29th Conference on Hurricanes and Tropical Meteorology


Can aerosols explain hurricane prediction errors?

Michal Clavner, The Hebrew University of Jerusalem, Jerusalem, Israel; and D. Rosenfeld

The hypothesis that aerosols weaken hurricanes by suppressing warm rain (Rosenfeld et al., 2007; Zhang et al., 2007) was tested positively when comparing the officially published 24-hours hurricane intensity predictions with the actual hurricane intensities at that time under various aerosol conditions. Two very different operational hurricane prediction models have been used: the Geophysical Fluid Dynamics Laboratory (GFDL) model and The Statistical Hurricane Intensity Prediction Scheme (SHIPS) model. Larger amounts of aerosols apparently caused weaker peak winds in hurricanes compared to the predicted intensities, because the model predictions did not take into account the suppressive effects of the aerosols. The Sahara Desert and Sahel are the primary sources of aerosols that are transported across the tropical Atlantic, therefore, high amounts of dust were assumed to be accompanied by drier desert air. The amount of precipitable water was used in order to understand if aerosols are only marking dry air or independently effecting hurricane cloud microphysics.

The dataset includes Atlantic hurricanes between 2001 and 2007. The values for smoke (black and organic carbon), dust, sea salt and sulfate aerosols were taken as Aerosol Optical Thickness (AOT) from the Goddard Chemistry Aerosol Radiation and Transport (GOCART) hindcast model. The precipitable water was obtained from the National Centers for Environment protection/ National Center for Atmospheric Research (NCEP/NCAR) reanalysis.

Statistically significant weakening of the peak wind intensities compared to the prediction (dVmax) was found independently for both desert dust and organic carbon, while an invigorating effect was found for black carbon. The values of the dust AOT and precipitable water were found to be negatively correlated. In order to disentangle the effects of these two variables the data was divided into five pentads of similar precipitable water values. A similar effect of dust on the prediction error was found within all the five sub-groups. This means that the suppression occurs by the dust itself and not due to its association with dry air. Furthermore, the smoke (black and organic carbon) aerosols were weakly positively correlated with the precipitable water, so that they could not be a marker for the suppressive effect of dry air.

The GOCART generated sulfate and sea salt AOT could not be used as independent variables, because their quantities depend on wind intensity above the ocean.. The sensitivity of the GFDL forecast error for the maximum wind velocities in the hurricanes was about -0.5 knots per 0.01 units of dust AOT, -8.3 knots per 0.01 units of organic carbon AOT and +26.1 knots per 0.01 units of black carbon AOT. The actual range of the AOT was up to 0.28 for the dust, 0.04 for organic carbon and 0.009 for black carbon. It is risky to extrapolate the effects linearly beyond these bounds.

The indicated enhancement of TCs by black carbon might be related to low level warming of the air by these light absorbing aerosols which contributes to the overall heat that is released in the storm, and hence invigorates it. The greater indicated sensitivity to organic carbon aerosols with respect to desert dust is consistent with their greater effectiveness as small CCN for the same AOT. This provides additional support to the hypothesis that the aerosols suppress hurricanes by their CCN activity that nucleate larger number of smaller cloud drops. The delayed rainout enhances the vigor of the clouds at the periphery of the hurricane on expense of the hurricane peak winds at the eyewall.


Rosenfeld D., A. Khain, B. Lynn, W.L. Woodley 2007: Simulation of hurricane response to suppression of warm rain by sub-micron aerosols. Atmos. Chem. Phys. Discuss., 7, 5647-5674.

Zhang H., G.M. McFarquhar, S.M. Saleeby and W.R. Cotton, 2007: Impacts of Saharan dust as CCN on the evolution of an idealized tropical cyclone. Geophys. Res. Lett., 34, L14812, doi: 10.2029/2007GL029876.

Session 2C, Hurricane Aerosol and Microphysics Program (HAMP)
Monday, 10 May 2010, 10:15 AM-12:00 PM, Arizona Ballroom 10-12

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page