Sent2Agri System Based Crop Type Mapping in Yellow River Irrigation Area

doi:10.11947/j.JGGS.2020.0411

Abstract

Abstract:

Agricultural monitoring is essential for adequate management of food production and distribution. Crop land and crop type classification, using remote sensing time series, form an important tool to capture the agricultural production information. The recently launched Sentinel-2 satellites provide unprecedented monitoring capacities in terms of spatial resolution, swath width, and revisit frequency. The Sentinel-2 for Agriculture (Sen2-Agri) system has been developed to fully exploit those capacities, by providing four relevant earth observation products for agricultural monitoring. Under the Dragon 4 Program, the crop mapping with various satellite images and a specific focus on the Yellow River irrigated agricultural area in the Ningxia Hui Autonomous Region in China was carried out with the Sentinel-2 for Agriculture system (Sent2Agri). 9 types of crops were classified and the crop type map in 2017 was produced based on 35 scenes Sentinel 2A/B images. The overall accuracy computed from the error confusion matrix is 88%, which includes the cropped and uncropped types. After the removal of the uncropped area, the overall accuracy for a cropped decrease to 73%. In order to further improve the crop classification accuracy, the training dataset should be further improved and tuned.

Key words: crop mapping; Dragon Program; Sentinel 2; Sent2Agri system

Jinlong FAN,Pierre DEFOURNY,Qinghan DONG,Xiaoyu ZHANG,Mathilde De VROEY,Nicolas BELLEMANS,Qi XU,Qiliang LI,Lei ZHANG,Hao GAO. Sent2Agri System Based Crop Type Mapping in Yellow River Irrigation Area[J]. Journal of Geodesy and Geoinformation Science, 2020, 3(4): 110-117.

Figures/Tables 10

Fig.1

Tab.1

Fig.2

Fig.3

Fig.4

Tab.2

Fig.5

Tab.3

Tab.4

Fig.6

References 16

[1]	REYNOLDS C A, YITAYEW M, SLACK D C, et al. Estimating crop yields and production by integrating the FAO Crop Specific Water Balance model with real-time satellite data and ground-based ancillary data[J]. International Journal of Remote Sensing, 2000,21(18):3487-3508.
[2]	LIU M W, OZDOGAN M, ZHU X. Crop type classification by simultaneous use of satellite images of different resolutions[J]. IEEE Transactions on Geoscience & Remote Sensing, 2014,52(6):3637-3649.
[3]	BADHWAR G B. Automatic corn-soybean classification using Landsat MSS data. II. Early season crop proportion estimation[J]. Remote Sensing of Environment, 1984,14(1-3):31-37.
[4]	SAKAMOTO T, GITELSON A A, ARKEBAUER T J. Near real-time prediction of U.S. corn yields based on time-series MODIS data[J]. Remote Sensing of Environment, 2014,147(147):219-231.
[5]	LUNETTA R S, JOHNSON D M, LYON J G, et al. Impacts of imagery temporal frequency on land-cover change detection monitoring[J]. Remote Sensing of Environment, 2004,89(4):444-454.
[6]	CARRO H, GONALVES P, CAETANO M. Contribution of multispectral and multitemporal information from MODIS images to land cover classification[J]. Remote Sensing of Environment, 2008,112(3):986-997.
[7]	BAUER M E, CIPRA J E, ANUTA P E, et al. Identification and area estimation of agricultural crops by computer classification of Landsat MSS data[J]. Remote Sensing of Environment, 1979,8(1):77-92.
[8]	OSMAN J, INGLADA J, DEJOUX J F. Assessment of a Markov logic model of crop rotations for early crop mapping[J]. Computers and Electronics in Agriculture, 2015,113:234-243.
[9]	GÓMEZ C, WHITE J C, WULDER M A. Optical remotely sensed time series data for land cover classification: A review[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016,116:55-72.
[10]	XIONG J, THENKABAIL P S, GUMMA M K, et al. Automated cropland mapping of continental Africa using Google Earth Engine cloud computing[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2017,126:225-244.
[11]	SENTINEL E S A. User Handbook[J]. ESA Standard Document, 2, 64.
[12]	MARTINS V S, BARBOSA C C F, DE CARVALHO L A S, et al. Assessment of atmospheric correction methods for Sentinel-2 MSI images applied to Amazon floodplain lakes[J]. Remote Sensing, 2017,9(4):322.
[13]	MANNHEIMER H. Landsat missions[J]. 1978.
[14]	ESA. Sentinel-2 for Agriculture [EB/OL]. [ 2015- 09- 18]. http://www.esa-sen2agri.org.
[15]	MCINERNEY D, KEMPENEERS P. Orfeotoolbox[M] //Open Source Geospatial Tools. Springer, Cham, 2015: 199-217.
[16]	ESA. ESA Sentinels—Sentinel-2: operations ramp-up phase[EB/OL].[2015-08-21]. https://sentinel.esa.int/web/sentinel/missions/sentinel-2/operations-ramp-up-phase.

Band number	Central wavelength /nm	Bandwidth /nm	Lref (reference radiance) (Wm^-2 sr^-1 μm^-1)	SNR@ Lref	Resolution /m
1	443	20	129	129	60
2	490	65	128	154	10
3	560	35	128	168	10
4	665	30	108	142	10
5	705	15	74.5	117	20
6	740	15	68	89	20
7	783	20	67	105	20
8	842	115	103	172	10
8a	865	20	52.5	72	20
9	945	20	9	114	60
10	1375	30	6	50	60
11	1610	90	4	100	20
12	2190	180	1.5	100	20

		Samples		Pixels(10m)
	Land Cover	Training	Validation		Training	Validation
Crops	fodder	19	39	433		1054
	Wheat	40	41	1241		1274
	Rice	121	111	3983		3697
	Corn	101	92	3925		3698
	Grapes	25	38	1397		2028
	Cabbage	8	12	208		336
	Tomatoes	3	4	53		70
	Watermelons	3	4	86		136
	Medlar	3	5	106		242
No crops	Plantations	67	60	2964		2273
	Grassland	6	5	7068		4700
	Forests	11	9	4054		1883
	Bare Soils	33	42	1576		5081
	Build-up	90	74	5167		5081
	Water bodies	43	35	3367		1904
	Total	573	571	35628		34424

Classified
		j=1	j=2	…	j=J
Reference	i=1	n₁₁	n₁₂		n₁_J
	i=2	n₂₁	n₂₂		n₂_J
	…
	i=J	n_J₁	n_J₂		n_JJ

Classified
	Fodder	Wheat	Maize	Rice	Grape	Cabbage	Tomato	Watermelon	Medlar	No Crop	Precision
Fodder	815	3	17	22	0	2	0	1	0	60	0.89
Wheat	16	678	22	114	29	6	0	0	6	298	0.58
Maize	45	264	3170	184	294	84	56	10	29	259	0.72
Rice	89	12	196	3334	0	13	4	0	0	44	0.90
Grape	2	17	21	0	942	0	0	1	3	755	0.54
Cabbage	2	2	1	0	0	214	0	48	0	17	0.75
Tomato	25	0	1	0	0	0	0	0	0	10	0.00
Watermelon	0	0	9	0	0	0	0	0	0	76	0.00
Medlar	0	0	1	0	8	0	0	0	0	204	0.00
No crop	6	84	387	114	58	35	0	0	4	21201	0.00
Recall	0.82	0.64	0.83	0.88	0.71	0.60	0.00	0.00	0.00	0.92
F1-score	0.85	0.61	0.77	0.89	0.61	0.67	0.00	0.00	0.00	0.95