Classification of Missed Layers in CALIPSO Products Based on U-Net Neural Network

Yilin Geng; Lin Zang; Feiyue Mao; Weiwei Xu; Wei Gong

doi:10.3788/AOS240893

Acta Optica Sinica, Volume. 44, Issue 24, 2401008(2024)

Classification of Missed Layers in CALIPSO Products Based on U-Net Neural Network

Yilin Geng¹, Lin Zang^2,3、*, Feiyue Mao¹, Weiwei Xu¹, and Wei Gong⁴

Author Affiliations

¹School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, Hubei , China

²Chinese Antarctic Center of Surveying and Mapping, Wuhan University, Wuhan 430079, Hubei , China

³Key Laboratory of Polar Environmental Monitoring and Public Governance (Wuhan University), Ministry of Education, Wuhan 430079, Hubei , China

⁴Electronic Information School, Wuhan University, Wuhan 430079, Hubei , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Cloud-aerosol classification flow chart of space-borne Lidar based on 2DMHT-UNet

Download full size

Fig. 2. Cases of daytime (2017-11-20T03:38:54ZD) and nighttime (2017-11-28T16:52:42ZN) cloud-aerosol classification over land. (a)(d) CALIPSO VFM cloud-aerosol classification; (b)(e) cloud-aerosol classification of U-Net classification model; (c)(f) difference between two classification results

Download full size

Fig. 3. Cases of daytime (2017-11-05T07:40:57ZD) and nighttime (2017-11-02T17:56:13ZN) cloud-aerosol classification over sea. (a)(d) CALIPSO VFM cloud-aerosol classification; (b)(e) cloud-aerosol classification of U-Net classification model; (c)(f) difference between two classification results

Download full size

$Case of 2DMHT-UNet cloud-aerosol classification over land (2017-11-20T03:38:54ZD). (a) Attenuation scattering ratio ASR; (b) CALIPSO VFM detection layer; (c) 2D-MHT detection layer; (d) differences between two layer detection algorithm; (e) Radar-Lidar joint detection cloud distribution; (f) CALIPSO VFM cloud-aerosol classification; (g) 2DMHT-UNet cloud-aerosol classification; (h) vertical distribution of cloud fraction$

Fig. 4. Case of 2DMHT-UNet cloud-aerosol classification over land (2017-11-20T03:38:54ZD). (a) Attenuation scattering ratio ASR; (b) CALIPSO VFM detection layer; (c) 2D-MHT detection layer; (d) differences between two layer detection algorithm; (e) Radar-Lidar joint detection cloud distribution; (f) CALIPSO VFM cloud-aerosol classification; (g) 2DMHT-UNet cloud-aerosol classification; (h) vertical distribution of cloud fraction

Download full size

$Case of 2DMHT-UNet cloud-aerosol classification over land (2017-11-05T07:40:57ZD). (a) Attenuation scattering ratio ASR; (b) CALIPSO VFM detection layer; (c) 2D-MHT detection layer; (d) differences between two layer detection algorithm; (e) Radar-Lidar joint detection cloud distribution; (f) CALIPSO VFM cloud-aerosol classification; (g) 2DMHT-UNet cloud-aerosol classification; (h) vertical distribution of cloud fraction$

Fig. 5. Case of 2DMHT-UNet cloud-aerosol classification over land (2017-11-05T07:40:57ZD). (a) Attenuation scattering ratio ASR; (b) CALIPSO VFM detection layer; (c) 2D-MHT detection layer; (d) differences between two layer detection algorithm; (e) Radar-Lidar joint detection cloud distribution; (f) CALIPSO VFM cloud-aerosol classification; (g) 2DMHT-UNet cloud-aerosol classification; (h) vertical distribution of cloud fraction

Download full size

Fig. 6. Verification of cloud classification accuracy of 2DMHT-UNet classification model based on Radar-Lidar joint observation. (a) Statistical comparison of cloud base height between CALIPSO VFM and Radar-Lidar over land; (b) statistical comparison of cloud base height between 2DMHT-UNet and Radar-Lidar over land; (c) statistical comparison of cloud base height between CALIPSO VFM and Radar-Lidar over sea; (d) statistical comparison of cloud base height between 2DMHT-UNet and Radar-Lidar over sea

Download full size

Table 1. Source and parameters of the product used in the study

View table

Table 1. Source and parameters of the product used in the study

Source	Product	Version	Parameter
CALIPSO	Level 1B Profile	4.10	Total attenuated backscatter at 532 nm， perpendicular attenuated backscatter at 532 nm， attenuated backscatter at 1064 nm， molecular number density， ozone number density
CALIPSO	Level 2 5 km VFM	4.20	Aerosol and cloud type
Radar-Lidar	2B-CLDCLASS-LIDAR	P1_R05	Cloud layer base height

Table 2. Performance evaluation of U-Net cloud-aerosol classification model

View table

Table 2. Performance evaluation of U-Net cloud-aerosol classification model

Performance	Type	Land			Sea
Performance	Type	Day	Night	All	Day	Night	All
$A$ /%	Comprehensive statistic	90.2	88.8	89.4	92.3	88.8	90.2
	November	90.0	88.7	89.3	92.0	89.2	90.4
	December	90.4	88.9	89.5	92.6	88.4	90.0
$P_{c}$ /%	Comprehensive statistic	94.4	91.2	92.5	94.2	92.5	93.1
	November	95.0	91.5	92.9	94.5	92.9	93.5
	December	93.9	91.0	92.2	94.0	92.1	92.7
$P_{a}$ /%	Comprehensive statistic	82.8	82.2	82.4	86.9	74.2	80.3
	November	81.6	81.5	81.5	85.0	74.3	79.8
	December	84.0	82.9	83.3	88.5	74.2	80.8
$R_{c}$ /%	Comprehensive statistic	88.5	88.5	88.5	92.8	91.5	92.0
	November	87.8	88.5	88.2	92.3	91.7	91.9
	December	89.1	88.5	88.7	93.3	91.4	92.0
$R_{a}$ /%	Comprehensive statistic	92.9	89.4	90.8	91.4	80.6	86.0
	November	93.4	89.1	90.9	91.5	81.6	86.8
	December	92.4	89.6	90.7	91.4	79.7	85.2

Tools

Get Citation

Copy Citation Text

Yilin Geng, Lin Zang, Feiyue Mao, Weiwei Xu, Wei Gong. Classification of Missed Layers in CALIPSO Products Based on U-Net Neural Network[J]. Acta Optica Sinica, 2024, 44(24): 2401008

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Atmospheric Optics and Oceanic Optics

Received: Apr. 23, 2024

Accepted: Jun. 18, 2024

Published Online: Dec. 16, 2024

The Author Email: Zang Lin (zanglin2018@whu.edu.cn)

DOI:10.3788/AOS240893

CSTR:32393.14.AOS240893

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology