刘智业,杨 群,凌琪涵,魏 勇,宁 强,孔发明,张跃强,,3,石孝均,,3,王 洁,,3
采用柑橘叶片功能性氮含量无损监测模型的调控施氮方法
刘智业1,杨 群1,凌琪涵2,魏 勇2,宁 强1,孔发明1,张跃强1,2,3,石孝均1,2,3,王 洁1,2,3※
(1. 西南大学长江经济带农业绿色发展研究中心,重庆 400715;2. 西南大学资源环境学院,重庆 400715;3. 国家紫色土肥力与肥料效益监测站,重庆 400716)
为实现柑橘氮素管理的定量化,该研究以5年生‘春见’橘橙为试验材料,设置不同对照施氮处理N0、N1、N2、N3(施氮量分别为0、50、100、200 g/株)和调控施氮处理Nr1、Nr2、Nr3(分别根据N1、N2、N3进行调控),在试验开展的第1年利用高光谱技术,分别建立柑橘果实膨大期和转色期的叶片功能性氮含量无损监测模型;第2年利用叶片功能性氮含量无损监测模型与追氮量公式计算调控施氮处理的实际追氮量,比较分析对照施氮和调控施氮对柑橘果实产量、品质及氮肥利用率的影响。结果表明,利用反向传播神经网络构建的叶片功能性氮含量模型精度较高,决定系数2为0.78(果实膨大期)和0.77(果实转色期)。调控施氮处理Nr1和Nr3比对照施氮N1和N3分别增产5.49和4.43 kg/株(增幅为48%和40%);相比于N1,调控施氮处理Nr1的单果质量和可溶性固形物含量显著增加(<0.05),果实横纵径、果形指数增幅不显著。相比于N3,调控施氮处理Nr3的氮肥偏生产力升高了103%;Nr1和Nr3的氮肥农学效率分别提高了290%和364%。Nr2和N2的产量、品质和氮肥利用率无显著差异(<0.05)。基于柑橘叶片功能性氮含量无损监测模型的调控施氮方法,能在一定程度上减少施氮不足或过量对柑橘产量、品质的影响,提高氮肥偏生产力和农学效率。
柑橘;高光谱;调控施氮;叶片功能性氮;无损监测
随着柑橘产业持续发展,中国已成为柑橘栽培面积和柑橘产量第一大国。影响柑橘产业绿色可持续发展的因素较多,明确柑橘的营养状况并进行科学施肥是保障柑橘正常生长发育、提高产量、改善品质、保护生态环境的基础[1]。中国柑橘主产区存在氮、磷、钾投入过量及不足并存的问题[2-3]。其中,氮肥投入过量及不足的问题尤为明显,严重限制了柑橘产业的绿色可持续发展[4]。因此,对柑橘种植生产进行实时、无损、精准的氮素监测以及基于监测结果进行调控施氮,成为柑橘种植的现实需求。
近年来,高光谱遥感和数据处理技术发展迅猛,光谱技术被广泛应用于作物氮素营养诊断[5]。利用光谱技术在玉米、水稻等粮食作物[6-7]和果树、棉花等经济作物[8-9]叶片和冠层氮含量无损监测研究取得一定进展。刘雪峰等[10]利用机载多光谱获取果实膨大期柑橘冠层光谱图像,提取光谱反射率利用支持向量机算法构建冠层氮含量的无损监测模型,精度可达0.80。易时来等[11]运用锦橙叶片全波段光谱和偏最小二乘回归建立叶片氮含量的预测模型,精度为0.90。MENESATTI等[12]测定果实转色期塔罗科血橙叶片的可见/近红外光谱反射率,建立叶片全氮含量无损监测模型,其决定系数达0.91,光谱对柑橘叶片全氮含量有较好的估测能力。前人较多是以测得的叶片全氮含量来调整当季氮肥用量[13],然而叶片全氮含量适宜值较宽,氮素营养诊断已经开始由叶片全氮含量到表征叶片生理生化的特征参数方向发展[14-16]。
根据植物对氮素的吸收利用特性,植株体内的氮素可以分为营养性氮、结构性氮和功能性氮三大类,植物体内三类形态氮素处于动态变化中,各组分在叶片中的含量与分布对植物叶片生理生化反应有一定的指示作用[17-18],近年来,对叶片生理生化特性的研究也多利用光谱化学计量法来建模。AINSWORTH等[19]研究表明,利用叶片可见/近红外光谱定量测定光合作用中最大羧化速率,模型精度2可达0.88。前人研究利用光谱技术构建作物氮营养无损监测模型、实施变量施氮能够在一定程度上增加作物产量,改善品质[20]。WANG等[21]利用高光谱技术构建梨树果实膨大期叶片全氮含量的无损监测模型并变量追氮,结果表明可见/近红外光谱技术能实现叶片全氮含量快速诊断并及时追施氮肥,可以在一定程度上缓解早期施氮不足或过量对梨果产量、品质的影响,增产20%以上。李旭[22]研究表明,氮肥施用不足时,影响果实膨大期和转色期产量、品质的形成,无核椪柑的产量减少1.0%~3.5%,可溶性固形物降低5.70%~11.51%;杨江波等[23]研究结果表明,氮肥施用过量时,对塔罗科血橙增产效果不显著,果实可溶性固形物含量、固酸比等增加不显著,肥料利用率显著降低。不少学者利用光谱技术实现对水稻、玉米等作物的氮营养诊断与变量施氮,提升作物产量和品质,提高氮肥利用效率[24-25]。
本研究利用高光谱技术分别在柑橘果实膨大期和果实转色期建立叶片功能性氮无损监测模型,通过模型监测功能性氮含量和追氮量公式,以期准确、连续对柑橘进行调控施氮,探讨调控施氮技术对柑橘产量、品质的影响以及氮肥利用效率的影响。以期为实现柑橘叶片功能性氮含量无损监测和调控施氮提供理论依据和技术支持。
试验地位于重庆市长寿区龙河镇八卦村,地理位置为东经107°13′,北纬29°59′,海拔406 m,属中亚热带湿润气候区,年平均气温17.7 ℃,年平均降水量1 165.2 mm,常年日照时数1 245.1 h。供试土壤为紫色土,基础理化性质:pH 6.38,有机质9.06 g/kg,全氮0.75 g/kg,全磷0.29 g/kg,全钾26.72 g/kg。
