signal yields valuable information. Correction techniques have also been developed for polarimetric radars (which are discussed subsequently). Correction for attenuation is a complicated process involving many assumptions (see Snyder et al., 2010 for a review of some commonly used techniques); consequently, there are a number of techniques, but none will be detailed here.
When the diameter of the targets distributed within the radar volume is small compared to the wavelength of the radar, the scattering is of the Rayleigh type, for which the intensity of the returned radiation varies as the sum of the sixth power of the diameter of all the scatterers. For an S-band radar, raindrops (most typically ∼ 0.5 mm -3 mm in diameter) are mainly within the Rayleigh range. Large hailstones (> ∼ 5 cm in diameter)
are not, and fall within the Mie range, for which there is not a simple relationship between reflectivity and the size of the scatterers. For X-band radars, however, even some large raindrops may be within the Mie range. It is therefore more difficult to estimate precipitation rate using C-band or X-band radars than using S-band radars, because the relationship between reflectivity and hydrometeor size does not necessarily vary monotonically with hydrometeor size.
One must also know the distribution of hydrometeor sizes within the radar volume, which can vary significantly depending on the type of precipitation (e. g. continental rain, tropical rain, drizzle, convective rain, stratiform rain, wet snow, dry snow, etc.). In addition, the radar volume must be as close to the surface as possible. In mountain areas, radar estimation of precipitation may be hampered by beam blockage. In this case, special radars must be located in valleys. Radar radiation does not trace a straight path, but rather is bent, owing to refraction associated with gradients in temperature and moisture. Radars are therefore most useful for probing low altitudes relatively near the radar, but at far ranges may reach a minimum in height (radar horizon) as much as halfway to the tropopause. So, when the radar is a long distance from the precipitation, the precipitation near the ground is also not detected (Fig. 5). For Doppler radars, which can detect tornadoes and strong straight-line winds (see Section 4.b.d.ii), tornadoes and strong winds in general also cannot be detected at the surface at long range. The strength of the backscattered radar signal divided by the strength of the transmitted signal is typically given in dB, which is 10 times the logarithm of the power ratio. Since radars must detect a wide dynamic range of signal intensities, it is useful to account logarithmically for the large variation in signal intensity due to the wide range of precipitation reflectivity. Radar reflectivity is commonly given in dBm as the received power ratio relative to one milliwatt. Accounting for distance to the radar, radar reflectivity factor (Z) is given by the sum of all the sixth power of all the scatterers within a cubic metre of volume. Ten times the logarithm of Z divided by the Z of a raindrop of 1 mm in diameter, in a volume of 1 m , dBZ, is the most useful and widely used measurement of the intensity of the backscattered signal because it is independent of range and provides a useful standard for comparison among different radars.
There have been many empirical studies relating radar reflectivity to precipitation rate (e.g. Anagoustou & Krajewski, 1999;Austin, 1987) and doing so is still a topic of extensive research. Estimating precipitation rates is critical for warnings and studies of flooding hazards, and for obtaining climate information on precipitation amounts over populated land areas on spatial scales smaller than that of surface observing networks and especially over ocean areas and sparsely populated land areas where there are few if any surface measurements of precipitation. However, one must be cautious in using these radar datasets owing to the many uncertainties inherent in relating radar reflectivity to precipitation rate. When possible, radar estimates of precipitation are calibrated against the available actual measurements from rain gauges at the ground.
Range folding must be eliminated, attenuation must be corrected for, and an accurate relationship between radar reflectivity and precipitation rate must be known. Finally, the radar itself must be well calibrated and the radar beam must pass as close to the ground as possible.
ii Doppler radars Doppler radars compute the line-of-sight component of the wind in a radar volume from the frequency shift of the backscattered signal. There are a number of techniques for doing this including direct calculation of Doppler spectra from a time series of measurements and via "pulse-pair processing." The former is relatively slow, but delivers the most information (shows the range of velocities within a volume), while the latter is faster and more efficient, but reveals less overall information. The maximum unambiguous Doppler velocity V (m s ) increases with PRF as follows:
where λ is the wavelength of the radar in metres. If the PRF is too small, then there may aliasing, also known as "velocity folding," which can be corrected objectively using algorithms that compare neighbouring measurements and by assuming that the Doppler velocity varies smoothly in space on the spatial scale the radar can make measurements. Otherwise, multiple PRFs may be needed to resolve the ambiguity.
When there is a tornado, for example, the Doppler wind speeds may vary so rapidly that the algorithms may not work well, so that either subjective de-aliasing is necessary or the use of multiple PRFs is employed.
Since both the maximum unambiguous range and Doppler velocities are functions of the PRF, they are both related to each other independently of the PRF as:
Therefore, there is an inverse relationship, known as the "Doppler dilemma," between the maximum unambiguous Doppler velocity and the maximum range for a given wavelength, independent of the PRF. Radars must be operated so that range folding is mitigated (radar echoes intense enough to be detected are not present beyond the maximum range) and that Doppler velocities that are aliased can be easily de-aliased.
The longer the wavelength, the greater flexibility there is with respect to R and V , but the larger the antenna must be. Very short-wavelength radars have rather inflexible choices for R and V , but they are typically limited in range, so that R can be sacrificed at the expense of V .
There are some research radars, however, that do not send out pulses, but rather send out radiation continuously. These radars have greater sensitivity than pulsed radars, because according to the radar equation the power of backscattered signal increases with pulse length and in this case the pulses are in effect infinitely long, but range information cannot be obtained. Radars that send out radiation continuously while the frequency changes monotonically with time periodically (frequency modulatedcontinuous wave: FM-CW, or "chirped") are known as pulse-compression radars, and they can provide range information. They are particularly amenable to space-based radar measurements because they offer high sensitivity at low power output, especially at relatively high frequencies and do not require large antennas. There are, however, some problems unique to pulse-compression radars, which require special techniques to mitigate (Kurdzo et al., 2014).
S-band and other UHF (ultra-high frequency) radars, and VHF (very-high frequency)
radars, can detect signals from spatial gradients in the index of refraction in "clear"
(having no scatterers such as hydrometeors or insects) air through a process known as Bragg scattering, which depends on the scale at which the index of refraction, due to temperature and moisture content, varies in space due to turbulence. Shorterwavelength radars such as those at C-band, X-band, etc, are not as efficient at producing useful Bragg scattering. "Doppler wind profilers" have therefore been used to measure clear-air wind profiles as a function of height (Balsley & Gage, 1982;Gage & Balsley, 1978), like rawinsondes, but with much higher temporal resolution (six times to once per hour, depending on the time averaging employed, as opposed to once every 12 h in the operational synoptic network). There are usually several radar beams, each set off a small angle from the vertical, so that the horizontal wind may be estimated from the geometric relationship between the angle of the radar beam and the range to the radar volume. For a number of years, a network of wind profilers (Weber et al., 1990) maintained by NOAA was operational over the central U.S., but it has since been abandoned for financial reasons. For UHF profilers, when precipitation is present, Rayleigh scattering is assumed instead of Bragg scattering. There is usually a dead zone near the radar for VHF radars, so that wind measurements below ∼ 500 m are not possible (Ecklund et al., 1988). Thus, VHF wind profilers cannot usually detect winds in the boundary layer with the desired spatial resolution. The highest the radar can make clear-air, Doppler-wind velocity, measurements depends on the temperature and moisture stratification and is greatest for VHF radars operating at relatively low frequencies.
Surveillance precipitation radars can also estimate the vertical profile of winds using the VAD (velocity azimuth display) technique. This technique works when the winds do not vary substantially over the domain of the radar; in this case, the Doppler velocity should vary systematically as a function of azimuth: When the antenna is pointed in the direction of the wind, the Doppler velocity is at a maximum in approaching velocities, while the Doppler velocity is at a minimum in receding velocities when the antenna is pointed in the opposite direction. The height at which a wind measurement is valid is a function of the range to the radar volume and the elevation angle of the radar beam.
VADs, like Doppler wind profilers, are a useful complement to rawinsonde data because they may be collected much more frequently than twice per day. Furthermore, they do double duty because they are also used to detect precipitation. Because they must share the time doing VADs with surveillance scans, they are not available continuously. In addition, there must be sufficient clear-air scatterers, such as insects in the boundary layer, or index-of-refraction gradients so that there can be Bragg scattering when there is clear air.
If the airflow is turbulent, then the range of Doppler velocities within a radar volume can be relatively large. The "spectrum width" is therefore a Doppler-radar variable that may be indicative of turbulence. Also, the vertical shear of the Doppler velocity may also be used as an indicator of turbulence, since high shear contributes to low Richardson numbers, a necessary condition for turbulence.
