楊 偉, 張樹(shù)文, 姜曉麗
1 太原師范學(xué)院地理科學(xué)學(xué)院, 晉中 030619 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 130102 3 太原師范學(xué)院城鎮(zhèn)與區(qū)域發(fā)展研究所, 晉中 030619
基于MODIS時(shí)序數(shù)據(jù)的黑龍江流域火燒跡地提取
楊 偉1, 張樹(shù)文2,*, 姜曉麗3
1 太原師范學(xué)院地理科學(xué)學(xué)院, 晉中 030619 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 130102 3 太原師范學(xué)院城鎮(zhèn)與區(qū)域發(fā)展研究所, 晉中 030619
火燒跡地信息是研究火災(zāi)的重要參數(shù)和基礎(chǔ)數(shù)據(jù),也是研究全球生態(tài)系統(tǒng)和碳循環(huán)擾動(dòng)的重要依據(jù)之一。以受森林火災(zāi)影響較為嚴(yán)重的黑龍江流域?yàn)檠芯繀^(qū),以MODIS時(shí)間序列數(shù)據(jù)為數(shù)據(jù)源建立了一個(gè)分為兩階段的火燒基地提取算法(即首先設(shè)定較為嚴(yán)格的提取條件對(duì)最有可能發(fā)生火災(zāi)的像元——核心像元進(jìn)行提取,然后設(shè)定較為寬松的閾值提取距離核心像元一定范圍內(nèi)的火燒像元),對(duì)2000—2011年的火燒跡地信息進(jìn)行了提取,生成了研究區(qū)長(zhǎng)時(shí)間序列火燒跡地分布圖,并對(duì)其時(shí)空分布特征進(jìn)行了分析。選擇黑龍江省為典型驗(yàn)證區(qū)對(duì)算法精度進(jìn)行了驗(yàn)證,結(jié)果顯示算法的整體精度較之以往的算法有了一定程度的提高。
火燒跡地; MODIS; GEMI; BAI
火災(zāi)是影響眾多生態(tài)系統(tǒng)(森林、草地等)的一個(gè)重要的擾動(dòng)因素?;馂?zāi)對(duì)于植被的結(jié)構(gòu)和組成具有顯著的作用,被認(rèn)為是一個(gè)重要的“地表管理工具”[1]。以森林生態(tài)系統(tǒng)為例,火災(zāi)是森林生態(tài)系統(tǒng)最為重要的干擾之一,全球平均每年約有1%的森林受到火災(zāi)的影響[2]。森林生態(tài)系統(tǒng)是全球碳循環(huán)的重要組成部分,火災(zāi)的發(fā)生通過(guò)改變森林生態(tài)系統(tǒng)的格局與過(guò)程,進(jìn)而改變整個(gè)生態(tài)系統(tǒng)的碳循環(huán)以及分配過(guò)程[3]。此外,森林火災(zāi)對(duì)于氣候變暖也有著重要的響應(yīng)[4- 6]。研究表明,北方森林的火災(zāi)發(fā)生范圍對(duì)于溫度的增加非常敏感,受氣候變化的影響較為顯著[7]。此外,還有研究認(rèn)為氣候變暖很有可能影響到傳統(tǒng)的火災(zāi)循環(huán),例如縮短火災(zāi)的周期、增大火災(zāi)尺度等[8-9]。
不論是對(duì)于碳循環(huán)的影響研究,還是森林火災(zāi)與氣候變暖的相關(guān)性研究,火燒跡地的空間信息都是一個(gè)重要的基礎(chǔ)參數(shù)。傳統(tǒng)的火燒跡地信息主要來(lái)源于統(tǒng)計(jì)數(shù)據(jù),難以覆蓋較大的區(qū)域,收集較為困難,且難以將數(shù)據(jù)進(jìn)行空間化。遙感技術(shù)的發(fā)展為解決這一問(wèn)題提供了很好的手段,特別是隨著遙感數(shù)據(jù)時(shí)空分辨率的提高,使得遙感數(shù)據(jù)能夠更為準(zhǔn)確的對(duì)地表過(guò)程進(jìn)行刻畫[10]。
在區(qū)域或者全球尺度下,為獲取長(zhǎng)時(shí)間序列的火燒跡地信息,中空間分辨率且具有高時(shí)間分辨率的遙感數(shù)據(jù)被認(rèn)為是最好的選擇。目前,應(yīng)用最為廣泛的為AVHRR(Advanced Very High Resolution Radiometer)數(shù)據(jù)[11- 13]與MODIS(Moderate-Resolution Imaging Spectroradia-meter) 數(shù)據(jù)[14-15]。前者由于發(fā)射時(shí)間較早,時(shí)間序列較長(zhǎng)而被使用。但研究表明AVHRR數(shù)據(jù)提取火燒跡地信息存在一定的潛在誤差,主要來(lái)源于輻射的不穩(wěn)定性、云污染以及輻射傳輸問(wèn)題等方面[16- 18]。較之AVHRR數(shù)據(jù),MODIS數(shù)據(jù)在這些方面都得到了很大的改善,但其局限在于數(shù)據(jù)僅從2000年開(kāi)始。
