劉煥軍,鮑依臨,徐夢(mèng)園,張新樂※,孟祥添,潘 越,楊昊軒,謝雅慧
基于SOM和NDVI的黑土區(qū)精準(zhǔn)管理分區(qū)對(duì)比
劉煥軍1,2,鮑依臨1,徐夢(mèng)園1,張新樂1※,孟祥添1,潘 越1,楊昊軒1,謝雅慧1
(1. 東北農(nóng)業(yè)大學(xué)公共管理與法學(xué)院,哈爾濱 150030:2. 中國(guó)科學(xué)院東北地理與農(nóng)業(yè)生態(tài)研究所,長(zhǎng)春 130012)
根據(jù)土壤養(yǎng)分的空間異質(zhì)性對(duì)耕地進(jìn)行分區(qū)是實(shí)施變量施肥管理的關(guān)鍵環(huán)節(jié),施肥的變量管理將減輕黑土區(qū)農(nóng)業(yè)面源污染和土壤退化問題。該文以典型黑土區(qū)黑龍江省海倫市某合作社地塊為研究對(duì)象,利用SPOT-6遙感影像提取歸一化植被指數(shù)(normalized differential vegetation index, NDVI)、插值計(jì)算土壤有機(jī)質(zhì)(soil organic matter,SOM),結(jié)合數(shù)字高程模型(digital elevation model,DEM),應(yīng)用面向?qū)ο蟮姆指罘椒ǎ瑢?duì)研究地塊進(jìn)行分區(qū),并應(yīng)用莫蘭(Morans)指數(shù)對(duì)分區(qū)結(jié)果進(jìn)行評(píng)價(jià),以期對(duì)比研究基于SOM空間插值與遙感信息的分區(qū)精度。結(jié)果表明:結(jié)合4期NDVI空間信息分區(qū)的精度最高;結(jié)合SOM、DEM、NDVI空間信息進(jìn)行分區(qū)的精度次之;結(jié)合地形與SOM空間信息分區(qū)精度較低;僅根據(jù)SOM空間插值進(jìn)行分區(qū)的精度最低。研究結(jié)果可為黑土區(qū)農(nóng)田精準(zhǔn)管理分區(qū)輸入量的選擇與多尺度分區(qū)提供思路,為實(shí)施田間精準(zhǔn)追肥提供科學(xué)依據(jù)。
遙感;評(píng)價(jià);空間插值;多源空間數(shù)據(jù);精準(zhǔn)管理分區(qū);面向?qū)ο?/p>
精準(zhǔn)管理分區(qū)是根據(jù)不同田塊的空間變異性和實(shí)際需求,把一個(gè)田塊分割成若干個(gè)不同均質(zhì)性的子田塊來(lái)調(diào)整土壤和作物的管理措施[1],分區(qū)的結(jié)果可以為精準(zhǔn)農(nóng)業(yè)的發(fā)展提供管理決策,以期最大限度地提升耕地資源潛力[2]。
目前在精準(zhǔn)管理分區(qū)輸入量主要以土壤養(yǎng)分為主。陳彥等利用土壤有機(jī)質(zhì)(soil organic matter,SOM)、堿解氮、速效磷等數(shù)據(jù)為變量,應(yīng)用空間插值法對(duì)新疆棉田進(jìn)行土壤養(yǎng)分精確管理分區(qū)研究[3]。Davatgar等通過主成分分析(principal component analysis,PCA)提取總氮、有效磷、速效鉀等土壤養(yǎng)分?jǐn)?shù)據(jù),插值后基于K模糊聚類法對(duì)水稻栽培區(qū)特定的管理區(qū)進(jìn)行了劃分[4]。以上研究,多采用實(shí)地采樣方式測(cè)定數(shù)據(jù),這種作業(yè)方法耗時(shí)耗力、成本高、時(shí)效性差;且基于單一數(shù)據(jù)源的空間劃分存在精度低、應(yīng)用范圍受限等局限性[5]。隨著遙感技術(shù)的發(fā)展,分區(qū)指標(biāo)的選取也逐漸從基于土壤養(yǎng)分?jǐn)?shù)據(jù)的空間插值過渡到利用多光譜影像進(jìn)行空間信息的提取[6-7]。SOM含量在一定程度上反映了土壤肥力,是影響土壤質(zhì)量的關(guān)鍵因素,所以土壤養(yǎng)分SOM的空間分布對(duì)精準(zhǔn)分區(qū)管理具有重要的意義[8]。Schillaci等通過模擬西西里島表層SOM含量,發(fā)現(xiàn)土壤中有機(jī)質(zhì)含量隨土地利用、降雨、侵蝕等多方面因素發(fā)生變化[9],SOM在小尺度范圍內(nèi)存在空間異質(zhì)性與空間關(guān)聯(lián)性[10]。黃魏等的分析驗(yàn)證了加入地形因素,進(jìn)行SOM的空間插值,會(huì)使全局預(yù)測(cè)精度2提升到0.75[11]。添加地形因素提升空間信息的精度,有利于更加精準(zhǔn)地對(duì)研究區(qū)域進(jìn)行劃分。此外,利用NDVI等植被信息,可以提高研究區(qū)的土壤類型的識(shí)別精度,以便于更好地歸類與分析[12-14]。