基于3种机器学习方法的农业干旱监测比较

doi:10.13866/j.azr.2022.01.31

摘要/Abstract

摘要：

频繁的旱灾对甘肃省经济、农业生产造成严重危害,因此利用先进方法建立准确可靠的农业干旱监测模型对该省防旱减灾十分重要。本文基于随机森林（RF）、BP神经网络和支持向量机（SVM）3种机器学习方法,利用甘肃省2002—2019年4—10月多源遥感数据得到的植被状态指数（VCI）、温度状态指数（TCI）、植被供水指数（VSWI）、降水状态指数（PCI）以及DEM、土壤有效含水量（AWC）和气候类型作为自变量,气象站点以3个月时间尺度的标准化降水蒸发指数（SPEI_3）为因变量,构建3种不同的农业干旱监测模型,分析比较出适用于监测甘肃省农业干旱的最佳模型,同时进一步探究了机器学习方法构建的模型在不同环境下的适用性。结果表明：构建的3种机器学习模型中,随机森林模型的R²平均值高（0.86）且误差小（RMSE为0.40,MAE为0.31）,农业干旱的监测效果要优于BP神经网络和支持向量机模型;干燥和湿润两种环境下分别构建的3种机器学习模型在湿润环境中监测能力表现均更优异（R²>0.82）,而随机森林模型在两种环境中监测干旱的表现比其他两种模型强。研究结果为甘肃省的农业干旱监测与评估提供了新的科学方法,对农业干旱研究具有重要意义。

关键词: 农业干旱, 机器学习, SPEI, MODIS

Abstract:

Frequent droughts have caused serious harm to the economy and agricultural production of Gansu Province. Therefore, it is very important for the province to use advanced methods to establish accurate and reliable agricultural drought monitoring models. Three machine learning methods (random forest, BP neural network, support vector machine) are used to construct an agricultural drought monitoring model using remote sensing factors and meteorological factors. Analysis of the best model for agricultural drought monitoring in Gansu Province. At the same time, the applicability of the model constructed by machine learning in different environments was further explored. The results show that among the models constructed by the three machine learning methods, the RF model has a high coefficient of determination (0.86) and small errors (RMSE 0.40, MAE 0.31). The monitoring effect of agricultural drought is better than the model constructed by BP and SVM; The three machine learning method models constructed in the two environments, respectively, performed better in the wet environment, while the model constructed by the random forest method performed better than the other two models in monitoring drought in the two environments. The research results provide a new scientific method for agricultural drought monitoring and evaluation in Gansu Province, and are of great significance to agricultural drought research.

Key words: agricultural drought, machine learning, SPEI, MODIS

王晓燕,李净,邢立亭. 基于3种机器学习方法的农业干旱监测比较[J]. 干旱区研究, 2022, 39(1): 322-332.

WANG Xiaoyan,LI Jing,XING Liting. Comparative agricultural drought monitoring based on three machine learning methods[J]. Arid Zone Research, 2022, 39(1): 322-332.

图/表 9

表1

干旱指数计算方法"

干旱指数	公式	文献来源
VCI植被状态指数	$(NDV I i - NDV I min) / (NDV I max - NDV I min)$	[22]
TCI温度状态指数	$(LS T max - LS T i) / (LS T max - LS T min)$	[22]
PCI降水指数	$(TRM M i - TRM M min) / (TRM M max - TRM M min)$	[22]
VSWI植被供水指数	$EVI / LST$	[23]

表1

图1

表2

图2

表3

3种机器学习方法对验证数据拟合结果统计"