本文提出了一種基于MODIS時(shí)序數(shù)據(jù)的火燒跡地提取方法,以此為基礎(chǔ)對(duì)黑龍江流域2000—2011年的火燒跡地信息進(jìn)行了提取,生成研究區(qū)長(zhǎng)時(shí)間序列火燒跡地分布圖,并對(duì)其分布特征進(jìn)行了分析。
1.1 研究區(qū)
研究區(qū)選擇位于西伯利亞北方森林南部的黑龍江流域(41°45′—53°33′N,115°13′—135°05′E),面積208×104km2,西起蒙古高原,包括蒙古(東方省、蘇赫巴托爾省等)、俄羅斯(阿穆?tīng)栔?、哈巴羅夫斯克、外貝加爾等)和中國(guó)(黑龍江省、吉林省、遼寧省、內(nèi)蒙古自治區(qū)等)的13個(gè)省及朝鮮的小部分,研究區(qū)植被覆蓋度較高(圖1),受森林火災(zāi)影響較為嚴(yán)重。流域的東部地區(qū)主要屬于溫帶濕潤(rùn)季風(fēng)氣候,這是全球季風(fēng)氣候的最北緣,西部主要受大陸性氣候的影響。全年平均氣溫在-8 ℃到6 ℃之間,但其時(shí)空分布差異顯著。同時(shí),流域內(nèi)降水量的時(shí)空分布也很不均衡,年平均降水量主要在250—800 mm,大約50%以上的降水量集中在最熱的夏季,而近7個(gè)月的干季(1—4月,10—12月)降水量?jī)H為25%;在空間上,降水主要集中在沿海地帶,向西逐漸遞減。
圖1 研究區(qū)位置及其土地覆被Fig.1 Location and land cover of Heilongjiang basin
1.2 MODIS產(chǎn)品數(shù)據(jù)
研究選擇數(shù)據(jù)為MODIS 8d合成地表反射率數(shù)據(jù)MOD09Q1(MODIS atmospherically- correct Level 3 8-Day composite Surface Reflectance products, 空間分辨率250 m)以及MODIS 8d合成火產(chǎn)品數(shù)據(jù)MOD14A2(MODIS Level 3 8-Day composite active fire products, 空間分辨率1 km)。其中,MODIS火產(chǎn)品數(shù)據(jù)為MODIS火情監(jiān)測(cè)算法下提取的火點(diǎn)信息[19-20],其算法依據(jù)主要為火災(zāi)發(fā)生時(shí)的熱學(xué)特性,代表火災(zāi)發(fā)生時(shí)的溫度異常。
MOD09Q1與MOD14A2產(chǎn)品均采用Sinusoidal投影系統(tǒng)發(fā)布,數(shù)據(jù)格式為HDF (Hierarchy Data Format)。研究區(qū)共涉及6景MODIS標(biāo)準(zhǔn)分幅數(shù)據(jù)。下載研究區(qū)2000—2011年的MODIS產(chǎn)品數(shù)據(jù)(每年34期×12年×6景)。時(shí)間選擇為每年的2月末至11月初,這一時(shí)間為火災(zāi)發(fā)生集中的時(shí)間段。對(duì)數(shù)據(jù)進(jìn)行投影轉(zhuǎn)換及裁剪處理。此外,由于數(shù)據(jù)空間分辨率不一致,需要將MOD14A2產(chǎn)品重采樣為250 m分辨率,與MOD09Q1產(chǎn)品相一致。
1.3 火燒跡地提取方法
基于遙感的火災(zāi)研究主要包括火點(diǎn)(active fire)監(jiān)測(cè)[21-22]與火燒跡地(burn scars)提取[23-24]兩類,兩者均可以產(chǎn)生火燒跡地?cái)?shù)據(jù)[25]?;瘘c(diǎn)監(jiān)測(cè)主要是基于火災(zāi)發(fā)生時(shí)的溫度異常,監(jiān)測(cè)衛(wèi)星過(guò)境時(shí)可能發(fā)生火災(zāi)的像元,對(duì)火點(diǎn)進(jìn)行實(shí)時(shí)觀測(cè)。這一過(guò)程的主要目的在于捕捉火災(zāi)發(fā)生的時(shí)間以及位置信息,雖然也可以產(chǎn)生火燒跡地信息,但結(jié)果并不可靠[26-27]?;馃E地提取通過(guò)對(duì)比火災(zāi)發(fā)生前后的光譜反射特征或者光譜指數(shù)特征變化來(lái)識(shí)別火燒跡地,從而達(dá)到提取火燒跡地的目的[28-29]。這一方法的缺陷在于,一些與火災(zāi)發(fā)生具有類似光譜特征的事件較難區(qū)分,如洪水、森林砍伐以及農(nóng)作物收獲等導(dǎo)致的光譜特征變化。本文的火燒跡地提取方法以后者為基礎(chǔ),并進(jìn)行了改進(jìn),以消除類似的混淆事件。
1.3.1 判別指數(shù)選取