Burry等通過對(duì)比NDVI與增強(qiáng)型植被指數(shù)(enhanced vegetation index,EVI ),驗(yàn)證了無(wú)論是對(duì)植被敏感性分析,或是樹木多樣性的指標(biāo)分析,NDVI都具有更強(qiáng)的識(shí)別能力[15]。Meera等根據(jù)NDVI的特性,明晰了NDVI在區(qū)域檢測(cè)中所起到的重大意義[16],為基于NDVI進(jìn)行精準(zhǔn)管理分區(qū)提供了有力支持。但基于單期NDVI空間數(shù)據(jù)所反映的信息有限[17],考慮的影響因素不夠全面,在精準(zhǔn)分區(qū)的精度上會(huì)有所降低。通過對(duì)比試驗(yàn)發(fā)現(xiàn),綜合2期影像信息的分區(qū)結(jié)果明顯優(yōu)于單期的分區(qū)[18]。
以上分區(qū)研究,多基于影像數(shù)據(jù)或土壤養(yǎng)分,數(shù)據(jù)源單一。為了更大限度融合多源數(shù)據(jù),提高分區(qū)的精度,本研究嘗試將與分區(qū)密切相關(guān)的SOM空間數(shù)據(jù)、地形數(shù)據(jù)和遙感影像植被指數(shù)等空間數(shù)據(jù)相結(jié)合,對(duì)典型黑土區(qū)田塊進(jìn)行分區(qū),并對(duì)分區(qū)結(jié)果評(píng)價(jià),比較空間插值與遙感影像數(shù)據(jù)分區(qū)差異,以期為黑土區(qū)精準(zhǔn)管理分區(qū)數(shù)據(jù)類型的選取提供借鑒,提高分區(qū)精度,為田間變量管理處方圖與智慧農(nóng)業(yè)發(fā)展提供支持。
研究區(qū)位于黑龍江省中部海倫市東興農(nóng)機(jī)合作社,地處松嫩平原東北端,小興安嶺西麓,平均海拔239 m,地塊中心經(jīng)緯度為126°55′E、47°25′N,面積約為37.6 hm2,屬寒溫帶大陸性氣候。地勢(shì)從東北到西南依次呈梯形逐漸降低,地貌是由小興安嶺山地向松嫩平原的過渡地帶。研究區(qū)位于典型黑土區(qū),漫川漫崗地帶,田塊地形起伏較大,土壤養(yǎng)分和內(nèi)部作物長(zhǎng)勢(shì)空間差異性顯著,因而作為本次研究的對(duì)象。研究區(qū)位置及樣點(diǎn)布置如圖1,2016年以種植大豆供試。
1.2.1 遙感與地形數(shù)據(jù)的獲取和預(yù)處理
訂購(gòu)大豆生長(zhǎng)期2016-06-09、2016-07-19、2016-08-08、2016-09-03 4期SPOT-6遙感數(shù)據(jù)(多光譜波段的空間分辨率為6 m)。在ENVI5.1中對(duì)SPOT-6數(shù)據(jù)進(jìn)行大氣校正、幾何校正。以上數(shù)據(jù)在Arcgis10.2中按照研究區(qū)范圍進(jìn)行裁剪,并提取SPOT-6 4個(gè)時(shí)期NDVI。
圖1 研究區(qū)位置及樣點(diǎn)布置
Fig 1 Location of study fields and sampling point distribution map
2016年5月27日,使用定位精度厘米級(jí)的海星達(dá)iRTK2對(duì)研究區(qū)地塊進(jìn)行實(shí)地測(cè)量,得到780個(gè)精確的坐標(biāo)和高程點(diǎn),利用Arcgis10.2將高程數(shù)據(jù)生成TIN圖層,將TIN圖層轉(zhuǎn)成空間分辨率為4 m的高精度DEM柵格數(shù)據(jù)。
1.2.2 土壤養(yǎng)分?jǐn)?shù)據(jù)的獲取
2016年4月21日,運(yùn)用GPS定位,依據(jù)研究區(qū)地形變化特征,平均每隔60 m預(yù)設(shè)一個(gè)采樣點(diǎn),共確定98個(gè)采樣點(diǎn),記錄采樣點(diǎn)的空間坐標(biāo),并記錄各點(diǎn)編號(hào)。用布袋將土樣帶回實(shí)驗(yàn)室,稱量樣本質(zhì)量后進(jìn)行研磨、風(fēng)干、過2 mm篩,用重鉻酸鉀容量法[19]測(cè)得SOM含量(表1)。
表1 土壤有機(jī)質(zhì)描述統(tǒng)計(jì)量
1.2.3 土壤養(yǎng)分空間變異性分析與SOM空間土壤屬性映射
根據(jù)塊金值與基臺(tái)值之比反映空間變異性,塊金值表示隨機(jī)部分的空間變異性,基臺(tái)值表示整體的變異程度[20]。將通過采樣獲取研究區(qū)范圍內(nèi)98個(gè)采樣點(diǎn)數(shù)據(jù)在地學(xué)統(tǒng)計(jì)軟件GS+中對(duì)空間變異性進(jìn)行分析,結(jié)果如表2所示。
