Current rotation detection algorithms are prone to poor performance due to high local variance in a Doppler radial velocity field. A new linear least squares derivatives (LLSD) technique has been developed which is less prone to problems from such variability and outputs a number of fields diagnosing the radial wind field, including azimuthal shear and radial convergence. We subjectively diagnosed rotation within thunderstorms from seven storm days which yielded eighty tornadoes. Threshold values and trends of azimuthal shear and radial convergence fields were extracted from the tornado producing storms. These values and trends will be used to train a new rotationdetection algorithm.