Precipitation Nowcasting Based on Dual-Flow 3D Convolution and Monitoring Images

Suhui Yang; Zhiwei Lin; Shaojun Lai; Jinfu Liu

doi:10.3788/LOP57.201011

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 20 >Page 201011 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 20, 201011 (2020)

Precipitation Nowcasting Based on Dual-Flow 3D Convolution and Monitoring Images

Suhui Yang¹, Zhiwei Lin^{1、3、4、*}, Shaojun Lai², and Jinfu Liu^{1、5、6、**}

Author Affiliations

¹College of Computer and Information Sciences, Fujian Agriculture and Forestry University, Fuzhou, Fujian 350002, China

²Fuzhou Meteorological Bureau, Fuzhou, Fujian 350014, China

³College of Forestry, Fujian Agriculture and Forestry University, Fuzhou, Fujian 350002, China

⁴Forestry Post-Doctoral Station, Fujian Agriculture and Forestry University, Fuzhou, Fujian 350002, China

⁵Cross-Strait Nature Reserve Research Center, Fujian Agriculture and Forestry University, Fuzhou, Fujian 350002, China

⁶Key Laboratory of Fujian Universities for Ecology and Resource Statistics, Fuzhou, Fujian 350002, China

show less

DOI: 10.3788/LOP57.201011 Cite this Article Set citation alerts

Suhui Yang, Zhiwei Lin, Shaojun Lai, Jinfu Liu. Precipitation Nowcasting Based on Dual-Flow 3D Convolution and Monitoring Images[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201011 Copy Citation Text

show less

Fig. 1. Dual-flow 3D convolution neural network

Download full size

Fig. 2. Convolution operation

Download full size

Fig. 3. Max pooling operation

Download full size

Fig. 4. Fully connection operation

Download full size

Fig. 5. Prediction effect. (a)Proposed network; (b)ConvLSTM network; (c)C3D network; (d) Asy 3D network

Download full size

Fig. 6. Accuracy and loss value under training. (a) Accuracy; (b) loss value

Download full size

Fig. 7. Accuracy and loss value under test. (a) Accuracy; (b) loss value

Download full size

Fig. 8. Characteristic graphs under different depth convolution modules

Download full size

Model			A	B	C	D	E
Single- flow	Single loss	No pooling	√
Dual-flow	Single loss	Max pooling		√
		No pooling			√
	Dual loss	Max pooling				√
		average pooling					√
A/%			82.19	81.00	84.23	84.20	84.46

Table 1. Ablation experiment

P_I/(mL·h^-1)	Proposed network					ConvLSTM						C3D					Asy 3D							Samplesize
P_I/(mL·h^-1)	P_FAR /%↓	P_POD /%↑		P_CSI /%↑		P_FAR /%↓		P_POD /%↑		P_CSI /%↑		P_FAR /%↓		P_POD /%↑		P_CSI /%↑	P_FAR /%↓		P_POD /%↑		P_CSI /%↑			Samplesize
0	0.11		0.96		0.92		0.24		0.89		0.82		0.12		0.95			0.91		0.18		0.86	0.84	1227
1	0.19		0.78		0.79		0.42		0.61		0.59		0.18		0.74			0.78		0.31		0.67	0.68	501
2	0.22		0.72		0.75		0.35		0.47		0.55		0.29		0.72			0.72		0.43		0.62	0.60	261
3	0.22		0.76		0.77		0.35		0.61		0.63		0.29		0.73			0.72		0.36		0.69	0.67	193
4	0.24		0.76		0.76		0.37		0.57		0.60		0.27		0.72			0.73		0.32		0.51	0.59	103
5	0.17		0.70		0.76		0.30		0.51		0.59		0.26		0.70			0.72		0.36		0.43	0.52	69
6	0.03		0.69		0.81		0.32		0.40		0.51		0.13		0.62			0.72		0.38		0.60	0.61	42
7	0.23		0.59		0.67		0.27		0.28		0.41		0.29		0.56			0.63		0.39		0.44	0.51	39
8	0.24		0.62		0.68		0.11		0.38		0.53		0.36		0.67			0.65		0.54		0.52	0.49	21
9	0.26		0.56		0.64		0.10		0.36		0.51		0.18		0.56			0.67		0.47		0.40	0.45	25
10	0.21		0.55		0.65		0.14		0.30		0.44		0.17		0.50			0.62		0.33		0.90	0.77	20
11	0.21		0.81		0.80		0.11		0.30		0.44		0.16		0.78			0.81		0.27		0.41	0.52	27
12	0.20		0.80		0.80		0.00		0.20		0.33		0.00		0.80			0.89		0.33		0.40	0.50	5
13	0.00		0.71		0.83		0.00		0.43		0.60		0.00		0.57			0.73		0.00		0.71	0.83	7
14	0.33		0.89		0.76		0.33		0.44		0.53		0.25		0.67			0.71		0.33		0.44	0.53	9
15	0.00		0.33		0.50		0.00		0.33		0.50		0.00		0.33			0.50		0.25		0.50	0.60	6
16	1.00		0.00		0.00		1.00		0.00		0.00		1.00		0.00			0.00		0.00		1.00	1.00	1
17	0.00		0.80		0.89		0.00		0.40		0.57		0.00		0.80			0.89		1.00		0.00	0.00	5
18	0.00		0.75		0.86		0.00		0.50		0.67		0.00		0.75			0.86		0.00		0.50	0.67	4
19	0.00		0.43		0.60		0.00		0.29		0.44		0.25		0.43			0.55		0.00		0.86	0.92	7
2	00.25		0.86		0.80		0.38		0.71		0.67		0.33		0.86			0.75		0.00		0.43	0.60	7
21	0.00		1.00		1.00		0.33		0.33		0.44		0.29		0.83			0.77		0.20		0.67	0.73	6
22	0.00		1.00		1.00		0.00		1.00		1.00		0.00		1.00			1.00		0.00		1.00	1.00	2
24	0.00		0.50		0.67		0.00		0.50		0.67		0.00		0.50			0.67		0.00		1.00	1.00	2
25	0.50		1.00		0.67		0.00		1.00		1.00		0.00		1.00			1.00		0.00		1.00	1.00	1
27	0.00		1.00		1.00		0.00		0.33		0.50		0.25		1.00			0.86		0.00		0.67	0.80	3
29	1.00		0.00		0.00		1.00		0.00		0.00		1.00		0.00			0.00		1.00		0.00	0.00	1
32	0.00		1.00		1.00		0.00		1.00		1.00		0.00		1.00			1.00		0.00		1.00	1.00	1
34	1.00		0.00		0.00		1.00		0.00		0.00		1.00		0.00			0.00		1.00		0.00	0.00	1
38	1.00		0.00		0.00		1.00		0.00		0.00		1.00		0.00			0.00		0.00		1.00	1.00	1
46	1.00		0.00		0.00		1.00		0.00		0.00		1.00		0.00			0.00		0.00		1.00	1.00	1
56	0.00		1.00		1.00		1.00		0.00		0.00		0.00		1.00			1.00		0.00		0.50	0.67	2

Table 2. Comparison of test results of each precipitation intensity

Suhui Yang, Zhiwei Lin, Shaojun Lai, Jinfu Liu. Precipitation Nowcasting Based on Dual-Flow 3D Convolution and Monitoring Images[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201011

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information