In order to measure the full, three-dimensional wind field, rather than just the along-thebeam wind component, the Doppler velocity, more than one radar observing volumes from different viewing angles at different locations is required (e.g. Armijo, 1969).
Bistatic techniques have also been developed which allow one active radar to be paired with one or more fixed-site receivers, each at different locations (e.g. Wurman, 1994).
There are also techniques for retrieving the three-dimensional wind (SDVR -Single Doppler Velocity Retrieval) by fitting single-Doppler data to dynamic and kinematic constraints (e.g. Liou et al., 2018).
iii Polarimetric radars Most surveillance Doppler radars such as the operational, National Weather Service, Sband radars in the U. S. (Doviak et al., 2000) now also have separate vertically and horizontally polarized channels. The relative amount of horizontally to vertically polarized backscattered signal depends on the shape of the scatterers. This effect is measured quantitatively by the differential reflectivity (Z ) given as:
where Z and Z are the radar reflectivity from the horizontal and vertical channels, respectively. For example, large raindrops that fall are flattened into ellipses, such that the horizontal axis is longer than the vertical axis. The amount of horizontally polarized radiation backscattered is therefore greater than the amount of vertically polarized radiation, so that Z is relatively large. For small raindrops, which are not flattened as much, Z is negligible.
Information about the shape of scatterers is also given by the co-polar cross correlation coefficient (ρ ), where The difference in phase between the transmitted and received signals ϕ provides information about the path length of liquid water traversed, since the speed of the radiation is slowed down by the accumulated liquid or solid water substance in its path.
More information about the nature of the hydrometeors is given by the rate of change of ϕ with respect to the range from the radar, K , which is known as the specific differential phase. It is useful because changes in phase are not dependent on the calibration of the radar. However, in practice, K can be rather noisy and extreme care must be taken when differentiating ϕ with range.
While Z , ρ , and K are the most commonly used polarimetric variables, there are also others such as the LDR, linear depolarization ratio, and the reader is referred to
Zhang (2016) for a discussion of the definition and uses of others. Polarimetric variables have been used to identify hydrometeor/scatterer type using fuzzy logic and also to correct for attenuation. They can be used to estimate drop-size distributions, improve on precipitation estimation based on using radar reflectivity of only one polarization (e.g.
A radiosonde balloon that is tracked to provide upper-air wind speed and direction is called a rawinsonde. At first, optical theodolites were used for tracking, which limited wind information to cloud base heights, but these were soon replaced by radio theodolites and radar tracking. This permitted soundings into the stratosphere except when high winds were overhead, creating elevation angles too low to permit tracking.
Radio navigation systems used by the aviation industry such as LORAN (LOng RAnge Navigation) and Omega were subsequently used, but have now been replaced by global navigation satellite systems such as the Global Positioning System (GPS), which permits the transmission of very high vertical resolution data at all levels.
A small number of wind observations per day are made by balloons without radiosonde instruments; these are called pilot balloons and are tracked by theodolites.
The commercial aircraft observations discussed in Section 3.a.3 usually contain wind speed and direction information. Wind observations from aircraft are calculated from the difference between the velocity of the aircraft with respect to the earth and the aircraft velocity with respect to the air (true airspeed). These velocities are obtained from data from the aircraft navigation system, which includes an inertial platform, magnetic compass and GPS, and its airspeed system, involving pitot static pressure and air temperature (WMO, 2018). In principle, these are 3D velocities, but since the vertical wind speed is typically three orders of magnitude less than the horizontal speed, the
The energy received, reflected, absorbed, and emitted are the components of the Earth's radiation budget. The radiation budget is the balance between incoming solar radiation and outgoing radiation, which is partly reflected solar radiation and partly radiation emitted from the Earth, including the atmosphere. Incoming ultraviolet, visible, and a limited portion of infrared energy (together referred to as "shortwave radiation") from the Sun drive the Earth's climate system. Some of this incoming radiation is reflected off clouds, some is absorbed by the atmosphere, and some passes through to the Earth's surface, where it may be absorbed or reflected by soil, vegetation, snow, and ice. Larger aerosol particles in the atmosphere interact with and absorb some of the radiation, causing the atmosphere to warm. The heat generated by this absorption is emitted as longwave infrared radiation, some of which radiates out into space.
