Saharan Dust Impacts on Hurricanes: From Process-Level Numerical Model to Data-Driven Machine Learning

Saharan dust outbreak events peak from June to September, which coincides with the peak of the Atlantic hurricane season. Therefore, Saharan dust has large potential of impacting the genesis and intensification of tropical cyclones by their modulating cloud hydrometeor content, diabatic heating distributions, and thermodynamic structure, as well as large-scale environmental parameters. We first utilize the dust-aware and cloud-resolving WRF model to investigate the relative importance of the microphysical and radiative effects of dust on two hurricanes (Earl and Danielle, 2010) at different life stages under similar dynamical conditions. The WRF model illustrates the dominant dust radiative effects on both intensity and track of the storm. Moreover, we find that dust suppresses the early-stage hurricane but provides favorable conditions for matured one by altering the mid-level ridge in the Central Atlantic, in agreement with the previous finding based on global climate model simulations. To further explore the long-term impacts of dust on hurricanes, we employ a XGBoost machine learning (ML) model and 19-year precipitation, dust, and meteorological records. The improved model capability of predicting hurricane related precipitation is identified when dust optical depth is taken into account in developing the model. The ML predicted relationship between dust and hurricane precipitation exhibits a general boomerang shape, with the precipitation first enhanced and then suppressed by dust, indicating a possible regime shift from microphysical to radiative effects.