基于作物生長(zhǎng)監(jiān)測(cè)診斷儀的雙季稻葉干重監(jiān)測(cè)模型

李艷大 曹中盛 舒時(shí)富 孫濱峰 葉 春 黃俊寶 朱 艷 田永超

基于作物生長(zhǎng)監(jiān)測(cè)診斷儀的雙季稻葉干重監(jiān)測(cè)模型

李艷大1,*曹中盛1舒時(shí)富1孫濱峰1葉 春1黃俊寶1朱 艷2田永超2

1江西省農(nóng)業(yè)科學(xué)院農(nóng)業(yè)工程研究所 / 江西省智能農(nóng)機(jī)裝備工程研究中心 / 江西省農(nóng)業(yè)信息化工程技術(shù)研究中心, 江西南昌 330200;2南京農(nóng)業(yè)大學(xué) / 國(guó)家信息農(nóng)業(yè)工程技術(shù)中心, 江蘇南京 210095

本文旨在驗(yàn)證作物生長(zhǎng)監(jiān)測(cè)診斷儀(crop growth monitoring and diagnosis apparatus, CGMD)監(jiān)測(cè)雙季稻長(zhǎng)勢(shì)指標(biāo)的準(zhǔn)確性, 建立基于CGMD的雙季稻葉干重監(jiān)測(cè)模型。通過實(shí)施8個(gè)不同早、晚稻品種和4個(gè)施氮水平的小區(qū)試驗(yàn), 采用CGMD獲取從分蘗期至灌漿期的冠層歸一化植被指數(shù)(normalized difference vegetation index, NDVI)、差值植被指數(shù)(differential vegetation index, DVI)和比值植被指數(shù)(ratio vegetation index, RVI), 同步采用高光譜儀(analytical spectral devices field-spec handheld 2, ASD FH2)獲取冠層光譜反射率計(jì)算NDVI、DVI和RVI; 分析2種光譜儀獲取的植被指數(shù)間的相關(guān)關(guān)系, 驗(yàn)證CGMD的測(cè)量精度, 建立基于CGMD的葉干重監(jiān)測(cè)模型, 并用獨(dú)立試驗(yàn)數(shù)據(jù)對(duì)模型進(jìn)行檢驗(yàn)。結(jié)果表明: 早、晚稻葉干重隨施氮水平的增加而增大, 隨生育進(jìn)程的推進(jìn)呈“低—高—低”動(dòng)態(tài)變化趨勢(shì); CGMD與ASD FH2獲取的NDVI、DVI和RVI呈極顯著相關(guān), 相關(guān)系數(shù)(correlation coefficient,)分別為0.9535~0.9972、0.9099~0.9948和0.9298~0.9926, 表明2種光譜儀獲取的植被指數(shù)具有高度的一致性, CGMD可替代價(jià)格昂貴的ASD FH2獲取NDVI、DVI和RVI。CGMD獲取的3個(gè)植被指數(shù)相比, RVICGMD與葉干重的相關(guān)性最高; 基于RVICGMD的冪函數(shù)模型可準(zhǔn)確地監(jiān)測(cè)葉干重, 模型建立的決定系數(shù)(determination coefficient,2)為0.8604~0.9216, 模型檢驗(yàn)的均方根誤差(root mean square error, RMSE)、相對(duì)均方根誤差(relative root mean square error, RRMSE)和分別為12.97~17.87 g m–2、4.88%~16.79%和0.9951~0.9992。與人工采樣測(cè)定葉干重相比, 利用CGMD可實(shí)時(shí)準(zhǔn)確地獲取雙季稻葉干重動(dòng)態(tài)變化, 在雙季稻長(zhǎng)勢(shì)精確診斷和豐產(chǎn)高效栽培中具有應(yīng)用價(jià)值。

雙季稻; 葉干重; 作物生長(zhǎng)監(jiān)測(cè)診斷儀; 植被指數(shù); 監(jiān)測(cè)模型

水稻是中國(guó)最重要的糧食作物之一, 其產(chǎn)量和面積的豐缺將直接影響國(guó)家糧食安全與社會(huì)穩(wěn)定。江西是中國(guó)水稻主產(chǎn)省, 水稻總產(chǎn)約200億千克, 種植面積約340萬公頃, 均居全國(guó)第三, 其中雙季稻占比約89%, 居全國(guó)之首[1-2], 發(fā)展江西雙季稻生產(chǎn)對(duì)國(guó)家糧食安全具有重要的保障作用。葉干重(leaf dry weight, LDW)是表征作物長(zhǎng)勢(shì)狀況和群體質(zhì)量?jī)?yōu)劣的重要指標(biāo), 對(duì)作物光合作用、物質(zhì)生產(chǎn)和產(chǎn)量形成具有重要作用[3-4]。因此, 實(shí)時(shí)定量監(jiān)測(cè)葉干重的動(dòng)態(tài)變化對(duì)于作物田間精確管理及豐產(chǎn)高效栽培具有重要的意義[5]。