火燒跡地識(shí)別方法是通過(guò)比較火災(zāi)發(fā)生前后的光譜特征變化來(lái)提取火燒面積。因而需要選取適合的光譜指數(shù)來(lái)進(jìn)行表征,如NDVI(Normalized Difference Vegetation Index)、BBFI(Burned Boreal Forest Index)、GEMI(Global Environmental Monitoring Index)以及BAI (Burned Area Index)等。其中,NDVI的應(yīng)用最為廣泛。NDVI能夠很好的對(duì)植被覆蓋進(jìn)行描述[30,31],但研究表明NDVI在植被覆蓋度較高的地區(qū)容易達(dá)到飽和[32],且在火燒跡地信息提取中存在較大的潛在誤差[33]。因此,選用GEMI作為識(shí)別火燒跡地的主要判別指數(shù),其計(jì)算公式如下:
GEMI=η×(1-0.25η)-(ρred-0.125)/(1-ρred)
η=(2(ρnir2-ρred2)+1.5ρnir+0.5ρred)/(ρnir+ρred+0.5)
(1)
式中,ρnir以及ρred為近紅外波段和紅光波段。火災(zāi)發(fā)生后GEMI表現(xiàn)出顯著的下降。
為了避免采用單一光譜指數(shù)所帶來(lái)的潛在誤差,選擇了另一個(gè)光譜指數(shù)BAI來(lái)作進(jìn)一步的限定,其計(jì)算公式如下:
BAI=1/((ρnir-ρcnir)2+(ρred-ρcred)2)
(2)
式中,ρcnir和ρcred分別被設(shè)定為0.06與0.1?;馂?zāi)發(fā)生后BAI值表現(xiàn)出顯著的上升。
除此之外,在比較火災(zāi)發(fā)生前后光譜特征變化的同時(shí),為了考慮火災(zāi)發(fā)生時(shí)的熱學(xué)特性,即溫度異常,將MODIS火產(chǎn)品數(shù)據(jù)作為一個(gè)輸入波段加入到判別流程,以提高判別精度。
1.3.2 判別流程
火燒跡地的識(shí)別流程主要分為兩個(gè)階段:首先,設(shè)定較為嚴(yán)格的判別閾值以提取火燒的核心像元——即火災(zāi)最有可能發(fā)生的像元。這一階段的主要目標(biāo)在于盡可能的減少錯(cuò)判誤差,因而需要對(duì)火災(zāi)發(fā)生前后的光譜指數(shù)變化設(shè)定嚴(yán)格的閾值,并且同時(shí)用使用MODIS火災(zāi)產(chǎn)品進(jìn)行篩選,以表示提取像元在相關(guān)植被指數(shù)變化前曾出現(xiàn)溫度異常。這一過(guò)程同時(shí)考慮了火災(zāi)發(fā)生前后地表植被的突變以及火災(zāi)發(fā)生時(shí)的溫度異常,從而可以與其他造成地表突變的因素相區(qū)分。其次,對(duì)第一階段提取的核心像元15公里范圍內(nèi)的光譜指數(shù)變化特征進(jìn)行判別,設(shè)定較為寬松的閾值,以盡可能減少漏判誤差。判別流程及判別條件如圖2所示。
圖2 算法流程圖Fig.2 Flowchart of algorithm
第一階段的提取過(guò)程以GEMI、BAI、以及MOD14A2產(chǎn)品為基礎(chǔ),具體的判別條件如下所示:
首先,火災(zāi)發(fā)生之前的GEMI值必須大于一定的閾值,以確保判別區(qū)域?yàn)橹脖桓采w。
GEMIt-1>0.170
(3)
式中,t為時(shí)間(下同)。選擇數(shù)據(jù)為MODIS 8d合成數(shù)據(jù),每年共34期數(shù)據(jù),因此t的范圍為:0 火災(zāi)發(fā)生后,GEMI值必須表現(xiàn)出顯著的下降,且這一下降過(guò)程必須持續(xù)一定的時(shí)間,以區(qū)分由云污染等造成的GEMI值的短暫下降。這一過(guò)程通過(guò)以下兩個(gè)判別條件來(lái)實(shí)現(xiàn): (GEMIt-GEMIt-1)/GEMIt<-0.1 (4) (GEMIt+2-GEMIt-1)/GEMIt+2<-0.1 (5) 然后,使用BAI指數(shù)來(lái)對(duì)火燒像元做進(jìn)一步的限定?;馂?zāi)發(fā)生后,BAI值顯著增加,其判別條件如下: BAIt>250且 BAIt-1>200 (6) 最后,使用MODIS 火產(chǎn)品數(shù)據(jù)來(lái)對(duì)火燒像元進(jìn)行掩膜,以保證光譜指數(shù)變化前,所提取像元表現(xiàn)出溫度異常的特征。 ρt>6 或者 ρt-1>6 (7) 式中,ρ為MODIS火產(chǎn)品數(shù)據(jù)像元值。 第二階段的判別過(guò)程以第一階段提取的核心像元為基礎(chǔ),采用較為寬松的閾值來(lái)對(duì)鄰近像元進(jìn)行判別。在對(duì)研究區(qū)的火災(zāi)發(fā)生特征進(jìn)行分析之后,距離核心像元的最大距離被設(shè)定為15 km。第二階段的火燒跡地信息提取,僅對(duì)核心像元15 km范圍內(nèi)像元進(jìn)行判別,判別條件如下: GEMIt-GEMIt-1<-0.03 (8) GEMIt+1-GEMIt-1<-0.02 (9) GEMIt+2-GEMIt-1<0 (10) GEMIt+1-GEMIt≤0 (11) BAIt>250 (12) 最后,將兩個(gè)階段的提取結(jié)果進(jìn)行合成。