供试柑橘品种为‘春见’橘橙[×(×)]。本试验设置4个不同施氮处理,分别为N0(0 g/株)、N1(50 g/株)、N2(100 g/株)、N3(200 g/株),其中,N1为优化施氮减量处理(减量50%),N2为柑橘优化施氮处理[26],N3是农户常规施氮量。以N1、N2、N3为基础,分别设置相应的3个调控施氮处理,即Nr1、Nr2、Nr3,调控施氮量由当年监测柑橘叶片功能性氮含量并结合追氮量公式[21,27](式(1))计算得出。每个处理3组重复,每组重复2棵树。氮磷钾分别用尿素(含46% N)、过磷酸钙(含12% P2O5)、硫酸钾(含51% K2O)提供。肥料运筹:分3次施用,萌芽肥(3月下旬)、果实膨大肥(7月下旬)和果实转色肥(10月下旬)。其中,萌芽肥施氮50%,施磷20%,施钾25%;果实膨大肥施氮20%,施磷20%,施钾50%;果实转色肥施氮30%,施磷60%,施钾25%。施肥方式为沿树冠滴水线穴施,肥料混匀后覆土。具体施肥用量和计算如表1。
式中、分别为果实膨大期和果实转色期调控施氮处理的施氮量调整值,g/株;N实际采用该时期的柑橘叶片功能性氮含量;N标统一采用该时期N2处理柑橘叶片功能性氮含量(果实膨大期:16.91 g/kg,果实转色期:18.82 g/kg);HDL为统计各施氮处理平均百叶质量:0.026 kg[28];Leaf为叶片数量,取400;%是柑橘叶片功能性氮含量占全氮含量的百分数:60%[28];F为肥料利用率,本文取30%。
表1 调控施氮与对照施氮处理氮肥施用情况
注:以优化施氮处理(N2)两个时期的尿素施用量43和65 g为基准,在此基础上进行调整值()计算得出调控施氮处理每个时期的尿素施用量;()的数字下标分别代表优化施氮减量处理、优化施氮处理和常规施氮处理的对应调控施氮处理。
Note: The urea application amount of 43 and 65 g in the two periods of optimal nitrogen application treatment (N2) was taken as the benchmark, and the urea application amount in each period of adjusted nitrogen application was calculated by using the adjusted value(). The numerical subscripts ofrepresent the corresponding adjusted N application treatments of optimal N reduction treatment, optimal N application treatment, and conventional N application treatment, respectively.
分别在柑橘果实膨大期和果实转色期采集当年生春梢叶片(由上往下第2~4片叶)并测定光谱值,每棵树按照“东南西北”4个方位随机取16片叶,利用美国ASD FieldSpec 4便携式地物光谱仪结合叶片夹持器测定其反射光谱,该仪器波段值为350~2 500 nm,其中350~1 000 nm光谱采样间隔为1.4 nm,光谱分辨率为3 nm;1 000~2 500 nm光谱采样间隔为2 nm,光谱分辨率为6.5~8.5 nm。利用植被探头配合叶片夹持器黑色背景板采集叶片光谱(采集面积为3.14 cm2),将叶片置于叶片夹的叶室中,然后夹紧叶片以保证叶片水平且被测探的面积相同;每个叶片样品采集正面、叶脉中部两端对称的两个点,每点记录5条光谱,以求平均值作为该叶片的光谱值,再使用The Unscrambler X 10.4完成光谱数据的标准正态化变换(SNV)光谱预处理[29]。
叶片营养性氮(N)含量测定[30]:N包括硝态氮、铵态氮、酰胺以及各种氨基酸等小分子含氮化合物。硝态氮使用苏州科铭生物技术有限公司生产的植物硝态氮测试盒测定。在浓酸条件下,硝酸根与水杨酸反应,生成硝基水杨酸,硝基水杨酸在碱性条件下(pH值>12)呈黄色,在一定范围内,其颜色深浅与含量成正比,可在410 nm波长下测定吸光度,通过以下计算式计算硝态氮含量(2=0.999 7):
=(Δ−0.007 3)·(/样总)/0.007 8
=128.2×(Δ−0.007 3)/(2)
式中为叶片中硝态氮(NO3--N)含量,mg/kg(鲜质量);Δ样品的吸光度减去空白的吸光度;样总为加入的提取液体积,mL;为样本质量,g。
铵态氮、氨基酸和酰胺态氮的测定:采用改良的茚三酮溶液比色法,-氨基酸与水合茚三酮溶液一起加热,经氧化脱氨变成相应的-酮酸,酮酸进一步脱羧变成醛,水合茚三酮则被还原,在弱酸环境中,还原型茚三酮、氨和另一分子水合茚三酮反应,缩合生成蓝紫色物质。根据蓝紫色的深浅,在570 nm波长下测定吸光值。本实验中在茚三酮试剂中添加乙二醇并补加正丁醇和丙醇,可以克服茚三酮的不稳定性。以亮氨酸的氮含量做标准曲线[31]。
叶片结构性氮(SN)含量测定:参考LIU等[30]的方法。称量约0.4 g叶片在液氮下磨碎,加1 mL磷酸钠缓冲液(Buffer)研磨,并转移到离心管中,重复2次。通过在4 ℃下15 000离心15 min,弃上清液。将1 mL含3%SDS的磷酸盐缓冲液添加到沉淀中,然后在90 ℃的水中加热5 min。将混合物以4 500 g离心10 min。重复3次,弃上清液。将沉淀用乙醇冲洗几遍,定量滤纸过滤,将沉淀和滤纸在50 ℃下烘干,以空白定量滤纸作为对照,凯氏法定氮。