The gases, ash, and dust particles lofted into the atmosphere during volcanic eruptions may have a profound impact on the climate. These particles cool the planet by blocking the incoming solar radiation and may even circle the Earth with the cooling effect lasting for months to years depending on the characteristics of the eruption as was the case with the catastrophic 1991 eruption of Mount Pinatubo in the Philippines (Robock, 2002).
The radiation budget is measured by Clouds and the Earth's Radiant Energy System (CERES). CERES on S-NPP and JPSS provides continuity of the CERES instrument on the NASA EOS AQUA satellite and extends the climate data record of Earth's radiation budget from the first Earth Radiation Budget Experiment (ERBE) (Wielicki et al., 1996).
CERES is a scanning radiometer system with total shortwave and longwave window channels with a 20 km footprint at nadir. CERES products include both solar-reflected and Earth-emitted radiation from the top of the atmosphere to the Earth's surface.
Cloud properties are determined using simultaneous measurements from VIIRS.
The first atmospheric measurements were made at the surface, beginning with qualitative assessments of wind, temperature, moisture and precipitation, followed by quantitative observations as instruments to measure these quantities were invented in the sixteenth and seventeenth centuries. The barometer was also invented during this period, often used in "Lagrangian mode" by scientists ascending hills to deduce that We start with the variables shown on a standard surface station plot as shown in Fig. 7.
Temperature is observed by a variety of thermometers -see above instrument texts for the many measurement techniques. Pressure is provided by a barometer -early versions were mercury based but most now are of the aneroid type. The actual station pressure measured is not usually shown -its value is reduced to sea level by hypsometric methods (Bluestein, 1992), which permits construction of surface pressure charts. The 3-hour pressure tendency is also indicated. Wind speeds are measured by anemometers, which usually contain wind vanes to provide wind direction. Among the many types of wind instruments (see referenced texts), the most precise one is the sonic anemometer, which is often used for turbulence measurements. It measures the individual components of the wind (u,v,w) by transmitting sound pulses in opposite directions along three orthogonal paths. Since sound waves travel with the atmospheric medium, a wind component can be obtained by the difference between the two transit times. The temperature can also be calculated by the sum of the two transit times using the speed of sound equation. By convention, most surface stations measure wind at 10 m, with all other variables measured 1.5-2 m above the ground. Fig. 8 shows a typical U.S. Automated Surface Observing System, which measures most of the 18 weather elements shown in Fig. 7. The moisture content of the air is expressed in surface reports as the dewpoint temperature, which can be measured in combination with the temperature via a chilledmirror hygrothermometer. The general term hygrometer refers to any instrument that measures water vapour content, but there are many kinds, depending on the variable desired and the physical principle being used to make the measurement. For example, absorption hygrometers and carbon hygristors (often used in radiosondes) measure relative humidity directly, psychrometers measure wet bulb temperature and dewpoint hygrometers measure dewpoint temperature. There are a variety of other moisture variables important to atmospheric scientists such as specific humidity, mixing ratio, water vapour pressure, saturation vapour pressure, and these can be obtained from the moisture variable that is measured (along with temperature and pressure) via Aerosols are defined as colloidal systems wherein the "dispersed phase" is either solid or liquid particles and the "dispersion medium" is gas (here, the atmosphere) (AMS online Glossary -https://glossary.ametsoc.org/wiki/Welcome). Atmospheric conditions such as fog, smoke, smog, mist and haze are considered aerosols, although haze may also contain photochemical smog (O , NO , hydrocarbons). They all cause obscurations to visibility, endangering aviation and transportation operations in general, whose measurement is described above. Smoke, comprised of small particulate matter caused by combustion, also causes obscurations as well as health hazards. The increase in wildfires and acreage burned in recent years has brought greater attention to "fire weather", for which it is important to monitor and predict precipitation trends, temperature, soil moisture or "wetness", fuel availability and dryness, and wind speed.
Cloud condensation nuclei and dust are also aerosols, most of which originate naturally (e.g. from soil, salt spray, smoke, volcanic eruptions, pollen and other organic material).
Many of these substances are not regularly measured (unless they affect visibility, health and/or air quality), but are done so during field research experiments, often in aircraft.
