束美艷,顧曉鶴,孫 林,朱金山,楊貴軍,王延倉
?
表征冬小麥倒伏強(qiáng)度敏感冠層結(jié)構(gòu)參數(shù)篩選及光譜診斷模型
束美艷1,2,3,4,顧曉鶴1,2,3※,孫 林4,朱金山4,楊貴軍1,2,3,王延倉5
(1. 農(nóng)業(yè)部農(nóng)業(yè)遙感機(jī)理與定量遙感重點(diǎn)實(shí)驗(yàn)室,北京農(nóng)業(yè)信息技術(shù)研究中心,北京 100097; 2. 國家農(nóng)業(yè)信息化工程技術(shù)研究中心,北京 100097;3. 北京市農(nóng)業(yè)物聯(lián)網(wǎng)工程技術(shù)研究中心,北京 100097; 4. 山東科技大學(xué)測繪科學(xué)與工程學(xué)院,青島 266590;5. 北華航天工業(yè)學(xué)院計(jì)算機(jī)與遙感信息技術(shù)學(xué)院,廊坊 065000)
針對倒伏脅迫下冬小麥冠層結(jié)構(gòu)變化規(guī)律不清、冠層光譜響應(yīng)機(jī)理不明的問題,以灌漿期倒伏冬小麥為研究對象,分析不同倒伏強(qiáng)度下冬小麥冠層結(jié)構(gòu)參數(shù)變化規(guī)律,通過光譜探測視場內(nèi)的莖、葉、穗面積比率與倒伏角度的相關(guān)性分析,篩選出表征倒伏強(qiáng)度的敏感冠層結(jié)構(gòu)參數(shù),采用傳統(tǒng)光譜變換方法與連續(xù)小波變換方法對倒伏冬小麥冠層高光譜數(shù)據(jù)進(jìn)行處理分析,篩選冠層結(jié)構(gòu)參數(shù)的敏感波段和小波系數(shù),采用偏最小二乘法構(gòu)建冠層結(jié)構(gòu)參數(shù)與高光譜特征參量的響應(yīng)模型,并利用野外實(shí)測樣本驗(yàn)證模型精度(建模集樣本28個(gè),驗(yàn)證集樣本13個(gè))。研究結(jié)果表明:倒伏后的冬小麥莖葉比與倒伏角度的相關(guān)性最高(-0.687,0.01),能較好地表征冬小麥倒伏強(qiáng)度,且莖葉比隨著倒伏角度的減小而增加;基于連續(xù)小波變換的冬小麥倒伏災(zāi)情診斷模型優(yōu)于常規(guī)光譜變換方法,檢驗(yàn)樣本的決定系數(shù)為0.632(<0.01);以冠層莖葉比預(yù)測結(jié)果進(jìn)行倒伏災(zāi)情等級劃分的精度可達(dá)84.6%。因此,不同倒伏強(qiáng)度的冠層莖葉比與冬小麥冠層光譜之間的響應(yīng)規(guī)律可以有效區(qū)分倒伏災(zāi)情等級,有助于為區(qū)域尺度的冬小麥倒伏災(zāi)情遙感監(jiān)測提供先驗(yàn)知識(shí)。
作物;災(zāi)害;預(yù)測;倒伏;冬小麥;高光譜;莖葉比;連續(xù)小波變換
近年來由于全球氣候變暖引起的極端天氣事件頻發(fā),導(dǎo)致作物倒伏發(fā)生概率也逐漸增加[1-4]。倒伏成為限制小麥高產(chǎn)優(yōu)質(zhì)的主要災(zāi)害脅迫之一。小麥倒伏后植株水分、養(yǎng)分的運(yùn)轉(zhuǎn)及光合作用都會(huì)降低,易誘發(fā)各種病蟲害,進(jìn)而對籽粒灌漿造成嚴(yán)重影響,嚴(yán)重影響冬小麥籽粒產(chǎn)量、品質(zhì)和機(jī)械收獲能力[5-8]。因此,客觀、快速和定量地監(jiān)測小麥倒伏的發(fā)生范圍和災(zāi)情程度尤為重要[9-10]。
近年來發(fā)展迅速的遙感技術(shù)為小麥倒伏災(zāi)情監(jiān)測及產(chǎn)量減損評估提供了科學(xué)的技術(shù)手段,可為小麥品種改良、田間水肥決策、農(nóng)業(yè)保險(xiǎn)快速理賠提供可靠的信息支撐。目前許多學(xué)者在利用遙感技術(shù)進(jìn)行農(nóng)作物倒伏監(jiān)測方面從不同角度進(jìn)行了探索研究。Kendall等[11]利用航空影像分析了不同倒伏時(shí)間和角度對英國油菜產(chǎn)量和含油量的影響,并在全國范圍內(nèi)評估倒伏對作物的綜合影響。Liu等[12]分析了倒伏水稻在無人機(jī)數(shù)碼影像上的顏色和紋理特征,構(gòu)建了粒子群算法和支持向量機(jī)相結(jié)合的水稻倒伏識(shí)別監(jiān)測模型。Berry等[13]通過設(shè)置不同倒伏角度分析倒伏強(qiáng)度對小麥冠層光合作用的影響,構(gòu)建小麥倒伏減產(chǎn)損失模型。Zhang等[14]利用地面高光譜數(shù)據(jù)和小波變換評估了倒伏對玉米品質(zhì)的影響。