Development of a statistical relationship between ground-based and remotely-sensed damage in windstorms

With rapid growth in technology, new methods of wind engineering research are being explored, and new tools are being utilized. In the past, ground-based surveys of windstorm damage were frequently performed with the aid of aerial photographs in some cases. Researchers have recently begun using remote-sensing data such as digitized satellite, aerial, and LIDAR imagery, to assess damage following natural and man-made disasters, in addition to, or instead of employing the older method of walking house-to-house for surveys. This research explores the relationship between the damage states observed at ground-level, and those observed from space using remote-sensing data, for both tornadoes and hurricanes, and for site-built one- or twofamily residences (FR12). From the Enhanced Fujita (EF) scale, "Degrees of Damage" (DOD) can be determined for ground-based observations. Damage states can also be assigned for remote-sensing imagery, using a modified version of Womble's Remote-Sensing (RS) Damage Scale. The completion of both types of assessments for a given dataset is necessary to develop the statistical relationship between the damage states, which will be based upon a simple linear regression. Several methods were utilized to fit a regression model to the data. Using both an alphabetic and alphanumeric RS damage state, models were fit to relate the ground-based and remotely-sensed damage. The model fitted with the alphabetic RS scale has the best fit. Models were also fitted using RS damage states identified by imagery from different satellite platforms, with different resolutions. The results indicate that the model fitted utilizing QuickBird imagery has a better fit to the data. Several additional methods and/or combinations of datasets will be utilized in future work to refine the models. In addition, the developed models will be tested against other datasets, such as Hurricanes Katrina and Ivan to verify their performance. We anticipate that, once fully developed, the relationship can be used to predict the ground-level damages state using remote-sensing imagery.