Finally, the composition of the atmosphere is monitored from both the surface and space, the latter covered in Section 3.c. The two main components, nitrogen and oxygen, are not monitored as closely as the minor constituents or trace gases such as water vapour, carbon dioxide (CO ), ozone (O ), methane (CH ), carbon monoxide (CO), nitrous oxides, etc. since the latter are more important to global radiative forcing, weather and human health. Water vapour is the most important greenhouse gas, as well as the source for the earth's hydrologic cycle and its measurement is described above; however, its total percentage of the atmosphere's mass is not changing very quickly.
While the first generations of satellites primarily provided observations on the synoptic and large-mesoscale, recent generations now measure the atmosphere and surface properties at the convective or storm-scale (0.25-5 km). This section will describe the current sounding and imaging capabilities of many of the instruments carried by the satellites shown in Fig. 2.
The spectral bands of the satellite imagers cover select bands in the visible and infrared portions of the electromagnetic spectrum.
The primary contribution of infrared and microwave soundings to mesoscale and high impact weather diagnostics, nowcasting, and short-range forecasting is through their assimilation into NWP models. Near real-time satellite sounding retrievals of temperature and moisture from LEO, available at the right place and time, can also aid in warnings of severe weather through measurements of (evolving) atmospheric stability conditions (Esmaili et al., 2020)
The Global Precipitation Measurement (GPM) core (Fig. 2) has two widely used primary
The cryosphere broadly includes snow, sea ice, lake and river ice, icebergs, glaciers, ice caps, ice sheets, ice shelves, permafrost, seasonally frozen ground, and solid b Radar
Weather radars were specifically developed with the purpose of detecting mesoscale and convective-scale precipitation systems because they can quickly scan mesoscale areas, which are difficult to sample sufficiently and quickly enough using in situ sensors mounted on instruments such as rawinsondes or on aircraft. S-band radar systems are best suited for surveillance of mesoscale and convective-scale features. These radars are useful for computing the climatology of precipitation, type of precipitation, and type of precipitating system over broad areas. The range of these S-band surveillance radars is typically 100-200 km or more, but is limited by the sensitivity of the radar, the nature of the scatterers, and the lowest altitude that can be scanned owing to the curvature of the Earth and refraction.
The radar-echo imagery from a network of radars may be composited into a mosaic on a much larger domain to show how precipitation varies across the country. Some composited products of three-dimensional radar structure have become available in the U. S. and are archived (e.g. Homeyer, 2014) and are very useful for climate studies. The MRMS programme, mentioned earlier, makes use of composited radar data to produce quantitative precipitation estimates over broad areas (Zhang et al., 2016). Radar resolution and the height above the ground are important considerations in producing these composites because the resolution at far ranges from radars is much less than that near radars and beyond a given range, there are no data below a given altitude (Fig. 5). There are also surveillance radar networks in Europe, in Asia, and elsewhere around the globe, but not all are free and open access, especially when the networks are run by the military and the data are considered a security issue.
C-band systems and X-band systems have also been used, the latter particularly in mesoscale networks such as CASA (Collaborative Adaptive Sensing of the Atmosphere) in the U. S. Advantages of CASA-like networks are that the atmosphere can be sampled near the ground (Fig. 5) and that high spatial resolution can be attained when viewing small-scale features such as tornadoes (McLaughlin et al., 2009).
There are many radar measurements unique to the mesoscale and convection scale.
These measurements include both non-severe and severe phenomena. The motions of convective storms, both severe and non-severe, may be determined by tracking cells using algorithms to detect local regions of enhanced reflectivity. Updrafts in severe convective storms may also be tracked by following WERs, BWERs, and Z columns, but they are not always detected in convective storms.
Lighting flashes in convective storms detected by radar may be pinpointed by arrays of lightning detectors that locate regions of radiation generated along lightning channels.
The incidence, location, frequency, and polarity of lightning flashes can be used to infer vertical motion and precipitation processes in convective storms.
Convection in tropical cyclones has been studied extensively by surveillance Dopplers near coastlines in the U. S. Outer rainbands, inner rainbands, and eyewall convection, and mini-supercells, some tornadic, in landfalling tropical cyclones. S-band radars are the best tools for studying hurricanes, owing to the extreme attenuation occurring in their intense precipitation regions.