采用一個(gè)3×3的變換核,對(duì)合成結(jié)果進(jìn)行濾波處理,消除提取過(guò)程中產(chǎn)生的小斑塊。 2.1 精度驗(yàn)證 由于研究區(qū)涉及境外地區(qū),驗(yàn)證數(shù)據(jù)難以獲取??紤]到以上提取方法的基礎(chǔ)為火燒前后植被指數(shù)的變化以及溫度的異常,使得該方法對(duì)火燒跡地的提取具有普適性,從而可以采用選擇典型驗(yàn)證區(qū)的方法對(duì)算法進(jìn)行驗(yàn)證。故以黑龍江省為典型研究區(qū),對(duì)提取結(jié)果進(jìn)行精度驗(yàn)證。驗(yàn)證數(shù)據(jù)來(lái)源于相關(guān)林業(yè)部門2000—2005年的火災(zāi)統(tǒng)計(jì)數(shù)據(jù),包括火災(zāi)發(fā)生的時(shí)間、地點(diǎn)、經(jīng)緯度信息以及過(guò)火面積等。 圖3 黑龍江省火災(zāi)發(fā)生位置(2000—2005年)Fig.3 Fire position of Heilongjiang province (2000—2005) 鑒于MODIS產(chǎn)品空間分辨率以及火燒跡地信息提取后進(jìn)行去除小斑塊的濾波處理的需要,對(duì)過(guò)火面積小于60 hm2(約3×3個(gè)像元)的火災(zāi)進(jìn)行剔除,最終得到黑龍江省2000—2005年的火災(zāi)驗(yàn)證數(shù)據(jù)(圖3)。 利用2000年—2005年火災(zāi)發(fā)生的經(jīng)緯度信息(圖3)對(duì)提取結(jié)果進(jìn)行錯(cuò)判以及漏判分析(表1)。由于驗(yàn)證數(shù)據(jù)僅提供了火災(zāi)發(fā)生的位置,因而不能對(duì)提取結(jié)果進(jìn)行空間化(逐像元)的誤差分析。以火災(zāi)發(fā)生的位置信息為參照,對(duì)提取結(jié)果進(jìn)行分析,兩者一致則認(rèn)為提取結(jié)果正確。如果在標(biāo)有火災(zāi)發(fā)生的位置沒(méi)有提取出火燒跡地信息,被認(rèn)為是漏判;相反,在沒(méi)有標(biāo)出火災(zāi)發(fā)生的位置,卻提取出火燒跡地信息,認(rèn)為是錯(cuò)判。 表1 火燒跡地提取驗(yàn)證表 此外,將提取結(jié)果的面積進(jìn)行匯總與驗(yàn)證數(shù)據(jù)進(jìn)行了比較(表1)。結(jié)果顯示,2000—2005年每年均有一定的漏判以及錯(cuò)判誤差存在,且提取的火燒跡地面積均小于驗(yàn)證數(shù)據(jù),總體精度為71%。其中,提取面積精度最高為2002年,達(dá)84%;提取面積精度最低為2003年,精度為61%。較之以往的研究[35],精度有所提高。 2.2 黑龍江流域火燒跡地分布特征分析 2.2.1 黑龍江流域火燒跡地信息提取 圖4 黑龍江流域火燒跡地分布(2000—2011) Fig.4 Distribution of burned area in Heilongjiang basin (2000—2011) 使用以上所驗(yàn)證的火燒跡地提取方法,以MOD09Q1數(shù)據(jù)以及MOD14A2數(shù)據(jù)為基礎(chǔ),對(duì)黑龍江流域2000—2011年的火燒跡地信息進(jìn)行了提取,得到研究區(qū)長(zhǎng)時(shí)間序列火燒跡地分布圖(圖4)。 2.2.2 火燒跡地特征分析 將黑龍江流域的火燒跡地面積進(jìn)行逐年匯總,從而得到流域逐年的火燒跡地面積統(tǒng)計(jì)數(shù)據(jù)(圖5)。 圖5 黑龍江流域火燒跡地面積變化Fig.5 Burned area dynamic in Heilongjiang basin 黑龍江流域2000—2011年受火災(zāi)影響較為嚴(yán)重,年均過(guò)火面積達(dá)53.21萬(wàn)hm2。火災(zāi)發(fā)生最嚴(yán)重的年份為2003年,面積為146.79萬(wàn)hm2;而受火災(zāi)影響最小的年份為2010年,過(guò)火面積僅有18.39萬(wàn)hm2,差距較大。火災(zāi)發(fā)生較為嚴(yán)重的年份還包括2008年,過(guò)火面積也超過(guò)了百萬(wàn)公頃,達(dá)119.41萬(wàn)hm2。其他年份受火災(zāi)影響較為平均,其中2005年相對(duì)較為嚴(yán)重,面積為62.41萬(wàn)hm2;其次為2001、2002、2004年以及2011年,火燒跡地面積分別為41.55萬(wàn)hm2、43.83萬(wàn)hm2、43.59萬(wàn)hm2以及43.23萬(wàn)hm2;最后為2000、2006、2007年以及2009年,火燒跡地面積分別為33.08萬(wàn)hm2、38.87萬(wàn)hm2、24.01萬(wàn)公頃以及23.41萬(wàn)hm2。 從火燒跡地的空間特征來(lái)看,火燒跡地的分布與森林覆被密切相關(guān),主要分布于黑龍江流域的中、高緯度地區(qū)。