作物葉干重的常規(guī)測(cè)量方法需要進(jìn)行田間破壞性取樣和室內(nèi)烘干稱量, 雖然結(jié)果準(zhǔn)確可靠, 但費(fèi)時(shí)耗力、時(shí)效性差, 不能滿足作物生長(zhǎng)實(shí)時(shí)無損監(jiān)測(cè)診斷需求[6-7]。近年來, 具有實(shí)時(shí)、無損、快速和準(zhǔn)確等優(yōu)點(diǎn)的光譜遙感技術(shù)發(fā)展迅速, 廣泛應(yīng)用于作物長(zhǎng)勢(shì)、養(yǎng)分、水分、病蟲害和產(chǎn)量等生長(zhǎng)指標(biāo)的定量監(jiān)測(cè)[8-12]。許多學(xué)者采用光譜信息量大、測(cè)量精度高的便攜式高光譜儀和無人機(jī)載高光譜儀提取作物生長(zhǎng)指標(biāo)的特征光譜波段, 建立基于不同高光譜植被指數(shù)的作物生物量、葉面積指數(shù)、葉綠素和葉片氮累積量等監(jiān)測(cè)模型[13-16], 為作物生長(zhǎng)指標(biāo)的實(shí)時(shí)無損獲取及田間精確管理提供了新的技術(shù)手段。但便攜式高光譜儀和無人機(jī)載高光譜儀測(cè)量操作繁瑣、價(jià)格昂貴及數(shù)據(jù)處理復(fù)雜, 導(dǎo)致其在生產(chǎn)中推廣應(yīng)用可行性不強(qiáng)。因此, 許多學(xué)者研發(fā)了基于不同植被指數(shù)和不同特征光譜波段的便攜式多光譜儀[17-21], 可實(shí)時(shí)準(zhǔn)確的獲取作物長(zhǎng)勢(shì)和氮素營(yíng)養(yǎng)指標(biāo)等信息, 具有操作簡(jiǎn)便、價(jià)格適中及良好的推廣應(yīng)用價(jià)值。

許多學(xué)者利用便攜式作物生長(zhǎng)監(jiān)測(cè)診斷儀[20](crop growth monitoring and diagnosis apparatus, CGMD)建立了玉米葉面積指數(shù)和莖粗[22-23]、小麥葉面積指數(shù)和葉干重[24]等長(zhǎng)勢(shì)指標(biāo)光譜監(jiān)測(cè)模型, 具有一定的實(shí)用價(jià)值, 但因研究對(duì)象不同, 以及作物葉干重和冠層光譜信息受不同生育時(shí)期冠層結(jié)構(gòu)、栽培管理措施和生態(tài)區(qū)域等因素的影響, 導(dǎo)致構(gòu)建的監(jiān)測(cè)模型形式、模型參數(shù)及適用性等均存在一定差異, 需要進(jìn)行本地化建模和驗(yàn)證完善, 進(jìn)而提高監(jiān)測(cè)模型的預(yù)測(cè)精度和可靠性。雙季稻生產(chǎn)季節(jié)緊、用工多、區(qū)域性強(qiáng), 實(shí)時(shí)無損獲取葉干重等長(zhǎng)勢(shì)指標(biāo)信息對(duì)雙季稻精確診斷與省工節(jié)本生產(chǎn)具有重要的現(xiàn)實(shí)意義。目前基于CGMD的雙季稻葉干重定量監(jiān)測(cè)的研究鮮有報(bào)道。為此, 本文以雙季稻為研究對(duì)象, 通過實(shí)施不同品種、施氮水平和生態(tài)點(diǎn)的田間小區(qū)試驗(yàn), 采用CGMD和高光譜儀(analytical spectral devices field-spec handheld 2, ASD FH2)同步獲取不同生育時(shí)期雙季稻的冠層植被指數(shù)與葉干重, 比較分析2種光譜儀獲取的冠層植被指數(shù)間的相關(guān)關(guān)系, 建立基于CGMD的雙季稻不同生育時(shí)期葉干重監(jiān)測(cè)模型, 旨在為雙季稻長(zhǎng)勢(shì)實(shí)時(shí)精確診斷和豐產(chǎn)高效栽培提供理論基礎(chǔ)與技術(shù)支持。

1 材料與方法

1.1 試驗(yàn)設(shè)計(jì)

試驗(yàn)I: 于2016年和2017年3月至11月在江西省南昌市南昌縣, 開展不同早、晚稻品種和施氮水平的小區(qū)試驗(yàn)。試驗(yàn)田耕作層土壤含有機(jī)質(zhì)26.50 g kg–1、全氮1.98 g kg–1、堿解氮145.60 mg kg–1、速效磷18.50 mg kg–1和速效鉀100.85 mg kg–1。采用裂區(qū)設(shè)計(jì), 主區(qū)為品種, 副區(qū)為氮肥。早、晚稻均設(shè)2個(gè)品種和4個(gè)施氮水平, 株行距14 cm×24 cm, 3本移栽, 南北行向, 小區(qū)間以埂相隔, 獨(dú)立排灌, 小區(qū)面積35 m2, 3次重復(fù)。早稻4個(gè)施氮水平分別為純氮0、75、150和225 kg hm–2(分別記為N0、N1、N2和N3), 供試早稻品種為“中嘉早17”(記為C1)和“潭兩優(yōu)83”(記為C2), 3月23日播種, 4月22日移栽, 7月21日收獲。晚稻4個(gè)施氮水平分別為純氮0、90、180和270 kg hm–2(分別記為N0、N1、N2和N3), 供試晚稻品種為“天優(yōu)華占”(記為C3)和“岳優(yōu)9113”(記為C4), 6月25日播種, 7月24日移栽, 10月28日收獲。早、晚稻各小區(qū)的鉀肥和磷肥施用量一致, 分別采用氯化鉀和鈣鎂磷肥, 用量分別為150 kg hm–2(K2O)和75 kg hm–2(P2O5); 氮肥采用尿素。氮肥和鉀肥均按基肥40%、分蘗肥30%和穗肥30%施用, 磷肥作基肥一次施用。其他栽培措施與當(dāng)?shù)馗弋a(chǎn)栽培一致。