叶片全氮(N)含量测定:称取烘干磨碎后的叶片干样约0.3 g,使用凯氏定氮法测定氮含量,每份样品测定2次,取其平均值。
叶片功能性氮(N)利用式(4)计算得出[30]。
N=N−N−N(3)
式中N为叶片功能性氮含量,N为叶片全氮含量,N为叶片营养性氮含量,N为叶片结构性氮含量,单位均为g/kg。
本研究选取偏最小二乘回归(partial least squares regression,PLSR)、支持向量机回归(support vector machine,SVM)、反向传播神经网络(back-propagation neural networks,BPNN)和随机森林(random forest,RF)尝试构建叶片功能性氮含量无损监测模型。PLSR集成了主成分分析和多元线性回归的优点,可降低高光谱数据的维度、提高了模型的运算效率;SVM和BPNN是机器学习的经典方法,SVM可以通过选择不同的核函数实现非线性分类或回归任务,能很好地处理高光谱中非线性问题,BPNN可以通过反向传播算法学习复杂的特征表示,从而实现对高光谱数据的有效分类和回归。随机森林(Random Forest,RF)是一种集成学习算法,能通过随机选择训练样本和特征来构建决策树,通过决策树的集成应用,减少过拟合、提高模型准确性和稳定性[32]。
建模样品的选择与划分:利用K-fold法将整个样本分为建模集和验证集两部分,建模集和验证集比例为7∶3。分别利用上述建模方法构建柑橘叶片功能性氮含量无损监测模型。在python 3.8的Anaconda3中采用Sklearn机器学习库并自编码建立PLSR和RF模型,采用TensorFlow 2.0学习库并自编码建立BPNN和SVM模型,分别探讨其建模精度[32]。
模型评价参数为决定系数(coefficient of determination,2)和均方根误差(root-mean-square error,RMSE)。2可衡量回归模型拟合度的统计量,反映自变量对因变量的解释程度,其取值范围为0~1,越接近1表示模型对数据的拟合越好;RMSE是衡量回归模型预测误差大小的统计量,其值越小模型精度越高[33]。
产量统计:挂果数×平均单果质量=产量(kg/株)。
品质测定:单果质量采用天平(精度0.01 g)测定;果实横纵径采用游标卡尺测定,利用横纵径比值获取果形指数;果实可溶性固形物含量采用ATAGO公司的PAL-1型电子折光仪测定。
氮肥利用率测定:氮肥偏生产力=施氮肥所获得的作物产量/化肥的投入量;氮肥农学效率=(施氮区产量−空白区产量)/施氮量。
柑橘叶片不同形态氮含量、果实产量、果实品质及氮肥利用率使用Microsoft Office Excel 16和IBM SPSS Statistics 23统计分析,建模分析在Python3.8完成上编程完成,利用Origin 9.0完成科学绘图和数据分析。
2.1.1 不同施氮处理叶片的高光谱特征
柑橘果实膨大期和果实转色期不同施氮处理叶片的反射光谱如图1所示,在可见光350~700 nm波段区域内,柑橘叶片光谱反射率随施氮量增加而降低。果实膨大期在550 nm的绿峰处,不同施氮处理柑橘叶片光谱反射率差异较大,N0处理和N1处理的叶片光谱反射率均为0.17,N2处理叶片光谱反射率为0.15,N3处理叶片光谱反射率为0.13;而在近红外700~1 350 nm波段区域内,柑橘叶片光谱反射率随施氮量增加而升高,不同施氮处理间柑橘叶片光谱反射率差异较小。果实转色期在550 nm的绿峰处,N0、N1、N2、N3处理的柑橘叶片光谱反射率分别为0.12、0.11、0.11和0.11,在近红外700~1 350 nm波段区域内,柑橘叶片光谱反射率随施氮量增加而升高。
2.1.2 建模方法对叶片功能性氮含量无损监测模型的影响
将柑橘果实膨大期的231个叶片样本数据和果实转色期179个叶片样本数据划分为建模集和验证集。其中,柑橘果实膨大期建模集包含161个样本,验证集包含70个样本;柑橘果实转色期建模集包含125个样本,验证集包含54个样本。具体数据如表2所示。
利用偏最小二乘回归(PLSR)、反向传播神经网络(BPNN)、随机森林(RF)和支持向量机(SVM)分别构建柑橘果实膨大期和果实转色期叶片功能性氮含量无损监测模型,模型2结果如表3所示。对比得出,RF建模集精度最高,但是验证集精度低;总体上,利用BPNN构建叶片功能性氮含量无损监测的拟合度最高,建模集和验证集结果如图2所示。
图1 不同施氮处理柑橘叶片反射光谱特征
表2 柑橘不同生育期叶片样本功能性氮含量建模集和验证集划分
表3 不同建模方法的柑橘叶片功能性氮含量监测结果
注:PLSR为偏最小二乘回归法;BPNN为反向传播神经网络;RF为随机森林;SVM为支持向量机。2为决定系数。下同。
Note: PLSR is partial least squares regression method; BPNN, backpropagation neural network; RF is random forest; SVM is support vector machine.2is coefficient of determination. Same below.
