The satellite constellation in the United States is called WAAS, which stands for Wide Area Augmentation System. It consists of 38 satellites, three ground master stations, and three geostationary satellites. Though the US GPS constellations works globally, other regions have developed similar constellations. The European equivalent, completed in 2016, is aptly named Galileo.
Source
How does WAAS improve the accuracy of the GPS?
In short
WAAS, wide-area augmentation system evaluates errors on GPS satellite signals, and broadcast applicable corrections to GPS receivers located in North and Central Americas. Errors detected include satellite clock errors, ephemeris (orbital elements) errors and atmospheric delays. The corrections are sent using satellites orbiting higher than GPS satellites, 36,000km vs. 20,200km, but using the same L1 carrier frequency (1.5GHz) as GPS, and the same modulation techniques.
A receiver listening on L1 receives both the GPS ranging messages and the WAAS correction messages. It applies corrections when calculating its position. Correcting ("augmenting") GPS with WAAS increases the GPS typical 30m-precision (95% limits) to about 3m in single-frequency receivers.
Delay is the most critical parameter. Waves are slowed down and refracted in the atmosphere layers (adding up to about 60m to the path of L1 wave). This makes satellites appear at distances larger than they are, and leads to erroneous calculations.
Corrections to be applied are determined on the ground using a network of 38 GPS reference stations (WRS) distributed over North and Central Americas. Corrections are uploaded to general-purpose geostationary satellites hosting a WAAS transmitter and, like GPS signals, corrections are blindly broadcast to whoever listens. The satellites cover all Americas, however WAAS signals are only beamed to the Northern hemisphere as there are no WAAS reference stations to evaluate the corrections in Latin America (ICAO currently evaluates SACCSA hosted on Immarsat for CAR/SAM regions).
WAAS is operational since 2003, it is one of the existing augmentation systems, designed to improve GPS (and other GNSS) use. Like GPS was the first GNSS, WAAS was the first implementation of ICAO SBAS, satellite-based augmentation system. This is how GNSS and SBAS cooperate:
(Source)
WAAS was intended to provide aviation with a signal-integrity guarantee (main goal) and an approach-class positioning. The WAAS signal received and demodulated by the GPS IC, most GPS IC are now WAAS-capable (e.g. popular SiRFstar sensors support WAAS since 2004), so WAAS use is nearly generalized today, from smartphones to hi-precision surveying stations.
For a receiver not located in North/Central Americas, another SBAS must be used, e.g. EGNOS provides corrections in Europe, for Galileo and GPS GNSS.
Do not confuse SBAS (WAAS, EGNOS, etc), the augmentation system, and GNSS (GPS, Galileo, etc), the positioning system. A SBAS cannot be used alone, it only evaluates and broadcast errors for one or more GNSS signals. Of course a satellite carrying a SBAS transmitter (SBAS messages) is not prevented to also transmit a GNSS signal (navigation message) on the same and/or a different frequency, and some do, but that's two different things.
Plenty of details follow.
Refresh on GNSS principle
A GNSS satellite is basically an atomic clock delivering a time service by radio. It emits 6-second-long data streams (subframes), data including the time the first bit of the next subframe will be precisely sent, allowing a receiver to measure the transit time of the next subframe. When merged, the subframes form a repeated 750-second digital navigation message containing, in addition of the time information, essentially fixed or slowly-changing data, like an identification and orbital elements (orbit eccentricity, inclination, etc). A GNSS receiver use the time-stamps and other data from the navigation message to determine:
The satellite position when the current subframe was sent using the orbital parameters and the departure time known from the previous subframe.
The distance between the satellite and the receiver from the time of departure and the time of arrival.
Its own position from the the satellites positions and distances.
This requires to determine X, Y, Z coordinates in WGS-84 coordinate reference system and the unknown bias between the receiver and the satellites clocks. To solve the four unknowns the receiver must have four satellites in sight. It determines four satellite positions and pseudo-ranges (a pseudo-range is a range computed using the biased receiver clock for the arrival time) to create a system of four linear equations.
In this process, any error on a satellite orbital element, on a satellite clock or in the assumed wave path translates in an error on the position calculated by the receiver.