郭翠花等[15]研究分析了不同產(chǎn)量水平下小麥倒伏與莖稈力學(xué)特性的關(guān)系,最終篩選出了“彎曲強(qiáng)度”和“彈性模量”2個(gè)能作為抗倒育種的參考指標(biāo)。劉忠陽等[16]通過設(shè)置與豎直方向成30°、60°、90°共3種倒伏情況探索了冬小麥后期倒伏對其干物質(zhì)分配及產(chǎn)量的影響,發(fā)現(xiàn)灌漿初期發(fā)生倒伏小麥產(chǎn)量降低最明顯。梁玉超等[17]研究施氮量對滴灌冬小麥莖部特征及其抗倒伏性的影響,發(fā)現(xiàn)莖稈基部的機(jī)械強(qiáng)度和抗倒伏指數(shù)隨著施氮量的增加逐漸降低,隨生育期的推進(jìn)呈降低的趨勢。在作物倒伏雷達(dá)監(jiān)測方面,楊浩等[18-19]利用Radarsat-2全極化影像數(shù)據(jù),提出用雷達(dá)極化指數(shù)監(jiān)測小麥倒伏的方法。韓東等[20]利用Sentinel-1雷達(dá)數(shù)據(jù)基于后向散射系數(shù)研究構(gòu)建能有效評價(jià)玉米倒伏程度的倒伏監(jiān)測模型,并取得了較高精度,準(zhǔn)確度可達(dá)100%。目前這些研究多集中于區(qū)域尺度上監(jiān)測作物倒伏受災(zāi)范圍和災(zāi)情程度,倒伏對作物品種、產(chǎn)量及經(jīng)濟(jì)收益造成的影響,對于作物倒伏遙感監(jiān)測大多為經(jīng)驗(yàn)回歸模型為主,缺乏對作物倒伏后的冠層光譜響應(yīng)機(jī)理研究。倒伏脅迫導(dǎo)致冬小麥冠層結(jié)構(gòu)產(chǎn)生較大改變,探測視場中的莖、葉、穗面積比率及各組分受光條件的變化直接反映于冠層光譜反射率,解析不同倒伏強(qiáng)度下的冬小麥冠層結(jié)構(gòu)變化及其冠層光譜響應(yīng)機(jī)理,有助于提升光學(xué)遙感技術(shù)在冬小麥倒伏災(zāi)情監(jiān)測中的實(shí)際應(yīng)用能力。
本文以2018年河北省藁城區(qū)實(shí)發(fā)倒伏冬小麥為研究對象,獲取不同倒伏強(qiáng)度的冬小麥樣本冠層高光譜數(shù)據(jù),通過人工目視解譯勾畫方法提取各樣本冠層結(jié)構(gòu)參數(shù)信息,將其與倒伏角度進(jìn)行相關(guān)性分析,篩選表征倒伏強(qiáng)度的最佳冠層結(jié)構(gòu)參數(shù);基于連續(xù)小波變換等多種算法對冠層高光譜數(shù)據(jù)進(jìn)行數(shù)學(xué)變換,篩選敏感光譜參量,利用偏最小二乘法構(gòu)建冬小麥倒伏光譜診斷模型;利用野外實(shí)測樣本驗(yàn)證模型精度,并根據(jù)冠層結(jié)構(gòu)參數(shù)光譜預(yù)測結(jié)果進(jìn)行冬小麥倒伏災(zāi)情等級劃分。
研究區(qū)位于河北省藁城區(qū),地理坐標(biāo)為北緯37.85°-38.31°,東經(jīng)114.64°-114.98°。藁城區(qū)耕地面積約549 km2,屬暖溫帶半濕潤大陸性季風(fēng)氣候,四季分明,表現(xiàn)為夏熱冬寒的特點(diǎn),年平均氣溫和年平均降水量分別為12.5 ℃和494 mm,土壤類型主要為褐土和潮土。藁城區(qū)是典型的冀中平原冬小麥-玉米輪作區(qū),也是河北省糧食生產(chǎn)大縣。
2018年5月中旬藁城區(qū)出現(xiàn)了季節(jié)性暴風(fēng)雨,導(dǎo)致研究區(qū)內(nèi)發(fā)生了大面積冬小麥倒伏。野外觀測試驗(yàn)于2018年5月29日進(jìn)行,時(shí)屬小麥灌漿期,共計(jì)采集了41個(gè)冬小麥倒伏樣本。采用美國ASD Fieldspec Pro FR2500光譜儀測定小麥冠層光譜數(shù)據(jù),其波譜覆蓋范圍為350~2 500 nm,光譜分辨率為1 nm,儀器質(zhì)量約為8.5 kg,其最快采集速度為100 ms。測量時(shí)間為10:00-14:00,天氣晴朗、無云無風(fēng)。測量時(shí)探頭垂直于冠層上方1 m處,測定前后均用白板進(jìn)行標(biāo)定,每個(gè)樣本區(qū)測量10次,取平均值作為該樣本的冠層光譜反射率。采用ASD數(shù)據(jù)處理軟件ViewSpecPro進(jìn)行高光譜數(shù)據(jù)預(yù)處理,剔除光譜反射率大于1的異常值(3個(gè)水汽吸收帶1 350~1 400 nm、1 800~1 950 nm、2 450~2 500 nm)。
