基于便携式蜜柚光谱仪的金沙柚叶片氮含量监测模型研究

李艳大1,3,曹中盛1,3,潘玉霞1,2,孙滨峰1,黄俊宝1,杨伟光1,彭忻怡1,舒时富1,叶 春1

1江西省农业科学院农业工程研究所·江西省农业信息化工程技术研究中心·江西省智能农机装备工程研究中心,南昌 330200;2江西农业大学,南昌 330045;3井冈山红壤研究所,江西吉安 343016)

摘 要:【目的】旨在检验便携式蜜柚光谱仪(PPS)监测金沙柚叶片氮含量(LNC)的准确性,构建基于PPS 的金沙柚LNC光谱监测模型。【方法】通过实施不同施氮量的试验,于幼果期和果实膨大期利用便携式蜜柚光谱仪(PPS)、ASD高光谱仪和RapidSCAN光谱仪测定冠层光谱反射率,计算得到归一化红边指数(NDRE)与归一化植被指数(NDVI),分析比较3种光谱仪测定的冠层植被指数变化规律与相关性,检验PPS的测定精度,构建基于PPS的金沙柚LNC光谱监测模型,采用不同试验点的数据检验模型。【结果】金沙柚LNC、NDVI和NDRE随着施氮量的增加表现为递增趋势;PPS 和ASD 测定的NDVI、NDRE 间拟合的决定系数(R2)依次为0.909 5 和0.900 5,PPS 和RapidSCAN 测定的NDVI、NDRE间拟合的R2依次为0.954 3和0.900 2,证明PPS的测定结果与ASD、RapidSCAN具有很高的一致性。幼果期和果实膨大期的光谱监测模型的监测效果比生长中期好;PPS测定的NDVIPPS与NDREPPS相比,NDVIPPS与LNC之间的相关性高于NDREPPS;基于NDVIPPS的幂函数模型能准确地监测LNC,模型构建R2介于0.821 0~0.847 2之间,模型检验的相关系数(r)、相对均方根误差(RRMSE)和均方根误差(RMSE)依次为0.895 2~0.933 3、4.4%~8.1%和0.1%~0.2%。【结论】与常规化学测试LNC相比,利用PPS可实时准确获取金沙柚LNC信息,在金沙柚轻简化种植与养分无损监测诊断中具有广阔应用前景。

关键词:金沙柚;叶片氮含量;便携式蜜柚光谱仪;光谱监测;模型

金沙柚是江西省吉安市种植的优良蜜柚品种之一,种植面积和产量持续增加,成为当地果农增收致富的主导产业[1]。氮素是保证果树正常生长发育及丰产稳产所需的大量营养元素[2]。叶片氮含量(leaf nitrogen content,LNC)是反映树体氮素营养盈亏和长势状况的重要指标,对果树的光合生产、物质代谢和产量形成等都具有重要的影响[3-4]。因此,叶片氮含量的快速无损监测对科学实时指导果树种植管理、营养盈缺诊断、确定合理的施氮量及发展果树轻简化种植具有非常重要的现实意义。果树叶片氮含量测定的常规方法为室内化学分析测试,具有测定结果重现性高、直观准确等优点,但需破坏性采样、时效性弱、测定成本高,难以满足果树快速无损的叶片氮含量实时监测需要[5];目视外观诊断法虽然便捷简单,但判定误差较大、经验性较强[6]。随着光谱监测技术的快速发展,现已成为果树长势、叶片营养元素等指标的无损定量监测手段[7]。很多研究者利用光谱数据丰富、准确可靠的进口高光谱仪测定果树叶片营养元素敏感光谱波段,建立基于特征植被指数的叶片磷含量、叶绿素含量、叶片氮含量等指标光谱监测模型[8-10],为果树生长指标的无损快速测定提供了有力技术支持,但进口高光谱仪价格贵、实用性不强。为此,许多研究者研制了实用性较强、价格便宜的便携式光谱仪,能快速准确地获取叶片营养元素等指标信息[11-13]。虽然前人利用便携式光谱仪开展了许多叶片营养元素的无损监测研究,建立了许多适用准确的光谱监测模型,但由于研究区域、研究对象及种植管理条件等不一致,使得建立的监测模型的适用性、模型形式与参数等具有较大的差异,开展本地化建模研究显得尤为必要,且已有便携式光谱仪主要适用于水稻、小麦、蔬菜等大田作物,对于株高2 m 以上的成年柚树适用性差、数据获取费时费力。因此,笔者在本文中通过实施4 个施氮量的金沙柚试验,利用课题组研发的便携式蜜柚光谱仪(portable pomelo spectrometer,PPS)、美国进口的ASD 高光谱仪和RapidSCAN 光谱仪同步测定冠层植被指数,综合分析比较3 种光谱仪测定的冠层植被指数变化规律与相关性,构建基于PPS 的叶片氮含量光谱监测模型,以期为金沙柚轻简化种植与养分无损监测诊断提供技术支持。

1 材料和方法

1.1 试验设计

于2021 年在江西省吉安市吉州区(Jizhou district,JZD)和吉水县(Jishui county,JSC)实施不同施氮量的金沙柚试验。吉州区和吉水县试验点的土壤类型为红壤,分别含速效钾128.44 和117.52 mg·kg-1、速效磷30.14和28.45 mg·kg-1、全氮1.31和1.16 g·kg-1、碱解氮119.34 和107.22 mg·kg-1、有机质22.17 和20.15 g·kg-1。供试金沙柚树龄8 a(年),种植行株距5 m×4 m,3 次重复,每个重复选择4 株长势一致的相邻柚树为1 个小区,共计12 个小区和48 株柚树。综合金沙柚生长发育对肥料的需求规律及前人研究[14-15],设置4 个施氮量,依次为纯氮0(N0)、0.5(N1)、1.0(N2)和1.5(N3)kg·棵-1,氮肥(选用尿素、商品有机肥)依次按照基肥40%、萌芽肥15%、保果肥15%和壮果肥30%施用;另配K2O 1.0 kg·棵-1、P2O5 0.6 kg·棵-1,钾肥(选用硫酸钾、商品有机肥)、磷肥(选用过磷酸钙、商品有机肥)均依次按照基肥70%、壮果肥30%施用。施肥时,在树冠滴水线两侧挖条状沟,沟深20 cm左右,将肥料和园土拌匀再回填沟中,前后两次的施肥位置要错开。

1.2 测定项目与方法

1.2.1 ASD植被指数的获取 利用ASD公司(Analytical Spectral Devices)研发的便携式高光谱仪(记为ASD,产品型号FieldSpec HandHeld 2,波长范围:325~1075 nm,从美国进口),于幼果期(花后59 d)和果实膨大期(花后120 d),选取晴天,在各小区选择1株长势健康的柚树于10:00—14:00 获取其冠层的反射率数据,仪器传感器垂直离冠层1 m,数据获取前用白板校正,每株柚树东南西北4个方位各测3个点,每个点测5次,计算均值作为该棵柚树的测定结果。选出670 nm(R,红光波段)、730 nm(RE,红边波段)及780 nm(NIR,近红外波段)的反射率数据,分别按公式(1)和(2)计算得到归一化植被指数NDVI 与归一化红边指数NDRE值。

式(1)~(2)中:NDVIASD 与NDREASD 依次表示ASD获取的NDVI与NDRE;R、RE与NIR依次表示670、730与780 nm的反射率值。

1.2.2 RapidSCAN 植被指数的获取 与ASD 植被指数的获取同步,利用RapidSCAN CS-45 光谱仪(记为Rapid,包含3 个波段,即670 nm 红光波段、730 nm 红边波段及780 nm 近红外波段,从美国进口),仪器可直接获取得到NDVI 与NDRE 值(分别记为NDVIRapid和NDRERapid)。冠层植被指数获取方法与ASD相同。

1.2.3 PPS 植被指数的获取 与ASD 植被指数的获取同步,利用便携式蜜柚光谱仪(记为PPS,波长范围:400~1100 nm,笔者课题组研发)。冠层植被指数获取方法与ASD 相同。选出670 nm 红光波段、730 nm红边波段及780 nm近红外波段的反射率数据,计算得到NDVI 和NDRE 值(分别记为NDVIPPS和NDREPPS)。

1.2.4 金沙柚LNC 的获取 与ASD 植被指数获取同步,在每棵采样柚树4个不同方位(东、西、南、北)的冠层中部选择外围春梢,从上向下采摘第3、第4片叶,每个方位采集10 枚叶片,每株柚树共采集40片叶作为一个测定样品。将采摘叶片运回实验室,洗净后放入烘箱,105 ℃下杀青30 min,80 ℃下烘至恒质量,利用凯氏定氮法测定叶片氮含量[16]

1.3 监测模型的建立及检验

利用吉州区试验点数据建立监测模型,吉水县试验点数据检验监测模型。将获取的金沙柚冠层植被指数(NDVI 与NDRE)和叶片氮含量依次作为模型拟合分析的自变量与因变量,采用Excel 软件进行幂函数、多项式及指数等拟合,进而筛选建立相关性最好的监测模型。选用相关系数r、相对均方根误差RRMSE 和均方根误差RMSE 作为检验模型预测精度的评价指标,分别按公式(3)、(4)和(5)计算得到rRRMSERMSE

