Torgny Lindström, Matematisk statistik, Matematikcentrum, Lunds
universitet

Analysis of LIDAR measurements using nonparametric kernel regression
methods

Abstract

The LIDAR (LIght Detection And Ranging) technique is an efficient tool
for remote monitoring of   the distribution of a number of atmospheric
species. We study  measurements of SO2 emitted from the Italian vulcano
Etna. In  this study we focus on the treatment of data and on the
procedure to   evaluate range resolved concentrations. In order to make
an in-depth   analysis we prepared the lidar system to store
measurements of   individual backscattered laser pulses. Utilizing these
repeated  mesurements we compare three different methods to average
the   returned signals.

In the evaluation process we use local polynomial regression to
estimate the range resolved concentrations. Here we calculate optimal
bandwidths based on the empirical-bias bandwidth selector (EBBS). We
also compare two different variance estimators for the path-integrated
curves: local polynomial variance estimation and variance estimation
based on Taylor approximations.