倒伏角度表示為倒伏冬小麥植株與地面的夾角,即植株傾斜程度,可表征倒伏脅迫強(qiáng)度。倒伏角度由量角器測得,其值越接近0,表示倒伏強(qiáng)度越大。將0.5 m×0.5 m的樣本框放置于倒伏小麥冠層,使用索尼A6300高清相機(jī)(有效像素為2 420萬)垂直小麥冠層50 cm處拍攝樣本框照片,每個(gè)樣本地塊隨機(jī)布設(shè)4次樣本框子區(qū)。采用人工目視解譯勾畫的方式提取每個(gè)樣本框照片中的莖、葉、穗和陰影,統(tǒng)計(jì)4個(gè)樣本子區(qū)內(nèi)各組分面積,取其均值計(jì)算該樣本莖葉穗所占比例。
本文內(nèi)主要技術(shù)流程如圖1所示。對去除噪聲后的41個(gè)倒伏小麥冠層光譜數(shù)據(jù)進(jìn)行倒數(shù)、對數(shù)、微分、倒數(shù)的對數(shù)、倒數(shù)的微分、對數(shù)的微分、倒數(shù)的對數(shù)的微分、弓曲差[21]、連續(xù)小波變換(continuous wavelet transform,CWT)9種數(shù)學(xué)變換。小波變換是源于傅里葉算法的信號處理技術(shù),可從時(shí)間與頻率上同時(shí)開展數(shù)據(jù)分析,因此可從數(shù)據(jù)中分離出更多有效信息[22]。經(jīng)過多年的發(fā)展,小波變換已在各行各業(yè)中有了廣泛的應(yīng)用,如信號處理[23-24]、遙感影像紋理[25-26]及高光譜數(shù)據(jù)降維[27]等。連續(xù)小波變換技術(shù)可將倒伏小麥冠層高光譜數(shù)據(jù)分解為不同尺度的小波系數(shù)。
圖1 冬小麥倒伏強(qiáng)度敏感冠層結(jié)構(gòu)參數(shù)篩選及光譜診斷模型構(gòu)建
式中φ,b為小波母函數(shù),為伸縮系數(shù),為平移系數(shù),為波段數(shù)。W(,)為小波系數(shù),其為由波長(350~2 500 nm)和分解尺度構(gòu)成的二維矩陣,()為冠層光譜數(shù)據(jù)。
針對9種數(shù)學(xué)變換后的冠層光譜數(shù)據(jù),篩選倒伏冠層結(jié)構(gòu)參數(shù)的敏感光譜參量,利用偏最小二乘回歸法(partial least squares regression,PLSR)構(gòu)建倒伏冠層結(jié)構(gòu)參數(shù)的光譜診斷模型。采用隨機(jī)抽樣法將樣本分為建模組與驗(yàn)證組,2/3樣本用于構(gòu)建模型,共計(jì)28個(gè)樣本,未參與建模的1/3樣本用于模型精度檢驗(yàn),共計(jì)13個(gè)樣本。采用決定系數(shù)和均方根誤差評價(jià)模型的精度。
流程具體包括:1)對野外樣本框的莖、葉、穗、陰影進(jìn)行人工目視解譯勾畫,根據(jù)莖、葉、穗面積比例構(gòu)建倒伏冠層結(jié)構(gòu)參數(shù);2)將各個(gè)倒伏冠層結(jié)構(gòu)參數(shù)與倒伏角度進(jìn)行相關(guān)性分析,篩出最適宜表征冬小麥倒伏強(qiáng)度的敏感冠層結(jié)構(gòu)參數(shù);3)利用傳統(tǒng)光譜變換及連續(xù)小波變換技術(shù)將冠層光譜數(shù)據(jù)進(jìn)行處理,通過光譜特征參量與冠層結(jié)構(gòu)參數(shù)的相關(guān)性分析,提取各光譜變換敏感于冠層結(jié)構(gòu)參數(shù)的波段及小波系數(shù);4)采用偏最小二乘法構(gòu)建倒伏冠層結(jié)構(gòu)參數(shù)的光譜診斷模型,并驗(yàn)證模型精度;5)基于冠層結(jié)構(gòu)參數(shù)光譜預(yù)測結(jié)果進(jìn)行冬小麥倒伏災(zāi)情等級劃分。
冬小麥灌漿期受到倒伏脅迫后冠層群體結(jié)構(gòu)發(fā)生了明顯改變,由于莖、葉、穗在光譜探測視場內(nèi)的貢獻(xiàn)比例發(fā)生變化,必然導(dǎo)致倒伏脅迫對小麥冠層光譜產(chǎn)生變化。選取倒伏脅迫下的原始光譜與光譜一階微分變換進(jìn)行不同脅迫強(qiáng)度的小麥倒伏光譜特征分析。
根據(jù)小麥實(shí)際倒伏情況,參照曹利萍等[28]的倒伏等級劃分,將小麥倒伏劃分為4個(gè)等級:重度倒伏(倒伏角度≤15°)、中度倒伏(15°<倒伏角度<45°)、輕度倒伏(45°≤倒伏角度<70°)和未倒伏(倒伏角度≥70°)。
