Longlong Wang

Biomass burning smoke (BBS) emitted from Southeast Asia can be transported through regional atmospheric circulation, exerting significant impacts on air quality and climate over southwestern China. However, continuous observations of the vertical structures and optical properties of transported BBS remain limited. This study used one year of continuous multi-wavelength polarization Raman lidar observations at Mengzi (23.37°N, 103.38°E) in Yunnan, China, to identify transboundary BBS transport events and characterize their optical properties. Eight transboundary BBS events originating from Southeast Asia were identified using a comprehensive data processing framework involving aerosol extinction coefficient, backscatter coefficient, lidar ratio, particle linear depolarization ratio (PDR), and Ångström exponent. The transported smoke layers were mainly distributed below 4 km, exhibiting significantly enhanced aerosol extinction coefficients of 0.22–0.32 km⁻¹ (compared with 0.02–0.06 km⁻¹ for background aerosols). The 532 nm lidar ratios of the smoke layers (49.0–68.3 sr) were higher than those of the background aerosols (31.6–39.0 sr). The particle linear depolarization ratios ranged from 0.050 to 0.097 at 532 nm and from 0.14 to 0.19 at 355 nm, revealing a pronounced wavelength dependence. These values indicate measurable particle depolarization but do not independently establish particle morphology or the degree of smoke aging. Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) backward trajectory analysis together with fire hotspot observations revealed that this smoke mainly originated from biomass burning regions over the Indochina Peninsula and was transported to Yunnan across northern Myanmar, Laos, and northern Vietnam. This study provides important observational evidence of transport pathways, vertical distribution, and optical properties of transboundary BBS aerosols transported from Southeast Asia to southwest China.

Cloud-layer identification is a key prerequisite for automated ground-based lidar processing, but remains challenging in Raman lidar networks because of incomplete near-range overlap, weak high-level cloud echoes, day-night signal-to-noise ratio (SNR) contrasts, energy fluctuations, and aerosol-cloud ambiguity. Leveraging the China Aerosol Raman Lidar Network (CARLNET), we propose a multi-channel cloud-layer identification framework that operates on 355/532/1064 nm elastic signals and 355/532 nm volume depolarization ratios, reducing dependence on absolute calibration and retrieved optical products. The framework combines an anisotropic encoder-decoder architecture with a three-stage training strategy, including self-supervised pretraining, traditional-algorithm-guided probabilistic distillation, and fine-tuning with limited expert refinement. Strict cross-site and cross-time evaluation on held-out sites shows that, without using any test-site observations during pretraining or test-site annotations during fine-tuning, the framework achieves an F1 score of 0.937. Compared with direct supervised training, the pretrained model yields comparable cloud-base accuracy (MAE: 59.1 versus 55.3 m) while reducing the detectable cloud-top MAE from 94.0 to 82.2 m. It also substantially outperforms the conventional VDE baseline in both cloud-base and cloud-top estimation. Independent evaluation based on a total of 47 matched lidar–radiosonde cases aggregated across the three held-out sites yields a structural consistency rate of 38.3 % with radiosonde-derived moist layers, compared with 34.0 % for VDE, providing an independent structural-consistency assessment across the two observing systems. These results demonstrate a cloud-layer identification technique that is transferable across held-out CARLNET sites and reduces dependence on absolute calibration, supporting network-scale Raman lidar processing.

The planetary boundary layer height (PBLH) is a critical physical parameter for characterizing atmospheric dynamics, thermodynamic structures, and pollutant dispersion. LiDAR, as a high-spatiotemporal-resolution remote sensing tool, is widely employed to retrieve PBLH by analyzing backscatter signals that reflect aerosol concentrations. However, in complex terrains such as valleys, the boundary layer structure exhibits significant non-homogeneity and complexity due to local terrain-induced dynamic effects. This complexity renders traditional retrieval methods susceptible to interference from multi-layer aerosols and low signal-to-noise ratios (SNR). To address these challenges, this paper proposes an automated identification method—the Physics-Constrained Adaptive K-means (PCA-Kmeans) clustering analysis—specifically tailored for complex valley environments. The PCA-Kmeans framework first constructs a multi-feature scoring mechanism—incorporating signal intensity, vertical gradients, temporal continuity, and vertical compactness—to apply physical constraints for screening clustering results. Subsequently, a gradient-weighted selection and median smoothing strategy are introduced to pinpoint the PBLH from the optimal cluster, effectively mitigating the interference of complex stratification induced by valley topography. The proposed method was validated using LiDAR observation data from the Vipava Valley, Slovenia, from 2018 to 2019. Using Radiosonde (RS) measurements as the reference benchmark, a comparative analysis was conducted between the PCA-Kmeans method, the Gradient Method (GRD), and the Standard Deviation Method (STD). The results indicate that in valley atmospheres influenced by thermally driven local circulations, the PCA-Kmeans method demonstrates superior robustness under low SNR and multi-layer aerosol conditions. The inversion results show higher consistency with RS data, effectively overcoming the identification biases of traditional algorithms caused by residual aerosol layers. This study provides a reliable and automated technical solution for the long-term monitoring of valley boundary layers under the influence of complex local circulations.