表2 土壤有機(jī)質(zhì)空間變異性分析
采用地統(tǒng)計(jì)學(xué)中最優(yōu)內(nèi)插法對(duì)空間內(nèi)其他未知區(qū)域的相同屬性進(jìn)行計(jì)算,表達(dá)SOM含量空間分布,在Arcgis10.2中進(jìn)行Kriging插值生成SOM含量的空間分布圖。
1.2.4 不同空間數(shù)據(jù)標(biāo)準(zhǔn)化
由于輸入量并不能用可比單位進(jìn)行度量,為了將研究對(duì)象的多指標(biāo)信息進(jìn)行綜合,需要將研究區(qū)SOM、DEM、NDVI等空間數(shù)據(jù)進(jìn)行標(biāo)準(zhǔn)化處理,將變量控制在0~1之間。NDVI與生物量、葉面積指數(shù)、產(chǎn)量的關(guān)系密切,作為長(zhǎng)勢(shì)評(píng)價(jià)的指標(biāo)經(jīng)常被用于解釋作物的營(yíng)養(yǎng)情況,且在NDVI處于最大時(shí),與產(chǎn)量關(guān)系最為明顯[21-22],為了更好將作物信息與土壤養(yǎng)分信息耦合,選擇大豆結(jié)莢鼓粒期NDVI(8月)空間信息進(jìn)行標(biāo)準(zhǔn)化處理,并加入了地形因素以提高分區(qū)的精準(zhǔn)度[23]。通過坐標(biāo)轉(zhuǎn)換構(gòu)建研究的具體模型。將轉(zhuǎn)換后的模型進(jìn)行空間重采樣,將輸入數(shù)據(jù)均采樣為6 m×6 m的像元大小,進(jìn)行空間數(shù)據(jù)的信息綜合。
1.2.5 基于像元值標(biāo)準(zhǔn)差與莫蘭(Morans)指數(shù)的分區(qū)評(píng)價(jià)
NDVI在一定程度上可反映作物的產(chǎn)量與長(zhǎng)勢(shì)等信息。根據(jù)大豆的生長(zhǎng)規(guī)律以及東北地區(qū)的物候特征,在7月下旬至8月初,大豆處于生長(zhǎng)最茂盛時(shí)期,提取SPOT-6影像8月8日的NDVI去驗(yàn)證基于4種空間信息下的分區(qū)結(jié)果。Morans指數(shù)通常用于計(jì)算空間相關(guān)性,其值越大,空間相關(guān)性越明顯[24]。
基于高分辨影像探究地理對(duì)象的詳細(xì)變化,建立空間對(duì)應(yīng)關(guān)系,細(xì)化對(duì)比度和緊致度,可以取得更好的分割效果[25]。為比較不同分割尺度的優(yōu)劣程度,定義分割對(duì)象的集合為(scale,shape,compactness),存在平均分割評(píng)價(jià)指數(shù)ASEI(average segmentation evaluation index)的最大值:
在一定尺度內(nèi)平均分割評(píng)價(jià)指數(shù)達(dá)到最大時(shí),對(duì)應(yīng)的分割尺度即最優(yōu)分割尺度,所對(duì)應(yīng)集合(scale,shape,compactness)為最優(yōu)分割尺度集合。
式中為某區(qū)域內(nèi)像元的個(gè)數(shù),G為第個(gè)區(qū)域的驗(yàn)證指標(biāo)的均值,G表示區(qū)域內(nèi)的驗(yàn)證指標(biāo)均值。為該區(qū)域附近相鄰區(qū)域的個(gè)數(shù),G為第個(gè)相鄰區(qū)域的驗(yàn)證指標(biāo)均值,G表示該區(qū)域的指標(biāo)均值。為當(dāng)前區(qū)域邊界的長(zhǎng)度,L為當(dāng)前區(qū)域與第個(gè)相鄰區(qū)域公共邊的長(zhǎng)度。
定義平均分割評(píng)價(jià)指數(shù)ASEI,對(duì)研究區(qū)域內(nèi)所有區(qū)塊的SEI取平均值。
式中表示區(qū)域的總面積,A表示第個(gè)區(qū)域的面積,表示被劃分區(qū)域的總數(shù)量,SEI表示第個(gè)區(qū)域的分割評(píng)價(jià)指數(shù)。
2.3.1 分區(qū)內(nèi)部的同質(zhì)性
依據(jù)分區(qū)內(nèi)部像元值的標(biāo)準(zhǔn)差,綜合考慮面積大小賦予相應(yīng)的權(quán)重,判斷分區(qū)內(nèi)部的同質(zhì)性,公式如下
式中s是分區(qū)范圍內(nèi)像元值的標(biāo)準(zhǔn)差,a是分區(qū)的面積。為研究區(qū)域內(nèi)分割后的分區(qū)總個(gè)數(shù)。越小,代表分區(qū)內(nèi)部的空間異質(zhì)性越低,即分區(qū)內(nèi)部具有良好的同質(zhì)性。
2.3.2 分區(qū)之間的異質(zhì)性
