The primary source of measurement error from the widely-used particulate matter (PM) PurpleAir sensors is ambient relative humidity (RH). Recently, the U.S. EPA developed a national correction model for PM<sub>2.5</sub> concentrations measured by PurpleAir sensors (Barkjohn model). However, their study included few sites in the Southeastern U.S., the most humid region of the country. To provide high-quality spatial and temporal data and inform community exposure risks in this area, our study developed and evaluated PurpleAir correction models for use in the warm-humid climate zones of the U.S. We used hourly PurpleAir data and hourly reference grade PM<sub>2.5</sub> data from the EPA Air Quality System database from January 2021 to August 2023. Compared with the Barkjohn model, we found improved performance metrics with error metrics decreasing by 16–23 % when applying a multi linear regression (MLR) model with RH and temperature as predictive variables. We also tested a novel semi-supervised clustering (SSC) method and found that a nonlinear effect between PM<sub>2.5</sub> and RH emerges around a RH of 50 % with slightly greater accuracy. Therefore, our results suggested that a clustering approach might be more accurate in high humidity conditions to capture the non-linearity associated with PM particle hygroscopic growth.