圖2a為不同倒伏強(qiáng)度下冬小麥冠層光譜反射率,可以看出,不同倒伏強(qiáng)度的冬小麥光譜曲線具有相似的變化特征,波谷和波峰所在波段大致相同。倒伏后的冬小麥冠層光譜反射率較正常小麥整體增高,近紅外波段(780~1300 nm)的反射率比可見光波段明顯增加,且倒伏強(qiáng)度越大反射率越高,說明冬小麥?zhǔn)艿狗{迫后,冠層光譜對于倒伏脅迫強(qiáng)度表現(xiàn)出較好的響應(yīng)特征。在倒伏脅迫下,冬小麥冠層光譜的紅邊位置發(fā)生藍(lán)移,紅邊(700~780 nm)幅值與紅邊面積(紅邊內(nèi)一階導(dǎo)數(shù)光譜的積分)呈增大趨勢。由于水分在1 450及1 940nm的強(qiáng)吸收特征,在中紅外波段形成2個(gè)主要反射峰,位于1 650和2 200 nm附近。在整個(gè)波段區(qū)間內(nèi),光譜反射率表現(xiàn)為:重度倒伏>中度倒伏>輕度倒伏>未倒伏。
有研究表明,對反射率進(jìn)行微分變換有助于限制低頻噪聲對目標(biāo)光譜的影響[29-32]??紤]到作物受到脅迫時(shí)會(huì)在紅邊位置產(chǎn)生較大光譜變化響應(yīng),選取不同倒伏強(qiáng)度的350~800 nm波段的原始光譜進(jìn)行一階微分變換。從圖2b可以看出,不同倒伏下的冬小麥一階微分光譜反射率隨倒伏程度增加而增大,說明倒伏越嚴(yán)重,冠層光譜原始反射率數(shù)據(jù)變化越顯著。一階微分值等于0對應(yīng)的原始光譜位置為曲線的波峰或波谷。在可見光波段,一階微分存在3個(gè)明顯的波峰,波峰分別出現(xiàn)在420、525和725 nm處。
圖2 不同倒伏強(qiáng)度的冬小麥冠層光譜反射率與光譜一階微分曲線
采用人工目視解譯勾畫的方式提取各個(gè)倒伏樣本的莖、葉、穗面積后,計(jì)算各個(gè)樣本的莖面積占比、葉面積占比、穗面積占比、莖葉比率、莖穗比率、葉穗比率、莖穗之和與葉的比率等7個(gè)倒伏冠層結(jié)構(gòu)參數(shù),利用SPSS軟件分析7個(gè)冠層結(jié)構(gòu)參數(shù)與倒伏角度的相關(guān)性,結(jié)果如表1所示。根據(jù)相關(guān)系數(shù)的大小,并考慮到各參數(shù)之間的冗余問題,最終選取與倒伏角度相關(guān)性最高的參數(shù)莖葉比(相關(guān)系數(shù)?0.687,<0.01),作為冬小麥倒伏光譜響應(yīng)解析的敏感冠層結(jié)構(gòu)參數(shù)。隨著倒伏角度越小降低,倒伏強(qiáng)度越高,而冠層莖葉比也隨著增加。
表1 倒伏角度與各冠層結(jié)構(gòu)參數(shù)相關(guān)性分析
注:**表示顯著水平小于0.01。下同。
Note: ** representsa significant levelat 0.01. The same as below.
將冠層光譜反射率及其各變換形式與倒伏樣本莖葉比進(jìn)行相關(guān)性分析,提取冠層結(jié)構(gòu)參數(shù)的敏感波段,表2為各變換形式下與倒伏樣本莖葉比的敏感波段及相關(guān)系數(shù)。并采用偏最小二乘回歸法構(gòu)建莖葉比光譜預(yù)測模型,表3為不同倒伏強(qiáng)度下的莖葉比預(yù)測模型。
表2 倒伏樣本各光譜變換下的敏感波段及 對應(yīng)反射率與莖葉比的相關(guān)性分析
注:表示冠層光譜反射率,表示光譜反射率的微分,Gqc表示弓曲差,CWT表示連續(xù)小波變換。
Note:is canopy reflectance,is first order differential of reflectance, Gqc is gong qu cha and CWT is continuous wavelet transform
表3 基于各光譜變換形式的冬小麥倒伏冠層莖葉比光譜診斷模型
注:1、2、3、4分別為敏感波段所對應(yīng)的光譜反射率及不同數(shù)學(xué)變換形式下的值,為冠層莖葉比,2為決定系數(shù),RMSE為均方根誤差。
Note:1,2,3, and4are the spectral reflectance and values in different mathematical transformation formscorresponding to sensitive bands, respectively;is canopy stem-leave ratio;2 is the determination coefficient and RMSE is the root mean square error.