利用柑橘叶片功能性氮含量无损监测模型反演得出柑橘果实膨大期和果实转色期叶片功能性氮含量,通过式(1)分别计算柑橘果实膨大期和果实转色期调控施氮处理需追施的尿素用量,结果如表4所示。对照施氮处理N1、N2和N3的需氮量50、100和200 g,而萌芽肥、果实膨大肥和果实转色肥用量为5∶2∶3,本试验氮肥为氮含量46%尿素提供,因此在果实膨大期分别施尿素量分别为22、43和87 g;根据模型得出,调控施氮处理Nr1、Nr2和Nr3的叶片功能氮含量分别为14.69、16.91和20.66 g/kg,调控施氮处理Nr1、Nr2和Nr3分别追施尿素57、43和13 g,较N1、N2、N3尿素用量相比,分别增加25 g、不变、减少74 g。在柑橘果实转色期,对照施氮处理N1、N2和N3施尿素量分别为32、65和130 g;根据模型和测算综合分析,调控施氮处理Nr1、Nr2和Nr3的叶片功能氮含量分别为18.60、18.82和19.73 g/kg,调控施氮处理Nr1、Nr2和Nr3分别追施尿素67、65和58 g。与对照施氮处理N1、N2、N3相比,分别增加35 g、不变、减少78 g。
注:实线为各自时期验证集的拟合线性方程,虚线为各自时期的1:1线。
表4 柑橘叶片功能性氮含量和施尿素量
具体施氮量如表5所示,对照施氮处理N1、N2和N3全年实际尿素施用量分别为108、217和434 g,调控施氮处理Nr1、Nr2和Nr3全年实际尿素施用量分别为178、217和288 g。N1果实膨大期和转色期尿素使用量分别为22和32 g,调控施氮处理Nr1的果实膨大尿素施用量和果实转色尿素施用量分别为57和67 g,Nr1的实际尿素使用量较N1多70 g。N3果实膨大期和转色期尿素使用量分别为87和130 g,调控施氮处理Nr3的果实膨大尿素施用量和果实转色尿素施用量分别为13和58 g,Nr3的实际尿素使用量较N3少146 g。N2为优化施氮处理,利用柑橘叶片功能性氮含量无损检测模型得出,Nr2功能性氮含量与N2功能性氮含量相差不大,因此N2和Nr2萌芽肥、果实膨大肥和果实转色肥的尿素施用量均为109、43和65 g。
表5 调控施氮与对照施氮处理氮肥实际施用情况
在试验开展第2年的柑橘果实成熟期统计各施氮处理果实产量,分析施氮量对于柑橘果实产量的影响,结果如表6所示。随着氮肥用量的增加,柑橘果实产量呈现先升高再降低的趋势,相比于N0处理,各施氮处理的柑橘果实产量均显著增加,其中N2处理果实产量最高,为17.80 kg/株,N0、N1和N3处理的产量分别为10.53、11.55和12.65 kg/株。调控施氮处理Nr1、Nr2和Nr3的产量分别为17.04、18.89和17.08 kg/株,相比于对照施氮处理N1和N3,调控施氮处理Nr1和Nr3的产量显著增加,分别增产5.49和4.43 kg/株(增幅为48%和40%);Nr2处理和N2处理均为优化施氮处理,果实产量无显著差异。
表6 不同施氮处理对柑橘果实品质的影响
注:在同列数据中,不同小写字母表示差异显著(<0.05)。下同。
Note: In each column, different lowercase letters show significant differences (<0.05). Same below.
在柑橘果实成熟期测定各施氮处理柑橘果实横纵径,从而计算各施氮处理柑橘果实果形指数,分析施氮量对于柑橘果实品质的影响;调控施氮处理Nr1、Nr2和Nr3的果实横径、纵径与对照施氮处理N1、N2和N3对比,均有略微增加,调控施氮处理Nr1、Nr2和Nr3的果实横径与对照施氮处理对比分别增加2.27、2.78和0.71 mm,纵径分别增加6.07、0.58和2.78 mm。如表6所示,调控施氮处理Nr1和Nr3的果形指数相比于对照施氮处理均有所增加,呈现更饱满的优质果形,调控施氮处理Nr2与对照施氮处理N2对比无显著变化。
表6所示,对照施氮处理N1、N2和N3的单果质量分别为202.89、238.86和242.86 g,调控施氮处理Nr1、Nr2和Nr3的单果质量分别为256.94、252.92和267.31 g,调控施氮处理Nr1、Nr2和Nr3与对照施氮处理相比,单果质量分别增加54.06、14.06和14.45 g。随着施氮量的增加,柑橘果实可溶性固形物含量增加,对照施氮处理N0、N1、N2和N3的果实可溶性固形物含量分别为9.77%、10.20%、11.37%和11.65%,N2和N3处理果实可溶性固形物含量显著高于N0和N1处理;调控施氮处理Nr1、Nr2和Nr3的果实可溶性固形物含量分别为10.97%、11.53%和11.60%,相比于对照施氮处理N1,调控施氮Nr1的果实可溶性固形物含量显著提高,Nr2和Nr3相比于N2和N3处理无显著变化。
调控施氮处理与对照施氮处理对氮肥偏生产力与氮肥农学效率的影响如表7所示。随着施氮量的增加,氮肥偏生产力降低。对照施氮处理下,N1、N2和N3的氮肥偏生产力分别为231.04、177.98和63.27 kg/kg,调控施氮处理Nr1、Nr2和Nr3的氮肥偏生产力分别为207.86、188.86和128.44 kg/kg,与对照施氮处理相比,调控施氮处理Nr1的氮肥偏生产力降低了10%,Nr2和Nr3分别升高了6%和103%。对照施氮处理N1、N2和N3的氮肥农学效率分别为20.35、72.63和10.60 kg/kg,调控施氮处理Nr1、Nr2和Nr3的氮肥农学效率分别为79.38、83.51和49.23 kg/kg,各施氮处理中,Nr2的氮肥农学效率最高,其次是Nr1处理。相比于对照施氮处理,调控施氮处理的氮肥农学效率均有提高,Nr1和Nr3处理相比于N1和N3处理分别提高了290%和364%,Nr2和N2差异不显著。
表7 不同施氮处理对氮肥偏生产力与农学效率的影响
对柑橘果实膨大期和果实转色期不同施氮处理的叶片高光谱进行综合分析,可见光350~700 nm波段区域内,柑橘叶片光谱反射率随施氮量增加而降低,这与WANG等[21]的研究关于梨树叶片光谱反射率在可见光波段范围内随着氮含量的增加而显著降低的结果相同。本研究柑橘叶片光谱反射率在近红外700~1 350 nm波段区域内,随施氮量增加而升高,这与岳学军等[34]在柑橘叶片近红外波段750~1 300 nm的光谱反射率随着氮含量的增加而提高的研究结果相同。本研究基于反向传播神经网络(BPNN)构建的柑橘叶片功能性氮含量无损监测模型精度较高,两个时期建模集和验证集的决定系数为0.77~0.78,与李金梦等[35]、黄双萍等[36]研究结果一致,即运用BPNN建立柑橘叶片氮含量无损监测模型精度最优。但与全东平[37]利用支持向量机(SVM)构建柑橘叶片氮含量预测模型精度优于BPNN的研究结果相反。