式(3)~(5)中:SIMiSIMiOBSiOBSin依次表示模拟值、模拟值均值、观测值、观测值均值和样本容量。

2 结果与分析

2.1 金沙柚叶片氮含量的动态变化特征

从图1 可以看出,施氮量会影响金沙柚的叶片氮含量。在不同生育期,金沙柚叶片氮含量随施氮量增加表现为递增趋势。如吉州区(JZD)试验点金沙柚果实膨大期N0、N1、N2 和N3 的叶片氮含量依次为2.36%、2.67%、2.82%和3.05%。不同生育期相比,果实膨大期的叶片氮含量比幼果期的叶片氮含量高。如吉水县(JSC)试验点金沙柚果实膨大期N2处理的叶片氮含量为2.99%,幼果期的叶片氮含量为2.71%,两者相差0.28个百分点。

图1 金沙柚叶片氮含量在不同生育期与施氮量下的动态变化特征
Fig.1 Dynamic change trends of LNC for Jinsha pomelo at different stages and nitrogen fertilizer application amounts

YFS.幼果期;FES.果实膨大期;LNC.叶片氮含量;JZD.吉州区;JSC.吉水县;N0.0 kg·棵-1;N1.0.5 kg·棵-1;N2.1.0 kg·棵-1;N3.1.5 kg·棵-1。下同。
YFS.Young fruit stage; FES. Fruit expanding stage. LNC. Leaf nitrogen content. JZD. Jizhou district ; JSC. Jishui county. N0. 0 kg·plant-1; N1.0.5 kg·plant-1;N2.1.0 kg·plant-1;N3.1.5 kg·plant-1.The same below.

2.2 金沙柚冠层植被指数的动态变化特征

从图2可以看出,利用ASD、PPS和RapidSCAN三种光谱仪测定的不同生育期金沙柚冠层植被指数随着施氮量的增加表现为递增趋势。如吉州区(JZD)试验点金沙柚幼果期N0、N1、N2和N3处理的NDVIASD依次为0.56、0.79、0.83和0.88,NDVIPPS依次为0.63、0.70、0.80和0.85,NDVIRapid依次为0.67、0.79、0.84和0.90。不同生育期相比,果实膨大期的冠层植被指数比幼果期的冠层植被指数高。如吉水县(JSC)试验点金沙柚果实膨大期N2 处理的NDREASD、NDREPPS 和NDRERapid 依次为0.22、0.22 和0.24,幼果期的NDREASD、NDREPPS和NDRERapid依次为0.17、0.20 和0.22,两者分别相差0.05、0.02 和0.02。

图2 金沙柚冠层植被指数在不同生育期与施氮量下的动态变化特征
Fig.2 Dynamic change trends of canopy vegetation indices for Jinsha pomelo at different stages and nitrogen fertilizer application amounts

NDVIASD.ASD 获取的NDVI;NDVIPPS.PPS 获取的NDVI;NDVIRapid.Rapid 获取的NDVI;NDREASD.ASD 获取的NDRE;NDREPPS.PPS 获取的NDRE;NDRERapid.Rapid 获取的NDRE。下同。
NDVIASD. NDVI obtained by ASD; NDVIPPS. NDVI obtained by PPS; NDVIRapid. NDVI obtained by Rapid; NDREASD. NDRE obtained by ASD;NDREPPS.NDRE obtained by PPS;NDRERapid.NDRE obtained by Rapid.The same below.

2.3 3种光谱仪获取的冠层植被指数之间的相关性

从图3可以看出,利用ASD、PPS和RapidSCAN三种光谱仪获取的金沙柚冠层NDVI和NDRE值依次为0.55~0.92和0.11~0.29。3种光谱仪获取的金沙柚冠层NDVI 和NDRE 间差异不显著。进一步将3种光谱仪获取的不同试验点、生育期的NDVI 与NDRE 进行相关性分析,结果显示,基于PPS 的NDVIPPS、NDREPPS 分别与基于ASD 的NDVIASD、NDREASD间线性拟合的R2依次为0.909 5、0.900 5;基于PPS 的NDVIPPS、NDREPPS分别与基于RapidSCAN 的NDVIRapid、NDRERapid间线性拟合的R2依次为0.954 3、0.900 2。证明PPS、ASD 和RapidSCAN 的测定结果一致性高。

图3 ASD、PPS 和RapidSCAN 三种光谱仪获取的冠层植被指数之间的相关性
Fig.3 Correlation between canopy vegetation indices obtained by ASD,PPS and RapidSCAN spectrometers

