基于PLS的不同水氮条件下带状套作玉米产量预测

doi:10.3864/j.issn.0578-1752.2022.06.006

摘要/Abstract

摘要：

【目的】利用高光谱数据构建一种“高光谱参数-光合色素-产量”模型间接估测套作玉米产量,为带状套作玉米产量无损预测提供技术手段。【方法】以不同年份、地点、品种、处理（氮肥、水分）的田间试验为基础,综合分析带状套作玉米各生育时期及全生育期光合色素参数与冠层高光谱参数和玉米产量的关系,明确玉米产量预测的最佳生育时期及光合色素参数,基于线性函数、二次函数和偏最小二乘（partial least squares,PLS）回归法构建产量估测模型。【结果】光合色素-产量预测模型中,冠层类胡萝卜素密度的PLS产量估测模型效果最佳（R²=0.882,RMSE=0.669 t·hm^-2）。光谱参数-光合色素分析中,抽雄期叶绿素含量与波段自由组合指数rRVI（534,546）相关性最好（r=0.927）。其余光合色素参数与对应光谱指数相关性均在0.797以上。在高光谱参数-光合色素-产量估测模型中,由叶绿素含量、类胡萝卜素含量、冠层叶绿素密度、冠层类胡萝卜素密度为连接点,并以光谱指数rNDVI（534,546）,rRVI（531,555）,rNDVI（532,546）,rNDVI（531,555）为自变量构建的PLS产量预测模型效果较好（R²=0.509,RMSE=1.352 t·hm^-2）。【结论】利用色素参数作为光谱数据和产量连接的桥梁,通过PLS回归法建立的预测模型,能够在带状套作玉米中,对玉米产量实现较好估测,为带状套作玉米的田间管理和生长监测提供理论和技术参考。

关键词: 套作, 玉米, 高光谱, 色素参数, 产量, 偏最小二乘回归

Abstract:

【Objective】 This study was designed mainly to provide technical means for non-destructive prediction of intercropped maize yield. The prime objective of our study was to construct a “hyperspectral parameter-photosynthetic pigment-yield” model from the hyperspectral data. 【Method】 Based on field trials of different years, locations, varieties, and treatments (nitrogen fertilizer, moisture), the relationships among photosynthetic pigment parameters, canopy hyperspectral parameters, and maize yield at each growth period and the entire growth period of intercropping maize were comprehensively analyzed. In addition, the optimal growth period and photosynthetic pigment parameters for maize yield prediction were also clarified. Then, the prediction model of yield was constructed based on linear function, quadratic function and partial least squares regression (PLS). 【Result】 Among the photosynthetic pigment-yield prediction models, the PLS prediction model of yield based on canopy carotenoid density had the best effect (R ²=0.882, RMSE=0.669 t·hm ^-2). In the spectral parameter-photosynthetic pigment analysis, the chlorophyll content during the tasseling stage had the best correlation with the band free combination index rRVI (534, 546) (r=0.927). The correlation between the other photosynthetic pigment parameters and the corresponding spectral index was above 0.797. In the hyperspectral parameter- photosynthetic pigment-yield prediction model, the chlorophyll content, carotenoid content, canopy chlorophyll density, and canopy carotenoid density were used as connection points, and using the spectral indices of rNDVI (534, 546), rRVI (531, 555), rNDVI (532, 546), and rNDVI (531, 555) as independent variables, the PLS output prediction model had better effect (R²=0.509, RMSE=1.352 t·hm ^-2). 【Conclusion】 In intercropping maize, the pigment parameters were used as a bridge between spectral data and yield. A prediction model was established through PLS regression, which could achieve a better estimation of maize yield and provide the theoretical and technical reference for field management and growth monitoring of maize in intercropping.

Key words: relay intercropping, maize, hyperspectral, pigment parameters, yield, partial least squares regression

谭先明,张佳伟,王仲林,谌俊旭,杨峰,杨文钰. 基于PLS的不同水氮条件下带状套作玉米产量预测[J]. 中国农业科学, 2022, 55(6): 1127-1138.

TAN XianMing,ZHANG JiaWei,WANG ZhongLin,CHEN JunXu,YANG Feng,YANG WenYu. Prediction of Maize Yield in Relay Strip Intercropping Under Different Water and Nitrogen Conditions Based on PLS[J]. Scientia Agricultura Sinica, 2022, 55(6): 1127-1138.

