Bringing Satellite Measurements Down to Earth
The modern satellite era has revolutionized our understanding of the earth system and our ability to study and monitor the atmosphere. It was not so long ago that atmospheric measurements and observational data were limited to areas where humans could readily travel, be it by ship, aircraft, or even dogsled, thus leaving vast swaths of the remote atmosphere unsampled and unstudied. However, with the advent of sophisticated satellite instrumentation capable of measuring even trace levels of atmospheric compounds and particles, and a network of satellites keeping regular watch around the globe, we have been launched into a new era of scientific knowledge.
But to do these measurements, and to do them well, requires a lot of work behind the scenes, and that is where Scientific Aviation comes in. For the last year, SciAv has been flying one of their instrumented Mooney aircraft to measure the vertical (column) profile of trace gases, from ground level to 25,000 feet, to provide ground truth data that NASA scientists rely on for calibrating and validating the satellite measurements obtained by the OCO-2 and Sentinel-5 Precursor (S5P) satellites.
OCO-2 is the Orbital Carbon Observatory, an American satellite that measures primarily atmospheric carbon dioxide (CO2) and is part of the A-Train constellation of environmental observation satellites. S5P is a European satellite that houses TROPOMI (the TROPOspheric Monitoring Instrument), which measures ozone (O3), nitrogen dioxide (NO2), sulfur dioxide (SO2), formaldehyde (HCHO), methane (CH4), carbon monoxide (CO), aerosols, and clouds.
At this point, there are at least two common questions that we often hear from folks who are not very familiar with satellite measurements: 1) What is ground truth? and 2) Why is this necessary?
Ground truth is a term commonly used in remote sensing (i.e., satellite measurements or other optical measurements taken from a distance) and refers to data that is collected in situ, which can be used to relate or calibrate the remotely “sensed” data to real features at the location of interest. In the case of satellite observations of atmospheric trace gases, the “real features” being measured for ground truth are the actual gas concentrations.
The driving motivations behind satellite measurements of atmospheric trace gases are to track and monitor air pollutants on a global scale, identify sources of air pollution (such as methane, CO2, or NO2) that adversely impact air quality and/or climate, and estimate the emission rates of air pollutants. For these applications, it is the gas concentrations in the troposphere (the air nearest the Earth’s surface where all living things reside) that we care about the most. The Earth’s surface is also where nearly all of the sources of air pollution are located. However, interpreting satellite retrievals to determine the distribution and concentrations of pollutants is not a trivial matter, nor is it straightforward.
Satellite instruments do not directly measure concentrations of trace gases in the troposphere; rather, they operate by detecting how much solar radiation reflected off of the Earth’s surface is absorbed by pollutant molecules throughout the entire column of the atmosphere (i.e., the entire distance between the satellite and the ground, which is a very long way!). The amount that the solar radiation is reduced is proportional to the number of pollutant molecules in the atmospheric column, and by extension, the concentration of the pollutant. Furthermore, the wavelength of the radiation that is absorbed is different for different pollutants. This allows scientists to separate and identify multiple individual pollutants (for example, NO2 absorbs in the 400 – 500 nm range).
But, that is still only part of the story. For the sunlight to reach the surface in the first place, it will have already passed through the atmosphere and been partially absorbed by pollutant molecules on the way down. This means that the reflected light that reaches the satellite will have passed through the atmosphere not once, but twice. We’ve already established that the amount of sunlight that is absorbed is a function of the number of pollutant molecules in the total column, but to even further complicate the situation, the height of the column itself (the distance that the sunlight has to travel through the atmosphere to reach the ground) varies with the angle of the sun according to time of day. So when the sun is directly overhead, it is actually passing through less atmosphere than in the morning or late afternoon, meaning that more radiation is reaching the surface. To avoid this added complication (because, let’s be honest, satellite measurements are complicated enough already), satellites like OCO-2 and S5P are placed into sun-synchronous orbit so that when the satellite passes over any given point of the Earth’s surface it does so at the same local mean solar time for every pass. Thus, every time the satellite is overhead, the illumination angle of the sun to the surface will be nearly the same.