采用Morans指數(shù)判斷空間相關(guān)性,根據(jù)相關(guān)性的大小來(lái)判斷分區(qū)間的異質(zhì)性高低,公式如下
通過土壤養(yǎng)分空間變異性分析(表2)可知,塊金值與基臺(tái)值的比值為0.048,說(shuō)明SOM具有較強(qiáng)的空間自相關(guān)性,受隨機(jī)因素影響小[27];SOM變程為114 m,本研究采樣點(diǎn)間隔60 m,說(shuō)明采樣點(diǎn)間土壤養(yǎng)分不具有明顯的空間異質(zhì)性,基于該試驗(yàn)數(shù)據(jù)進(jìn)行空間插值具有進(jìn)行精準(zhǔn)理分區(qū)的意義。
如圖2所示:相同或近似顏色代表一定區(qū)域內(nèi)的土壤屬性相同,而面向?qū)ο蠓指罘椒ǖ姆指钜罁?jù)則是按照樣區(qū)的色彩差異進(jìn)行分割[28-29],將以上不同分區(qū)結(jié)果進(jìn)行分區(qū)計(jì)算,依據(jù)平均分割評(píng)價(jià)指數(shù)的大小確定最優(yōu)分區(qū)尺度,對(duì)4種不同輸入量的區(qū)域進(jìn)行精準(zhǔn)劃分。將該尺度代入面向?qū)ο蠓指钪械玫较鄳?yīng)分區(qū)的個(gè)數(shù)、均值、標(biāo)準(zhǔn)差,并依據(jù)上述公式計(jì)算得到平均分割評(píng)價(jià)指數(shù)。
從圖2可以看到4種不同數(shù)據(jù)信息下的分區(qū)情況,由于輸入量的差異,分區(qū)的結(jié)果在表現(xiàn)形式上明顯不同。通過將分區(qū)結(jié)果進(jìn)行對(duì)比,發(fā)現(xiàn)基于相同土壤養(yǎng)分,加入地形因素的分區(qū)結(jié)果更為規(guī)整,較未考慮地形因素的分區(qū)更易于管理與劃分?;?種空間信息的分區(qū)個(gè)數(shù)依次為55、55、48、47(見表3),從耕作單元角度出發(fā),分區(qū)的結(jié)果符合當(dāng)代耕作單元的發(fā)展進(jìn)程,利于實(shí)施耕作[30]。
圖2 最優(yōu)尺度下的分區(qū)情況
Fig 2 Zoning at optimal scale
通過表3可知,結(jié)合地形信息的SOM空間信息的分區(qū)尺度與分區(qū)個(gè)數(shù)同基于SOM空間信息的分區(qū)一致,且個(gè)數(shù)都多于基于其他空間信息下的分區(qū),評(píng)價(jià)分割指數(shù)是用來(lái)計(jì)算同一輸入量下的最優(yōu)分割結(jié)果,因此,表中數(shù)據(jù)僅用于反映某一輸入量下的分割結(jié)果,并不能用于不同輸入量的對(duì)比。結(jié)合分區(qū)效果來(lái)看,基于空間插值信息的分區(qū)形狀更多的是與SOM空間分布有關(guān),個(gè)數(shù)多,不規(guī)整,耕作上存在局限性。
對(duì)分區(qū)結(jié)果本著各區(qū)域內(nèi)部空間同質(zhì)性高,分區(qū)之間異質(zhì)性高為原則進(jìn)行評(píng)價(jià),即分區(qū)內(nèi)像元值標(biāo)準(zhǔn)差越低、分區(qū)間Morans指數(shù)數(shù)值越低,代表分區(qū)效果越好。分區(qū)評(píng)價(jià)結(jié)果(表4)顯示:基于4期NDVI空間信息的分區(qū)精度最高,無(wú)論是對(duì)分區(qū)內(nèi)部或者分區(qū)間差異的評(píng)價(jià),均具有最優(yōu)的效果;基于SOM、DEM、NDVI空間信息的分區(qū)精度略低于4期NDVI的分區(qū),但分區(qū)的結(jié)果更易實(shí)施,分區(qū)更為工整,便于操作與管理。將地形因子融入SOM空間信息后的分區(qū)精度顯著提升,無(wú)論是在區(qū)域間的差異性,還是區(qū)域內(nèi)部的均一性,效果均好于基于SOM空間插值分布圖的分區(qū)。
表3 4種空間數(shù)據(jù)輸入量最優(yōu)分割尺度、分區(qū)數(shù)、平均分割評(píng)價(jià)指數(shù)
表4 基于不同分區(qū)信息的分區(qū)評(píng)價(jià)