月份	样本	随机森林			BP神经网路			支持向量机
月份	样本	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE
4	1	0.78	0.79	0.66	0.70	1.00	0.79	0.73	0.81	0.66
	2	0.91	0.43	0.29	0.86	0.45	0.32	0.83	0.50	0.37
	3	0.84	0.50	0.39	0.80	0.58	0.42	0.85	0.49	0.39
	4	0.86	0.56	0.42	0.82	0.53	0.39	0.85	0.49	0.39
	5	0.90	0.36	0.31	0.87	0.40	0.32	0.87	0.37	0.31
	平均值	0.86	0.53	0.41	0.81	0.59	0.45	0.82	0.53	0.42
5	1	0.84	0.56	0.40	0.80	0.68	0.53	0.82	0.58	0.40
	2	0.95	0.33	0.23	0.91	0.34	0.24	0.91	0.35	0.24
	3	0.85	0.37	0.28	0.81	0.44	0.31	0.84	0.42	0.30
	4	0.88	0.36	0.28	0.81	0.47	0.37	0.84	0.38	0.27
	5	0.82	0.37	0.31	0.77	0.41	0.33	0.78	0.41	0.31
	平均值	0.87	0.40	0.30	0.82	0.47	0.36	0.84	0.43	0.30
6	1	0.80	0.38	0.29	0.77	0.39	0.30	0.76	0.40	0.32
	2	0.91	0.24	0.18	0.87	0.29	0.22	0.91	0.25	0.19
	3	0.87	0.23	0.18	0.85	0.35	0.28	0.85	0.27	0.22
	4	0.85	0.31	0.26	0.82	0.29	0.23	0.82	0.38	0.32
	5	0.82	0.36	0.29	0.82	0.35	0.27	0.79	0.35	0.28
	平均值	0.85	0.30	0.24	0.83	0.33	0.26	0.83	0.33	0.26
7	1	0.86	0.36	0.26	0.84	0.39	0.30	0.82	0.51	0.39
	2	0.90	0.38	0.27	0.85	0.41	0.31	0.87	0.39	0.29
	3	0.81	0.42	0.32	0.78	0.43	0.34	0.81	0.41	0.35
	4	0.85	0.51	0.37	0.80	0.48	0.36	0.81	0.48	0.40
	5	0.88	0.36	0.28	0.83	0.42	0.33	0.85	0.41	0.33
	平均值	0.86	0.41	0.30	0.82	0.43	0.33	0.83	0.44	0.35
8	1	0.85	0.49	0.39	0.81	0.52	0.40	0.83	0.50	0.45
	2	0.88	0.40	0.29	0.80	0.47	0.35	0.81	0.48	0.38
	3	0.90	0.34	0.29	0.87	0.38	0.31	0.88	0.36	0.29
	4	0.87	0.40	0.36	0.85	0.45	0.36	0.83	0.50	0.41
	5	0.89	0.39	0.29	0.82	0.48	0.39	0.86	0.43	0.29
	平均值	0.88	0.40	0.32	0.83	0.46	0.36	0.84	0.45	0.36
9	1	0.85	0.44	0.33	0.82	0.48	0.35	0.80	0.53	0.41
	2	0.84	0.45	0.34	0.80	0.46	0.31	0.82	0.46	0.33
	3	0.89	0.32	0.23	0.87	0.35	0.26	0.90	0.32	0.21
	4	0.89	0.34	0.28	0.82	0.43	0.32	0.86	0.42	0.31
	5	0.83	0.44	0.32	0.81	0.46	0.36	0.80	0.50	0.38
	平均值	0.86	0.40	0.30	0.82	0.44	0.32	0.84	0.45	0.33
10	1	0.84	0.44	0.32	0.76	0.54	0.44	0.83	0.46	0.35
	2	0.90	0.39	0.28	0.83	0.46	0.34	0.86	0.42	0.32
	3	0.87	0.36	0.30	0.83	0.49	0.38	0.85	0.37	0.28
	4	0.89	0.36	0.29	0.85	0.39	0.31	0.87	0.41	0.32
	5	0.86	0.37	0.26	0.82	0.42	0.33	0.84	0.41	0.31
	平均值	0.87	0.38	0.29	0.82	0.46	0.36	0.85	0.41	0.32