图/表 12

表1

表2

表3

表4

表5

图1

表6

图2

表7

表8

光合色素参数与光谱指数的线性非线性回归分析"

色素参数 Pigment parameter	光谱指数 Spectral index	模型 Model	函数 Function	建模集 Calibration set (n=90)		验证集 Test set (n=40)
色素参数 Pigment parameter	光谱指数 Spectral index	模型 Model	函数 Function	R²	RMSE	R²	RMSE
Ct (mg·cm^-2)	rDVI (1009,972)	线性 Linear	y=0.0753x-0.0694	0.779	0.033	0.792	0.033
	rDVI (1009,972)	二次Quadratic	y = -0.0043x²+ 0.0976x -0.0952	0.780	0.032	0.790	0.031
	rRVI (534,546)	线性 Linear	y=5.181x-4.612	0.859	0.026	0.820	0.028
	rRVI (534,546)	二次Quadratic	y = 39.363x² - 66.684x + 28.183	0.866	0.025	0.812	0.029
	rNDVI (534,546)	线性 Linear	y=9.4714x+0.5494	0.857	0.026	0.820	0.028
	rNDVI (534,546)	二次Quadratic	y = 145.48x² + 22.759x +0.8462	0.867	0.025	0.813	0.029
CCD (g·m^-2)	rDVI (1009,972)	线性 Linear	y=3.7883x-4.2894	0.690	2.058	0.695	1.930
	rDVI (1009,972)	二次Quadratic	y = -0.0206x²+ 3.8945x -4.4126	0.690	2.191	0.538	2.476
	rRVI (531,555)	线性 Linear	y=168.25x-138.36	0.831	1.520	0.813	1.556
	rRVI (531,555)	二次Quadratic	y = 1211.3x²- 1899.1x + 743.25	0.847	1.445	0.859	1.377
	rNDVI (533,539)	线性 Linear	y=697.07x+25.798	0.829	1.529	0.820	1.459
	rNDVI (533,539)	二次Quadratic	y = 14598x² + 1548.5x + 37.863	0.837	1.493	0.839	1.391
Car (mg·cm^-2)	rDVI (1011,962)	线性 Linear	y=0.0046x+0.0008	0.727	0.003	0.667	0.003
	rDVI (1011,962)	二次Quadratic	y = -0.0008x² + 0.0088x -0.0041	0.749	0.003	0.677	0.003
	rRVI (532,546)	线性 Linear	y=0.3379x-0.2864	0.824	0.002	0.775	0.003
	rRVI (532,546)	二次Quadratic	y = -0.0011x²+ 0.3399x - 0.2873	0.824	0.002	0.789	0.003
	rNDVI (532,546)	线性 Linear	y=0.5999x+0.0493	0.824	0.002	0.776	0.003
	rNDVI (532,546)	二次Quadratic	y = 0.8472x² + 0.704x + 0.0524	0.824	0.002	0.790	0.003
CCarD (mg·m^-2)	rDVI (1009,975)	线性 Linear	y=328.64x-259.68	0.635	186.720	0.641	175.510
	rDVI (1009,975)	二次Quadratic	y = 17.113x² + 243.8x - 164.28	0.637	186.426	0.671	174.162
	rRVI (531,555)	线性 Linear	y=13824x-11244	0.801	137.950	0.797	136.940
	rRVI (531,555)	二次Quadratic	y = 43623x² - 60627x + 20505	0.804	136.888	0.824	130.002
	rNDVI (531,555)	线性 Linear	y=23718x+2432.8	0.800	138.340	0.792	151.290
	rNDVI (531,555)	二次Quadratic	y = 156703x²+ 48613x + 3400.1	0.804	136.771	0.819	129.656

