Developing a decision tool to assist management of prescribed fires in the Flint Hills region to reduce smoke impact on ambient ozone

Abstract

Prescribed burning is a long-standing practice to maintain the ecosystem in the Flint Hills region. Landowners use fires as economically efficient practices on rangeland to maintain ecosystem health and function, facilitate favorable grazing, and enhance livestock performance. Due to these benefits, prescribed fire is used to increase and maintain the diversity of tallgrass prairie species while suppressing the invasion of woody. However, the intensive burning impacts air quality and constitutes public health concerns. Smoke from the fires contributes to the violation of the ambient ozone (O₃) standard, formed through the reaction of nitrogen oxide (NOx) and volatile organic compounds (VOCs) emitted from the fires. The ultimate goal of this research is to establish practical forecasting models and decision tools for smoke and air quality management in the Flint Hills region, which will encourage and enable the continuous practice of prescribed rangeland burning and maintenance of the ecosystem in a manner that minimizes adverse air quality and social impacts. The specific research objectives of this dissertation include: (1) identify the spatial and temporal distributions and patterns of daily prescribed fire activities in the Flint Hills region using the combination of MODIS resolution imagery and standard active fire data; (2) characterize meteorological parameter sensitivities connected to daily fire activities and propose optimum weather conditions for mitigating smoke pollution associated with prescribed fires; and (3) develop regression models to simulate O₃ contributions from prescribed burning in order to assist collaborative management of prescribed fires and reduce smoke impact on ambient O₃. To map the daily burned area in the Flint Hills region from 2003 to 2019, correlations between the accumulated burned area from Moderate Resolution Imaging Spectroradiometer (MODIS) resolution imagery and standard active fire data from satellites were developed as a scale model. Spatial and temporal distributions and patterns of daily fire activities were performed. The results showed that fire activity was concentrated on a few days in mid-April. The annual burned percentage (ranged from 11% to 52%) and average fire size (ranging from 160 to 400 ha) were higher in the Flint Hills region's center than in the south and at the edge. All ground-level O₃ concentrations exceeding 70 ppb were recorded in April, coinciding with intensive fires in the Flint Hills region and on days with high solar radiation. Overall, the results provided a solid foundation for the development of O₃ regression models using daily burned area and weather factors as predictor variables. The relative importance of meteorological driving factors of prescribed burning in the Flint Hills region was evaluated with machine learning techniques using random forest (RF) with Shapley additive explanation (SHAP) values and Pearson correlation to identify weather impact on landowner burn decisions and optimum weather for prescribed burning based on historic data from 2003-2019. The three most important weather variables that affected the subjectivity of landowner decisions in prescribed burning were solar radiation, cloud cover, and relative humidity (RH). The data indicated that less cloud cover demonstrated superior predictive power in landowner decisions for prescribed burning and largely resulted in elevated O₃. About 62% of all the heavy-fire days had cloud cover less than 10⁺%, which was not in the Kansas Department of Health (KDHE) recommended range of optimum burning conditions (30~50%). Optimal weather conditions occurred on average six days per burning season when cloud cover varied from 10⁺ to 55%. Transitioning from current burning practices by landowners to the proposed optimal weather conditions could lead to a remarkable reduction in the occurrence of 70⁺ ppb O₃ levels, by reducing the probability from 35% to 9% on days characterized by heavy fires. Finally, a practical burning decision tool to assist management of prescribed fires in the Flint Hills region was developed to reduce smoke impact on ambient O₃. Pearson correlation and machine learning techniques were used to evaluate predictor variables' influence on O₃ mixing ratios. Results show high sensitivity of daily O₃ mixing ratio to cloud cover, the previous day's O₃ mixing ratio, daily burned area and RH. Proposed optimal burning conditions demonstrate reduced O₃ sensitivity to fire activities, averaging six days per season. Further recommendations were offered about burning under second-class conditions (cloud cover = 10~30%) and marginal conditions (cloud cover =0~10%). The detailed findings from this research can assist land managers in planning burning activities for improved air quality, reducing smoke impacts, and enhancing land management capacity, thereby promoting safer, effective, and sustainable prescribed burning practices. These results highlight the rising concern about smoke impacts on air quality during the prescribed burning season in the Flint Hills region of the U.S. and suggest further research to address this challenge.