2.4 金沙柚叶片氮含量监测模型的构建

从表1 可以看出,金沙柚NDVIPPS与LNC 相关性最显著的模型为幂函数,模型R2介于0.821 0~0.847 2 之间;金沙柚NDREPPS与LNC 相关性最显著的模型为多项式,模型R2介于0.781 3~0.792 0之间;NDVIPPS和NDREPPS相比,NDVIPPS与LNC 间的相关性更显著。将金沙柚生长中期(幼果期和果实膨大期)的数据进行综合分析,表明生长中期的植被指数与LNC之间的相关性较幼果期和果实膨大期差,R2小于0.81。图4 为基于NDVIPPS的LNC 监测模型拟合曲线图,从图4可知,基于NDVIPPS的LNC幂函数模型R2介于0.801 8~0.847 2 之间,LNC 监测模型在不同生育期都具有较好的拟合效果。

表1 基于PPS 的金沙柚叶片氮含量监测模型
Table 1 Monitoring models of LNC for Jinsha pomelo based on PPS

植被指数Vegetation index归一化植被指数NDVIPPS归一化红边指数NDREPPS生育期Growth stage幼果期Young fruit stage,YFS果实膨大期Fruit expanding stage,FES生长中期Middle growth stage,MGS幼果期Young fruit stage,YES果实膨大期Fruit expanding stage,FES生长中期Middle growth stage,WGS监测模型Monitoring model LNC=3.235 9×NDVIPPS0.771 8 LNC=3.344 5×NDVIPPS0.814 7 LNC=3.302 3×NDVIPPS0.805 9 LNC=-56.966×NDREPPS2+28.079×NDREPPS-0.577 1 LNC=-9.735 5×NDREPPS2+10.926×NDREPPS+0.877 3 LNC=-12.679×NDREPPS2+11.533×NDREPPS+0.901 1决定系数R2 0.847 2 0.821 0 0.801 8 0.792 0 0.781 3 0.779 3

图4 不同生育期金沙柚叶片氮含量监测模型
Fig.4 Monitoring models of LNC for Jinsha pomelo at different stages

MGS.生长中期。下同。
MGS.Middle growth stage.The same below.

2.5 金沙柚叶片氮含量监测模型的检验

利用吉水县试验点获取的试验数据对本文建立的基于NDVIPPS的金沙柚叶片氮含量光谱监测模型进行检验。从图5 可知,模型对不同生育期的叶片氮含量具有较好的预测效果,模拟值和观测值间表现为较高的一致性,基于NDVIPPS的幂函数模型预测金沙柚LNC 的rRRMSERMSE 依次为0.895 2~0.933 3、4.4%~8.1%和0.1%~0.2%。

图5 不同生育期金沙柚叶片氮含量观测值和模拟值的比较
Fig.5 Comparison of simulated and observed LNC for Jinsha pomelo at different stages

3 讨 论

蜜柚是人们喜食的水果之一,广泛种植于中国南方丘陵山区[17-18]。江西吉安地区种植的金沙柚是中国南方红壤丘陵区名特优柚类品种之一。轻简化种植是当前金沙柚生产栽培的重要目标,对金沙柚叶片氮素营养盈亏状况进行无损定量监测诊断是实现此目标的前提条件之一,而LNC是表征金沙柚植株氮素营养状况和生长特征的重要指标。因此,快速、无损获取LNC 信息,对于实现金沙柚轻简化栽培及确定合理追氮处方具有十分重要的意义。

笔者在本文中开展了不同施氮量的观测试验,利用便携式蜜柚光谱仪(PPS)、ASD 高光谱仪和RapidSCAN 光谱仪获取不同处理的金沙柚冠层NDVI 和NDRE,同步采样获取叶片氮含量,分析比较3种仪器获取的NDVI、NDRE的动态变化规律与相关性。结果显示,金沙柚的LNC与冠层植被指数均随着施氮量的增加表现为递增趋势,这与前人[14]结果一致。表明不同施氮量对金沙柚LNC 具有显著的调控效应,间接影响金沙柚的产量品质形成。因此,氮肥按需投入是金沙柚获得丰产优质的关键栽培措施之一。

近年来,具有便捷高效、信息丰富和无损准确特点的光谱监测技术普遍应用于果树生长及叶片营养元素定量监测中[19-21]。虽然国外进口的光谱仪具有数据丰富、观测精度高、结果可靠等优点,可准确揭示果树长势、叶片氮素营养指标与光谱植被指数之间的规律性关系,但其售价较高、手持式,在采集株高较高的果树冠层光谱时需要借助梯子等平台获取数据,操作不便、费时费力。通过分析比较笔者课题组研发的便携式蜜柚光谱仪(PPS,其优势是可以调整纵向高度和横向距离来方便快捷地采集不同高度和方位的井冈蜜柚冠层光谱数据,且成本分别是ASD 高光谱仪和RapidSCAN 光谱仪的12%和60%),与美国进口的ASD高光谱仪、RapidSCAN光谱仪获取的金沙柚冠层植被指数间的定量结果表明,3 种光谱仪获取的NDVI 和NDRE 间差异不显著。将PPS和ASD获取的NDVI和NDRE分别进行相关分析,线性拟合的R2依次为0.909 5 和0.900 5;将PPS和RapidSCAN获取的NDVI和NDRE分别进行相关分析,线性拟合的R2依次为0.954 3和0.900 2,结果表明,3种光谱仪的测定结果一致性高,PPS 的测定结果准确可靠,能取代售价高的进口ASD 和RapidSCAN获取金沙柚冠层NDVI和NDRE数据。