表8

图3

表9

参考文献 31

[1]	董守坤, 赵坤, 刘丽君, 孙聪姝, 郭茜茜, 阮英慧, 王利彬. 干旱胁迫对春大豆叶绿素含量和根系活力的影响. 大豆科学, 2011,30(6):949-953.
	DONG S K, ZHAO K, LIU L J, SUN C S, GUO Q Q, RUAN Y H, WANG L B. Effect of drought stress on chlorophyll content and root activity of spring soybean. Soybean Science, 2011,30(6):949-953. (in Chinese)
[2]	宋贺, 蒋延玲, 许振柱, 周广胜. 玉米光合生理参数对全生育期干旱与拔节后干旱过程的响应. 生态学报, 2019,39(7):2405-2415.
	SONG H, JIANG Y L, XU Z Z, ZHOU G S. Response of photosynthetic physiological parameters of maize to drought during the whole growth period and after the jointing stage. Acta Ecologica Sinica, 2019,39(7):2405-2415. (in Chinese)
[3]	赵成凤, 王晨光, 李红杰, 郑学慧, 杨梅, 张仁和. 干旱及复水条件下外源褪黑素对玉米叶片光合作用的影响. 生态学报, 2021,41(4):1431-1439.
	ZHAO C F, WANG C G, LI H J, ZHENG X H, YANG M, ZHANG R H. Effects of exogenous melatonin on photosynthesis of maize leaves under drought stress and rewatering. Acta Ecologica Sinica, 2021,41(4):1431-1439. (in Chinese)
[4]	ADRIENN V S, JANOS N. Effects of nutrition and water supply on the yield and grain protein content of maize hybrids. Australian Journal of Crop Science, 2012,6(3):381-390.
[5]	CAVIGLIA O P, MELCHIORI R J M, SADRAS V O. Nitrogen utilization efficiency in maize as affected by hybrid and N rate in late-sown crops. Field Crops Research, 2014,168:27-37. doi: 10.1016/j.fcr.2014.08.005
[6]	杨峰, 范亚民, 李建龙, 钱育蓉, 王艳, 张洁. 高光谱数据估测稻麦叶面积指数和叶绿素密度. 农业工程学报, 2010,26(2):237-243.
	YANG F, FAN Y M, LI J L, QIAN Y R, WANG Y, ZHANG J. Estimating LAI and CCD of rice and wheat using hyperspectral remote sensing data. Transactions of the Chinese Society of Agricultural Engineering, 2010,26(2):237-243. (in Chinese)
[7]	薛利红, 杨林章. 采用不同红边位置提取技术估测蔬菜叶绿素含量的比较研究. 农业工程学报, 2008,24(9):165-169.
	XUE L H, YANG L Z. Comparative study on estimation of chlorophyll content in spinach leaves using various red edge position extraction techniques. Transactions of the Chinese Society of Agricultural Engineering, 2008,24(9):165-169. (in Chinese)
[8]	SCHLEMMER M, GITELSON A, SCHEPERS J, FERGUSON R, PENG Y, SHANAHAN J, RUNDQUIST D. Remote estimation of nitrogen and chlorophyll contents in maize at leaf and canopy levels. International Journal of Applied Earth Observation & Geoinformation, 2013,25(1):47-54.
[9]	田永超. 基于冠层反射光谱的水稻水分及稻麦生长监测[D]. 南京: 南京农业大学, 2003.
	TIAN Y C. Monitoring water status in rice and growth parameters in rice and wheat with canopy spectral reflectance[D]. Nanjing: Nanjing Agricultural University, 2003. (in Chinese)
[10]	ZHANG L L, ZHANG Z, LUO Y C, CAO J, TAO F L. Combining optical, fluorescence, thermal satellite, and environmental data to predict county-level maize yield in China using machine learning approaches. Remote Sensing, 2020,12(1):21. doi: 10.3390/rs12010021. doi: 10.3390/rs12010021
[11]	XUE L H, CAO W X, YANG L Z. Predicting grain yield and protein content in winter wheat at different N supply levels using canopy reflectance spectra. Pedosphere, 2007,17(5):646-653. doi: 10.1016/S1002-0160(07)60077-0