表3

表4

表5

图3

图4

参考文献 29

[1]	方秀琴, 郭晓萌, 袁玲, 等. 随机森林算法在全球干旱评估中的应用[J]. 地球信息科学学报, 2021, 23(6): 1040-1049. doi: 10.12082/dqxxkx.2021.200474
	[Fang Xiuqin, Guo Xiaomeng, Yuan Ling, et al. Application of random forest algorithm in global drought assessment[J]. Journal of Geo-Information Science, 2021, 23(6): 1040-1049. ] doi: 10.12082/dqxxkx.2021.200474
[2]	陈少丹, 张利平, 汤柔馨, 等. 基于SPEI和TVDI的河南省干旱时空变化分析[J]. 农业工程学报, 2017, 33(24): 126-132.
	[Chen Shaodan, Zhang Liping, Tang Rouxin, et al. Analysis on temporal and spatial variation of drought in Henan province based on SPEI and TVDI[J]. Transactions of the Chinese Society of Agricultural Engineering, 2017, 33(24): 126-132. ]
[3]	Tian L Y, Yuan S S, Quiring S M. Evaluation of six indices for monitoring agricultural drought in the south-central United States[J]. Agricultural and Forest Meteorology, 2018, 249: 107-119. doi: 10.1016/j.agrformet.2017.11.024
[4]	陶然, 张珂. 基于PDSI的1982—2015年我国气象干旱特征及时空变化分析[J]. 水资源保护, 2020, 36(5): 50-56.
	[Tao Ran, Zhang Ke. PDSI-based analysis of characteristics and spatiotemporal changes of meteorological drought in China from 1982 to 2015[J]. Water Resources Protection, 2020, 36(5): 50-56. ]
[5]	Wu X, Wang P, Huo Z, et al. Crop Drought Identification Index for winter wheat based on evapotranspiration in the Huang-Huai-Hai Plain, China[J]. Agriculture Ecosystems & Environment, 2018, 263: 18-30. doi: 10.1016/j.agee.2018.05.001
[6]	张钢, 滕济端, 张立杰, 等. 综合干旱指数在崇左市的适用性分析与改进[J]. 人民珠江, 2020, 41(3): 23-29.
	[Zhang Gang, Teng Jiduan, Zhang Lijie, et al. Feasibility analysis and improvement of comprehensive meteorological drought index in Chongzuo[J]. Pearl River, 2020, 41(3): 23-29. ]
[7]	王劲松, 郭江勇, 倾继祖. 一种K干旱指数在西北地区春旱分析中的应用[J]. 自然资源学报, 2007, 22(5): 709-717.
	[Wang Jinsong, Guo Jiangyong, Qing Jizu. Application of a kind of K drought index in the spring drought analysis in Northwest China[J]. Journal of Natural Resources, 2007, 22(5): 709-717. ]
[8]	林慧, 王景才, 黄金柏, 等. 基于SPI和SPEI的淮河中上游流域气象干旱时空分布特征对比研究[J]. 水资源与水工程学报, 2019, 30(6): 59-67.
	[Lin Hui, Wang Jingcai, Huang Jinbai, et al. Comparative study on spatial and temporal distribution characteristics of meteorological drought in the upper and middle reaches of Huai river basin based on SPI and SPEI[J]. Journal of Water Resources and Water Engineering, 2019, 30(6): 59-67. ]
[9]	张璐, 朱仲元, 席小康, 等. 基于SPEI的锡林河流域干旱演化特征分析[J]. 干旱区研究, 2020, 37(4): 819-829.
	[Zhang Lu, Zhu Zhongyuan, Xi Xiaokang, et al. Analysis of drought evolution in the Xilin River basin based on standardized precipitation evapotranspiration index[J]. Arid Zone Research, 2020, 37(4): 819-829. ]
[10]	徐一丹, 任传友, 马熙达, 等. 基于SPI/SPEI指数的东北地区多时间尺度干旱变化特征对比分析[J]. 干旱区研究, 2017, 34(6): 1250-1262.
	[Xu Yidan, Ren Chuanyou, Ma Xida, et al. Change of drought at multiple temporal scales based on SPI/SPEI in Northeast China[J]. Arid Zone Research, 2017, 34(6): 1250-1262. ]
[11]	史晓亮, 尚雨, 陈冲, 等. 淮河流域植被NDVI与干旱条件的相关性[J]. 西安科技大学学报, 2019, 39(6): 1033-1040, 1064.
	[Shi Xiaoliang, Shang Yu, Chen Chong, et al. Correlation of vegetation NDVI and drought conditions in Huaihe river basin[J]. Journal of Xi’an University of Science and Technology, 2019, 39(6): 1033-1040, 1064. ]
[12]	李维娇, 王云鹏. 基于VCI的2003—2017年广东省干旱时空变化特征分析[J]. 华南师范大学学报(自然科学版), 2020, 52(3): 85-91.
	[Li Weijiao, Wang Yunpeng. An Analysis of the spatial-temporal characteristics of drought in Guangdong based on vegetation condition index from 2003 to 2017[J]. Journal of South China Normal University (Natural Science Edition), 2020, 52(3): 85-91. ]
[13]	张静, 魏伟, 庞素菲, 等. 基于FY-3C和TRMM数据的西北干旱区干旱监测与评估[J]. 生态学杂志, 2020, 39(2): 690-702.
	[Zhang Jing, Wei Wei, Pang Sufei, et al. Monitoring and assessment of drought in arid area in Northwest China based on FY-3C and TRMM data[J]. Chinese Journal of Ecology, 2020, 39(2): 690-702. ]
[14]	张红卫, 陈怀亮, 周官辉, 等. 归一化多波段干旱指数在农田干旱监测中的应用[J]. 科技导报, 2009, 27(11): 23-26.
	[Zhang Hongwei, Chen Huailiang, Zhou Guanhui, et al. Application of normalized multiband drought index method in cropland drought monitoring[J]. Science & Technology Review, 2009, 27(11): 23-26. ]