通过特征波段组合构成的光谱植被指数能有效消除仪器本身及周围环境条件等的因子影响[22]。笔者在本文中通过获取金沙柚不同生育期和施氮量下的冠层植被指数与LNC 数据,建立了基于NDVIPPS与NDREPPS的金沙柚LNC 光谱监测模型,采用不同试验点的数据检验了LNC 光谱监测模型。结果显示,利用PPS 光谱仪获取的NDVIPPS 与NDREPPS 相比,NDVIPPS与LNC 间的相关性更显著;NDVIPPS与LNC之间的相关关系可用幂函数模型来定量表达,建模R2 介于0.821 0~0.847 2 之间,检验模型的rRRMSERMSE依次为0.895 2~0.933 3、4.4%~8.1%和0.1%~0.2%,比前人在温州蜜橘中采用双波段植被指数TBVI(811,856)[6]对LNC的预测结果更准确可靠,表明采用PPS可精确预测金沙柚的LNC。与常规果园采样室内化学分析测定LNC法[23]相比,笔者利用PPS 测定冠层植被指数建立LNC 光谱监测模型,从而计算金沙柚LNC,具有数据获取快捷准确、时效性强等优势,可较好克服常规测试方法费时耗工、成本高、时效性弱等不足。此外,PPS可适用于金沙柚及其他柑橘品种的叶片叶绿素含量、叶片磷含量等指标的无损监测诊断,在柑橘轻简化种植与养分无损监测诊断中具有应用前景。

笔者只利用吉水县的试验数据检验了金沙柚LNC 光谱监测模型,虽然模型预测效果表现较好,但还需要获取吉安市不同试验点、不同生育期的数据对模型进行检验评价,进而提高模型的预测效果与准确性。另外,笔者在本文中未考虑镁、铁、锰等元素对叶片光谱的影响,这有待今后进一步研究。

4 结 论

金沙柚LNC、NDVI 和NDRE 随施氮量增加表现为递增趋势;3 种光谱仪获取的植被指数具有高度的一致性,PPS 获取的NDVIPPS与LNC 相关性更显著,基于NDVIPPS的幂函数模型可较准确地监测LNC;与常规室内化学分析法相比,采用PPS可实时准确获取LNC,在生产中具有应用前景。

参考文献 References:

[1] 王芳飞.吉安市井冈蜜柚产业发展现状及建议[J].现代农业科技,2020(8):94-95.

WANG Fangfei. Development status and suggestions of Jinggang pomelo industry in Ji’an city[J]. Modern Agricultural Science and Technology,2020(8):94-95.

[2] 徐凯悦,王晓华,宋彪,郭世伟,郑朝元,郭九信,吴良泉,苏达.不同时期追施氮肥对成熟期蜜柚树体氮素分配的影响[J].植物营养与肥料学报,2021,27(4):553-564.

XU Kaiyue,WANG Xiaohua,SONG Biao,GUO Shiwei,ZHENG Chaoyuan, GUO Jiuxin,WU Liangquan,SU Da. Effects of nitrogen topdressing at different stages on nitrogen distribution in mature pomelo trees[J]. Journal of Plant Nutrition and Fertilizers,2021,27(4):553-564.

[3] SETE P B,COMIN J J,CIOTTA M N,…,BRUNETTO G.Nitrogen fertilization affects yield and fruit quality in pear[J].Scientia Horticulturae,2019,258:108782.

[4] 陈倩,丁宁,彭玲,葛顺峰,姜远茂. 供氮水平对矮化苹果15N-尿素吸收、利用、损失及产量和品质的影响[J].应用生态学报,2017,28(7):2247-2253.

CHEN Qian,DING Ning,PENG Ling,GE Shunfeng,JIANG Yuanmao. Effects of different nitrogen application rates on15N-urea absorption,utilization,loss and fruit yield and quality of dwarf apple[J]. Chinese Journal of Applied Ecology,2017,28(7):2247-2253.

[5] 易时来,邓烈,何绍兰,郑永强,姚珍珍,韦献果,李松伟,简水仙. 应用多光谱图像技术进行锦橙叶片氮含量监测[J].植物营养与肥料学报,2012,18(1):176-181.

