Detection of particle layers in backscatter profiles: application to Antarctic lidar measurements
Abstract. A detection method is proposed and studied to infer the presence of hidden signals in a statistical way. It is applied here to the detection of Polar Stratospheric Cloud (PSC) layers in lidar backscatter profiles measured over the Dumont D'Urville station (Antarctica). PSCs appear as layers with enhanced variance in non stationary, heteroscedastic signal profiles, between two unknown altitudes to be estimated. The method is based on a three step algorithm. The first step is the stationarization of the signal, the second performs the maximum likelihoods estimation of the signal (PSC altitude range and variance inside and outside the PSC layer). The last step uses a Fisher-Snédécor test to decide whether the detection of PSC layer is statistically significant. Performances and robustness of the method are tested on simulated data with given statistical properties. Bias and detection limit are estimated. The method is then applied to lidar backscatter profiles measured in 2008. No PSC are detected during seasons when PSCs are not expected to form. As expected, PSC layers are detected during the austral winter and early spring. The effect of time averaging of the profiles is investigated. The best compromise for detection of PSC layers in lidar backscatter profiles acquired at Dumont D'Urville is a time averaging window of 1 h typically.