[15]	谷佳贺, 薛华柱, 董国涛, 等. 归一化水体指数用于河南省干旱监测适用性分析[J]. 干旱地区农业研究, 2020, 38(6): 209-217.
	[Gu Jiahe, Xue Huazhu, Dong Guotao, et al. Applicability analysis of NDWI for drought monitoring in Henan Province[J]. Agricultural Research in Arid Areas, 2020, 38(6): 209-217. ]
[16]	王立丽, 张安兵. 基于温度供水干旱指数的京津冀春旱监测及时空分布分析[J]. 测绘与空间地理信息, 2021, 44(4): 72-75.
	[Wang Lili, Zhang Anbing. Spring drought monitoring in Beijing-Tianjin-Hebei based on temperature water supply drought index[J]. Geomatics & Spatial Information Technology, 2021, 44(4): 72-75. ]
[17]	吴志勇, 程丹丹, 何海, 等. 综合干旱指数研究进展[J]. 水资源保护, 2021, 37(1): 36-45.
	[Wu Zhiyong, Cheng Dandan, He Hai, et al. Research progress of composite drought index[J]. Water Resources Protection, 2021, 37(1): 36-45. ]
[18]	Zhang A Z, Jia G S. Monitoring meteorological drought in semiarid regions using multi-sensor microwave remote sensing data[J]. Remote Sensing of Environment, 2013, 134: 12-23. doi: 10.1016/j.rse.2013.02.023
[19]	杜瑞麒, 张智韬, 巨娟丽, 等. 基于波文比和降水的综合干旱指数的构建及应用[J]. 节水灌溉, 2020, 45(8): 63-74.
	[Du Ruiqi, Zhang Zhitao, Ju Juanli, et al. Construction and application of comprehensive drought index based on Bowen ratio and precipitation[J]. Water Saving Irrigation, 2020, 45(8): 63-74. ]
[20]	韩兰英, 张强, 赵红岩, 等. 甘肃省农业干旱灾害损失特征及其对气候变暖的响应[J]. 中国沙漠, 2016, 36(3): 767-776.
	[Han Lanying, Zhang Qiang, Zhao Hongyan, et al. The characteristics of agricultural drought disaster loss and response to climate warming in Gansu, China[J]. Journal of Desert Research, 2016, 36(3): 767-776. ]
[21]	Gupta S C, Larson W E. Estimating soil water retention characteristics from particle size distribution, organic mat-ter percent, and bulk density[J]. Water Resources Research, 1979, 15(6): 1633-1635. doi: 10.1029/WR015i006p01633
[22]	胡鹏飞, 李净, 王丹, 等. 基于MODIS和TRMM数据的黄土高原农业干旱监测[J]. 干旱区地理, 2019, 42(1): 172-179.
	[Hu Pengfei, Li Jing, Wang Dan, et al. Monitoring agricultural drought in the Loess Plateau using MODIS and TRMM data[J]. Arid Land Geography, 2019, 42(1): 172-179. ]
[23]	杨莎. 基于植被供水指数的锡林浩特市干旱监测分析[J]. 草原与草业, 2021, 33(3): 42-45.
	[Yang Sha. Drought monitoring analysis in Xilinhot City based on vegetation water supply index[J]. Grassland and Prataculture, 2021, 33(3): 42-45. ]
[24]	刘冀, 张特, 魏榕, 等. 基于随机森林偏差校正的农业干旱遥感监测模型研究[J]. 农业机械学报, 2020, 51(7): 170-177.
	[Liu Ji, Zhang Te, Wei Rong, et al. Development of agricultural drought monitoring model using remote sensing based on bias-correcting random forest[J]. Transactions of the Chinese Society of Agricultural Machinery, 2020, 51(7): 170-177. ]
[25]	李晓辉, 杨勇, 杨洪伟. 基于BP神经网络与灰色模型的干旱预测方法研究[J]. 沈阳农业大学学报, 2014, 45(2): 253-256.
	[Li Xiaohui, Yang Yong, Yang Hongwei. Combining BP neural network with gray model to achieve drought predicting[J]. Journal of Shenyang Agricultural University, 2014, 45(2): 253-256. ]
[26]	赵国羊, 涂新军, 王天, 等. 基于人工神经网络和支持向量回归机的干旱预测[J]. 人民珠江, 2021, 42(4): 1-9.
	[Zhao Guoyang, Tu Xinjun, Wang Tian, et al. Drought prediction based on artificial neural network and support vector regression machine[J]. Pearl River, 2021, 42(4): 1-9. ]
[27]	沈润平, 郭佳, 张婧娴, 等. 基于随机森林的遥感干旱监测模型的构建[J]. 地球信息科学学报, 2017, 19(1): 125-133. doi: 10.3724/SP.J.1047.2017.00125
	[Shen Runping, Guo Jia, Zhang Jingxian, et al. Construction of a drought monitoring model using the random forest based remote sensing[J]. Journal of Geo-Information Science, 2017, 19(1): 125-133. ] doi: 10.3724/SP.J.1047.2017.00125
[28]	董婷, 任东, 邵攀, 等. 基于多源遥感数据和随机森林的综合旱情指标构建[J]. 农业机械学报, 2019, 50(8): 200-212.
	[Dong Ting, Ren Dong, Shao Pan, et al. Construction of integrated drought condition index based on multi-sensor remote sensing and random forest[J]. Transactions of the Chinese Society of Agricultural Machinery, 2019, 50(8): 200-212. ]
[29]	Feng P Y, Wang B, Liu D L, et al. Machine learning-based integration of remotely-sensed drought factors can improve the estimation of agricultural drought in South-Eastern Australia[J]. Agricultural Systems, 2019, 173: 303-316. doi: 10.1016/j.agsy.2019.03.015