從表3中可以看出,原始光譜、對數(shù)、倒數(shù)等變換形式所建模型的決定系數(shù)較低,其中倒數(shù)的對數(shù)建立的預(yù)測模型精度最低(2=0.116),故可視為無效模型。一階微分構(gòu)建的預(yù)測模型較好,建模2=0.603,驗(yàn)證2=0.659,驗(yàn)證2結(jié)果略高于建模,這可能是將樣本隨機(jī)分成建模組和驗(yàn)證組造成的。從模型的精度和穩(wěn)定性綜合來看,基于連續(xù)小波變換構(gòu)建的模型為最優(yōu)模型,精度較高,且具有較好的穩(wěn)定性,建模樣本的2為0.654,RMSE為0.034。這是由于小波技術(shù)采用特定小波基對光譜信號從時(shí)域、頻率2個(gè)視角進(jìn)行分解,時(shí)域信息為光譜隨波長變化的規(guī)律,主要包括均值、方差、峰度等。頻域信息是信號在各個(gè)頻率上的能量分布,信息主要為頻率與譜值。光譜信息經(jīng)多尺度小波技術(shù)分解后,原光譜信息被一分為多,有助于原光譜信息內(nèi)部信息的凸顯。從模型分析可知多種微分變換的建模精度與驗(yàn)證精度均較高,表明對小麥倒伏敏感的信息多為細(xì)微信息,而小波技術(shù)又具有凸顯細(xì)微信息的優(yōu)勢且基于小波變換的模型的建模精度最高,綜上可知小波技術(shù)具有凸顯原光譜內(nèi)細(xì)微信息的優(yōu)勢。小麥倒伏樣本建模組與驗(yàn)證組的散點(diǎn)圖為圖3所示,從圖中可以看出,建模樣本與驗(yàn)證樣本均分布1:1線兩側(cè),檢驗(yàn)樣本的2為0.632,RMSE為0.034。
圖3 莖葉比實(shí)測圖與預(yù)測圖的散點(diǎn)圖
倒伏角度可有效表征倒伏對冬小麥植株的脅迫程度,但倒伏角度不能直接進(jìn)行光譜定量分析。冠層光譜反射率來自于探測視場內(nèi)的莖葉穗組合貢獻(xiàn)的反射光能量,而倒伏角度對于冠層光譜不提供直接貢獻(xiàn),無法用冠層光譜直接診斷倒伏角度。本文以冠層結(jié)構(gòu)參數(shù)為中間量,來開展光譜響應(yīng)解析,進(jìn)而劃分倒伏災(zāi)情等級。
以倒伏角度為自變量,以前文中野外建模樣本的冠層莖葉比為因變量建立線性回歸模型,如式(3),其中建模決定系數(shù)2=0.5,均方根誤差RMSE=0.039,計(jì)算各倒伏等級的莖葉比閾值區(qū)間,確定倒伏等級劃分范圍。最終確定重度倒伏(莖葉比≥0.22)、中度倒伏(0.13<莖葉比<0.22)、輕度倒伏(0.06<莖葉比≤0.13)和未倒伏(莖葉比≤0.06)。
=?0.0028+0.26 (3)
根據(jù)上述倒伏災(zāi)情程度的莖葉比劃分區(qū)間,對莖葉比光譜預(yù)測結(jié)果進(jìn)行倒伏災(zāi)情等級劃分,并利用未參與建模的野外樣本進(jìn)行精度驗(yàn)證,如表3所示。
表3 驗(yàn)證樣本的倒伏等級預(yù)測
從表中3中可以看出,基于冠層莖葉比的倒伏災(zāi)情診斷模型的精度可達(dá)84.6%,預(yù)測倒伏等級基本上與實(shí)測倒伏等級一致。說明基于莖葉比高光譜診斷方法可以有效診斷冬小麥灌漿期的伏強(qiáng)度等級。
由于冬小麥莖稈、葉片、麥穗具有較大光譜差異,當(dāng)冬小麥正常生長時(shí)處于直立狀態(tài),光譜反射率主要由視場范圍內(nèi)的冠層葉片結(jié)構(gòu)決定,當(dāng)冬小麥植株發(fā)生倒伏時(shí),冠層視場范圍內(nèi)的莖、葉、穗面積比例和重疊程度發(fā)生變化,促使倒伏冬小麥的冠層光譜也隨著變化,且不同倒伏強(qiáng)度的光譜特征具有較好的響應(yīng)規(guī)律。因此冬小麥倒伏冠層光譜變化特征與倒伏角度災(zāi)情程度密切相關(guān),而倒伏嚴(yán)重度可用倒伏角度來定量劃分,采用冠層光譜數(shù)據(jù)可用來評估冬小麥倒伏冠層結(jié)構(gòu)變化特征,進(jìn)而劃分災(zāi)情等級。衛(wèi)星遙感直接監(jiān)測與冬小麥倒伏受災(zāi)嚴(yán)重程度密切相關(guān)的倒伏角度、類型及生育期等信息有較大的難度,如:有些遙感影像的空間分辨率和時(shí)間分辨率不能夠滿足倒伏災(zāi)情監(jiān)測的精度和時(shí)間要求,但可以通過定量反演倒伏前后的作物冠層結(jié)構(gòu)參量變化信息來間接實(shí)現(xiàn)倒伏災(zāi)情程度的監(jiān)測。在葉綠素含量較高的生育期,如:返青期、孕穗期、開花期及灌漿初期,株高較低、莖稈柔韌性較好,冬小麥發(fā)生倒伏的可能性較低;在穗質(zhì)量較大的灌漿后期,莖稈負(fù)荷較大,且大風(fēng)暴雨天氣增多,較易發(fā)生倒伏災(zāi)害,此時(shí)莖、葉中的葉綠素含量較低且較為穩(wěn)定,冠層光譜變化主要來源于冠層結(jié)構(gòu)變化,特別是莖葉比率的變化對冠層光譜具有較大貢獻(xiàn),因此本文暫未重點(diǎn)分析莖葉葉綠素濃度對倒伏脅迫冠層結(jié)構(gòu)光譜響應(yīng)的影響。
本文雖然在冬小麥倒伏冠層光譜響應(yīng)機(jī)理解析方面取得了一定的進(jìn)展,但仍存在不足之處。采用目視解譯及人工數(shù)字化的方法來提取倒伏樣本的莖葉穗各組分面積,該方法雖然精度較高,但工作量較大,在今后的研究中有必要探索一種高精度且方便快捷提取冠層結(jié)構(gòu)參數(shù)的方法。本研究所構(gòu)建的冬小麥倒伏災(zāi)情光譜診斷模型效果較好,但研究范圍僅限河北省藁城地區(qū),在更大范圍內(nèi)研究分析倒伏小麥冠層結(jié)構(gòu)與倒伏強(qiáng)度、冠層光譜的響應(yīng)關(guān)系,有助于提升冬小麥倒伏災(zāi)情高光譜診斷的穩(wěn)定性和普適性。
通過對不同倒伏強(qiáng)度下冬小麥冠層結(jié)構(gòu)特征變化的研究,發(fā)現(xiàn)不同倒伏強(qiáng)度下的冬小麥冠層光譜差異較大;將視場圖像中莖、葉、穗面積所構(gòu)建的冠層結(jié)構(gòu)參數(shù)與倒伏角度進(jìn)行相關(guān)性分析,發(fā)現(xiàn)冠層莖葉比與倒伏角度的相關(guān)性最佳,相關(guān)系數(shù)為?0.678(<0.01);采用傳統(tǒng)光譜技術(shù)與連續(xù)小波變換技術(shù)對冠層光譜進(jìn)行處理,利用偏最小二乘法構(gòu)建冠層莖葉比與冠層高光譜特征參量的響應(yīng)模型,基于連續(xù)小波變換所構(gòu)建的冬小麥倒伏光譜診斷模型的精度和穩(wěn)定性最佳,建模2=0.654,RMSE=0.034,驗(yàn)證2=0.632,RMSE=0.032;以莖葉比劃分的倒伏等級范圍為:重度倒伏(莖葉比≥0.22)、中度倒伏(0.13<莖葉比<0.22)、輕度倒伏(0.06<莖葉比≤0.13)和未倒伏(莖葉比≤0.06);采用冠層莖葉比高光譜診斷結(jié)果進(jìn)行倒伏災(zāi)情等級劃分,識(shí)別精度為84.6%。本文對不同倒伏強(qiáng)度下冬小麥冠層結(jié)構(gòu)變化及其光譜響應(yīng)規(guī)律的探究,并建立的基于冠層莖葉比的冬小麥倒伏災(zāi)情光譜診斷模型,可為后續(xù)構(gòu)建大范圍遙感監(jiān)測中的倒伏災(zāi)情指數(shù)提供先驗(yàn)知識(shí),有助于客觀、定量地表征小麥?zhǔn)艿狗{迫的等級程度。
