Characterization and source apportionment of atmospheric organic and elemental carbon during fall and winter of 2003 in Xi'an, China

A rooftop sampling station in Xi'an during the fall of 2003 used filter paper to draw in the city's air, hour by hour, capturing whatever was suspended there. It was a quiet, methodical act: a pump, a filter, a timer. Then the filters went to the lab, and what came back was a number that stops you cold. Organic carbon alone averaged sixty-one point nine micrograms per cubic meter that winter. To put that in context, the air in Xi'an, by weight, was carrying more carbon than most cities in Asia had ever measured. Cao and colleagues set out to answer two questions: what is in Xi'an's air and where does it come from? They focused on the carbonaceous fraction of fine particulate matter — organic carbon, or OC, and elemental carbon, or EC. OC is the complex mixture of carbon-containing compounds from combustion and atmospheric chemistry. EC is the black, light-absorbing carbon — essentially soot — that is often equated with black carbon in climate studies. Both are significant. EC absorbs sunlight and heats the atmosphere, while OC scatters and absorbs light and damages lungs. The paper cites estimates that roughly one quarter of global EC emissions originate from China, which gives the measurement of any major Chinese city significance beyond its borders. Xi'an was an insightful place to do this work. With five million people, it is the largest city in northwestern China, located on the southern margin of the Loess Plateau. Most carbonaceous aerosol studies in China had focused on coastal megacities. Xi'an offered something different: a city where residential coal combustion for winter heating was common, motor traffic was growing quickly, and almost no continuous measurements existed. Cao's team ran their samplers continuously from September 2003 through February 2004, during the high-pollution fall and winter seasons, collecting PM2.5 samples, which are particles smaller than 2.5 micrometers in diameter, daily, and PM10 samples, particles smaller than 10 micrometers, every third day. The method for separating OC from EC was thermal and optical reflectance, following the Interagency Monitoring of Protected Visual Environments protocol, using a DRI Model 2001 analyzer. The logic is straightforward: you heat the sample in stages under different gas mixtures, and different carbon fractions volatilize at various temperatures. Four organic carbon fractions come off under helium, while three elemental carbon fractions come off under a low-oxygen mixture. A laser monitors reflectance throughout, and when the reflected signal recovers after oxygen is introduced, the instrument marks the boundary between OC and EC. Interlaboratory checks showed agreement within five percent for total carbon, which supports confidence in the numbers. The numbers are striking. Average PM2.5 OC concentrations were thirty-four point one micrograms per cubic meter in the fall and jumped to sixty-one point nine in winter — nearly double. EC moved much less, from eleven point three to twelve point three micrograms per cubic meter. The divergence between those two trajectories provides the first clue. When you calculate total carbonaceous aerosol — OC multiplied by 1.6 to account for the non-carbon mass attached to organic molecules, plus EC — it accounted for about forty-nine percent of PM2.5 mass in fall and forty-six percent in winter. Roughly half the fine particle burden in Xi'an was carbon. The OC-to-EC ratio sharpens the clue into an argument. In fall, the average OC to EC ratio was three point three. In winter, it rose to five point one. Every individual sample exceeded a ratio of two point zero. High OC to EC ratios point toward sources that emit relatively more organic carbon — coal combustion is one of them. The OC-EC correlation also changed: it was very tight in fall, with correlation coefficients above zero point ninety-five, suggesting a common, stable set of sources, and weaker in winter, at zero point eighty-one, suggesting something new had entered the mix. That something, the source apportionment would confirm, was residential coal. To find out exactly what was driving what, the team used absolute principal component analysis, or APCA, applied to all eight thermally-derived carbon fractions simultaneously. The intuition is this: different combustion sources leave different patterns across the eight carbon fractions. Gasoline exhaust burns relatively hot and clean; it has a distinct fingerprint. Diesel exhaust has another. Coal combustion and biomass burning have their own. APCA finds the combinations of fractions that move together across samples — those groupings correspond to source types. The method first converts each fraction's concentrations into Z-scores, standardizing them so fractions measured in very different quantities can be compared. Then it extracts principal components — statistical groupings — and regresses total carbon against them to estimate how much each source contributed, sample by sample. In fall, three factors explained sixty-eight percent, fourteen percent, and ten percent of the variance. In winter, the top three explained fifty-five percent, twenty-one percent, and thirteen percent. The results from that analysis are where the story lands. In fall, gasoline engine exhaust accounted for seventy-three percent of total carbon, diesel exhaust for twenty-three percent, and biomass burning for four percent. Coal contributed negligibly. In winter, the picture changes completely. Gasoline engine exhaust and residential coal burning each accounted for forty-four percent of total carbon, with biomass burning at nine percent and diesel at just three percent. Coal went from a rounding error to nearly half the total carbon load. That is the dramatic finding — a seasonal flip, driven by residents lighting their coal stoves as temperatures dropped. The shift is internally consistent across every measure in the paper. The OC to EC ratio rising from three point three to five point one aligns with coal's known emission profile. The weakening of the OC-EC correlation in winter aligns with a second major source entering the picture alongside motor vehicles. The winter APCA factor loadings — concentrated in the OC2 through OC4 and EC1 fractions — are consistent with coal combustion and motor vehicle exhaust mixed together. The numbers reinforce each other from multiple directions, which provides the apportionment credibility. It's important to consider what this tells you about the structure of the problem. In fall, if you wanted to reduce carbonaceous aerosol in Xi'an, the answer was clear: vehicle exhaust, particularly from gasoline engines, was the dominant source. But come winter, addressing only vehicles would leave roughly half the carbon — the coal half — untouched. Conversely, addressing only coal in winter would leave the other forty-four percent from gasoline exhaust in place. The pollution genuinely has dual sources in the cold months, and any policy that treats it as a single source will underperform. Cao and colleagues also note that OC and EC levels in Xi'an exceed those measured in most urban Asian cities, placing this dataset near the top of the distribution. Most of the OC and EC mass was concentrated in the fine PM2.5 fraction, meaning it reaches deep into the lungs. The visibility impacts of that concentration would have been severe, as EC absorbs light directly, and high OC loads scatter it. What Cao and colleagues built, by standing on a rooftop and collecting air for six months, was a seasonal fingerprint of a city's combustion life. The act of catching air on a filter and resolving it into fractions gave them something precise: not just that Xi'an's air was bad, but why it was bad and how the answer changed between September and February. Carbonaceous particles made up nearly half the fine particle mass. The source mix rotated sharply with the heating season. Effective action required targeting both vehicles and residential coal at the same time during winter, rather than treating them as separate problems to be solved in sequence. That is what the numbers from a rooftop in Xi'an in 2003 actually say. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.