The numerical predictions of historical storm surges in which the low pressure meteorological system remains well offshore requires a factor three, or more, wind stress than that provided using the standard wind stress coefficients found in the literature. The literature values are from the ‘air side’, where wind stress coefficients are determined either from the measured vertical velocity gradient of the wind over the water or more commonly from wind Reynolds stresses. On the other hand, the wind stress required to match the observed storm surges is consistent with the momentum transfer observed during wind-wave growth experiments (i.e., the inferred wind stress coefficients from experiments measuring wave height growth due to wind). Both the wind-wave growth data and our storm surge modelling are consistent with field and laboratory measurements of Reynolds stresses of wind-driven waves. The possibility that there is more momentum being transferred downwards urges the development of a storm surge model capable of using the wind stress inferred from the wind-wave growth data. This requires a Reynolds stress model covering deep through to shallow water depths and at arbitrary levels (3D model implementation), which has been formulated herein. Application of this wind stress to Tropical Cyclone Roger using the steady shallow water equations qualitatively explains the observed storm surge. This new approach also raises the question of why storm surge estimates are generally acceptable for weather systems that cross the coastline despite the use of momentum transfer based on air side wind stress.