東北黑土區(qū)是中國(guó)重要的商品糧基地,由于過量施肥造成的土壤退化與耕地質(zhì)量下降,對(duì)該區(qū)的精準(zhǔn)管理分區(qū)與減肥減藥的研究意義重大。研究選取了典型黑土區(qū)海倫合作社為試驗(yàn)地,土壤中黏粒含量豐富,保水保肥能力強(qiáng);SOM較其他土壤養(yǎng)分而言更具有穩(wěn)定性,受隨機(jī)因素的影響小,由此進(jìn)行的分區(qū)更具有研究意義。本文訂購(gòu)了4期空間分辨率為6 m的研究區(qū)SPOT影像以提取高精度的NDVI,在輸入量的選取方面,分別選擇了4種指標(biāo)要素進(jìn)行劃分:SOM空間插值圖,綜合考慮地形因素的SOM空間分布圖,綜合SOM、DEM、8月NDVI信息圖,4期NDVI空間分布圖;從數(shù)據(jù)處理的角度而言,綜合考慮多時(shí)期的NDVI信息與多源空間數(shù)據(jù)的疊加分析是本研究的特殊之處,這比基于單一數(shù)據(jù)源的劃分更為精準(zhǔn),在本質(zhì)上更具有說(shuō)服力[19];同時(shí)相比于格網(wǎng)采樣法更加省時(shí)高效[31]。在研究方法上,本文通過計(jì)算確定了最優(yōu)分區(qū)尺度,消除了人為劃分的紕漏給試驗(yàn)結(jié)果帶來(lái)的影響,選擇面向?qū)ο蟮姆指钍址▽?duì)多種輸入量進(jìn)行劃分,最后,運(yùn)用內(nèi)部標(biāo)準(zhǔn)差評(píng)價(jià)方法與Morans指數(shù)法分別對(duì)分區(qū)后子區(qū)域內(nèi)部與各子區(qū)域之間的結(jié)果進(jìn)行評(píng)價(jià)。經(jīng)對(duì)比,綜合考慮4期NDVI的空間信息進(jìn)行精準(zhǔn)管理分區(qū)精度更高。通過對(duì)比發(fā)現(xiàn)基于SOM、DEM、NDVI(8月)空間信息的分區(qū)較結(jié)合了地形因素的SOM空間信息更為精準(zhǔn),在追肥時(shí)期考慮大豆結(jié)莢期的空間信息可使分區(qū)精度有著較大提升,基于該輸入量的分區(qū)操作性更強(qiáng),更易于實(shí)現(xiàn)田塊上的管理與規(guī)劃;僅通過空間插值獲取的SOM空間分布分區(qū)精度最低,評(píng)價(jià)結(jié)果證實(shí)了加入地形因素可提升分區(qū)精度。與以往對(duì)黑土區(qū)的研究不同[12],本研究將地形因素融入到空間信息,并結(jié)合多源空間數(shù)據(jù)進(jìn)行劃分。在數(shù)據(jù)的選取上,對(duì)比了基于土壤養(yǎng)分分區(qū)與遙感影像分區(qū)的精度,證實(shí)了遙感影像的像元大小比土壤采樣間距而言具有顯著優(yōu)勢(shì)。由于本文僅考慮了高程因素的影響,未評(píng)價(jià)其他地形因子對(duì)研究區(qū)的作用權(quán)重,因此,接下來(lái)的研究中,將綜合考慮多種地形因素的作用;同時(shí),在作物生長(zhǎng)期的不同環(huán)節(jié)考慮更多的影響因子,如水分、溫度、生物量等,從而提升分區(qū)精度仍然是研究的重點(diǎn);其次,綜合分析不同地區(qū)地形要素及土壤空間差異,尋求一個(gè)更具有普適性、高精度的分區(qū)方式。
本文以海倫市為研究區(qū),利用6 m空間分辨率的時(shí)間序列SPOT-6影像提取歸一化植被指數(shù)(NDVI),空間插值土壤有機(jī)質(zhì)(SOM)空間分布特征,結(jié)合高精度數(shù)字高程模型(DEM),疊加不同空間數(shù)據(jù)進(jìn)行精準(zhǔn)管理分區(qū),利用8月NDVI進(jìn)行評(píng)價(jià)。研究結(jié)果表明:1)經(jīng)對(duì)比發(fā)現(xiàn),相對(duì)于傳統(tǒng)基于空間插值的精準(zhǔn)管理分區(qū),基于遙感影像的劃分無(wú)論是在數(shù)據(jù)獲取上,還是精度上都更具有優(yōu)勢(shì);2)加入了地形因素的多源空間數(shù)據(jù)分區(qū)精度較基于SOM空間插值結(jié)果進(jìn)行分區(qū)精度高;3)基于多源空間數(shù)據(jù)的優(yōu)勢(shì)在于可以綜合考慮多種因素,比單一數(shù)據(jù)的劃分精準(zhǔn)。該方法進(jìn)行分區(qū)比傳統(tǒng)格網(wǎng)采樣分區(qū)省時(shí)省力,更加高效。通過對(duì)分區(qū)結(jié)果評(píng)價(jià),證明了綜合4期NDVI空間信息的分區(qū)精度最高;其次是基于SOM、DEM、NDVI(8月)空間信息的分區(qū)。該研究可為今后精準(zhǔn)管理分區(qū)的數(shù)據(jù)選取提供思路,分區(qū)結(jié)果有望在黑土區(qū)田塊尺度分區(qū)和管理得到推廣。
[1] Aasen H, Bolten A. Multi-temporal high-resolution imaging spectroscopy with hyperspectral 2D imagers: From theory to application[J]. Remote Sensing of Environment, 2018, 205: 374-389.