Accurate constraints on aerosol complex refractive index (CRI), single scattering albedo (SSA), and lidar ratio (LR) are essential for improving lidar-based aerosol retrievals and reducing radiative-effect uncertainties. However, consistent derivation from in situ observations remains challenging in complex mixed-aerosol environments, where size distribution, absorption, and humidity effects are strongly coupled. Here we present a decoupled inversion framework that integrates SMPS–OPC size distributions, AE33-derived absorption coefficients, and ambient relative humidity observations from the Vipava Valley (Slovenia) in April 2016 to derive dry-state homogeneous-equivalent CRI and RH-corrected lidar-relevant SSA and LR at 355, 532, and 1064 nm. The framework reconstructs a continuous particle size distribution through instrument-response-aware SMPS–OPC geometric alignment and retrieves the reference real refractive index. AE33-derived absorption constraints and a singly subtractive Kramers–Kronig (SSKK) relation are then used to derive a spectrally consistent dry state CRI, after which SSA and LR are recalculated through an RH-corrected wet-state forward optical calculation. The retrieved dry-state mean n values were 1.448, 1.441, and 1.437, and the mean k values were 0.0312, 0.0238, and 0.0192 at 355, 532, and 1064 nm, respectively. The RH-corrected mean SSA values were 0.866, 0.852, and 0.774, and the corresponding mean LR values were 86.4, 58.4, and 39.2 sr. Sensitivity tests showed that the AE33 effective multiple-scattering correction affects absorption-sensitive products, with C_eff = 5 used as a site- and period-specific baseline. The framework provides a practical pathway for linking dry in situ CRI retrievals with RH-corrected multi-wavelength lidar-relevant aerosol optical products.

Accurate constraints on aerosol complex refractive index (CRI), single scattering albedo (SSA), and lidar ratio (LR) are essential for improving lidar-based aerosol retrievals and reducing radiative-effect uncertainties. However, consistent derivation from in situ observations remains challenging in complex mixed-aerosol environments, where size distribution, absorption, and humidity effects are strongly coupled. Here we present a decoupled inversion framework that integrates SMPS–OPC size distributions, AE33-derived absorption coefficients, and ambient relative humidity observations from the Vipava Valley (Slovenia) in April 2016 to derive dry-state homogeneous-equivalent CRI and RH-corrected lidar-relevant SSA and LR at 355, 532, and 1064 nm. The framework reconstructs a continuous particle size distribution through instrument-response-aware SMPS–OPC geometric alignment and retrieves the reference real refractive index. AE33-derived absorption constraints and a singly subtractive Kramers–Kronig (SSKK) relation are then used to derive a spectrally consistent dry state CRI, after which SSA and LR are recalculated through an RH-corrected wet-state forward optical calculation. The retrieved dry-state mean n values were 1.448, 1.441, and 1.437, and the mean k values were 0.0312, 0.0238, and 0.0192 at 355, 532, and 1064 nm, respectively. The RH-corrected mean SSA values were 0.866, 0.852, and 0.774, and the corresponding mean LR values were 86.4, 58.4, and 39.2 sr. Sensitivity tests showed that the AE33 effective multiple-scattering correction affects absorption-sensitive products, with C_eff = 5 used as a site- and period-specific baseline. The framework provides a practical pathway for linking dry in situ CRI retrievals with RH-corrected multi-wavelength lidar-relevant aerosol optical products.

Accurate constraints on aerosol complex refractive index (CRI), single scattering albedo (SSA), and lidar ratio (LR) are essential for improving lidar-based aerosol retrievals and reducing radiative-effect uncertainties. However, consistent derivation from in situ observations remains challenging in complex mixed-aerosol environments, where size distribution, absorption, and humidity effects are strongly coupled. Here we present a decoupled inversion framework that integrates SMPS–OPC size distributions, AE33-derived absorption coefficients, and ambient relative humidity observations from the Vipava Valley (Slovenia) in April 2016 to derive dry-state homogeneous-equivalent CRI and RH-corrected lidar-relevant SSA and LR at 355, 532, and 1064 nm. The framework reconstructs a continuous particle size distribution through instrument-response-aware SMPS–OPC geometric alignment and retrieves the reference real refractive index. AE33-derived absorption constraints and a singly subtractive Kramers–Kronig (SSKK) relation are then used to derive a spectrally consistent dry state CRI, after which SSA and LR are recalculated through an RH-corrected wet-state forward optical calculation. The retrieved dry-state mean n values were 1.448, 1.441, and 1.437, and the mean k values were 0.0312, 0.0238, and 0.0192 at 355, 532, and 1064 nm, respectively. The RH-corrected mean SSA values were 0.866, 0.852, and 0.774, and the corresponding mean LR values were 86.4, 58.4, and 39.2 sr. Sensitivity tests showed that the AE33 effective multiple-scattering correction affects absorption-sensitive products, with C_eff = 5 used as a site- and period-specific baseline. The framework provides a practical pathway for linking dry in situ CRI retrievals with RH-corrected multi-wavelength lidar-relevant aerosol optical products.