[1] Tripathi A, Tripathi D K, Chauhan D K, et al. Paradigms of climate change impacts on some major food sources of the world: A review on current knowledge and future prospects[J]. Agriculture Ecosystems & Environment, 2016, 216:356-373.
[2] Tao F, Zhao Z. Impacts of climate change as a function of global mean temperature: Maize productivity and water use in China[J].Climatic Change, 2011, 105(3/4):409-432.
[3] Shah F, Huang J, Cui K, et al. Impact of high-temperature stress on rice plant and its traits related to tolerance[J]. Journal of Agricultural Science, 2011, 149(5):545-556.
[4] Wu W, Ma B L. Assessment of canola crop lodging under elevated temperatures for adaptation to climate change[J]. Agricultural & Forest Meteorology, 2018, 248:329-338.
[5] Berry P M, Sterling M, Baker C J, et al. A calibrated model of wheat lodging compared with field measurements[J]. Agricultural & Forest Meteorology, 2003, 119(3):167-180.
[6] Foulkes M J, Slafer G A, Davies W J, et al. Raising yield potential of wheat. III. Optimizing partitioning to grain while maintaining lodging resistance[J]. Journal of Experimental Botany, 2011, 62(2):469-486.
[7] Acreche M M, Slafer G A. Lodging yield penalties as affected by breeding in Mediterranean wheats[J]. Field Crops Research, 2011,122(1):40-48.
[8] Pi?era-Chavez F J, Berry P M, Foulkes M J, et al. Avoiding lodging in irrigated spring wheat. I. Stem and root structural requirements[J]. Field Crops Research, 2016, 196:325-336.
[9] Wei W, Ma B. A new method for assessing plant lodging and the impact of management options on lodging in canola crop production[J]. Scientific Reports, 2016, 6(1):31890.
[10] Peake A S, Huth N I, Carberry P S, et al. Quantifying potential yield and lodging-related yield gaps for irrigated spring wheat in sub-tropical Australia[J]. Field Crops Research, 2014, 158(2):1-14.
[11] Kendall S L, Holmes H, White C A, et al. Quantifying lodging-induced yield losses in oilseed rape[J]. Field Crops Research, 2017, 211:106-113.
[12] Liu T, Li R, Zhong X, et al. Estimates of rice lodging using indices derived from UAV visible and thermal infrared images[J]. Agricultural &Forest Meteorology, 2018, 252:144-154.
[13] Berry P M, Spink J. Predicting yield losses caused by lodging in wheat[J]. Field Crops Research, 2012, 137(3):19-26.
[14] Zhang J, Gu X, Wang J, et al. Evaluating maize grain quality by continuous wavelet analysis under normal and lodging circumstances[J]. NJAS-Wageningen Journal of Life Sciences, 2012, 10(1/2):580-585.
[15] 郭翠花,高志強(qiáng),苗果園. 不同產(chǎn)量水平下小麥倒伏與莖稈力學(xué)特性的關(guān)系[J]. 農(nóng)業(yè)工程學(xué)報(bào),2010,26(3):151-155.
GuoCuihua, GaoZhiqiang, Miao Guoyuan. Relationship between lodging and mechanical characteristics of winterwheat stem under different yield levels[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2010, 26(3):151-155. (in Chinese with English abstract)
[16] 劉忠陽,陳懷亮,胡程達(dá),等. 后期倒伏對冬小麥干物質(zhì)分配和產(chǎn)量的影響[J]. 中國農(nóng)業(yè)氣象,2017,38(5):321-329.