[2] 滿衛(wèi)東,王宗明,劉明月,等. 1990-2013年?yáng)|北地區(qū)耕地時(shí)空變化遙感分析[J]. 農(nóng)業(yè)工程學(xué)報(bào), 2016,32(7):1-10. Man Weidong, Wang Zongming, Liu Mingyue, et al. Spatio-temporal dynamics analysis of cropland in Northeast China during 1990-2013 based on remote sensing[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2016, 32(7): 1-10. (in Chinese with English Abstract)
[3] 陳彥,呂新. 基于FCM的綠洲農(nóng)田養(yǎng)分管理分區(qū)研究[J]. 中國(guó)農(nóng)業(yè)科學(xué),2008,41(7):2016-2024. Chen Yan, Lü Xin. Definition of management zones of soil nutrients based on FCM algorithm in oasis field[J].Scientia Agricultura Sinica,2008, 41(7): 2016-2024. (in Chinese with English Abstract)
[4] Davatgar N, Neishabouri M R, Sepaskhah A R. Delineation of site specific nutrient management zones for a paddy cultivated area based on soil fertility using fuzzy clustering[J]. Geoderma, 2012, 173-174(2): 111-118.
[5] Tripathi R, Nayak A K, Shahid M, et al. Delineation of soil management zones for a rice cultivated area in eastern India using fuzzy clustering[J]. Catena, 2015, 133: 128-136.
[6] Kratzer S, Therese Harvey E, Philipson P. The use of ocean color remote sensing in integrated coastal zone manage-ment: A case study from Himmerfj?rden, Sweden[J]. Marine Policy, 2014, 43: 29-39.
[7] Guo Y, Wan Z, Liu D. Dynamics of dissolved organic carbon in the mires in the Sanjiang Plain, Northeast China [J]. Journal of Environmental Sciences, 2010, 22(1): 84-90.
[8] Powlson D S, Brookes P C, Whitmore A P, et al. Soil carbon sequestration to mitigate climate change: A critical re-examination to identify the true and the false[J]. European Journal of Soil Science, 2011, 62(1): 1-4.
[9] Schillaci C, Acutis M, Lombardo L, et al. Spatiotemporal topsoil organic carbon mapping of a semiarid Mediterranean region: The role of land use, soil texture, topographic indices and the influence of remote sensing data to modelling[J]. Science of the Total Environment, 2017, 601-602: 821-832.
[10] Mirzaee S, Ghorbani-Dashtaki S, Mohammadi J, et al. Spatial variability of soil organic matter using remote sensing data[J]. Catena, 2016, 145: 118-127.
[11] 黃魏,韓宗偉,羅云,等. 基于地形單元的土壤有機(jī)質(zhì)空間變異研究[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2015,46(4):161-167. Huang Wei, Han Zongwei, Luo Yun, et al.Spatial Distribution of soil organic matter based on topographic unit[J]. Transactions of the Chinese Society for Agricultural Machinery, 2015, 46(4): 161-167. (in Chinese with English Abstract)
[12] Pang G, Wang X, Yang M. Using the NDVI to identify variations in, and responses of, vegetation to climate change on the Tibetan Plateau from 1982 to 2012[J]. Quaternary International, 2016, 444: 87-96.
[13] Tong X, Brandt M, Hiernaux P, et al. Revisiting the coupling between NDVI trends and cropland changes in the Sahel drylands: A case study in western Niger[J]. Remote Sensing of Environment, 2017, 191: 286-296.
[14] 吳才武,張?jiān)聟?,夏建? 基于地統(tǒng)計(jì)與遙感反演相結(jié)合的有機(jī)質(zhì)預(yù)測(cè)制圖研究[J]. 土壤學(xué)報(bào),2016,53(6):1568-1575. Wu Caiwu, Zhang Yuecong, Xia Jianxin. Prediction and mapping of soil organic matter based on geostatistics and remote sensing inversion[J]. Acta Pedologica Sinica, 2016, 53(6): 1568-1575. (in Chinese with English Abstract)
[15] Burry L S, Palacio P I, Somoza M, et al. Dynamics of fire, precipitation, vegetation and NDVI in dry forest environ-ments in NW Argentina. Contributions to environmental archaeology[J]. Journal of Archaeological Science: Reports, 2017, 18: 747-757.
