From spectrogram to range-weighting function: a unified 2-D convolution model for coherent Doppler wind lidar
Abstract. The range-weighting function (RWF) of a pulsed coherent Doppler wind lidar is commonly identified with the emitted pulse intensity, giving a range resolution of half the pulse length. We show that this identification is the last step of a hierarchy of approximations whose assumptions are rarely stated.
Starting from the discrete sliced-atmosphere model, the expected Doppler spectrogram is a two-dimensional convolution between the atmosphere and a range–frequency point-spread function. Its frequency marginal is a one-dimensional range kernel—the cross-correlation of the squared analysis window with the pulse intensity—that governs integrated power and any scalar gate-by-gate velocity estimator when the spectrum is symmetric. This window-completed kernel is wider than the pulse alone—by 31 % for the Gaussian pulse and Hann window examined here at Nw ≃ Nτ. When backscatter varies across the range gate (cloud edges, aerosol layers), the velocity is additionally weighted by the local carrier-to-noise ratio. When the spectrum is asymmetric (chirp) or multi-modal (severe turbulence, wakes, vortices), no one-dimensional weighting is sufficient and the full two-dimensional model is required, either for interpretation or for spectrogram inversion.
We tabulate each model with its assumptions and domain of validity, so that practitioners can pick the simplest model consistent with the observable of interest and identify when classical one-dimensional models silently fail.