What we get from the satellite measurement then is the total column measurement for a pollutant. Deriving the tropospheric pollutant concentrations and determining how the pollutant is distributed in the troposphere (the vertical profile) is a challenging endeavor that requires removing any contribution from pollutant molecules located in the stratosphere or above and then unraveling the vertical structure from the ground to the tropopause (the separation point between the troposphere and the stratosphere). This vertical profile component is critical for air quality purposes since a pollutant molecule located near the ground versus one at 20,000 feet will have different impacts on atmospheric chemistry (e.g., smog or particulate matter formation) and public health.
Many of the air quality pollutants that we are interested in studying are emitted or formed near ground level, which means that the majority of the measurement signal is already coming from the troposphere (e.g., NO2 and HCHO). This is not the case for ozone. The ozone levels in the stratospheric ozone layer are many times greater than surface levels, and the stratosphere contains about 90% of the total column ozone. In the case of CO2, there are significantly greater concentrations near the surface due to combustion sources, but the very long lifetime of CO2 means that the stratospheric concentrations, while lower, are still appreciable.
Teasing out the tropospheric component and the vertical distribution of a pollutant relies on atmospheric chemistry / transport models that incorporate weather simulation, forecasting, tropopause height (to delineate between tropospheric and stratospheric air), and simulated pollutant vertical distribution estimates derived from prior knowledge of sources and chemical behavior. These simulated vertical profiles act almost like an initial guess for the retrieval algorithm, and they are also one of the major sources of uncertainty in tropospheric column satellite retrievals.
The remaining critical piece of this puzzle is validation. How do we know how accurate the retrieval algorithms and models are at deriving the atmospheric profile of CO2 or methane or NO2? For this, we have to compare the derived profile to the actual atmospheric profile at the exact same time and location of the satellite observation.
This is where we get to the fun part: enter Scientific Aviation.
A satellite’s orbit is very regular and very predictable, so we know exactly where it will be at any given time. Scientific Aviation’s job is to collect vertical profile measurements of CO2, CH4, NO2, and O3 from the ground to 25,000 feet by spiraling up and down during a time window beginning 15 minutes before through to 15 minutes after the satellite overpass. These direct observations are then used to compare against the satellite retrievals, inform where there are weaknesses or biases in the retrievals, and ultimately reduce the uncertainty and improve the value of satellite-based observations.
Every couple of weeks, when the sky is sufficiently cloud-free enough to permit a high-quality satellite observation, SciAv takes to the skies to rendezvous with the distant satellites. Taking off from Lincoln Regional Airport near Sacramento, their first stop is at the NASA Ames Research Center / Moffett Field, where the instruments are carefully calibrated against known concentration chemical standards to ensure accurate measurements. After calibration, they fly to one of three measurement sites located in central California, central Nevada, or offshore over the Pacific near the California coast.
The use of these three different sites is strategic, for each provides a slightly different test of the satellite retrievals. The difference between the two over-land sites is the ground cover: the California site is covered in vegetation while the Nevada site is bare desert sand. This difference in ground cover means a different albedo, or reflectivity, of the ground. Vegetated areas reflect ~10-25% of solar radiation, whereas desert sand reflects ~30-45% (in contrast, snow can reflect up to ~85-90% of solar radiation, which can cause a condition known as “snow blindness” if you don’t protect your eyes on sunny winter day). The albedo of the ground surface will thus impact the amount of solar radiation reaching the satellite instrument, and so this is a factor that must be considered in the retrieval algorithm.
For these over land measurements, the satellite is looking straight down at the ground and collecting the portion of solar radiation that is reflected straight up. This downward-viewing geometry is termed “nadir.” Water, however, has a very low albedo (<10%), so the nadir viewing geometry does not work over water; it just looks dark to the satellite. But, water does reflect the sun at an oblique angle. This angled light reflection is known as “sun-glint”, and sun-glint is what causes a lake to sparkle in the afternoon on a sunny day. To catch the sun glint, the satellite must look at the water at the precise angle that the light is reflected off the water. The conditions necessary for a good quality sun-glint measurement don’t occur as often as for an over-land nadir measurement (and the ocean often has cloud cover), and this is part of the reason that satellite measurements over the ocean are more difficult, and more sparse, than land measurements.
The power of satellite measurements for observing and studying the atmosphere on a global scale is immense; perhaps equally immense is the work that goes into it, and this article has only scratched the surface. If it takes a village to launch a satellite, then it takes an entire city to carefully calibrate, validate, and process all of the collected data into the beautiful maps and images that we see in the final product. We at SciAv are proud to be a part of this effort.