Atmospheric Boundary Layer Height Monitoring Using a Kalman Filter and Backscatter Lidar Returns

Abstract

A solution based on a Kalman filter to trace the evolution of the atmospheric boundary layer (ABL) sensed by a ground-based elastic-backscatter tropospheric lidar is presented. An erf-like profile is used to model the mixing-layer top and the entrainment-zone thickness. The extended Kalman filter (EKF) enables to retrieve and track the ABL parameters based on simplified statistics of the ABL dynamics and of the observation noise present in the lidar signal. This adaptive feature permits to analyze atmospheric scenes with low signal-to-noise ratios (SNRs) without the need to resort to long-time averages or range-smoothing techniques, as well as to pave the way for future automated detection solutions. First, EKF results based on oversimplified synthetic and experimental lidar profiles are presented and compared with classic ABL estimation quantifiers for a case study with different SNR scenarios.Index Terms-Adaptive kalman filtering, laser radar, remote sensing, signal processing.

Performance Evaluation of the Boundary-Layer Height from Lidar and the Weather Research and Forecasting Model at an Urban Coastal Site in the North-East Iberian Peninsula

“…The technique builds upon previous work (Rocadenbosch et al 1998(Rocadenbosch et al , 1999. Lange et al (2013) found that the main advantages of the extended Kalman filter are the ability to time-track the PBL height without need for long time averaging and range smoothing and the ability to perform well under low signal-to-noise ratio. The extended Kalman filter technique benefits from the knowledge of past PBL height estimates and statistical covariance information to predict present-time estimates.…”

“…An adaptive approach utilizing an extended Kalman filter (Brown and Hwang 1982) has been developed and tested in the UPC Remote Sensing Laboratory to trace the evolution of the PBL (Lange et al 2013). The technique builds upon previous work (Rocadenbosch et al 1998(Rocadenbosch et al , 1999.…”

“…Visual inspection methods (Quan et al 2013) are infrequently used as a subjective approach, which adds levels of ambiguity that may possibly lead to poor results. Finally, an objective approach using an adaptive extended Kalman filter has recently been developed and tested (Lange et al 2013).…”

Mixing layer height as an indicator for urban air quality?

“…The same is true for sophisticated multi-wavelength lidars (e.g., Baars et al, 2008), sodars (e.g., Beyrich, 1995), combinations of instruments (e.g., Cohn and Angevine, 2000) and combinations of models and measurements (e.g., Bachtiar et al, 2014). A large number of studies relying on lidar data have been published introducing different methodologies to determine MLH: among others, Endlich et al (1979) and Flamant et al (1997) used algorithms based on first derivatives of the backscatter signal; Menut et al (1999) used second derivatives; Hooper and Eloranta (1986) the temporal variance; Cohn and Angevine (2000), Brooks (2003) and Baars et al (2008) applied wavelet covariance transforms; de Bruine et al (2017) used graph theory, Caicedo et al (2017) cluster analysis; and statistical methods were used by Eresmaa et al (2006) and Lange et al (2014). With recent upgrades of the hardware those methodologies can also be applied to ALC, and with the implementation of networks they have become more attractive as they provide continuous monitoring and good spatial coverage.…”

Mixing Layer Height Retrievals From MiniMPL Measurements in the Chiang Mai Valley: Implications for Particulate Matter Pollution

“…A variety of techniques for ML height detection has been suggested, which includes the application of a gradient technique on the range corrected signal (RCS) (Flamant et al, 1997;García-Franco et al, 2018) and the logarithm of the RCS (He et al, 2006), maximum of the RCS temporal variance (Hooper and Eloranta, 1986), wavelet analysis of the RCS profile (Cohn and Angevine, 2000), extended Kalman filter method (Lange et al, 2014), and the cubic root gradient of the RCS profile (Yang et al, 2017). Although fully automated, ML height detection is restricted by the limitations of various techniques Ware et al, 2016) in clearly identifying different features in the lower troposphere.…”