Given the HWRF model output tendencies arising from each component of model's physics it is possible to extract statistically the error distributions for hurricane forecasts made by the model. This is a statistical regression procedure where the total observed tendencies are regressed against the sum of the individual tendencies from the HWRF model. The least square regression provides multipliers in three dimensional space and model forecast time as well. This statistics is based on as many as 210 recent past hurricane forecasts made by the HWRF model covering several years. This statistics is next used to illustrate the structure of errors arising from different components of model physics. In this presentation the structure of errors arising from cumulus convection, planetary boundary layer physics and long and short wave radiation are illustrated. Overall we noted small errors for the radiative transfer terms for the five day forecasts. The planetary boundary layer moisture fluxes were also not very far off and the errors were generally small. The largest errors were noted in the heating from the pararameterized cumulus convection where we found large displacements in the vertical for the modeled heating rates and those magnitudes were also overestimated by the model. The corrected tendencies from the use of the statistical multipliers show a major improvement .Future work on how this information could lead to improved HWRF model will be outlined.