Liu Zhongyang, Chen Huailiang, Hu Chengda, et al. Effects of lodging at the late growth stage on dry matter distribution and yield of winter wheat[J]. Chinese Journal of Agrometeorology, 2017, 38(5):321-329. (in Chinese with English abstract)
[17] 梁玉超,張永強(qiáng),石書兵,等. 施氮量對滴灌冬小麥莖部特征及其抗倒伏性的影響[J]. 麥類作物學(xué)報(bào),2017,37(8):1078-1086.
Liang Yuchao, Zhang Yongqiang, Shi Shubing, et al. Effect of nitrogen fertilizer rate on stem morphology characteristics and lodging resistance in winter wheat with drip irrigation[J]. Journal of Triticeae Crops, 2017, 37(8):1078-1086. (in Chinese with English abstract)
[18] 楊浩,楊貴軍,顧曉鶴,等. 小麥倒伏的雷達(dá)極化特征及其遙感監(jiān)測[J]. 農(nóng)業(yè)工程學(xué)報(bào),2014,30(7):1-8.
Yang Hao, Yang Guijun, GuXiaohe, et al. Radar polarimetric response features and remote sensing monitoringof wheat lodging[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2014, 30(7): 1-8. (in Chinese with English abstract)
[19] Yang H, Chen E, Li Z, et al. Wheat lodging monitoring using polarimetric index from RADARSAT-2 data[J].International Journal of Applied Earth Observation &Geoinformation, 2015, 34(1):157-166.
[20] 韓東,楊浩,楊貴軍,等. 基于Sentinel-1雷達(dá)影像的玉米倒伏監(jiān)測模型[J]. 農(nóng)業(yè)工程學(xué)報(bào),2018,34(3):166-172.
Han Dong, Yang Hao, Yang Guijun, et al. Monitoring model of maize lodging based on Sentinel-1 radar image[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(3): 166-172. (in Chinese with English abstract)
[21] 王延倉,顧曉鶴,朱金山,等. 利用反射光譜及模擬多光譜數(shù)據(jù)定量反演北方潮土有機(jī)質(zhì)含量[J]. 光譜學(xué)與光譜分析,2014,34(1):201-206.
Wang Yancang, Gu Xiaohe, Zhu Jinshan, et al. Inversion of organic matter content of the north fluvo-aquic soil based on hyperspectral and multi-spectra[J]. Spectroscopy and Spectral Analysis, 2014, 34(1): 201-206. (in Chinese with English abstract)
[22] 王延倉,楊貴軍,朱金山,等. 基于小波變換與偏最小二乘耦合模型估測北方潮土有機(jī)質(zhì)含量[J]. 光譜學(xué)與光譜分析,2014,34(7):1922-1926.
Wang Yancang, Yang Guijun, Zhu Jinshan, et al. Estimation of organic matter content of north fluvo-aquic soil based on the coupling model of wavelet transform and partial least aquares[J]. Spectroscopy and Spectral Analysis, 2014, 34(7): 1922-1926. (in Chinese with English abstract)
[23] 楊秀芳,張偉,楊宇祥. 基于提升小波變換的雷達(dá)生命信號去噪技術(shù)[J]. 光學(xué)學(xué)報(bào),2014,34(3):292-297.
Yang Xiufang, Zhang Wei, Yang Yuxiang, et al. Denoisingtechnology of radar life signal based on lifting wavelet transform[J]. ActaOpticaSinica, 2014, 34(3): 292-297. (in Chinese with English abstract)
[24] 楊秀芳,王若嘉,王佩佩,等. 基于提升小波改進(jìn)型閾值函數(shù)的雷達(dá)生命信號去噪技術(shù)[J]. 光子學(xué)報(bào),2016,45(7):119-124.
Yang Xiufang, Wang Ruojia, Wang Peipei, et al. De-noising technology of radar life signal based on lifting wavelet transform and improved soft threshold function[J]. ActaPhotonicaSinica, 2016, 45(7): 119-124. (in Chinese with English abstract)
[25] 陳瀟,邢立新,高志勇,等. 基于小波變換的遙感影像紋理信息提取[J]. 安徽農(nóng)業(yè)科學(xué),2015(4):363-366.
Chen Xiao, Xing Lixin, GaoZhiyong, et al. Information extraction of remote sensing image texture based on wavelet transform[J]. Journal of Anhui Agricultural Sciences, 2015(4): 363-366. (in Chinese with English abstract)
[26] 黃昕,張良培,李翠琳. 基于小波變換的影像紋理特征提取試驗(yàn)[J]. 測繪地理信息,2005,30(6):7-9.
Huang Xin, Zhang Liangpei, Li Cuilin. Experiments to extract texture features of images based on wavelet[J]. Journal of Geomatics, 2005, 30(6): 7-9. (in Chinese with English abstract)
[27] 于雷,洪永勝,周勇,等. 連續(xù)小波變換高光譜數(shù)據(jù)的土壤有機(jī)質(zhì)含量反演模型構(gòu)建[J]. 光譜學(xué)與光譜分析,2016,36(5):1428-1433.
Yu Lei, Hong Yongsheng, Zhou Yong, et al. Inversion of soil orangic matter content using hyperspectral data based on continuous wavelet transformation[J]. Spectroscopy and Spectral Analysis, 2016, 36(5): 1428-1433. (in Chinese with English abstract)
[28] 曹利萍,王君杰,雷夢林,等. 高光譜對冬小麥倒伏的響應(yīng)[J]. 山西農(nóng)業(yè)科學(xué),2017,45(12):1930-1932.