YI Shilai,DENG Lie,HE Shaolan,ZHENG Yongqiang,YAO Zhenzhen,WEI Xianguo,LI Songwei,JIAN Shuixian. Monitoring nitrogen contents of Jincheng leaves using multi-spectrum digital image analysis technique[J]. Plant Nutrition and Fertilizer Science,2012,18(1):176-181.

[6] 王巧男,叶旭君,李金梦,肖宇钊,何勇.基于双波段植被指数(TBVI)的柑橘冠层含氮量预测及可视化研究[J].光谱学与光谱分析,2015,35(3):715-718.

WANG Qiaonan,YE Xujun,LI Jinmeng,XIAO Yuzhao,HE Yong. Estimation and visualization of nitrogen content in citrus canopy based on two band vegetation index (TBVI)[J].Spectroscopy and Spectral Analysis,2015,35(3):715-718.

[7] 岳学军,全东平,洪添胜,WEI Xiang,刘永鑫,王健.不同生长期柑橘叶片磷含量的高光谱预测模型[J].农业工程学报,2015,31(8):207-213.

YUE Xuejun,QUAN Dongping,HONG Tiansheng,WEI Xiang,LIU Yongxin,WANG Jian. Prediction model of phosphorus content for citrus leaves during different growth periods based on hyperspectrum[J]. Transactions of the Chinese Society of Agricultural Engineering,2015,31(8):207-213.

[8] 杨福芹,冯海宽,李振海,杨贵军,戴华阳.基于可见光-近红外光谱特征参数的苹果叶片氮含量预测[J].农业机械学报,2017,48(9):143-151.

YANG Fuqin,FENG Haikuan,LI Zhenhai,YANG Guijun,DAI Huayang. Prediction for nitrogen content of apple leaves using spectral features parameters from visible and near infrared lights[J]. Transactions of the Chinese Society for Agricultural Machinery,2017,48(9):143-151.

[9] 黄双萍,洪添胜,岳学军,吴伟斌,蔡坤,黎蕴玉.基于高光谱的柑橘叶片磷含量估算模型实验[J].农业机械学报,2013,44(4):202-207.

HUANG Shuangping,HONG Tiansheng,YUE Xuejun,WU Weibin,CAI Kun,LI Yunyu.Hyperspectral estimation model of total phosphorus content for citrus leaves[J].Transactions of the Chinese Society for Agricultural Machinery,2013,44(4):202-207.

[10] 岳学军,全东平,洪添胜,王健,瞿祥明,甘海明.柑橘叶片叶绿素含量高光谱无损检测模型[J].农业工程学报,2015,31(1):294-302.

YUE Xuejun,QUAN Dongping,HONG Tiansheng,WANG Jian,QU Xiangming,GAN Haiming. Non- destructive hyperspectral measurement model of chlorophyll content for citrus leaves[J].Transactions of the Chinese Society of Agricultural Engineering,2015,31(1):294-302.

[11] YAO L L,WU R S,WU S,JIANG X P,ZHU Y,CAO W X,NI J.Design and testing of an active light source apparatus for crop growth monitoring and diagnosis[J]. IEEE Access,2020,8:206474-206490.

[12] KIPP S,MISTELE B,SCHMIDHALTER U. The performance of active spectral reflectance sensors as influenced by measuring distance,device temperature and light intensity[J]. Computers and Electronics in Agriculture,2014,100:24-33.

[13] 杨彪,杜荣宇,杨玉,朱德宽,郭文川,朱新华.便携式植物叶片叶绿素含量无损检测仪设计与试验[J].农业机械学报,2019,50(12):180-186.

YANG Biao,DU Rongyu,YANG Yu,ZHU Dekuan,GUO Wenchuan,ZHU Xinhua. Design of portable nondestructive detector for chlorophyll content of plant leaves[J]. Transactions of the Chinese Society for Agricultural Machinery,2019,50(12):180-186.

[14] 戴祥生,赵晓东,张绩,罗辉,黎鑫林,刘江华,易时来.施肥处理对桃溪蜜柚养分吸收、产量及品质的影响[J]. 安徽农业科学,

2019,47(21):148-150.DAI Xiangsheng,ZHAO Xiaodong,ZHANG Ji,LUO Hui,LI Xinlin,LIU Jianghua,YI Shilai.Effects of fertilizer treatment on nutrient absorption,yield and quality of Taoxi pomelo[J].Journal of Anhui Agricultural Sciences,2019,47(21):148-150.

[15] 位高生,胡承孝,谭启玲,朱东煌,李潇彬.氮磷减量施肥对琯溪蜜柚果实产量和品质的影响[J].植物营养与肥料学报,2018,24(2):471-478.

WEI Gaosheng,HU Chengxiao,TAN Qiling,ZHU Donghuang,LI Xiaobin.The effect of nitrogen and phosphorus fertilizer reduction on yield and quality of Guanxi pomelo[J]. Journal of Plant Nutrition and Fertilizers,2018,24(2):471-478.