Refraction and atmospheric delay
Errors mentioned above are labeled (1) and (3) in this error recap:
(Source)
Propagation delays (3) are observed because:
$c$ velocity is less in air than in vacuum, and varies with air composition and altitude. This introduces a ranging error of about 30m.
The wave crosses different layers of air with different and opposed gradients of refraction index. This slightly bends the path, in one direction, then the opposite. At L1 frequency (1.5GHz), the distance traveled can be increased by 30m. The delay vary with the elevation angle, hence with the receiver location.
Atmospheric effects vary with frequency, dual-frequency receivers like military and land surveyors ones are able to infer the current delays and can correct a satellite range themselves but single-frequency receivers cannot, this is where WAAS comes on.
WAAS and SBAS
WAAS is the SBAS for North/Central Americas providing corrections and alerts regarding the GPS, the US GNSS. It consists in a network of 38 GPS ground stations in North and Central Americas (Wide-area Reference Stations) receiving the GPS L1 signal and determining for each satellite in sight:
Europe is covered by EGNOS, Japan by MSAS, India by GAGAN, China by BDSBAS, etc.
(Source)
These systems implement the same ICAO SBAS standard, described in RTCA MOPS DO-229-C minimum operational performance standards for global positioning system/wide area augmentation system airborne equipment:
All SBAS stations have their actual coordinates precisely surveyed in ITRF using multiple methods. Each station also
Corrections for atmospheric delays, which vary between the different SBAS stations, are transmitted as a 5x5° grid of interpolated values. Each receiver selects the square the closest to its position.
A new version of the standard, dual-frequency, multi-constellation SBAS (DFMC SBAS) will be applicable to new SBAS started next year, in order to create a seamless global coverage of complementary SBAS, more precise than the current version, for lower decision heights.
SBAS transmitters can also transmit a navigation message similar to a GNSS navigation message. This navigation message can be used for positioning. This service is not used in aviation with the current version of SBAS, but with DFMC it will be included.
Other ICAO augmentation systems: GBAS and LAAS/GLS
An equivalent, but more precise system, known as GBAS, ground based augmentation system, uses a network of ground stations distributed on a smaller area to compute more precise corrections.
Corrections are broadcast in VHF by ground transmitters instead of satellite-hosted transmitters. To get the corrections, the GNSS receiver must use a VHF reception module in addition of the GNSS IC. GBAS-capable receivers are more complex, more expensive and have more reception constraints.
LAAS, local-area augmentation system, also known as GLS, GBAS landing system, is a GBAS with reference stations in the vicinity of the airport, it is used for landing.
DGNSS and RTK
SBAS (including WAAS), GBAS and LAAS/GLS are actually differential GNSS using range/code measurement to determine corrections, their very fine details are found in differential GNSS descriptions.
There is an alternate way of using GNSS signals ignoring the navigation message data and ranges. It uses the phase of the carrier used to send the message. It is much more precise, but for a single-frequency receiver, it requires a delicate initialization phase followed by a permanent reception of the carrier, two impractical constraints in aircraft.
RTK, real time kinematic is an example of positioning by phase measurement.
Phase measurements, common for precise applications, are usually done this way:
One or more GNSS receivers on the field (called rovers) record phase values at different locations.
At the same time, a reference station, which position is known, usually located at the office, performs identical phase measurements.
When a rover returns to the office (or when its data are downloaded), rover and reference measurement are mixed to determine the locations surveyed by the rover.
In contrast RTK uses a link from the reference station to the rovers, allowing on-the-flight (OTF) positioning, removing the post-processing step. This is what real time means in RTK
RTK increases the precision to the centimeter. It is widely used by surveyors, but also in agricultural and autonomous vehicle applications. However the base station signal is usually a paid service, though some free NTRIP "casters" are available on the model of private ADS-B receivers used on flight-tracking sites.
In RTK static applications, long measurement periods are possible. Millimeter precision is reached by averaging the measures. Such precision allows to use RTK sensors, e.g. to track deformations on structures like bridges.
Basically, the GPS knows were the ground station is so it knows the height, position etc. So you have a way to verify the calculation of the satellites.
The calculation passes, then, through the WAAS master station and removes any errors. With this, is possible to maintain the integrity and accuracy of the GPS data recevied by the aircraft.