[16] Gandhi G M, Parthiban S, Thummalu N, et al. NDVI: Vegetation change detection using remote sensing and gis: A case study of Vellore District[J]. Elsevier Proceedia, International Conference on Recent Trends in Computing, 2015, 57: 1199-1210.
[17] 李敏,趙庚星,蔡明慶,等. 縣域棉花信息遙感提取與棉田精確化管理分區(qū)研究[J]. 自然資源學(xué)報(bào),2012, 27(11):1971-1980. Li Min, Zhao Gengxing, Cai Mingqing, et al. Extraction of cotton information using remote sensing and precision management zoning at county scale[J].Journal of Natural Resources , 2012, 27(11): 1971-1980. (in Chinese with English Abstract)
[18] 劉煥軍,邱政超,孟令華,等. 黑土區(qū)田塊尺度遙感精準(zhǔn)管理分區(qū)[J]. 遙感學(xué)報(bào),2017,21(3):470-478. Liu Huanjun, Qiu Zhengchao, Meng Linghua, et al. Site-specific management zone of field scale based on remote sensing image in a black soil area[J]. Journal of Remote Sensing, 2017, 21(3): 470-478. (in Chinese with English Abstract)
[19] Chang D, Zhang J, Zhu L, et al. Delineation of management zones using an active canopy sensor for a tobacco field[J]. Computers and Electronics in Agriculture, 2014, 109: 172-178.
[20] 李翔,潘瑜春,馬景宇,等. 基于多種土壤養(yǎng)分的精準(zhǔn)管理分區(qū)方法研究[J]. 土壤學(xué)報(bào),2007,44(1):14-20. Li Xiang, Pan Yuchun, Ma Jingyu, et al. Soil nutrients-based zoning for management of precision agriculture[J]. Acta Pedologica Sinica, 2007, 44(1): 14-20. (in Chinese with English Abstract)
[21] Tong X, Brandt M, Hiernaux P, et al. Revisiting the coupling between NDVI trends and cropland changes in the Sahel drylands: A case study in western Niger[J]. Remote Sensing of Environment, 2017, 191: 286-296.
[22] Xiaojun L, Richard F, Hengbiao Z, et al. Using an active-optical sensor to develop an optimal NDVI dynamic model for high-yield rice production (Yangtze, China)[J]. Sensors, 2017, 17(4): 672-685.
[23] Rasel S. Proxies for soil organic carbon derived from remote sensing[J]. International Journal of Applied Earth Observa-tion&Geoinformation, 2017, 59: 157-166.
[24] 蔡進(jìn),廖和平,邱道持,等. 重慶市農(nóng)村耕地資源貧困測(cè)度及空間格局研究[J]. 農(nóng)業(yè)工程學(xué)報(bào),2017,33(18):251-259. Cai Jin, Liao Heping, Qiu Daochi, et al. Study on poverty measure of farmland resources and spatial pattern in Chongqing[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2017, 33(18): 251-259. (in Chinese with English Abstract)
[25] Zhang X, Xiao P, Feng X, et al. Separate segmentation of multi-temporal high-resolution remote sensing images for object-based change detection in urban area[J]. Remote Sensing of Environment, 2017, 201: 243-255.
[26] Eisank C, Smith M, Hillier J. Assessment of multiresolution segmentation for delimiting drumlins in digital elevation models[J]. Geomorphology, 2014, 214: 452-464.
[27] 王紹強(qiáng),朱松麗,周成虎. 中國(guó)土壤土層厚度的空間變異性特征[J]. 地理研究,2001,20(2):161-169. Wang Shaoqiang, Zhu Songli, Zhou Chenghu. Characteristics of spatial variability of soil thickness in China[J]. Geographical Research, 2001, 20(2): 161-169. (in Chinese with English Abstract)
[28] Dronova I, Gong P, Wang L, et al. Mapping dynamic cover types in a large seasonally flooded wetland using extended principal component analysis and object-based classifi-ca-tion[J]. Remote Sensing of Environment, 2015, 158: 193-206.
[29] Mao J, Jain A K. Texture classification and segmentation using multiresolution simultaneous autoregressive models[J]. Pattern Recognition, 1992, 25(2): 173-188.