[12]	LI Z, JIN X, ZHAO C, WANG J, XU X, YANG G, LI C, SHEN J. Estimating wheat yield and quality by coupling the DSSAT-CERES model and proximal remote sensing. European Journal of Agronomy, 2015,71:53-62. doi: 10.1016/j.eja.2015.08.006
[13]	RODRIGUES F A, BLASCH G, DEFOURNY P, ORTIZ MONASTERIO J I, SCHULTHESS U, ZARCO-TEJADA P J, TAYLOR J A, GERARD B. Multi-temporal and spectral analysis of high-resolution hyperspectral airborne imagery for precision agriculture: Assessment of wheat grain yield and grain protein content. Remote Sensing, 2018,10(6):930. doi: 10.3390/rs10060930. doi: 10.3390/rs10060930
[14]	DELIA B, KINIRY J R, MAAS S J, TISCARENO M L, JAIME M C, MENDOZA J L, RICHARDSON C W, JAIME S G, MANJARREZ J R. Large-area maize yield forecasting using leaf area index based yield model. Agronomy Journal, 2005,97(2):418-425. doi: 10.2134/agronj2005.0418
[15]	冯伟, 朱艳, 田永超, 姚霞, 郭天财, 曹卫星. 基于高光谱遥感的小麦籽粒产量预测模型研究. 麦类作物学报, 2007,27(6):1076-1084.
	FENG W, ZHU Y, TIAN Y C, YAO X, GUO T C, CAO W X. Model for predicting grain yield with canopy hyperspectal remote sensing in wheat. Journal of Triticeae Crops, 2007,27(6):1076-1084. (in Chinese)
[16]	冯伟. 基于高光谱遥感的小麦氮素营养及生长指标监测研究[D]. 南京: 南京农业大学, 2007.
	FENG W. Monitoring nitrogen status and growth characters with canopy hyperspectal remote sensing in wheat[D]. Nanjing: Nanjing Agricultural University, 2007. (in Chinese)
[17]	陈鹏飞, 王吉顺, 潘鹏, 徐于月, 姚凌. 基于氮素营养指数的冬小麦籽粒蛋白质含量遥感反演. 农业工程学报, 2011,27(9):75-80.
	CHEN P F, WANG J S, PAN P, XU Y Y, YAO L. Remote detection of wheat grain protein content using nitrogen nutrition index. Transactions of the Chinese Society of Agricultural Engineering, 2011,27(9):75-80. (in Chinese)
[18]	孙锐, 朱平, 王志敏, 丛艳霞, 勾玲, 方立锋, 赵明. 春玉米叶面积系数动态特征的密度效应. 作物学报, 2009,35(6):1097-1105. doi: 10.3724/SP.J.1006.2009.01097
	SUN R, ZHU P, WANG Z M, CONG Y X, GOU L, FANG L F, ZHAO M. Effect of plant density on dynamic characteristics of leaf area index in development of spring maize. Acta Agronomica Sinica, 2009,35(6):1097-1105. (in Chinese) doi: 10.3724/SP.J.1006.2009.01097
[19]	KONG D X, LI Y Q, WANG M L, BAI M, ZOU R, TANG H, WU H. Effects of light intensity on leaf photosynthetic characteristics, chloroplast structure, and alkaloid content of Mahonia bodinieri (Gagnep.) Laferr. Acta Physiologiae Plantarum, 2016,38(5):120. doi: 10.1007/s11738-016-2147-1. doi: 10.1007/s11738-016-2147-1
[20]	DAUGHTRY C S T, WALTHALL C L, KIM M S, COLSTOUN E B D, III M M. Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance. Remote Sensing of Environment, 2000,74(2):229-239. doi: 10.1016/S0034-4257(00)00113-9
[21]	GAMON J A, PENUELAS J, FIELD C B. A narrow-waveband spectral index that tracks diurnal changes in photosynthetic efficiency. Remote Sensing of Environment, 1992,41(1):35-44. doi: 10.1016/0034-4257(92)90059-S
[22]	林同保, 孟战赢, 王志强, 支园园. 土壤水分对夏玉米水分代谢及产量的影响. 河南农业大学学报, 2008,42(2):135-139.