試驗(yàn)II: 于2017年3月至11月在江西省吉安市新干縣, 開展不同早、晚稻品種和施氮水平的小區(qū)試驗(yàn)。試驗(yàn)田耕作層土壤含有機(jī)質(zhì)25.50 g kg–1、全氮1.78 g kg–1、堿解氮135.50 mg kg–1、速效磷15.40 mg kg–1和速效鉀95.55 mg kg–1。采用裂區(qū)設(shè)計(jì), 主區(qū)為品種, 副區(qū)為氮肥。早、晚稻均設(shè)2個(gè)品種和4個(gè)施氮水平。供試早稻品種為“株兩優(yōu)1號(hào)”(記為C5)和“淦鑫203”(記為C6), 3月25日播種, 4月24日移栽, 7月17日收獲。供試晚稻品種為“五豐優(yōu)T025”(記為C7)和“泰優(yōu)398”(記為C8), 7月1日播種, 7月30日移栽, 10月28日收獲。早、晚稻4個(gè)施氮水平、株行距、行向、小區(qū)面積、重復(fù)數(shù)、氮磷鉀肥類型和用量均與試驗(yàn)I相同。其他栽培措施與當(dāng)?shù)馗弋a(chǎn)栽培一致。

1.2 測(cè)定項(xiàng)目與方法

1.2.1 ASD FH2植被指數(shù)測(cè)定 利用美國(guó)生產(chǎn)的ASDFH2高光譜儀(analytical spectral devices field-spec handheld 2), 具體參數(shù)為: 波長(zhǎng)范圍325~1075 nm, 視場(chǎng)角25°, 采樣間隔1.4 nm, 分辨率3 nm。于分蘗期、拔節(jié)期、孕穗期、抽穗期和灌漿期, 選擇晴朗、微風(fēng)或無風(fēng)天氣測(cè)定每個(gè)小區(qū)的冠層光譜反射率, 測(cè)量時(shí)間為10:00—14:00。測(cè)量時(shí)探頭垂直向下, 距離冠層1 m, 每個(gè)小區(qū)測(cè)定前使用標(biāo)準(zhǔn)白板進(jìn)行校正, 每個(gè)小區(qū)測(cè)量3個(gè)點(diǎn), 每點(diǎn)記錄5個(gè)采樣光譜, 取平均值作為該小區(qū)測(cè)量值。測(cè)定后的光譜分辨率經(jīng)儀器自帶軟件重采樣為1 nm, 提取810 nm和720 nm波段光譜反射率值, 計(jì)算歸一化植被指數(shù)(normalized difference vegetation index, NDVI)、差值植被指數(shù)(differential vegetation index, DVI)和比值植被指數(shù)(ratio vegetation index, RVI), 計(jì)算方法如下:

NDVIASD=(810–720)/(810+720) (1)

DVIASD=810–720(2)

RVIASD=810/720(3)

式中, NDVIASD、DVIASD和RVIASD分別為ASDFH2測(cè)量計(jì)算的NDVI、DVI和RVI;810和720分別為810 nm和720 nm波段光譜反射率。

1.2.2 CGMD植被指數(shù)測(cè)定 與ASDFH2植被指數(shù)測(cè)定同步, 利用南京農(nóng)業(yè)大學(xué)研發(fā)的作物生長(zhǎng)監(jiān)測(cè)診斷儀[20](記為CGMD), 具體參數(shù)為: 有810 nm和720 nm 2個(gè)波段, 視場(chǎng)角27°。測(cè)定值包括NDVI、DVI和RVI (分別記為NDVICGMD、DVICGMD和RVICGMD)。測(cè)量時(shí)探頭垂直向下, 距離冠層1 m, 各小區(qū)觀測(cè)3個(gè)點(diǎn), 每個(gè)點(diǎn)重復(fù)測(cè)量5次, 取平均值作為該小區(qū)測(cè)量值。

1.2.3 葉干重測(cè)定 與ASDFH2植被指數(shù)測(cè)定同步, 通過測(cè)定株高和莖蘗數(shù)等方式, 在每個(gè)小區(qū)取樣長(zhǎng)勢(shì)一致的代表性稻株4株, 根據(jù)植株器官發(fā)育情況, 將樣品植株分離為葉、莖鞘和穗, 在105℃殺青30 min, 80℃烘干48 h至恒重, 將各器官分別稱量, 根據(jù)密度計(jì)算單位土地面積上的葉干重(LDW, g m–2)。

1.3 數(shù)據(jù)處理分析及模型建立與檢驗(yàn)

Microsoft Excel 2010進(jìn)行數(shù)據(jù)整理, 利用SAS 8.0軟件進(jìn)行相關(guān)分析和方差分析; 利用ViewSpec Pro軟件對(duì)ASD FH2獲取的冠層光譜反射率進(jìn)行預(yù)處理。對(duì)2016年和2017年的試驗(yàn)數(shù)據(jù)進(jìn)行差異顯著性分析表明, 2年試驗(yàn)數(shù)據(jù)差異不顯著, 試驗(yàn)結(jié)果趨勢(shì)相同。因此, 圖1和表1僅列出試驗(yàn)I的2017年數(shù)據(jù)。試驗(yàn)I的2016年觀測(cè)數(shù)據(jù)用于葉干重監(jiān)測(cè)模型的建立, 試驗(yàn)I和試驗(yàn)II的2017年觀測(cè)數(shù)據(jù)用于監(jiān)測(cè)模型的檢驗(yàn)。以植被指數(shù)為自變量、葉干重為因變量, 利用Microsoft Excel 2010軟件將植被指數(shù)與葉干重分別進(jìn)行指數(shù)、線性、冪函數(shù)、對(duì)數(shù)和多項(xiàng)式擬合分析, 篩選建立相關(guān)性最高的LDW監(jiān)測(cè)模型。采用國(guó)際上常用的均方根誤差(root mean square error, RMSE)、相對(duì)均方根誤差(relative root mean square error, RRMSE)和相關(guān)系數(shù)(correlation coefficient,) 3個(gè)指標(biāo)來評(píng)價(jià)模型的監(jiān)測(cè)精度和可靠性, 并繪制觀測(cè)值和預(yù)測(cè)值間的1∶1關(guān)系圖直觀顯示模型擬合度和預(yù)測(cè)效果。RMSE、RRMSE和的計(jì)算方法如下:

2 結(jié)果與分析

2.1 早、晚稻葉干重的變化特征

由圖1可知, 施氮水平對(duì)早、晚稻葉干重均有顯著影響。不同生育期早、晚稻品種的葉干重均隨施氮水平的增加而增大, 同一品種不同施氮水平間差異顯著。如早稻品種“中嘉早17”(C1)拔節(jié)期N0、N1、N2和N3的葉干重分別為99.14、148.11、181.04和215.65 g m–2。N0處理由于不施氮肥, 葉干重較小, 不利于光合產(chǎn)物的積累; 而N3處理的葉干重均顯著高于其他處理, 因施氮量偏高, 易導(dǎo)致營(yíng)養(yǎng)生長(zhǎng)期延長(zhǎng)及貪青晚熟。在同一施氮水平下, 早、晚稻品種的葉干重均隨生育進(jìn)程的推進(jìn)呈“低—高—低”的動(dòng)態(tài)變化趨勢(shì), 即在生長(zhǎng)前期(分蘗期至拔節(jié)期)較低, 中期(孕穗期)達(dá)到最大值, 后期(抽穗期至灌漿期)逐漸降低。如晚稻品種“岳優(yōu)9113”(C4) N2處理分蘗期、拔節(jié)期、孕穗期、抽穗期和灌漿期的葉干重分別為119.48、232.86、307.30、260.83和232.12 g m–2。

2.2 早、晚稻CGMD植被指數(shù)與ASD FH2植被指數(shù)間的相關(guān)關(guān)系

由表1可知, CGMD與ASD FH2 2種光譜儀測(cè)定的不同生育期早、晚稻植被指數(shù)NDVI、DVI和RVI間差異不顯著,測(cè)驗(yàn)所得統(tǒng)計(jì)概率值均大于0.05 (表1); 不同生育期CGMD測(cè)定的NDVICGMD、DVICGMD、RVICGMD分別與ASD FH2測(cè)定的NDVIASD、DVIASD、RVIASD間均達(dá)極顯著相關(guān), 早、晚稻不同生育期NDVI、DVI和RVI的相關(guān)系數(shù)分別介于0.9535~0.9972、0.9099~0.9948和0.9298~0.9926。表明CGMD與ASD FH2測(cè)定的早、晚稻植被指數(shù)具有高度的一致性, CGMD具有較高的測(cè)量精度, 可替代價(jià)格昂貴的ASD FH2高光譜儀獲取早、晚稻植被指數(shù)NDVI、DVI和RVI。

表1 不同生育期早、晚稻CGMD植被指數(shù)與ASD FH2植被指數(shù)間的差異顯著性概率P和相關(guān)關(guān)系

NDVI: 歸一化植被指數(shù); DVI: 差值植被指數(shù); RVI: 比值植被指數(shù);**表示在0.01水平上顯著相關(guān)。

NDVI: normalized difference vegetation index; DVI: differential vegetation index: RVI: ratio vegetation index;**indicates significantly correlation at the 0.01 probability level.

TS: 分蘗期; JS: 拔節(jié)期; BS: 孕穗期; HS: 抽穗期; FS: 灌漿期。C1: 中嘉早17; C2: 潭兩優(yōu)83; C3: 天優(yōu)華占; C4: 岳優(yōu)9113。N0: 0 kg hm–2; N1: 早稻75 kg hm–2, 晚稻90 kg hm–2; N2: 早稻150 kg hm–2, 晚稻180 kg hm–2; N3: 早稻225 kg hm–2, 晚稻270 kg hm–2。

TS: tillering stage; JS: jointing stage; BS: booting stage; HS: heading stage; FS: filling stage. C1: Zhongjiazao 17; C2: Tanliangyou 83; C3: Tianyouhuazhan; C4: Yueyou 9113. N0: 0 kg hm–2; N1: 75 kg hm–2in early rice, 90 kg hm–2in late rice; N2: 150 kg hm–2in early rice, 180 kg hm–2in late rice; N3: 225 kg hm–2in early rice, 270 kg hm–2in late rice.

表2 基于CGMD植被指數(shù)的不同生育期早、晚稻葉干重監(jiān)測(cè)模型

LDW: 葉干重?？s略詞同表1。

LDW: leaf dry weight. Abbreviations are the same as those given in Table 1.

2.3 早、晚稻葉干重監(jiān)測(cè)模型的建立

由表2可知, 早、晚稻NDVICGMD與LDW相關(guān)性最高的模型為指數(shù)模型, 模型決定系數(shù)(2)為0.8235~0.8932; 早、晚稻DVICGMD與LDW相關(guān)性最高的模型為線性模型, 模型2為0.8162~ 0.8445; 早、晚稻RVICGMD與LDW相關(guān)性最高的模型為冪函數(shù)模型, 模型2為0.8604~0.9216; NDVICGMD、DVICGMD和RVICGMD3個(gè)植被指數(shù)相比, RVICGMD與LDW的擬合模型2值更大, 說明兩者相關(guān)性更高。

ER: 早稻; LR: 晚稻; 縮略詞同圖1。

ER: early rice; LR: late rice; Abbreviations are the same as those given in Fig. 1.

