Temperatures retrieved from NOAA-15 AMSU-A during the period from July to September in 2001 and 2002 are used to develop an algorithm for estimating the intensity of tropical cyclones (TCs) over the western North Pacific. The variance R² explained by the algorithm is 76.9%, and the mean absolute error MAE (root-mean-square error, RMSE) is 5.6 (7.5) m s⁻¹, all of which are comparable to the published results for Atlantic and Eastern Pacific TCs. In addition to the maximum temperature anomaly over the TC center, the height of the warm core, represented by the uppermost position of a certain temperature anomaly contour (1.0 and 0.5 K), is found to be another important predictor in the final equation.
Both the jackknife method, and an independent test, are applied to verify the algorithm. Results show that R² deteriorates slightly to 72.3% using the jackknife method, while the other three statistic, i.e., MAE, RMSE and standard deviation of residuals, increase by ∼0.5-1 m s⁻¹ in both the jackknife and independent datasets. Most of large underestimations occur for small but strong TCs, while significant overestimations are either due to the lag of an upper-level warm core following the weakening of the surface circulation, or due to upper-level warming ahead of the surface circulation development.