Cao Liping, Wang Junjie, Lei Menglin, et al. Response of canopy spectra on the winter wheat lodging[J]. Journal of Shanxi Agricultural Sciences, 2017, 45(12): 1930-1932. (in Chinese with English abstract)
[29] Qu Y, Liu Z. Dimensionality reduction and derivative spectral feature optimization for hyperspectral target recognition[J]. Optik-International Journal for Light and Electron Optics, 2017, 130:1349-1357.
[30] Bao J, Chi M, Benediktsson J A. Spectral derivative features for classification of hyperspectral remote sensing images: experimental evaluation[J]. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2013, 6(2): 594-601.
[31] Wang X, Zhang J, Ren G, et al. Yellow River Estuary typical wetlands classification based on hyperspectral derivative transformation[J]. Proceedings of SPIE - The International Society for Optical Engineering, 2014, 9142(1): 91421O-91421O-12.
[32] Plaza A, Benediktsson J A, Boardman J W, et al. Recent advances in techniques for hyperspectral image processing [J]. Remote Sensing of Environment, 2009, 113(1): S110-S122.
Selection of sensitive canopy structure parameters and spectral diagnostic model for lodging intensity of winter wheat
Shu Meiyan1,2,3,4, Gu Xiaohe1,2,3※, Sun Lin4, Zhu Jinshan4, Yang Guijun1,2,3, Wang Yancang5
(1.100097,; 2.100097,; 3.100097,; 4.,266590,; 5.,065000,)
At present, the changes of canopy structure and response mechanism of canopy spectral are not clear on winter wheat under lodging stress. Therefore, in this paper lodging winter wheat at the filling stage was taken as study object, the canopy structural parameters derived from the ratio of canopy stem, leaf and ear with different lodging strength were extracted. The correlation between the canopy structural parameters and lodging angle was analyzed, and the sensitive canopy structural parameters were selected to express lodging strength. The traditional spectral transform and the continuous wavelet transform were adopted to process the canopy hyperspectral data of lodging winter wheat. The bands and wavelet coefficients sensitive to canopy structural parameters were selected. The response model between canopy structural parameters of lodging winter wheat and hyperspectral characteristics parameters were constructed by partial least squares regression (PLSR) method, and the accuracy of the model was verified by field samples (28 samples for the modeling set, and 13 samples for the verification set). The results showed that the spectral curves of winter wheat with various lodging strengths had similar variation characteristics, and the wavelength bands of the troughs and peaks were roughly the same. Throughout the band interval, the spectral reflectance was expressed as: severe lodging > moderate lodging > mild lodging >not lodging. The first-order differential spectral reflectance of winter wheat increased with the increase of lodging degree, which indicated that the more severe the lodging, the more significant the change of the original reflectance data of the canopy spectrum. The correlation between stem-leaf ratio and lodging angle was the highest (=-0.687,<0.01), which could be used to characterize the lodging strength of winter wheat. The stem-leaf ratio increased with the decrease of lodging angle. The diagnostic model of lodging disaster of winter wheat based on continuous wavelet transform was superior to that based on the traditional transform, and the determination coefficient of the test samples was 0.632 (<0.01). The accuracy of lodging disaster classification based on the prediction results of canopy stem-leaf ratio could reach 84.6%. Therefore, the contribution proportion of stems, leaves and ears of the winter wheat canopy changed regularly in the sight of spectrometer after lodging. The stem-leaf ratio of winter wheat canopy could effectively characterize the changes of canopy structure under lodging stress, and had a good relationship with the lodging strength. The difference in the spectral reflectance of stem, leaf and ear and the variation in canopy structure after lodging were directly reflected in the canopy spectral difference of lodging wheat. The response rule between stem-leaf ratio with different lodging strength and canopy spectrum of winter wheat canopy can effectively distinguish the level of lodging disaster degree. It is helpful to provide a priori knowledge for remote sensing monitoring of winter wheat lodging disaster at regional scale.
crops; disasters; prediction;lodging; winter wheat; hyperspectral; stem-leaf ratio; continuous wavelet transform
束美艷,顧曉鶴,孫 林,朱金山,楊貴軍,王延倉. 表征冬小麥倒伏強(qiáng)度敏感冠層結(jié)構(gòu)參數(shù)篩選及光譜診斷模型[J]. 農(nóng)業(yè)工程學(xué)報(bào),2019,35(4):168-174. doi:10.11975/j.issn.1002-6819.2019.04.021 http://www.tcsae.org
Shu Meiyan, Gu Xiaohe, Sun Lin, Zhu Jinshan, Yang Guijun, Wang Yancang. Selection of sensitive canopy structure parameters and spectral diagnostic model for lodging intensity of winter wheat[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2019, 35(4): 168-174. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2019.04.021 http://www.tcsae.org
2018-10-10
2019-02-22
國家自然科學(xué)基金(41571323);北京市自然科學(xué)基金(6172011);院創(chuàng)新能力建設(shè)專項(xiàng)(KJCX20170705);河北省青年基金(D2017409021)
束美艷,主要從事農(nóng)業(yè)定量遙感研究。Email:2448858578@qq.com
顧曉鶴,博士,副研究員,主要從事農(nóng)業(yè)遙感相關(guān)研究。 Email:guxh@nercita.org.cn
10.11975/j.issn.1002-6819.2019.04.021
S127
A
1002-6819(2019)-04-0168-07