2.4 早、晚稻葉干重監(jiān)測(cè)模型的檢驗(yàn)

由圖2可知, 早、晚稻不同生育期的監(jiān)測(cè)模型對(duì)葉干重的預(yù)測(cè)效果好。早、晚稻不同生育期的葉干重觀測(cè)值和預(yù)測(cè)值之間具有一致性, 基于RVICGMD的冪函數(shù)模型預(yù)測(cè)早、晚稻不同生育期葉干重的RMSE、RRMSE和分別為12.97~17.87 g m–2、4.88%~16.79%和0.9951~0.9992。

3 討論

本研究基于不同早、晚稻品種和施氮水平的田間小區(qū)試驗(yàn), 采用CGMD和ASD FH2兩種光譜儀獲取不同生育期的冠層植被指數(shù)和葉干重?cái)?shù)據(jù), 比較分析了2種光譜儀獲取的植被指數(shù)間的相關(guān)關(guān)系。結(jié)果表明, 早、晚稻的LDW隨施氮水平的增加而增大, 隨生育進(jìn)程的推進(jìn)呈“低—高—低”動(dòng)態(tài)變化趨勢(shì), 這與前人研究結(jié)論一致[4]。這是由于在生育前期主要以營(yíng)養(yǎng)生長(zhǎng)為主, 隨著施氮水平的增加, 早、晚稻植株大量吸收氮肥, LDW逐漸增大, 孕穗期之后下部葉片開始衰亡及養(yǎng)分開始向籽粒轉(zhuǎn)運(yùn), LDW逐漸減小。說明不同施氮量對(duì)雙季稻LDW具有顯著的調(diào)控作用, 進(jìn)而影響葉片對(duì)光能的吸收利用和光合生產(chǎn)。因此, 氮肥的科學(xué)合理施用對(duì)定向調(diào)控合理的冠層結(jié)構(gòu)十分重要。

近年來, 具有實(shí)時(shí)、無損和準(zhǔn)確的高光譜技術(shù)廣泛應(yīng)用于作物長(zhǎng)勢(shì)和養(yǎng)分等生長(zhǎng)指標(biāo)的定量監(jiān)測(cè)中[25–27]。高光譜儀具有光譜信息量大、精度高和波段帶寬小等優(yōu)勢(shì), 可快速精確監(jiān)測(cè)作物生長(zhǎng)指標(biāo)動(dòng)態(tài)變化特征, 但其價(jià)格貴、數(shù)據(jù)獲取與分析處理繁瑣, 在實(shí)際生產(chǎn)應(yīng)用中具有一定的局限性。本研究分析了國(guó)產(chǎn)多光譜儀CGMD和美國(guó)進(jìn)口高光譜儀ASD FH2獲取的早、晚稻不同處理冠層植被指數(shù)間的相關(guān)關(guān)系, 結(jié)果表明, 2種光譜儀測(cè)定的NDVI、DVI和RVI間差異不顯著; 進(jìn)一步將2種光譜儀測(cè)定的NDVI、DVI和RVI進(jìn)行相關(guān)分析表明, 2種光譜儀測(cè)定的NDVI、DVI和RVI間均達(dá)極顯著相關(guān), 早、晚稻不同生育期NDVI、DVI和RVI的相關(guān)系數(shù)分別為0.9535~0.9972、0.9099~0.9948和0.9298~ 0.9926。說明CGMD與ASD FH2測(cè)定的植被指數(shù)具有高度的一致性, CGMD具有較高的測(cè)量精度, 可替代價(jià)格昂貴的ASD FH2高光譜儀測(cè)定早、晚稻植被指數(shù)NDVI、DVI和RVI, 這與前人研究結(jié)論相似[23]。

本研究采用CGMD多光譜儀獲取不同品種和不同施氮水平的試驗(yàn)數(shù)據(jù), 建立了基于NDVICGMD、DVICGMD和RVICGMD的早、晚稻LDW監(jiān)測(cè)模型, 并用獨(dú)立試驗(yàn)數(shù)據(jù)對(duì)模型進(jìn)行了檢驗(yàn)。結(jié)果表明, NDVICGMD與LDW相關(guān)性最高的模型為指數(shù)模型, DVICGMD與LDW相關(guān)性最高的模型為線性模型, RVICGMD與LDW相關(guān)性最高的模型為冪函數(shù)模型, 克服了前人研究[13,24]只進(jìn)行線性擬合分析的不足。CGMD獲取的3個(gè)植被指數(shù)相比, RVICGMD與LDW的相關(guān)性最高; 基于RVICGMD的LDW監(jiān)測(cè)模型的2為0.8604~0.9216, 模型檢驗(yàn)的RMSE、RRMSE和分別為12.97~17.87 g m–2、4.88%~16.79%和0.9951~0.9992, 比前人采用RVI(810, 560)對(duì)小麥LDW的監(jiān)測(cè)精度更高[4]。說明基于CGMD植被指數(shù)能快速準(zhǔn)確的反演早、晚稻的LDW, 建立的LDW監(jiān)測(cè)模型具有計(jì)算方便和準(zhǔn)確實(shí)用等優(yōu)點(diǎn), 是對(duì)前人研究[20]在江西雙季稻區(qū)的本地化應(yīng)用, 具有推廣應(yīng)用價(jià)值。此外, 與傳統(tǒng)人工采樣測(cè)定LDW相比, 本研究利用基于RVICGMD的LDW監(jiān)測(cè)模型, 可快速準(zhǔn)確的計(jì)算不同生育期的雙季稻LDW, 克服傳統(tǒng)人工采樣測(cè)定法費(fèi)時(shí)耗力、取樣誤差大的缺點(diǎn)。