	LIN T B, MENG Z Y, WANG Z Q, ZHI Y Y. Influence of different soil moisture on water metabolism and yield of summer corn. Journal of Henan Agricultural University, 2008,42(2):135-139. (in Chinese)
[23]	谷晓博, 李援农, 黄鹏, 杜娅丹, 方恒, 陈朋朋. 种植方式和施氮量对冬油菜产量与水氮利用效率的影响. 农业工程学报, 2018,34(10):113-123.
	GUO X B, LI Y N, HUANG P, DU Y D, FANG H, CHEN P P. Effects of planting patterns and nitrogen application rates on yield, water and nitrogen use efficiencies of winter oilseed rape (Brassica napus L.). Transactions of the Chinese Society of Agricultural Engineering, 2018,34(10):113-123. (in Chinese)
[24]	隋鹏祥, 有德宝, 安俊朋, 张文可, 田平, 梅楠, 王美佳, 王沣, 苏思慧, 齐华. 秸秆还田方式与施氮量对春玉米产量及干物质和氮素积累、转运的影响. 植物营养与肥料学报, 2018,24(2):316-324.
	SUI P X, YOU D B, AN J P, ZHANG W K, TIAN P, MEI N, WANG M J, WANG F, SU S H, QI H. Effects of straw management and nitrogen application on spring maize yield, dry matter and nitrogen accumulation and transfer. Journal of Plant Nutrition and Fertilizers, 2018,24(2):316-324. (in Chinese)
[25]	张苏. 叶绿素密度遥感反演与冬小麦单产估算研究[D]. 西安: 西安科技大学, 2014.
	ZHANG S. Retrieval of canopy chlorophyll density and yield prediction for winter wheat based on remote sensing[D]. Xi’an: Xi’an University of Science and Technology, 2014. (in Chinese)
[26]	CHEW B P, PARK J S. Carotenoid action on the immune response. Journal of Nutrition, 2004,134(1):257S-261S. doi: 10.1093/jn/134.1.257S
[27]	KRINSKY N I, JOHNSON E J. Carotenoid actions and their relation to health and disease. Molecular Aspects of Medicine, 2006,26(6):459-516. doi: 10.1016/j.mam.2005.10.001
[28]	LI L T, REN T, MA Y, WEI Q Q, WANG S Q, LI X K, CONG R H, LIU S S, LU J W. Evaluating chlorophyll density in winter oilseed rape (Brassica napus L.) using canopy hyperspectral red-edge parameters. Computers and Electronics in Agriculture, 2016,126:21-31. doi: 10.1016/j.compag.2016.05.008
[29]	谌俊旭, 黄山, 范元芳, 王锐, 刘沁林, 杨文钰, 杨峰. 单作套作大豆叶片氮素积累与光谱特征. 作物学报, 2017,43(12):1835-1844. doi: 10.3724/SP.J.1006.2017.01835
	CHEN J X, HUANG S, FAN Y F, WANG R, LIU Q L, YANG W Y, YANG F. Remote detection of canopy leaf nitrogen status in soybean by hyperspectral data under monoculture and intercropping systems. Acta Agronomica Sinica, 2017,43(12):1835-1844. (in Chinese) doi: 10.3724/SP.J.1006.2017.01835
[30]	王仲林, 谌俊旭, 程亚娇, 范元芳, 冯伟, 李昊宸, 吕金灿, 杨文钰, 杨峰. 干旱胁迫下玉米叶片可溶性糖光谱估测研究. 四川农业大学学报, 2018,36(4):436-443.
	WANG Z L, CHEN J X, CHENG Y J, FAN Y F, FENG W, LI H C, LÜ J C, YANG W Y, YANG F. Assessing the soluble sugar of maize leaves in drought stress based on hyperspectral data. Journal of Sichuan Agricultural University, 2018,36(4):436-443. (in Chinese)
[31]	JIA F F, LIU G S, LIU D S, ZHANG Y Y, FAN W G, XING X X. Comparison of different methods for estimating nitrogen concentration in flue-cured tobacco leaves based on hyperspectral reflectance. Field Crops Research, 2013,150:108-114. doi: 10.1016/j.fcr.2013.06.009