[30] 李曉婭,董雪嬌,徐艷. 我國(guó)耕作單元演變特征及影響因素研究[J]. 中國(guó)農(nóng)業(yè)資源與區(qū)劃,2016,37(5):20-26. Li Xiaoya, Dong Xuejiao, Xu Yan. Study on the evolution characteristics and influencing factors of farming unit in China[J]. Chinese Journal of Agricultural Resources and Regional Planning, 2016, 37(5): 20-26. (in Chinese with English Abstract)
[31] 白由路,金繼運(yùn),楊俐蘋,等. 基于GIS的土壤養(yǎng)分分區(qū)管理模型研究[J]. 中國(guó)農(nóng)業(yè)科學(xué),2001,34(1):1-4. Bai Youlu, Jin Jiyun, Yang Liping, et al. Research on the subarea management model of soil nutrients by GIS[J]. Scientia Agricultura Sinica, 2001, 34(1): 1-4. (in Chinese with English Abstract)
Comparison of precision management zoning methods in black soil area based on SOM and NDVI
Liu Huanjun1,2, Bao Yilin1, Xu Mengyuan1, Zhang Xinle1※, Meng Xiangtian1, Pan Yue1, Yang Haoxuan1, Xie Yahui1
(1.,150030,; 2.,130012,)
Cultivated land allocation is the key link to implement variable fertilization management. According to spatial heterogeneity, a field is divided into several sub-field blocks with different homogeneity to adjust soil and crop management measures. The explanate Machinery Cooperative of Heilongjiang Province is taken as a research object in the typical black soil area, and the SPOT-6 remote sensing images from June to September are obtained. With the support of Arcgis, crop growth can be simulated well with, such as the Normalized Difference Vegetation Index (NDVI); the soil organic matter (SOM) content is calculated according to the spatial interpolation method; and the field sample information is measured with iRTK2 and converted into the digital elevation model (DEM) raster data. Based on the spatial SOM distribution information, the SOM spatial information with the topographical factors, the spatial information of SOM with both DEM and NDVI in August, and spatial information with 4 phases of NDVI(in June, July, August, and September) are used as input. Since the inputs of this study are different from the previous single soil nutrient information, the synthesis of multiple spatial information can reflect the spatial difference of the study area in many aspects, which is more consistent with the actual influencing factors. The object-oriented segmentation method is used to divide the study area according to the principle of high homogeneity within the partition and high heterogeneity between partitions. In order to find the index elements that can better reflect the actual growth, the partition accuracy under different inputs is evaluated by two standard indicators, pixel standard deviation and Morans index, which reflect the suitability and accuracy of the partition. When the internal standard deviation of pixels is small, which proves that the soil physical and chemical properties and vegetation growth of each field are more similar to the reality; when the Morans index between the partitions is small, which shows that the differences between the partitions are large, and the spatial similarity is not obvious; which conforms to the principle of division of precise management partitions. The results show that the precision of the precise management partition based on spatial information with the 4 phases of NDVI is the highest, the internal standard deviation of the partition and the Morans index are 0.010 and 0.065, respectively. The partition accuracy for spatial information of SOM with both DEM and NDVI is the secondly, with standard deviation of 0.011 and the Morans index of 0.072 respectively. The accuracy for the SOM spatial information considering the topographical factors is relatively lower, with the internal standard deviation of 0.014 and the Morans index of 0.192. The accuracy of the partition based on only the SOM spatial information has the lowest accuracy, which internal pixel standard deviation and the Morans index are 0.015 and 0.223 respectively. Compared with the traditional spatial interpolation in precision management partition, the remote sensing image has advantages in both data acquisition and precision. In addition, the advantage of multi-source spatial data is that multiple factors can be considered comprehensively, which is more accurate than single data. This method saves a lot of time and more efficient than traditional grid sampling partitioning. The zoning results are expected to promoted field division and management in future research.
remote sensing; evaluation; spatial interpolation; multi-source spatial data; precision management partition; object-oriented
10.11975/j.issn.1002-6819.2019.13.020
S127;TP79
A
1002-6819(2019)-13-0177-07
2018-09-21
2019-06-29
國(guó)家自然科學(xué)基金(41671438);中國(guó)科學(xué)院東北地理與農(nóng)業(yè)生態(tài)研究所“引進(jìn)優(yōu)秀人才”項(xiàng)目吉林省科技發(fā)展計(jì)劃項(xiàng)目(20170301001NY)
劉煥軍,黑龍江穆棱人,副教授,博士生導(dǎo)師,主要從事精準(zhǔn)農(nóng)業(yè)及土壤遙感。Email:huanjunliu@yeah.net
張新樂,黑龍江雞西人,副教授,博士,主要研究方向?yàn)樯鷳B(tài)遙感。Email:xinlezhang@yeah.net
劉煥軍,鮑依臨,徐夢(mèng)園,張新樂,孟祥添,潘 越,楊昊軒,謝雅慧.基于SOM和NDVI的黑土區(qū)精準(zhǔn)管理分區(qū)對(duì)比 [J]. 農(nóng)業(yè)工程學(xué)報(bào),2019,35(13):177-183. doi:10.11975/j.issn.1002-6819.2019.13.020 http: //www.tcsae.org
Liu Huanjun, Bao Yilin, Xu Mengyuan, Zhang Xinle, Meng Xiangtian, Pan Yue,Yang Haoxuan, Xie Yahui.Comparison of precision management zoning methods in black soil area based on SOM and NDVI [J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2019, 35(13): 177-183. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2019.13.020 http://www.tcsae.org