俄羅斯境內(nèi)的火燒跡地分布較為均勻,原因在于其高的植被覆蓋度。中國(guó)境內(nèi)火燒跡地主要分布于大、小興安嶺以及長(zhǎng)白山地區(qū),其中以黑龍江省受影響最為嚴(yán)重。此外,蒙古和朝鮮境內(nèi)也有少部分火燒跡地分布。 長(zhǎng)時(shí)間序列火燒跡地?cái)?shù)據(jù)是區(qū)域或者全球尺度下森林火災(zāi)相關(guān)研究的重要基礎(chǔ)信息。本文以黑龍江流域?yàn)檠芯繀^(qū),利用MODIS時(shí)序數(shù)據(jù)對(duì)其2000—2011年的火燒跡地信息進(jìn)行了提取,主要結(jié)論如下: 低空間分辨率高時(shí)間分辨率的遙感數(shù)據(jù)是區(qū)域或者全球尺度下火燒跡地信息提取的主要數(shù)據(jù)源,通過(guò)建立相應(yīng)的算法,可以實(shí)現(xiàn)長(zhǎng)時(shí)間序列火燒跡地信息的提取過(guò)程。 綜合考慮火災(zāi)發(fā)生前后的植被變化(光譜指數(shù)變化)與火災(zāi)發(fā)生時(shí)的熱學(xué)特征能夠更為有效的對(duì)火燒跡地信息進(jìn)行提取,提高提取精度。 黑龍江流域2000—2011年受火災(zāi)影響較為嚴(yán)重,年平均產(chǎn)生火燒跡地53.21萬(wàn)hm2,受火災(zāi)影響最大年份與最小年份之間的火燒跡地面積差距較大。 本文基于MODIS數(shù)據(jù)空間分辨率以及提取結(jié)果濾波處理的需求,將過(guò)火面積大于60 hm2的森林火災(zāi)作為驗(yàn)證數(shù)據(jù),較之Emilio等使用大于200 hm2的火災(zāi)[34]作為驗(yàn)證數(shù)據(jù),提高了對(duì)算法精度的要求。結(jié)果顯示主要的漏判誤差仍來(lái)源于100 hm2左右的森林火災(zāi)。證明由于遙感數(shù)據(jù)空間分辨率的局限,算法對(duì)于面積較小的火燒跡地提取具有一定難度。而空間分辨率相對(duì)較高的遙感數(shù)據(jù),如TM數(shù)據(jù),其時(shí)間分辨率卻難以滿足火燒跡地信息提取的要求。遙感數(shù)據(jù)融合可以較好的解決這一問(wèn)題,如何采用融合之后的高空間分辨率以及時(shí)間分辨率的遙感數(shù)據(jù)進(jìn)行更為細(xì)致的火燒跡地提取有待深入研究。 [1] Dubinin M, Potapov P, Lushchekina A, Radeloff V C. Reconstructing long time series of burned areas in arid grasslands of southern Russia by satellite remote sensing. Remote Sensing of Environment, 2010, 114(8): 1638- 1648. [2] Fraser R H, Li Z. Estimating fire-related parameters in boreal forest using SPOT VEGETATION. Remote Sensing of Environment, 2002, 82(1): 95- 110. [3] 呂愛(ài)鋒, 田漢勤, 劉永強(qiáng). 火干擾與生態(tài)系統(tǒng)碳循環(huán). 生態(tài)學(xué)報(bào), 2005, 25(10): 2734- 2743. [4] Amiro B D, Orchansky A L, Barr A G, Black T A, Chambers S D, Chapin III F S, Goulden M L, Litvak M, Liu H P, McCaughey J H, McMillan A, Randerson J T. The effect of post-fire stand age on the boreal forest energy balance. Agricultural and Forest Meteorology, 2006, 140(1/4): 41- 50. [5] Kasischke E S, Turetsky M R. Recent changes in the fire regime across the North American boreal region—Spatial and temporal patterns of burning across Canada and Alaska. Geophysical Research Letters, 2006, 33(9): 1- 5. [6] Morton D C, Collatz G J, Wang D, Randerson J T, Giglio L, Chen Y. Satellite-based assessment of climate controls on US burned area. Biogeosciences, 2013, 10(1): 247- 