[16] 张志良,瞿伟菁,李小方.植物生理学实验指导[M].北京:高等教育出版社,2009:48-50.

ZHANG Zhiliang,QU Weijing,LI Xiaofang. Experimental Guidance on plant physiology[M]. Beijing:Higher Education Press,2009:48-50.

[17] 罗自威,陶晶霞,侯凯捷,张利军,陈晓辉,王玉雯,廖文强,吴良泉,李延,郭九信.养分优化管理实现蜜柚高产高效和降低碳排放[J].植物营养与肥料学报,2022,28(4):688-700.

LUO Ziwei,TAO Jingxia,HOU Kaijie,ZHANG Lijun,CHEN Xiaohui,WANG Yuwen,LIAO Wenqiang,WU Liangquan,LI Yan,GUO Jiuxin. Optimized nutrient management improves fruit yield and fertilizer use efficiency and reduces carbon emissions in pomelo production[J]. Journal of Plant Nutrition and Fertilizers,2022,28(4):688-700.

[18] 郭文武,叶俊丽,邓秀新.新中国果树科学研究70 年:柑橘[J].果树学报,2019,36(10):1264-1272.

GUO Wenwu,YE Junli,DENG Xiuxin. Fruit scientific research in New China in the past 70 years:Citrus[J].Journal of Fruit Science,2019,36(10):1264-1272.

[19] 刘雪萍,周小成,崔雅君,肖祥希.基于多特征优选的平和县蜜柚园遥感提取与扩张分析[J].贵州大学学报(自然科学版),2022,39(3):52-61.

LIU Xueping,ZHOU Xiaocheng,CUI Yajun,XIAO Xiangxi.Spatio-temporal change analysis of pomelo orchards in Pinghe county based on muti-feature optimization[J]. Journal of Guizhou University(Natural Sciences),2022,39(3):52-61.

[20] 栗方亮,孔庆波,张青,庄木来.琯溪蜜柚叶片氮素含量多种高光谱估算模型对比研究[J].果树学报,2022,39(5):882-891.

LI Fangliang,KONG Qingbo,ZHANG Qing,ZHUANG Mulai.Comparative study on several hyperspectral estimation models of nitrogen contents in Guanxi honey pomelo leaves[J]. Journal of Fruit Science,2022,39(5):882-891.

[21] 李金梦,叶旭君,王巧男,张初,何勇.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,ZHANG Chu,HE Yong. 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.

[22] 李哲,张飞,陈丽华,张海威.光谱指数的植物叶片叶绿素含量估算模型[J].光谱学与光谱分析,2018,38(5):1533-1539.

LI Zhe,ZHANG Fei,CHEN Lihua,ZHANG Haiwei. Research on spectrum variance of vegetation leaves and estimation model for leaf chlorophyll content based on the spectral index[J].Spectroscopy and Spectral Analysis,2018,38(5):1533-1539.

[23] 余璇,文婷,马青龄,刘德春,胡威,刘勇,杨莉.金沙柚果实品质与土壤、叶片、果实矿质养分的相关性分析[J].江西农业大学学报,2021,43(1):70-81.

YU Xuan,WEN Ting,MA Qingling,LIU Dechun,HU Wei,LIU Yong,YANG Li.Correlation between fruit quality and mineral nutrients in soil,leaf and fruit of‘Jinsha’pomelo[J]. Acta Agriculturae Universitatis Jiangxiensis,2021,43(1):70-81.

A model for monitoring leaf nitrogen content of Jinsha pomelo based on portable pomelo spectrometer

LI Yanda1, 3, CAO Zhongsheng1, 3, PAN Yuxia1, 2, SUN Binfeng1, HUANG Junbao1, YANG Weiguang1,PENG Xinyi1,SHU Shifu1,YE Chun1
(1Institute of Agricultural Engineering,Jiangxi Academy of Agricultural Sciences/Jiangxi Province Engineering Research Center of Information Technology in Agriculture/Jiangxi Province Engineering Research Center of Intelligent Agricultural Machinery Equipment,Nanchang 330200, Jiangxi, China;2Jiangxi Agricultural University, Nanchang 330045, Jiangxi, China;3Jinggangshan Institute of Red Soil,Jian 343016,Jiangxi,China)