當(dāng)然, 本研究?jī)H利用有限的試驗(yàn)數(shù)據(jù)對(duì)雙季稻LDW監(jiān)測(cè)模型進(jìn)行了檢驗(yàn), 且植被指數(shù)因品種類型、栽培措施和生態(tài)區(qū)域等不同而有所差異, 從而可能導(dǎo)致建立的LDW監(jiān)測(cè)模型的準(zhǔn)確性和適用性不廣泛。因此, 今后需進(jìn)一步通過采集不同雙季稻區(qū)、不同年份和不同處理的試驗(yàn)數(shù)據(jù)對(duì)模型進(jìn)行校正完善, 以提高監(jiān)測(cè)模型的可靠性, 從而推動(dòng)作物精確管理技術(shù)在雙季稻區(qū)的推廣應(yīng)用。

4 結(jié)論

作物生長(zhǎng)監(jiān)測(cè)診斷儀(CGMD)和高光譜儀(ASD FH2)獲取的植被指數(shù)差異不顯著, 具有高度的一致性。CGMD獲取的3個(gè)植被指數(shù)相比, RVICGMD與葉干重的相關(guān)性最高; 基于RVICGMD的冪函數(shù)模型可較準(zhǔn)確地監(jiān)測(cè)雙季稻葉干重。與傳統(tǒng)人工采樣法相比, 利用CGMD可實(shí)時(shí)準(zhǔn)確地獲取雙季稻葉干重信息, 在雙季稻豐產(chǎn)高效栽培中具有推廣應(yīng)用價(jià)值。

[1] 李艷大, 舒時(shí)富, 陳立才, 葉春, 黃俊寶, 孫濱峰, 王康軍, 曹中盛. 基于便攜式作物生長(zhǎng)監(jiān)測(cè)診斷儀的江西雙季稻氮肥調(diào)控研究. 農(nóng)業(yè)工程學(xué)報(bào), 2019, 35(2): 100–106.

Li Y D, Shu S F, Chen L C, Ye C, Huang J B, Sun B F, Wang K J, Cao Z S. Regulation of nitrogen fertilizer based on portable apparatus for crop growth monitoring and diagnosis in Jiangxi double cropping rice., 2019, 35(2): 100–106 (in Chinese with English abstract).

[2] 黃慶海, 李大明, 柳開樓, 丁小滿, 葉會(huì)財(cái), 胡志華, 余喜初, 胡秋萍, 胡惠文, 徐小林. 江西水稻清潔生產(chǎn)理論與技術(shù)實(shí)踐. 江西農(nóng)業(yè)學(xué)報(bào), 2020, 32(1): 7–12.

Huang Q H, Li D M, Liu K L, Ding X M, Ye H C, Hu Z H, Yu X C, Hu Q P, Hu H W, Xu X L. Theory and technical practice of clean production in double-cropping rice in Jiangxi province., 2020, 32(1): 7–12 (in Chinese with English abstract).

[3] 吳芳, 李映雪, 張緣園, 張雪紅, 鄒曉晨. 基于機(jī)器學(xué)習(xí)算法的冬小麥不同生育時(shí)期生物量高光譜估算. 麥類作物學(xué)報(bào), 2019, 39: 217–224.

Wu F, Li Y X, Zhang Y Y, Zhang X H, Zou X C. Hyperspectral estimation of biomass of winter wheat at different growth stages based on machine learning algorithms., 2019, 39: 217–224 (in Chinese with English abstract).

[4] 馮偉, 朱艷, 姚霞, 田永超, 曹衛(wèi)星. 基于高光譜遙感的小麥葉干重和葉面積指數(shù)監(jiān)測(cè). 植物生態(tài)學(xué)報(bào), 2009, 33: 34–44.

Feng W, Zhu Y, Yao X, Tian Y C, Cao W X. Monitoring leaf dry weight and leaf area index in wheat with hyperspectral remote sensing., 2009, 33: 34–44 (in Chinese with English abstract).

[5] 付元元, 王紀(jì)華, 楊貴軍, 宋曉宇, 徐新剛, 馮海寬. 應(yīng)用波段深度分析和偏最小二乘回歸的冬小麥生物量高光譜估算. 光譜學(xué)與光譜分析, 2013, 33: 1315–1319.

Fu Y Y, Wang J H, Yang G J, Song X Y, Xu X G, Feng H K. Band depth analysis and partial least square regression based winter wheat biomass estimation using hyperspectral measurements., 2013, 33: 1315–1319 (in Chinese with English abstract).

[6] Walter J, Edwards J, McDonald G, Kuchel H. Photogrammetry for the estimation of wheat biomass and harvest index., 2018, 216: 165–174.

[7] 馬浚誠(chéng), 劉紅杰, 鄭飛翔, 杜克明, 張領(lǐng)先, 胡新, 孫忠富. 基于可見光圖像和卷積神經(jīng)網(wǎng)絡(luò)的冬小麥苗期長(zhǎng)勢(shì)參數(shù)估算. 農(nóng)業(yè)工程學(xué)報(bào), 2019, 35(5): 183–189.

Ma J C, Liu H J, Zheng F X, Du K M, Zhang L X, Hu X, Sun Z F. Estimating growth related traits of winter wheat at seedling stages based on RGB images and convolutional neural network., 2019, 35(5): 183–189 (in Chinese with English abstract).

[8] Cheng T, Yang Z W, Inoue Y, Zhu Y, Cao W X. Preface: recent advances in remote sensing for crop growth monitoring., 2016, 8: 116.

[9] Prey L, Schmidhalter U. Sensitivity of vegetation indices for estimating vegetative N status in winter wheat., 2019, 19: 3712.

[10] Zhang L Y, Zhang H H, Niu Y X, Han W T. Mapping maize water stress based on UAV multispectral remote sensing., 2019, 11: 605.