BPNN模型预测结果优于PLSR和SVM模型,原因可能是BPNN模型能够解释光谱变量与叶片功能性氮含量间存在的非线性关系,而PLSR是一种线性算法,没有考虑光谱变量中某些潜在的非线性信息[38];SVM该方法在小样本下有较好的建模能力,然而数据集规模较大会影响核函数确定,进而影响模型精度[39]。BPNN模型用非线性输入输出数据误差逆向传播算法训练多层前馈网络,对随机产生的权值进行优化,提高了模型的精度、稳定性及泛化能力[40]。
本文试验结果表明,随着施氮量的增加,柑橘果实的产量先增加再降低;对比调控施氮与对照施氮处理的柑橘果实产量,调控施氮处理Nr1和Nr3显著高于对照施氮处理N1(优化施氮减氮处理)和N3(常规施氮处理)的产量,分别增产48%和40%。杨宇等[41]研究结果表明,利用化学方法测定柑橘叶片氮含量并调控施氮管理能增产4.7%,WANG等[21]利用光谱技术实现无损、快速的梨叶片氮含量诊断并调控施氮管理,梨果实产量提高27%。基于光谱技术的营养诊断和调控施氮技术相较于经验施氮和化学方法测定指导施肥,能够更加快速、精确诊断树体氮营养状况并指导施肥,从而提高果树产量、品质。
在柑橘果实品质方面,各调控施氮处理的柑橘果实横纵径增加不显著,果形指数差异不显著,这与李旭[22]对研究结果相似,即不同氮处理显著影响树体挂果量,对果形指数影响差异不显著。本研究结果表明,调控施氮处理的柑橘单果质量与对照施氮处理对比均有增加,其中Nr1增幅最大(26.6%),这与张磊等[13]基于叶片氮营养诊断的苹果精准施肥模型研究的结果相似,即模型施肥处理比经验施肥处理单果质量增加15.5%。本研究结果表明,随着施氮量增加,柑橘果实可溶性固形物含量增加。相比于对照施氮处理N1,调控施氮Nr1的果实可溶性固形物含量显著提高1.20个百分点,Nr3相比于N3处理无显著差异。杨江波等[23]探究不同施氮对塔罗科血橙果实可溶性固形物的影响研究结果表明,果实可溶性固形物随施氮量增加而增加,然而过量施氮对果实可溶性固形物增加不显著。
Nr1和Nr2的氮肥偏生产力对比N1和N2差异不显著。然而相比于对照施氮处理N3,Nr3的氮肥偏生产力提高了103%(65.17 kg/kg),通过柑橘叶片功能性氮含量无损监测模型并调控施氮后,能减少Nr3的施氮量、同时提高产量,氮肥偏生产力增幅较大,与赵帅翔等[42]比较不同技术的氮肥偏生产力结果相似,采用减氮增效配套技术后苹果产量提高,施氮量降低,从而氮肥偏生产力普遍较高。调控施氮处理Nr1和Nr3的氮肥农学效率分别提高了290%(59.03 kg/kg)、364%(38.63 kg/kg),与韩佳乐等[43]研究结果相似,基于生长模型和15N示踪的优化减氮处理能提高苹果氮肥利用率(高于常规高氮处理84.92%~178.35%)。调控施氮能够有效地提高氮肥利用率,本文基于柑橘叶片功能性氮含量无损监测模型的调控施氮方法,为柑橘氮素管理提供有效的技术手段。
本文利用高光谱技术构建柑橘叶片功能性氮含量无损监测模型,结果表明:
1)反向传播神经网络对柑橘果实膨大期和果实转色期叶片功能性氮含量预测精度较高,两个期2分别为0.78和0.77,RMSE分别为0.82和1.04 g/kg。
2)随着施氮量的增加,柑橘果实的产量先增加再降低;调控施氮处理Nr1和Nr3(分别依据的优化施氮减氮处理N1和常规施氮处理N3进行调控,调控前施氮量分别为50和200 g/株)分别增产48%和40%。相比于对照施氮处理,调控施氮处理显著提高单果质量和可溶性固形物含量,但果实横纵径和果形指数差异不显著。
3)调控施氮Nr3处理的氮肥偏生产力与对照施氮(N3)对比增加103%,调控施氮处理Nr1和Nr3的氮肥农学效率与对照施氮处理(N1和N3)对比增加290%和364%。
基于高光谱的柑橘叶片功能性氮无损监测模型的调控施氮方法,能在一定程度上减少萌芽期施氮不足或过量对柑橘产量和品质的影响,提高氮肥偏生产力和农学效率。
[1] 鲁剑巍. 湖北省柑橘园土壤—植物养分状况与柑橘平衡施肥技术研究[D]. 武汉:华中农业大学,2003.
LU Jianwei. Study on Soil and Plant Nutrient Status and Balanced Fertilization Technology of Citrus Orchards in Hubei[D]. Wuhan: Huazhong Agricultural University, 2003. (in Chinese with English abstract)
[2] 周鑫斌,石孝均,孙彭寿,等. 三峡重庆库区柑橘园土壤养分丰缺状况研究[J]. 植物营养与肥料学报,2010,16(4):817-823.
ZHOU Xinbin, SHI Xiaojun, SUN Pengshou, et al. Status of soil fertility in citrus orchards of Chongqing Sanxia Reservoir Area[J]. Plant Nutrition and Fertilizer Science, 2010, 16(4): 817-823. (in Chinese with English abstract)
[3] LI Y J, YANG M, ZHANG Z Z, et al. An ecological research on potential for zero-growth of chemical fertilizer use in citrus production in China[J]. Ekoloji, 2019, 28(107): 1049-1059.
[4] 武松伟,梁珊珊,胡承孝,等. 我国柑橘园“因土补肥”与化肥减施增效生态分区[J]. 华中农业大学学报(自然科学版),2022,41(2):9-19.