Abstract:【Objective】Jinsha pomelo is one of the high-quality honey pomelo varieties planted in Ji’an city, Jiangxi province, China. The leaf nitrogen content (LNC) of Jinsha pomelo is an important index for quantitative monitoring and diagnosis of the nitrogen nutrition status and growth characteristics.Traditional LNC measurement,which mainly relies on the laboratory chemical testing,can obtain accurate and reliable results, but the applications are time-consuming and restricted to relatively small spatial scales.The spectral monitoring technology is recognized as one of the most popular means for real-time,fast and non-destructive monitoring fruit tree growth status in orchard scale.It can provide an effective technical mean for accurately analyzing the quantitative relationship between the LNC and spectral vegetation indices.Consequently,the spectral monitoring models of the LNC were established to provide a basis for non-destructive, convenient and quantitative estimation of the LNC. The previous studies proved the application of spectral monitoring technology for estimating fruit tree growth indices based on the spectral vegetation indices. Nevertheless, which spectral vegetation indices is strongly related to the LNC of Jinsha pomelo remians unclear. The objective of this study was to test the precision of the portable pomelo spectrometer (PPS, wavelengths range 400-1100 nm, which was developed by our research group) in monitoring the LNC of Jinsha pomelo, and establish the spectral monitoring model of the LNC of Jinsha pomelo.【Methods】The orchard experiments were conducted in Jizhou district and Jishui county in Ji'an city,Jiangxi province in 2021,where eight year old Jinsha pomelo with four nitrogen fertilizer application amounts(0,0.5,1.0 and 1.5 kg·plant-1)and three replications were conducted.Three spectrometers,i.e.,PPS,ASD(wavelengths range 325-1075 nm,which is made in USA)and RapidSCAN(including 670,730 and 780 nm spectral bands,which is made in USA)were set at a height of 1 m above the top of the canopy to measure canopy spectral reflectance at young fruit stage and fruit expanding stage under clear sky conditions at 10:00 to 14:00 of Beijing local time. The NDRE (normalized difference red edge) and NDVI (normalized difference vegetation index) were calculated by the canopy spectral reflectance.The canopy vegetation indices change trends were compared between PPS,ASD and RapidSCAN,and their correlations were analyzed.The LNC were measured by the traditional chemical testing in the laboratory based on the leaf samples,which were collected from the canopy middle periphery of the each sampled pomelo tree in the four directions (including south, north, east and west. Then, the correlations between the canopy vegetation indices and LNC were analyzed. The PPSbased monitoring models of the LNC were established from orchard experimental data set at Jizhou district, which was tested with orchard experimental data set at Jishui county.【Results】The LNC, NDVI and NDRE of Jinsha pomelo were increased the increase of nitrogen fertilizer application amount at different stages.The R2(determination coefficient)of fitting for the NDVI and NDRE from PPS and ASD were 0.909 5 and 0.900 5,respectively.The R2 of fitting for the NDVI and NDRE from PPS and Rapid-SCAN were 0.954 3 and 0.900 2, respectively. This result proved that the measurement results of PPS were highly consistent with ASD and RapidSCAN.The model performance for the young fruit stage or fruit expanding stage was better than that for the middle growth stages. Comparison the NDVIPPS and NDREPPS were measured by PPS, the correlation between the NDVIPPS and LNC was higher than that of the NDREPPS.The power function equation NDVIPPS-based could exactly predict the LNC with the R2 between 0.821 0-0.847 2,and the r(correlation coefficient),RRMSE(relation root mean square error)and RMSE(root mean square error)of the model test were between 0.895 2-0.933 3,4.4%-8.1%and 0.1%-0.2%,respectively.【Conclusion】In this study,the LNC and canopy vegetation indices of Jinsha pomelo presented an increasing trend the increase of the nitrogen fertilizer application amount. The canopy vegetation indices collected by the three spectrometers were highly consistent, and the power function model based on NDVIPPS could monitor the LNC of Jinsha pomelo more accurately.Compared with the traditional chemical test methods,the use of PPS could conduct in real-time and accurately get the LNC information of Jinsha pomelo. It would have wide application prospects for the high-efficient cultivation and non-destructive monitoring and diagnosis of the nutrients in Jinsha pomelo production.

Key words: Jinsha pomelo; Leaf nitrogen content; Portable pomelo spectrometer; Spectral monitoring;Model

中图分类号:S666.3

文献标志码:A

文章编号:1009-9980(2023)04-0797-10

DOI:10.13925/j.cnki.gsxb.20220415

收稿日期:2022-08-17

接受日期:2022-11-20

基金项目:江西省重大科技研发专项课题(20203ABC28W014-4);江西省重点研发计划重点项目(20212BBF61013);江西省国家级高层次人才创新创业项目;国家红壤改良工程技术研究中心开放基金资助项目(2020NETRCRSI-16);江西现代农业科研协同创新专项项目(JXXTCXQN202214)

作者简介:李艳大,男,研究员,博士,研究方向为农业信息与果树精确栽培。Tel:15870656983,E-mail:liyanda2008@126.com

*通信作者Author for correspondence.Tel:15870656983,E-mail:liyanda2008@126.com