The currently best spatial resolution of any LEO AQ satellite is obtained from the Tropospheric Monitoring Instrument (TROPOMI) aboard the Sentinel-5 Precursor (S5P) satellite [Geffen et al., 2020] with a nadir pixel size of ~5.5km × 3.5km. Its temporal resolution is typically one overpass per day at low and moderate latitudes and two at high latitudes. The spatial resolution is in the same order of magnitude for existing and upcoming GEO AQ missions such as GEMS (Geostationary Environment Monitoring Spectrometer) [Kim et al., 2020] with 3.5km x 8km, TEMPO (Tropospheric Emissions: Monitoring of Pollution) with 2.1km x 4.7km and Sentinel 4 [Veihelmann et al., 2015] with 8km x 8km. The temporal resolution of the GEOs is about once per hour during daytime over their respective section of the world.
These AQ satellite resolution values are good enough to resolve the spatial and temporal variability of trace gases such as NO2 over regions of the world without sources, e.g. the oceans. However they fail in capturing the variability over most landmasses, especially over urban and complex (e.g. mountainous) terrains.
Are there options to increase the spatial and temporal resolution for space based observations? Obviously one choice would be to have a GEO satellite, or even a satellite at another distant orbit like e.g. at the Lagrange 1 point, with higher resolution than the ones mentioned above. However this would certainly mean to fly a rather costly mission with a massive instrument. We believe there is also another option using LEO satellites, which would be at significantly lower costs and risks. The basic concept of such a mission is described here:
We assume that a 2D-imaging system with one spatial component (across-track) and one spectral component is used. A region of about 50km x 50km should be mapped in order to include entire large cities in one observation sequence. Assuming a height of 400km, the field of regard (FOR) of the optical system of 50km would correspond to 7°. To obtain a ground resolution of ~50m, we need a detector with ~1000 spatial pixels. If we want approximately the same spatial coverage and resolution in along-track direction, 1000 images must be taken in a measurement sequence. Our experience shows that an optimized observation system could go as low as 150ms exposure time to reach a signal to noise ratio (SNR) around 300:1 and hence obtain meaningful NO2 observations without binning spatial pixels. With this number and knowing that a typical LEO ground speed is ~7km/s, the satellite tracks ~1km on ground during the exposure time. Hence to obtain a pixel size along-track of about 50m, the only option is to slew the optical system backwards relative to the flight direction while measuring. This principle is called “Target mode” and is already used for some satellite missions such as OCO-2 (Orbiting Carbon Observatory-2) and GHGSat (Greenhouse Gas Satellite - Demonstrator) [Jervis et al., 2021], both designed for measuring GHGs. It is illustrated in figure A2.
Note that the optics may also look to the side in case the target object (e.g. a city) is not exactly underneath the orbit, which is of course the most likely case. The pixel size on ground for such a system depends on the angle the satellite points off the nadir position. In nadir view it is about 50m x 50m and in extreme off-nadir view, for images 1 or 1000 and the case of a target at 500km left or right of the orbit, it grows to about 80m x 80m.
The suggested measurement sequence takes all together up to 200s if one allows time for the system to slew forward again to get ready for the next sequence. In the 200s the satellite covered ~1400km on ground, which is also the minimum spatial distance between two sequences. This means the system cannot measure two cities along the orbit, which are closer than 1400km apart. Hence if one wants to map out e.g. Brussels and Amsterdam, one city has to be measured by one satellite in one orbit and the other one either by another satellite or by the same satellite in another orbit.
With a repetition rate of 200s, one could produce 15 high resolution maps inside the ~50min flight over the sunlit portion of the Earth of a LEO. In practice it will rather be <10 maps due to “waiting times'' in between the sequences for two possible reasons: 1) the next target (city) might not be necessarily positioned in such a way, that one can start the sequence exactly when the previous one has finished and 2) one might need to allow for nadir-position periods in between the sequences to recharge the battery sufficiently. Considering that the satellite is over ocean for ⅔ of the time anyway, one will probably use the time over the ocean for optimized battery recharging (except for possibly the tracking of ship-lines) and over land measure as quickly as possible.
With the suggested system one could also make limb measurements by simply pointing the satellite forward. One image would cover an azimuth range of 7° (1000 pixels at 0.007° each) with a FOR in the elevation angle of 0.007°. In order to have a limb scan one would also slew the satellite, but to a very different rate as for the (near) nadir observations.
The measurement mode described above “solves” the problem of spatial resolution. The described target mode is doable for a modern spacecraft. In order to improve the temporal resolution one could build a fleet of such satellites flying in a constellation with different equator crossing times. Since the satellite instrument does not need to be very big, the obvious choice would be to use cubesats, which could be launched all at once and so significantly reduce the mission costs. Such a project could be compared to the GHGSat mission, but measuring reactive trace gases such as NO2, O3, SO2 instead of GHGs.