260. [7] Flannigan M D, Logan K A, Amiro B D, Skinner W R, Stocks B J. Future area burned in Canada. Climatic Change, 2005, 72(1/2): 1- 16. [8] IPCC. Climate Change 2007: The Physical Science Basis. Cambridge: Cambridge University Press, 2007. [9] Gillet N P, Weaver A J, Zwiers F W, Flannigan M D. Detecting the effect of climate change on Canadian forest fires. Geophysical Research Letters, 2004, 31(18), L18211, doi: 18210.11029/12004GL02 0876. [10] Mouillot F, Schultz M G, Yue C, Cadule P, Tensey K, Ciais P, Chuvieco E. Ten years of global burned area products from spaceborne remote sensing-A review: Analysis of user needs and recommendations for future developments. International Journal of Applied Earth Observation and Geoinformation, 2014, 26: 64- 79. [11] Vafeidis A T, Drake N A, Wainwright J. A proposed method for modelling the hydrologic response of catchments to burning with the use of remote sensing and GIS. Catena, 2007, 70(3): 396- 409. [12] Pu R L, Li Z Q, Gong P, Csiszar I, Fraser R, Hao W M, Kondragunta S, Weng F Z. Development and analysis of a 12-year daily 1-km forest fire dataset across North America form NOAA/AVHRR data. Remote Sensing of Environment, 2007, 108(2): 198- 208. [13] Sukhinin A I, French N H F, Kasischke E S, Hewson J H, Soja A J, Csiszar I A, Hyer E J, Loboda T, Gonrad S G, Romasko V I, Pavichenko E A, Miskiv S I, Slinkina O A. AVHRR-based mapping of fires in Russia: New products for fire management and carbon cycle studies. Remote Sensing of Environment, 2004, 93(4): 546- 564. [14] Oadilla M, Stehman S V, Litago J, Chuvieco E. Assessing the temporal stability of the accuracy of a time series of burned area products. Remote Sensing, 2014, 6(3): 2050- 2068. [15] Roy D P, Boschetti L, Justice C O, Ju J. The collection 5 MODIS burned area product—Global evaluation by comparison with the MODIS active fire product. Remote Sensing of Environment, 2008, 112(9): 3690- 3707. [16] Barbosa P M, Gregoire J M, Pereira J M C. An algorithm for extracting burned areas from time series of AVHRR GAC data applied at a continental scale. Remote Sensing of Environment, 1999, 69(3): 253- 263. [17] Martin M P, Chuvieco E. Mapping and evaluation of burned land from multi-temporal analysis of AVHRR NDVI