[11] Shi Y, Huang W J, Luo J H, Huang L S, Zhou X F. Detection and discrimination of pests and diseases in winter wheat based on spectral indices and kernel discriminant analysis., 2017, 141: 171–180.

[12] Zhang K, Ge X K, Shen P C, Li W Y, Liu X J, Cao Q, Zhu Y, Cao W X, Tian Y C. Predicting rice grain yield based on dynamic changes in vegetation indexes during early to mid-growth stages., 2019, 11: 387.

[13] 賀佳, 劉冰峰, 郭燕, 王來剛, 鄭國(guó)清, 李軍. 冬小麥生物量高光譜遙感監(jiān)測(cè)模型研究. 植物營(yíng)養(yǎng)與肥料學(xué)報(bào), 2017, 23: 313–323.

He J, Liu B F, Guo Y, Wang L G, Zheng G Q, Li J. Biomass estimation model of winter wheat (L.) using hyperspectral reflectances., 2017, 23: 313–323 (in Chinese with English abstract).

[14] 韓康, 于靜, 石曉華, 崔石新, 樊明壽. 不同光譜指數(shù)反演馬鈴薯葉片氮累積量的研究. 作物學(xué)報(bào), 2020, 46: 1979–1990.

Han K, Yu J, Shi X H, Cui S X, Fan M S. Inversion of nitrogen accumulation in potato leaf with different spectral indices., 2020, 46: 1979–1990 (in Chinese with English abstract).

[15] Zou X B, Shi J Y, Hao L M, Zhao J W, Mao H P, Chen Z W, Li Y X, Holmes M.noninvasive detection of chlorophyll distribution in cucumber () leaves by indices based on hyperspectral imaging., 2011, 706: 105–112.

[16] 高林, 楊貴軍, 李長(zhǎng)春, 馮海寬, 徐波, 王磊, 董錦繪, 付奎. 基于光譜特征與PLSR結(jié)合的葉面積指數(shù)擬合方法的無人機(jī)畫幅高光譜遙感應(yīng)用. 作物學(xué)報(bào), 2017, 43: 549–557.

Gao L, Yang G J, Li C C, Feng H K, Xu B, Wang L, Dong J H, Fu K. Application of an improved method in retrieving leaf area index combined spectral index with PLSR in hyperspectral data generated by unmanned aerial vehicle snapshot camera., 2017, 43: 549–557 (in Chinese with English abstract).

[17] 鄭文剛, 孫剛, 申長(zhǎng)軍, 黃文江, 周建軍. 可見-近紅外作物氮素光電測(cè)量?jī)x開發(fā). 農(nóng)業(yè)工程學(xué)報(bào), 2010, 26(3): 178–182.

Zheng W G, Sun G, Shen C J, Huang W J, Zhou J J. Development of a visible-infrared photoelectric instrument for measuring crop nitrogen., 2010, 26(3): 178–182 (in Chinese with English abstract).

[18] 李修華, 李民贊, 崔笛, 苗宇新. 基于雙波段作物長(zhǎng)勢(shì)分析儀的東北水稻長(zhǎng)勢(shì)監(jiān)測(cè). 農(nóng)業(yè)工程學(xué)報(bào), 2011, 27(8): 206–210.

Li X H, Li M Z, Cui D, Miao Y X. Monitoring of rice plant growth in Northeast China using dual-wavebands crop growth analyzer., 2011, 27(8): 206–210 (in Chinese with English abstract).

[19] Klaus E, Bodo M, Urs S. Comparison of active and passive spectral sensors in discriminating biomass parameters and nitrogen status in wheat cultivars., 2011, 124: 74–84.

[20] 倪軍, 姚霞, 田永超, 曹衛(wèi)星, 朱艷. 便攜式作物生長(zhǎng)監(jiān)測(cè)診斷儀的設(shè)計(jì)與試驗(yàn). 農(nóng)業(yè)工程學(xué)報(bào), 2013, 29(6): 150–156.

Ni J, Yao X, Tian Y C, Cao W X, Zhu Y. Design and experiments of portable apparatus for plant growth monitoring and diagnosis., 2013, 29(6): 150–156 (in Chinese with English abstract).

[21] Kipp S, Mistele B, Schmidhalter U. The performance of active spectral reflectance sensors as influenced by measuring distance, device temperature and light intensity., 2014, 100: 24–33.

[22] Chen Q C, Zhang Z L, Liu P F, Wang X M, Jiang F. Monitoring of growth parameters of sweet corn using CGMD302 spectrometer., 2015, 16: 364–368.

[23] 賀佳, 郭燕, 王利軍, 張彥, 趙犇, 王來剛. 基于作物生長(zhǎng)監(jiān)測(cè)診斷儀的玉米LAI監(jiān)測(cè)模型研究. 農(nóng)業(yè)機(jī)械學(xué)報(bào), 2019, 50(12): 187–194.

He J, Guo Y, Wang L J, Zhang Y, Zhao B, Wang L G. Monitor model of corn leaf area index based on CGMD-402., 2019, 50(12): 187–194 (in Chinese with English abstract).

[24] 周曉楠, 黃正來, 張文靜, 馬尚宇. 基于雙波段光譜儀CGMD-302 的小麥葉面積指數(shù)和葉干重監(jiān)測(cè). 中國(guó)農(nóng)業(yè)大學(xué)學(xué)報(bào), 2017, 22(1): 102–111.

Zhou X N, Huang Z L, Zhang W J, Ma S Y. Monitoring leaf area index and leaf dry weight of winter wheat with dual-wavebands spectrometer CGMD-302., 2017, 22(1): 102–111 (in Chinese with English abstract).

