Patching C n²time series data holes using principal component analysis

Abstract

Measurements of C 2 n time series using unattended commericial scintillometers over long time intervals inevitably lead to data drop-outs or degraded signals. We present a method using Principal Component Analysis (also known as Karhunen-Loève decomposition) that seeks to correct for these event-induced and mechanicallyinduced signal degradations. We report on the quality of the correction by examining the Intrinsic Mode Functions generated by Empirical Mode Decomposition.