images. EARSeL Advances in Remote Sensing, 1995, 4(3): 7- 13. [18] Pereira J M C. A comparative evaluation of NOAA/AVHRR vegetation indexes for burned surface detection and mapping. IEEE Transactions on Geoscience and Remote Sensing, 1999, 37(1): 217- 226. [19] Giglio L, Descloitres J, Justice C O, Kaufman Y J. An enhanced contextual fire detection algorithm for MODIS. Remote Sensing of Environment, 2003, 87(2/3): 273- 282. [20] Dozier J. A method for satellite identification of surface temperature fields of subpixel resolution. Remote Sensing of Environment, 1981, 11: 221- 229. [21] Movaghati S, Samadzadegan F, Azizi A. An agent-based approach for regional forest fire detection using MODIS data. Journal of Applied Sciences, 2009, 9(20): 3672- 3681. [22] 王彩云, 扎西頓珠, 陳濤. EOS/MODIS數(shù)據(jù)在西藏森林火災(zāi)中的應(yīng)用. 高原山地氣象研究, 2010, 30(3): 66- 69. [23] Ruiz J A M, Riano D, Arbelo M, French N H F, Ustin S L, Whiting M L. Burned area mapping time series in Canada (1984- 1999) from NOAA-AVHRR LTDR: A comparison with other remote sensing products and fire perimeters. Remote Sensing of Environment, 2012, 117: 407- 414. [24] Giglio L, Loboda T, Roy D P, Quayle B, Justice C O. An active-fire based burned area mapping algorithm for the MODIS sensor. Remote Sensing of Environment, 2009, 113(2): 408- 420. [25] Zhang X Y, Kondragunta S, Quayle B. Estimation of biomass burned areas using multiple-satellite-observed active fires. IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(11): 4469- 4482. [26] Kasischke E S, Hewson J H, Stock B, van der Werf G, Randerson J. The use of ATSR active fire counts for estimating relative patterns of biomass burning-a study from the boreal forest region. Geophysical Research Letters, 2003, 30(18), doi: 10.1029/2003GL017859. [27] Giglio L, Van der Werf G R, Randerson J T, Collatz G J, Kasibhatla P. Global estimation of burned area using MODIS active fire observations. Atmospheric Chemistry and Physics, 2006, 6(4): 957- 974. [28] Key C H. Ecological and sampling constraints on defining landscape fire severity. Fire Ecology, 2006, 2(2): 34- 59. [29] Roy D P, Jin Y, Lewis P E, Justice C O. Prototyping a golbal algorithm for systematic fire-affected area mapping using MODIS time series