Having large velocity gradients in the atmosphere can induce large errors in Jmean. Typically when Vz changes per grid cell more than the Doppler with of the line, errors will appear. I have added velocity gradients into the depth-optimization procedure. Also, sometimes the depth-optimization does not suffice to converge in some pixels. Smoothing the detected gradients in the optimization routine can help. I have added a new keyword in keyword.input called 15D_SMOOTH_OPTIMIZED_GRADIENTS=integer that defines the width of the smoothing window (zero by default).
Having large velocity gradients in the atmosphere can induce large errors in Jmean. Typically when Vz changes per grid cell more than the Doppler with of the line, errors will appear. I have added velocity gradients into the depth-optimization procedure. Also, sometimes the depth-optimization does not suffice to converge in some pixels. Smoothing the detected gradients in the optimization routine can help. I have added a new keyword in keyword.input called 15D_SMOOTH_OPTIMIZED_GRADIENTS=integer that defines the width of the smoothing window (zero by default).