WU Songwei, LIANG Shanshan, HU Chengxiao, et al. Ecological region division of soil based supplementary fertilization and decrement fertilization in China citrus orchards[J]. Journal of Huazhong Agricultural University, 2022, 41(2): 9-19. (in Chinese with English abstract)
[5] FENG D Y, XU W H, HE Z M, et al. Advances in plant nutrition diagnosis based on remote sensing and computer application[J]. Neural Computing and Applications, 2020, 32: 16833-16842.
[6] 魏鹏飞,徐新刚,李中元,等. 基于无人机多光谱影像的夏玉米叶片氮含量遥感估测[J]. 农业工程学报,2019,35(8):126-133.
WEI Pengfei, XU Xingang, LI Zhongyuan, et al. Remote sensing estimation of nitrogen content in summer maize leaves based on multispectral images of UAV[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2019, 35(8): 126-133. (in Chinese with English abstract)
[7] 王玉娜,李粉玲,王伟东,等. 无人机高光谱的冬小麦氮素营养监测[J]. 农业工程学报,2020,36(22):31-39.
WANG Yuna, LI Fenling, WANG Weidong, et al. Monitoring of winter wheat nitrogen nutrition based on UAV hyperspectral images[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(22): 31-39. (in Chinese with English abstract)
[8] 洪波,张泽,张强,等. 基于数码图像的棉花叶片氮含量估测研究[J]. 中国农学通报,2022,38(9):49-55.
HONG Bo, ZHANG Ze, ZHANG Qiang, et al. The nitrogen content in cotton leaves: Estimation based on digital image[J]. Chinese Agricultural Science Bulletin, 2022, 38(9): 49-55. (in Chinese with English abstract)
[9] 吴彤,李勇,葛莹,等. 利用Stacking集成学习估算柑橘叶片氮含量[J]. 农业工程学报,2021,37(13):163-171.
WU Tong, LI Yong, GE Ying, et al. Estimation of nitrogen contents in citrus leaves using Stacking ensemble learning[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2021, 37(13): 163-171. (in Chinese with English abstract)
[10] 刘雪峰,吕强,何绍兰,等. 柑橘植株冠层氮素和光合色素含量近地遥感估测[J]. 遥感学报,2015,19(6):1007-1018.
LIU Xuefeng, LYU Qiang, HE Shaolan, et al. Estimation of nitrogen and pigments content in citrus canopy by low-altitude remote sensing[J]. Journal of Remote Sensing, 2015, 19(6): 1007-1018. (in Chinese with English abstract)
[11] 易时来,邓烈,何绍兰,等. 锦橙叶片氮含量可见近红外光谱模型研究[J]. 果树学报,2010,27(1):13-17.
YI Shilai, DENG Lie, HE Shaolan, et al. Research on nitrogen content of leaf of Jincheng orange cultivar using visible near infrared spectroscopy model[J]. Journal of Fruit Science, 2010, 27(1): 13-17. (in Chinese with English abstract)
[12] MENESATTI P, ANTONUCCI F, PALLOTTINO F, et al. Estimation of plant nutritional status by Vis-NIR spectrophotometric analysis on orange leaves [(L) Osbeck cv Tarocco][J]. Biosystems Engineering, 2010, 105(4): 448-454.
[13] 张磊,张宏建,孙林林,等. 基于叶片营养诊断的苹果园果树精准施肥模型研究[J]. 中国土壤与肥料,2019(6):212-222.
ZHANG Lei, ZHANG Hongjian, SUN Linlin. Research on precision fertilization model in apple orchard based on leaf nutrition diagnosis[J]. Soil and Fertilizer in China, 2019(6): 212-222. (in Chinese with English abstract)
[14] ZHANG C, LIU F, KONG W, et al. Application of visible and near-infrared hyperspectral imaging to determine soluble protein content in oilseed rape leaves[J]. Sensors, 2015, 15(7): 16576-16588.
[15] MISHRA P, SHAHRIMIE M, ASAARI M, et al. Close range hyperspectral imaging of plants: A review[J]. Biosystems Engineering, 2017, 164: 49-67.
[16] BARÁNKOVÁ B, LAZÁR D, NAUŠ J. Analysis of the effect of chloroplast arrangement on optical properties of green tobacco leaves[J]. Remote Sensing of Environment, 2016, 174: 181-196.
[17] 杨铁钢. 不同品种和氮素水平下小麦高效氮素利用的生理机制研究[D]. 南京:南京农业大学,2007.
YANG Tiegang. Study on Physiological Mechanism for High Efficient Nitrogen Use in Wheat[D]. Nanjing: Nanjing Agricultural University, 2007. (in Chinese with English abstract)
[18] 张锡洲,吴沂珀,李廷轩. 不同施氮水平下不同氮利用效率小黑麦植株氮素积累分配特性[J]. 中国生态农业学报,2014,22(2):151-158.
ZHANG Xizhou, WU Xinpo, LI Tingxuan. Accumulation and distribution of nitrogen in triticale varieties with different nitrogen utilization efficiencies under different nitrogen application levels[J]. Chinese Journal of Eco-Agriculture, 2014, 22(2): 151-158. (in Chinese with English abstract)
[19] AINSWORTH E A, SERBIN S P, SKONECZKA J A, et al. Using leaf optical properties to detect ozone effects on foliar biochemistry[J]. Photosynthesis Research, 2014, 119(1/2): 65-76.
[20] 陈青春,田永超,姚霞,等. 基于冠层反射光谱的水稻追氮调控效应研究[J]. 中国农业科学,2010,43(20):4149-4157.
CHEN Qingchun, TIAN Yongchao, YAO Xia, et al. Effects of nitrogen fertilizer top-dressing based on canopy reflectance spectra in rice[J]. Scientia Agricultura Sinica, 2010, 43(20): 4149-4157. (in Chinese with English abstract)
[21] WANG J, SHI X J, XU Y C, et al. Nitrogen management based on visible/near infrared spectroscopy in pear orchards[J]. Remote Sensing, 2021, 13: 927.
[22] 李旭. 减氮施肥对柑橘树体氮素含量、果实品质产量和氮肥利用的影响[D]. 武汉:华中农业大学,2020.