[25] 陶惠林, 徐良驥, 馮海寬, 楊貴軍, 苗夢(mèng)珂, 林博文. 基于無人機(jī)高光譜長(zhǎng)勢(shì)指標(biāo)的冬小麥長(zhǎng)勢(shì)監(jiān)測(cè). 農(nóng)業(yè)機(jī)械學(xué)報(bào), 2020, 51(2): 180–191.

Tao H L, Xu L J, Feng H K, Yang G J, Miao M K, Lin B W. Monitoring of winter wheat growth based on UAV hyperspectral growth index., 2020, 51(2): 180–191 (in Chinese with English abstract).

[26] He J Y, Zhang X B, Guo W T, Pan Y Y, Yao Y, Cheng T, Zhu Y, Cao W X, Tian Y C. Estimation of vertical leaf nitrogen distribution within a rice canopy based on hyperspectral data., 2020, 10: 1–15.

[27] Guo B B, Zhu Y J, Feng W, He L, Wu Y P, Zhou Y, Ren X X, Ma Y. Remotely estimating aerial N uptake in winter wheat using red-edge area index from multi-angular hyperspectral data., 2018, 9: 1–14.

Model for monitoring leaf dry weight of double cropping rice based on crop growth monitoring and diagnosis apparatus

LI Yan-Da1,*, CAO Zhong-Sheng1, SHU Shi-Fu1, SUN Bin-Feng1, YE Chun1, HUANG Jun-Bao1, ZHU Yan2, and TIAN Yong-Chao2

1Institute of Agricultural Engineering, Jiangxi Academy of Agricultural Sciences / Jiangxi Province Engineering Research Center of Intelligent Agricultural Machinery Equipment / Jiangxi Province Engineering Research Center of Information Technology in Agriculture, Nanchang 330200, Jiangxi, China;2Nanjing Agricultural University / National Engineering and Technology Center for Information Agriculture, Nanjing 210095, Jiangsu, China

The quantitative, convenient and non-destructive monitoring of leaf dry weight (LDW) is critical for precise management in double cropping rice production. The objective of this study is to verify the accuracy of crop growth monitoring and diagnosis apparatus (CGMD, a passive multi-spectral sensor containing 810 nm and 720 nm wavelengths) in monitoring growth index of double cropping rice, and establish the model for monitoring LDW of double cropping rice based on CGMD. Plot experiments were conducted in Jiangxi province in 2016 and 2017, including eight early and late rice cultivars and four nitrogen application rates. The normalized difference vegetation index (NDVI), differential vegetation index (DVI), and ratio vegetation index (RVI) were measured at tillering, jointing, booting, heading and filling stages with two spectrometers, CGMD and analytical spectral devices field-spec handheld 2 (ASD FH2, a passive hyper-spectral sensor containing 325 nm to 1075 nm wavelengths). In order to verify the measurement precision of CGMD, the correlation relationship of vegetation indices between CGMD and ASD FH2 was analyzed. The LDW monitoring models of double cropping rice were established based on CGMD from an experimental dataset and then validated using an independent dataset involving different early and late rice cultivars and nitrogen application rates. The results indicated that the LDW of early and late rice were increased with the increase of nitrogen application rate at different growth stages, and exhibited “l(fā)ow–high–low” dynamic variation trend with early and late rice development progress. The NDVI, DVI, and RVI from CGMD and ASD FH2 were significantly correlation. The correlation coefficient () of NDVI, DVI, and RVI from CGMD and ASD FH2 were 0.9535–0.9972, 0.9099–0.9948, and 0.9298–0.9926, respectively. This result indicated that there was highly consistent of vegetation indices from CGMD and ASD FH2, and the CGMD could replace expensive ASD FH2 to measure NDVI, DVI and RVI. Compared with the three vegetation indices based on CGMD, the correlation between RVICGMDand LDW was the highest. The power function model based on RVICGMDcould accurate monitoring LDW with a determination coefficient (2) in the range of 0.8604–0.9216, the root mean square error (RMSE), relative root mean square error (RRMSE), andof model validation in the range of 12.97–17.87 g m–2, 4.88%–16.79%, and 0.9951–0.9992, respectively. Compared with the manual sampling measure LDW, CGMD method can timely and accurately measure the LDW dynamic variation of double cropping rice, which had a potential to be widely applied for growth precision diagnosis and high yield and high efficiency cultivation in double cropping rice production.

double cropping rice; leaf dry weight; crop growth monitoring and diagnosis apparatus; vegetation index; monitoring model

10.3724/SP.J.1006.2021.02077

本研究由國(guó)家重點(diǎn)研發(fā)計(jì)劃項(xiàng)目(2016YFD0300608), “萬人計(jì)劃”青年拔尖人才項(xiàng)目, 國(guó)家自然科學(xué)基金項(xiàng)目(31260293), 江西省科技計(jì)劃項(xiàng)目(20182BCB22015, 20202BBFL63044, 20192BBF60052), 江西省“雙千計(jì)劃”項(xiàng)目和江西省“遠(yuǎn)航工程”項(xiàng)目資助。

This study was supported by the National Key Research and Development Program of China (2016YFD0300608), the National Program for Support of Top-notch Young Professionals, the National Natural Science Foundation of China (31260293), the Jiangxi Science and Techno-logy Program (20182BCB22015, 20202BBFL63044, 20192BBF60052), and the Jiangxi Province “Double Thousand Plan” Program, and the Jiangxi Voyage Project.

李艷大, E-mail: liyanda2008@126.com

2020-11-16;

2021-01-13;

2021-02-19.

URL: https://kns.cnki.net/kcms/detail/11.1809.S.20210219.1523.008.html