月份	SPEI_1				SPEI_3				SPEI_6
月份	VCI	TCI	PCI	VSWI	VCI	TCI	PCI	VSWI	VCI	TCI	PCI	VSWI
4	0.06	0.38^**	0.54^**	0.30^*	0.09	0.23^**	0.47^**	0.34^**	0.09	0.28^**	0.56^**	0.30^**
5	0.09	0.28^**	0.61^**	0.29^*	0.07	0.24^**	0.60^**	0.26^**	0.08	0.29^**	0.53^**	0.21^**
6	0.12^**	0.35^**	0.65^**	0.34^**	0.10^**	0.34^**	0.51^**	0.30^*	0.21	0.33^**	0.53^**	0.25^*
7	0.15^**	0.34^**	0.76^**	0.40^*	0.28^**	0.25^**	0.68^**	0.29^**	0.19^**	0.32^**	0.78^**	0.32^*
8	0.16^**	0.24^**	0.78^**	0.27^**	0.17^**	0.27^**	0.65^**	0.28^**	0.20^**	0.25^**	0.62^**	0.23^**
9	-0.02	0.31^**	0.49^*	0.24^**	0.04^*	0.23^**	0.45^*	0.29^*	0.02	0.26^**	0.43^*	0.26^**
10	0.11^*	0.32^**	0.55^**	0.37^**	0.25^*	0.27^**	0.50^**	0.41^**	0.17^**	0.30^**	0.56^**	0.32^*

分类	站点	年平均降水量/mm	平均温度/℃
聚类1	酒泉、张掖、武威、民勤、靖远、永昌、景泰、皋兰、高台、敦煌、瓜州、玉门镇、山丹	155.93	7.97
聚类2	崆峒、临夏、榆中、临洮、环县、西峰、岷县、武都、麦积、华家岭、会宁、天水	483.96	8.60