LI Xu. Effects of Reduced Nitrogen Fertilization on Nitrogen Content, Fruit Yield and Nitrogen Fertilizer Utilization of Citrus Trees[D]. Wuhan: Huazhong Agricultural University, 2020. (in Chinese with English abstract)
[23] 杨江波,张绩,李俊杰,等. 三峡重庆库区施氮水平对塔罗科血橙树体养分、产量品质及土壤理化性质的影响[J]. 中国农业科学,2019,52(5):893-908.
YANG Jiangbo, ZHANG Ji, LI Junjie, et al. Effects of nitrogen application levels on nutrient, yield and quality of tarocco blood orange and soil physicochemical properties in the three gorges area of Chongqing[J]. Scientia Agricultura Sinica, 2019, 52(5): 893-908. (in Chinese with English abstract)
[24] ATA-UL-KARIM S, LIU X, LU Z, et al. Estimation of nitrogen fertilizer requirement for rice crop using critical nitrogen dilution curve[J]. Field Crops Research, 2017, 201: 32-40.
[25] 于丰华,曹英丽,许童羽,等. 基于高光谱遥感处方图的寒地分蘖期水稻无人机精准施肥[J]. 农业工程学报,2020,36(15):103-110.
YU Fenghua, CAO Yingli, XU Tongyu, et al. Precision fertilization by UAV for rice at tillering stage in cold region based on hyperspectral remote sensing prescription map[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 35(15): 103-110. (in Chinese with English abstract)
[26] 张福锁. 作物施肥图解[M]. 北京:中国农业出版社与农村读物出版社,2006.
[27] LUKINA E V, FREEMAN K W, WYNN K J, et al. Nitrogen fertilization optimization algorithm based on in-season estimates of yields and plant nitrogen uptake[J]. Journal of Plant Nutrition, 2001, 24: 885-898.
[28] 杨群. 基于可见/近红外光谱的柑橘叶片不同形态氮含量无损监测研究[D]. 重庆:西南大学,2022.
YANG Qun. Nondestructive Monitoring of Different Forms of Nitrogen Content in Citrus Leaves Based on Visible/Near Infrared Spectroscopy[D]. Chongqing: Southwest University, 2022. (in Chinese with English abstract)
[29] BI Y M, YUAN K L, XIAO W Q, et al. A local pre-processing method for near-infrared spectra, combined with spectral segmentation and standard normal variate transformation[J]. Analytica Chimica Acta, 2016, 909: 30-40.
[30] LIU T, REN T, WHITE P, et al. Storage nitrogen co-ordinates leaf expansion and photosynthetic capacity in winter oilseed rape[J]. Journal of Experimental Botany, 2018, 69(12): 2995-3007.
[31] 郑超,夏海鸣,周佳骋,等. 茚三酮分光光度法测定水中氨氮[J]. 分析试验室,2015,34(3):359-361.
ZHENG Chao, XIA Haiming, ZHOU Jiapin, et al. A novel method for determination of ammonia nitrogen in water[J]. Chinese Journal of Analysis Laboratory, 2015, 34(3): 359-361. (in Chinese with English abstract
[32] LIU X F, LYU Q, HE S L, et al. Estimation of carbon and nitrogen contents in citrus canopy by low-altitude remote sensing[J]. International Journal of Agricultural and Biological Engineering, 2016, 9(5): 149-157.
[33] VERRELST J, MALENOVSKY Z, VAN D, et al. Quantifying vegetation biophysical variables from imaging spectroscopy data: A review on retrieval methods[J]. Surveys in Geophysics, 2019, 40: 589-629.
[34] 岳学军,全东平,洪添胜,等. 基于流形学习算法的柑橘叶片氮含量光谱估测模型[J]. 农业机械学报,2015,46(6):244-250.
YUE Xuejun, QUAN Dongping, HONG Tiansheng, et al. Estimation model of nitrogen content for citrus leaves by spectral technology based on manifold learning algorithm[J]. Transactions of the Chinese Society for Agricultural Machinery, 2015, 46(6): 244-250. (in Chinese with English abstract
[35] 李金梦,叶旭君,王巧男,等. 高光谱成像技术的柑橘植株叶片含氮量预测模型[J]. 光谱学与光谱分析,2014,34(1):212-216.
LI Jinmeng, YE Xujun, WANG Qiaonan, et al. Development of prediction models for determining n content in citrus leaves based on hyperspectral imaging technology[J]. Spectroscopy and Spectral Analysis, 2014, 34(1): 212-216. (in Chinese with English abstract)
[36] 黄双萍,洪添胜,岳学军,等. 基于高光谱的柑橘叶片氮素含量多元回归分析[J]. 农业工程学报,2013,29(5):132-138.
HUANG Shuangping, HONG Tiansheng, YUE Xuejun, et al. Multiple regression analysis of citrus leaf nitrogen content using hyperspectral technology[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2013, 29(5): 132-138. (in Chinese with English abstract)
[37] 全东平. 基于高光谱的柑橘叶片氮磷钾含量检测模型[D]. 广州:华南农业大学,2016.
QUAN Dongping. Measurement Model of Nitrogen, Phosphorus and Potassium Content for Citrus Leaves Based on Hyperspectrum[D]. Guangzhou: South China Agricultural University, 2016. (in Chinese with English abstract)
[38] 胡越,罗东阳,花奎,等. 关于深度学习的综述与讨论[J]. 智能系统学报,2019,14(1):1-19.
HU Yue, LUO Dongyang, HUA Kui, et al. Overview on deep learning[J]. CAAI Transactions on Intelligent Systems, 2019, 14(1): 1-19. (in Chinese with English abstract)
[39] 王霞,董永权,于巧,等. 结构化支持向量机研究综述[J]. 计算机工程与应用,2020,56(17):24-32.
WANG Xia, DONG Yongquan, YU Qiao, et al. Review of structural support vector machines[J]. Computer Engineering and Applications, 2020, 56(17): 24-32. (in Chinese with English abstract)
[40] 邹德秋,王家强,张冬冬,等. 基于光谱指数的胡杨叶片氮含量估算[J]. 森林与环境学报,2022,42(6):623-630.