植被指数及光谱指数 Vegetation index and spectral index	计算公式 Calculation	文献来源 Reference
叶绿素反演植被指数 Transformed chlorophyll absorption in reflectance index, TCARI	3×[(R₇₀₀-R₆₇₀)-0.2×(R₇₀₀-R₅₅₀)×(R₇₀₀/R₆₇₀)]	[20]
绿度植被指数 Green normalized difference vegetation index, GNDVI	(R₈₀₀-R₅₅₀)/(R₈₀₀+R₅₅₀)	[20]
光化学植被指数 photochemical reflectance index, PRI	(R₅₃₁-R₅₇₀)/(R₅₃₁+R₅₇₀)	[21]
rDVI（R_i, R_j） Random difference vegetation index	R_i-R_j	—
rRVI（R_i, R_j） Random ratio vegetation index	R_i/R_j	—
rNDVI（R_i, R_j） Random normalized difference vegetation index	(R_i-R_j)/(R_i+R_j)	—

处理 Treatment	拔节期Exp.1 Jointing stage	灌浆期Exp.1 Filling stage	拔节期Exp.4 Jointing stage	抽雄期Exp.4 Tasseling stage	灌浆期Exp.4 Filling stage	试验2 Exp.2	试验3 Exp.3	试验5 Exp.5
CK(N1)	7.113±0.825a	5.112±0.549a	4.483±0.124a	5.573±0.206a	4.296±0.466b	—	8.829±0.332a	7.021±0.075a
L(N2)	6.473±0.495a	4.739±0.657a	3.780±0.125bc	4.639±0.079a	5.343±0.679ab	7.409±0.505b	8.978±0.316a	7.119±0.518a
M(N3)	5.914±0.230ab	4.494±0.676a	4.102±0.222b	5.250±0.483a	5.907±0.148a	9.700±0.863a	9.854±0.477a	8.237±0.662a
S(N4)	4.355±0.232b	4.493±0.582a	3.496±0.138c	3.627±0.258b	5.919±0.146a	9.956±0.171a	—	8.181±0.476a

生育时期 Growth period	试验组成 Test composition	Ct (mg·cm^-2)	CCD (g·m^-2)	Car (mg·cm^-2)	CCarD (mg·m^-2)
拔节期 Jointing stage	Exp.(1-5) n=54	0.056±0.026	0.544±0.426	0.008±0.007	84.778±98.435
抽雄期 Tasseling stage	Exp(3-5) n=33	0.107±0.068	4.615±3.588	0.012±0.005	502.782±299.183
灌浆期 Filling stage	Exp(1-2,4-5) n=40	0.077±0.064	3.255±3.334	0.009±0.005	349.077±279.695
总样本 Total sample	Exp.(1-5) n=127	0.076±0.056	2.457±3.127	0.009±0.006	276.637±286.108

生育时期 Growth period	试验组成 Test composition	Ct	CCD	Car	CCarD
拔节期 Jointing stage	Exp.(1-5) n=54	-0.111	-0.240	-0.321*	-0.358**
抽雄期 Tasseling stage	Exp(3-5) n=33	-0.729**	-0.663**	-0.616**	-0.565**
灌浆期 Filling stage	Exp(1-2,4-5) n=40	-0.275	0.283	-0.342*	-0.288
总样本 Total sample	Exp.(1-5) n=127	-0.330**	-0.295**	-0.373**	-0.287**

模型 Model	色素参数 Pigment parameter	函数 Function	R²	RMSE (t·hm^-2)	R²	RMSE (t·hm^-2)
线性非线性模型 Linear and non- linear model	Ct (mg·cm^-2)	y = -22.183x + 9.450	0.560	1.349	0.472	1.401
	Ct (mg·cm^-2)	y = -280.49x²+ 47.191x + 6.5568	0.669	1.170	0.545	1.338
	CCD (g·m^-2)	y = -0.383x + 8.8655	0.467	1.485	0.387	1.517
	CCD (g·m^-2)	y = 0.006x² - 0.4563x + 8.9976	0.468	1.484	0.392	1.510
	Car (mg·cm^-2)	y = -238.06x + 9.9226	0.399	1.577	0.298	1.616
	Car (mg·cm^-2)	y = -45375x² + 887.63x + 4.2694	0.646	1.209	0.477	1.437
	CCarD (mg·m^-2)	y = -0.0039x + 9.079	0.345	1.646	0.269	1.657
	CCarD (mg·m^-2)	y = -3E-06x² + 0.0002x + 8.1775	0.364	1.636	0.264	1.691
偏最小二乘模型 PLS-M	Ct (mg·cm^-2), CCD (g·m^-2) Car (mg·cm^-2), CCarD (mg·m^-2)	y=-79.7789x₁-0.6549x₂+529.075x₃+0.0123x₄+6.002 (x₁:Ct, x₂:CCD, x₃:Car, x₄:CCarD)	0.859	0.768	0.882	0.669