Comment on “Impacts of Sugarcane Fires on Air Quality and Public Health in South Florida”

Nowell et al.¹ evaluated potential community impacts from sugarcane harvesting related to fine particulate matter [PM $\le 2.5\mathrm{\mu }\mathrm{m}$ in aerodynamic diameter (${\mathrm{PM}}_{2.5}$)] exposure. However, in an attempt to fill numerous measurement gaps, they made erroneous assumptions and misapplied technical approaches, undermining their conclusions.

The assumption of a seasonal increase in community ${\mathrm{PM}}_{2.5}$ air concentration from sugarcane harvesting is supported by neither the data the authors relied on nor their analysis of those data. They reported a confidence interval in ${\mathrm{PM}}_{2.5}$ increase that was not statistically significant, yet they used the assumed increase to justify calculating sugarcane harvest emissions.

We also disagree with the authors’ use of the $t$-test to compare the harvest and non-harvest seasons. At a minimum, they should have conducted a regression that accounted for yearly fixed effects and clustered standard errors because the data they used were likely not independent—a requirement for conducting a $t$-test. Similarly, by pooling each year’s harvest season data as a single measure and non-harvest season data as a separate single measure, the authors failed to evaluate whether there were consistent yearly ${\mathrm{PM}}_{2.5}$ differences between each harvest and non-harvest season. In addition, although the authors’ publication was submitted in 2022, the data set they relied upon spanned only 2009 to 2018. No explanation was provided for why they used an incomplete data set for their modeled calculation of observed ${\mathrm{PM}}_{2.5}$ differences over time.

Finally, the use of months to categorize the data into harvest vs. non-harvest seasons is not appropriate for causal inference. This is because the temporal specifics of actual harvest activities are not included in their analysis, and their analysis does not account for other confounding seasonal factors that may affect ${\mathrm{PM}}_{2.5}$ concentrations (e.g., meteorological conditions, other ${\mathrm{PM}}_{2.5}$ sources).

The authors’ calculation of sugarcane harvest emissions demonstrates numerous instances of model inputs with errors, large uncertainties, and high bias that cannot justify the analysis precision reported and the certainty with which they stated their conclusion. These model issues occurred at nearly every step in the authors’ methodology: in the fire area and location² and the fuel loading factor they drew from Pouliot et al.,³ in the ${\mathrm{PM}}_{2.5}$ emission factors they drew from McCarty,⁴ in the HYSPLIT model plume rise,^5,6 and in the secondary particle formation they drew from Yokelson et al.⁷ In addition, by adjusting the satellite ${\mathrm{PM}}_{2.5}$ data to identify apparent increases in ${\mathrm{PM}}_{2.5}$ concentrations in the sugar-growing region, the authors increased uncertainty and bias through their choice of a limited sample of ground-based air monitors and the temporal mismatch between the monthly satellite data set and daily ground-based data set. Finally, although the authors concede that burning activity occurs discontinuously during the harvest season, they applied their differential ${\mathrm{PM}}_{2.5}$ exposure amounts for community exposure throughout the entire year.

Ultimately, correcting any of these highlighted deficiencies would have supported the null hypothesis. The corrected analysis would thus demonstrate there is no scientific basis to assert community health impacts associated with an increase in ${\mathrm{PM}}_{2.5}$ from sugarcane harvesting.

Refers to https://doi.org.10.1289/EHP9957