ZOU Deqiu, WANG Jiaqiang, ZHANG Dongdong, et al. Estimation of nitrogen content in Populus euphratica leaves based on spectral index[J]. Journal of Forestry and Environment, 2022, 42(6): 623-630. (in Chinese with English abstract)
[41] 杨宇,邓正春,彭永胜,等. 柑橘叶片营养诊断施肥技术研究[J]. 湖南农业科学,2013(15):183-184.
YANG Yu, DENG Zhengchun, PENG Yongsheng, et al. Study on nutrient diagnostic fertilization of citrus leaves[J]. Hunan Agricultural Sciences, 2013(15): 183-184. (in Chinese with English abstract)
[42] 赵帅翔,张卫峰,姜远茂,等. 黄土高原苹果过量施氮因素分析[J]. 植物营养与肥料学报,2017,23(2):484-491.
ZHAO Shuaixiang, ZHANG Weifeng, JIANG Yuanmao, et al. Factor analysis of excessive nitrogen application in apples on Loess Plateau[J]. Plant Nutrition and Fertilizer Science, 2017, 23(2): 484-491. (in Chinese with English abstract)
[43] 韩佳乐,冯涛,朱志军,等. 生长模型和15N示踪评价施肥处理对苹果树氮肥利用的影响[J]. 农业工程学报,2021,37(15):96-104.
HAN Jiale, FENG Tao, ZHU Zhijun, et al. Evaluating the effects of fertilization treatments on the nitrogen use efficiency of apple trees using allometric model and15N tracer[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2021, 37(15): 96-104. (in Chinese with English abstract)
Adjusted nitrogen application using non-destructive monitoring model of citrus leaf functional nitrogen content
LIU Zhiye1, YANG Qun1, LING Qihan2, WEI Yong2, NING Qiang1, KONG Faming1, ZHANG Yueqiang1,2,3, SHI Xiaojun1,2,3, WANG Jie1,2,3※
(1.400715,; 2400715,; 3,400716,)
The concentration and distribution of functional nitrogen (N) in citrus leaves can be significant indicators for the formation and transportation of fruit assimilation. A non-destructive monitoring model can be used for the functional nitrogen concentration in the leaves. The N application can also be adjusted to quantify the citrus nitrogen using hyperspectral technology. The five-year ‘Chunjian’ orange was taken as the experimental material in the Changshou District of Chongqing in China. The control treatments of nitrogen application with the different gradients were set: N0, N1, N2, and N3(Nitrogen application qualities were 0, 50, 100, and 200 g/plant, respectively). The adjusted nitrogen treatments were named Nr1, Nr2,and Nr3, according to the non-destructive monitoring model for the functional nitrogen concentration in the citrus leaf. In the first year of the experiment, the leaves of the spring shoot (the second to fourth leaves from the top to the bottom) were collected at the fruit expansion and color-changed period, respectively. Sixteen leaves were randomly selected from each tree, according to the four directions of “south, east, north, west”, where the spectral values were determined simultaneously. A non-destructive monitoring model was then established for the functional nitrogen concentration in the citrus fruit leaves at the fruit expansion and color-changed period by the hyperspectral technique. In the second year, the leaf functional nitrogen concentration (LFNC) model and topdressing formula were used to calculate the actual nitrogen application ratio. The fertilizer of the actual nitrogen application ratio was applied in the adjusted N application treatments at the fruit expansion and color-changed period. A comparison was made to clarify the effects of control and adjusted nitrogen application on the yield, fruit quality, and nitrogen use efficiency. The results show that the LFNC model performed the higher accuracy using the back propagation neural network, where the2were 0.78 (fruit expansion period) and 0.77 (fruit color-changed period). The Nr1and Nr3treatments increased the yield by 5.49, and 4.43 kg/ plant with the rate of increments of 48% and 40%, respectively, compared with the N1and N3. Compared with the N1, the single fruit weight and soluble solid content of the citrus increased significantly by the adjusted N treatment Nr1. However, there was no change in the transverse and longitudinal diameter of the citrus fruits and fruit shape index between the control and adjusted N treatments. The partial factor productivity of applied (PFP-N) of adjusted N application treatments with the Nr1was 10% lower than that of the control with the N1. There was only a little change in the fruit shape index and soluble solids of Nr3. Specifically, the single fruit weight increased compared with the N3. compared with the N3. The agronomic efficiency of the Nr2and Nr3increased by 290% and 364%, compared with the N1and N3, respectively. There was no significant difference in the yield, quality, and nitrogen use efficiency between the Nr2and N2. In conclusion, the adjusted nitrogen application using the non-destructive monitoring model of the citrus leaf functional nitrogen concentration can be expected to reduce the effects of insufficient or excessive nitrogen application on the citrus yield and quality, in order to improve the nitrogen partial productivity and agronomic efficiency. The finding can provide the theoretical basis and technical support to realize the non-destructive monitoring of functional nitrogen concentration in the citrus leaves and adjusted nitrogen application.
citrus; hyperspectral; adjusted nitrogen application; leaf functional nitrogen concentration; non-destructive monitoring
2023-01-26
2023-03-29
国家自然科学基金项目(31801932)
刘智业,研究方向为植物光谱监测。Email:lzy20000124@126.com
王洁,博士,讲师,研究方向为基于近地遥感技术的植物营养无损诊断、果树养分资源管理、智慧农业系统。Email:mutouyu@swu.edu.cn
10.11975/j.issn.1002-6819.202301083
S666
A
1002-6819(2023)-07-0167-09
刘智业,杨群,凌琪涵,等. 采用柑橘叶片功能性氮含量无损监测模型的调控施氮方法[J]. 农业工程学报,2023,39(7):167-175. doi:10.11975/j.issn.1002-6819.202301083 http://www.tcsae.org
LIU Zhiye, YANG Qun, LING Qihan, et al. Adjusted nitrogen application using non-destructive monitoring model of citrus leaf functional nitrogen content[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2023, 39(7): 167-175. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.202301083 http://www.tcsae.org