Tropical cyclone intensity estimation from satellite images

Accurate intensity estimation of tropical cyclones (TC) is an important topic of research due to their economic impact and public safety concerns. An accurate measure of the current wind strength is a must to accurately predict TC intensity. Wind measurement is obtained by aircraft flying through the cyclones, however routine flights occur only in the North Atlantic Ocean. Meteorologists also use satellite images to infer the wind strength. We have developed and tested automated method to estimate TC intensity based on the existing historical satellite images. The Hurricane Satellite data (HURSAT–B1) is used to develop the algorithm. The intensity estimation algorithm uses satellite images for intensity analysis. The expected intensity is estimated using age of the cyclone and satellite imagery from the current time and, 6, 12 and 24 hours prior as predictors. The algorithm is trained and validated using aircraft reconnaissance-based data. Instead of regression techniques, the 10 closest analogs (determined using a K-nearest-neighbor algorithm) are averaged to estimate the intensity. Several tests are implemented to statistically justify the proposed algorithm using k-fold cross-validation. The resulting average Mean Absolute Error is 11 kt (50% of points are within 10 kt) and its accuracy is at least parallel with current objective techniques. Simplicity, objectivity and consistency aspect of the proposed technique makes it superior to other techniques.