The typical location is a highway on-ramp where you are accelerating, i.e. you’re making a significant amount of power (and therefore a useful amount of exhaust). The measurement is optical: the system is sending a light beam across the roadway, roughly through the region where your car will be spewing exhaust, and a receiver on the far side of the road measures the attenuation of specific wavelengths of the beam by the components of your exhaust. The higher the concentration of criteria pollutants in your exhaust, the more certain wavelengths of the beam are attenuated.
Formal compliance testing of a vehicle or engine in an EPA or manufacturer test cell is done under exacting conditions to assure repeatability. The vehicle is “soaked” at a specified temperature for a specified time before the test, the test cell is kept at a specific temperature, the vehicle/engine is driven through specific sequences of speed/distance, the the exhaust is handled in a very specific manner (e.g. it’s passed through a “dilution tunnel” to simulate atmospheric mixing/chemistry before measurement), and so on. (if you want to know ALL the requirements for how to conduct a vehicle test, well, here ya go.) By comparison, as you might expect, a drive-by emissions test isn’t terribly precise because it doesn’t control for many of these factors. OTOH, it doesn’t need to be that precise, because when cars go bad WRT emissions, they tend to go really bad. The drive-by test isn’t looking for cars that are 5-10% over the limit, it’s looking for cars that are 5-10 times over the limit, and that kind of discrepancy is relatively easy to spot with a drive-by test.
The primary value of the drive-by test is that it reduces the inconvenience to most drivers by eliminating the need to go somewhere just for a smog check - unless the drive-by test specifically shows that your vehicle probably has an issue. If it does, then you are required to go to a mechanic shop for a more accurate emissions check. The drive-by test can also show model-by-model trends that can indicate a broader model-related problem that might justify an investigation/recall. For example, after the Volkwagen emissions scandal several years ago, a check of Colorado’s aggregated drive-by-emissions data clearly showed that certain VW models emitted a lot more on the road than they did in a test cell.
The limited extent of the testing (esp. the limitation to the Twin Cities metro area) was likely related to a failure of the region to meet one or more National Ambient Air Quality Standards (NAAQS). When this happens, the state is required to submit a State Implementation Plan (SIP) describing actions the state will take in order to bring local air quality into compliance. Vehicle emissions testing, restricted to just the metro area, was likely one aspect of their SIP. If they’ve eliminated this test program, it likely means that they’ve achieved NAAQS compliance and had reason to believe the program was no longer necessary.
“Environmentally friendly” depends on what aspect of the environment is being considered.
There are two different aspects of a car’s emissions that are kind of in tension with each other. One is the carbon footprint, which seems to be all the rage these days. The carbon footprint does matter - for your great-grandchildren and their descendants. But the other aspect is the immediately noxious emissions, most notably raw hydrocarbons, NOx, and particulate matter (PM), and these will adversely affect your health today. Engines/vehicles could be made more efficient (lower carbon footprint) in a heartbeat if we eliminated the restrictions on HC/NOx/PM emissions, but present-day air quality would quickly turn to shit. In fact, even with current emissions standards, air quality in many cities in the US (e.g. Los Angeles) is already a problem. Cars are vastly cleaner than they were 50 years ago, but there are also far more of them now, and we drive them much more than we used to.