Cloud-layer identification is a key prerequisite for automated ground-based lidar processing, but remains challenging in Raman lidar networks because of incomplete near-range overlap, weak high-level cloud echoes, day–night signal-to-noise ratio (SNR) contrasts, energy fluctuations, and aerosol–cloud ambiguity. Leveraging the China Aerosol Raman Lidar NETwork (CARLNET), we propose a multi-channel cloud-layer identification framework that operates on 355/532/1064 nm elastic signals and 355/532 nm volume depolarization ratio, reducing dependence on absolute calibration and retrieved optical products. The framework combines an anisotropic encoder–decoder architecture with a three-stage training strategy, including self-supervised pretraining, traditional-algorithm-guided probabilistic distillation, and fine-tuning with limited expert refinement. Strict cross-site and cross-time evaluation on held-out sites shows that, without using any test-site labels, the framework achieves an F1 score of 0.9371 and reduces the mean absolute errors of cloud-base and cloud-top heights to 113 m and 213 m, respectively, outperforming training without pretraining by approximately 50 m and 70 m and the conventional baseline for cloud-top height by about 400 m. A labeled-data-size ablation further shows improved label efficiency, with near-plateau performance reached at about 40 labeled days. Consistency checks against co-located radiosonde moist-layer indications support the realism of identified high-level weak-echo clouds and the robustness of cross-site deployment. These results demonstrate a transferable and calibration-decoupled cloud-layer identification technique for network-scale Raman lidar processing.

Cloud-layer identification is a key prerequisite for automated ground-based lidar processing, but remains challenging in Raman lidar networks because of incomplete near-range overlap, weak high-level cloud echoes, day–night signal-to-noise ratio (SNR) contrasts, energy fluctuations, and aerosol–cloud ambiguity. Leveraging the China Aerosol Raman Lidar NETwork (CARLNET), we propose a multi-channel cloud-layer identification framework that operates on 355/532/1064 nm elastic signals and 355/532 nm volume depolarization ratio, reducing dependence on absolute calibration and retrieved optical products. The framework combines an anisotropic encoder–decoder architecture with a three-stage training strategy, including self-supervised pretraining, traditional-algorithm-guided probabilistic distillation, and fine-tuning with limited expert refinement. Strict cross-site and cross-time evaluation on held-out sites shows that, without using any test-site labels, the framework achieves an F1 score of 0.9371 and reduces the mean absolute errors of cloud-base and cloud-top heights to 113 m and 213 m, respectively, outperforming training without pretraining by approximately 50 m and 70 m and the conventional baseline for cloud-top height by about 400 m. A labeled-data-size ablation further shows improved label efficiency, with near-plateau performance reached at about 40 labeled days. Consistency checks against co-located radiosonde moist-layer indications support the realism of identified high-level weak-echo clouds and the robustness of cross-site deployment. These results demonstrate a transferable and calibration-decoupled cloud-layer identification technique for network-scale Raman lidar processing.

Cloud-layer identification is a key prerequisite for automated ground-based lidar processing, but remains challenging in Raman lidar networks because of incomplete near-range overlap, weak high-level cloud echoes, day–night signal-to-noise ratio (SNR) contrasts, energy fluctuations, and aerosol–cloud ambiguity. Leveraging the China Aerosol Raman Lidar NETwork (CARLNET), we propose a multi-channel cloud-layer identification framework that operates on 355/532/1064 nm elastic signals and 355/532 nm volume depolarization ratio, reducing dependence on absolute calibration and retrieved optical products. The framework combines an anisotropic encoder–decoder architecture with a three-stage training strategy, including self-supervised pretraining, traditional-algorithm-guided probabilistic distillation, and fine-tuning with limited expert refinement. Strict cross-site and cross-time evaluation on held-out sites shows that, without using any test-site labels, the framework achieves an F1 score of 0.9371 and reduces the mean absolute errors of cloud-base and cloud-top heights to 113 m and 213 m, respectively, outperforming training without pretraining by approximately 50 m and 70 m and the conventional baseline for cloud-top height by about 400 m. A labeled-data-size ablation further shows improved label efficiency, with near-plateau performance reached at about 40 labeled days. Consistency checks against co-located radiosonde moist-layer indications support the realism of identified high-level weak-echo clouds and the robustness of cross-site deployment. These results demonstrate a transferable and calibration-decoupled cloud-layer identification technique for network-scale Raman lidar processing.