Data assimilation (DA) and modelling systems can be used to assess the benefit (or lack thereof) of individual and combined observing systems. Observing System Experiments (OSE) (often called data denial experiments) are employed to evaluate current observations. Forecasts made using all available data are compared with ones made with one or more observing systems absent, thus assessing the impact of the denied data. An example of a regional OSE study of several of the ground-based observing systems mentioned in this paper is given by James and Benjamin (2017), and an OSE for the global scale is presented by Bouttier and Kelly (2001). Note that OSEs can also be used to see how effectively the DA system uses a particular set of observations. To assess the potential value of future observing systems, Observing System Simulation Experiments (OSSE) are used. Here, both current and future observations (and their errors) have to be simulated from a "nature run", which is usually a very high-resolution forecast (made by a different model than used in the experiments) designed to simulate the atmosphere as accurately as possible. Data denial experiments are then conducted to assess the impact of the future observations. OSSEs can also be used help design new observing networks and their sampling strategies, as well to assess whether a new observing capability can help improve our characterization of the earth system or our understanding of the processes that govern the system. A recent essay on current and past applications of OSSEs and recommendations for their future use is given by Zeng et As in Fig. 12, observation impact is taken to be the difference in the error measure Observing Laboratory also manages a widely used Field Catalog service (http://catalog.e ol.ucar.edu) offering a comprehensive collection of field project related documents and data sets. Each Field Catalog is customized as needed for each campaign and continues to be available on-line long after the field campaign has ended, so data can be used for post-project analysis.
NASA and NOAA and other data centres are actively moving data to the cloud. For example, the Global Hydrology Resource Center (GHRC) was the first of the 12 NASA Distributed Active Archive Centers (DAACs) to migrate all its data holdings to a commercial cloud provider (Amazon Web Services) mapping out a pathway for the other DAACs (https://earthdata.nasa.gov/learn/articles/tools-and-technology-articles/eosdis-cu mulus-project). The GHRC manages data from satellites, aircraft campaigns, in-situ sensors, and computer models. Previously, it might have taken weeks to download, reformat, and restructure the data needed to do a time-series analysis, a task that will eventually be completed in minutes. New tools for data exploitation can eliminate the need to download large volumes of data to a local data repository or computer for research analysis. The general public can access data and tools made available through the Cloud Service Providers (CSPs) for free, and the data are available to the public without restriction on use. These data will include the original and unaltered data content, and may additionally be provided using any broadly accessible and open format or formats. NOAA's Big Data Program (https://www.noaa.gov/organization/information-t echnology/big-data-program) works with three infrastructure-as-a-service (IaaS) providers to broaden access to NOAA's data resources. These partnerships are designed to not only facilitate full and open data access at no net cost to the taxpayer, but also foster innovation by bringing together the tools necessary to make NOAA's data more readily accessible. See also https://earthdata.nasa.gov/eosdis/cloud-evolution.
All branches of science require observations to inform their discipline and atmospheric science is no exception. As has documented in this paper, the atmosphere is very well monitored, with terabytes of information per day being generated from satellites, radars and hundreds of other observing systems and instruments. This does not mean that the atmosphere is perfectly observed or that we are making optimal use of the data that we do have. Every advance in numerical weather prediction and climate simulations caused by new observations, better data assimilation, higher resolution and improved physics motivates a desire to increase our capabilities in all four areas. For example, research has shown that climate models with convection-allowing resolution have superior results over coarser models -to accomplish this we'll need additional observations to provide more detailed characterization of the earth's surface and atmospheric composition, observations to inform improved model physics, and, since climate models are becoming well-coupled with ocean, sea ice, land-surface, air chemistry and hydrologic models, we'll need more observations in those domains as well. On the storm scale, the NWS has the "Warn-on-Forecast" vision, in which sub-kilometer resolution models provide reliable probabilities for hazardous weather, including tornadoes, hours before they occur. This will require far more rapid scanning, such as might be provided international atmospheric science and space agencies have teams working on components of the future satellite constellation, which will feature advanced instruments with higher spectral and spatial resolution in both LEO and GEO orbits.
These agencies as well as the private sector are also developing and launching swarms of Cubesats that may revolutionize the future of satellite architectures and how data are distributed and shared. We will see an explosive growth in new approaches that combine the radiances and retrieved information from passive and active sensors from multiple platforms, and increasingly employ Artificial Intelligence/Machine Learning (AI/ML) techniques, data analytics, visualization, and cloud data and software for an integrated depiction of the atmosphere and the Earth as a System.
No potential conflict of interest was reported by the author(s).