data. Remote Sensing of Environment, 2005, 97(2): 137- 162. [30] Fraser R H, Li Z, Cihlar J. Hotspot and NDVI differencing synergy (HANDS): a new technique for burned area mapping over boreal forest. Remote Sensing of Environment, 2000, 74(3): 362- 376. [31] Kucera J, Yasuoka Y, Dye D G. Creating a forest fire database for the Far East Asia using NOAA/AVHRR observation. International Journal of Remote Sensing, 2005, 26(11): 2423- 2439. [32] Cai H Y, Zhang S W, Bu K, Yang J C, Chang L P. Intergrating geographical data and phenological characteristics derived from MODIS data for improving land over mapping. Journal of Geographical Sciences, 2011, 21(4): 705- 718. [33] Chuvieco E, Martín M P, Palacios A. Assessment of different spectral indices in the red-near-infrared spectral domain for burned land discrimination. International Journal of Remote Sensing, 2002, 23(23): 5103- 5110. [34] Chuvieco E, Englefield P, Trishchenko A P, Luo Y. Generation of long time series of burn area maps of the boreal forest from NOAA-AVHRR composite data. Remote Sensing of Environment, 2008, 112(5): 2381- 2396. Burned area mapping for Heilongjiang basin based on MODIS time series data YANG Wei1, ZHANG Shuwen2,*, JIANG Xiaoli3 1SchoolofGeographicalSciences,TaiyuanNormalUniversity,Jinzhong030619,China2NortheastInstituteofGeographyandAgroecology,ChineseAcademyofScience,Changchun130102,China3InstituteofUrbanandDistrictDevelopment,TaiyuanNormalUniversity,Jinzhong030619,China Fire is an important natural disturbance that affects several ecosystems and is also one of the main factors of the terrestrial carbon cycle. As fire modifies the structure and composition of vegetation, it is considered to be an important land management tool. Burned area mapping is an essential step in forest fire research to investigate the relationship between forest fire and climate change and the effect of forest fire on carbon budgets. Traditional data collection of forest fires in field- which are statistically recorded are difficult to manipulate over a large area. The development of the remote sensing technique provides a labor-efficient method for research of land surface processes. At the regional or global scale, in order to obtain a long-time series of burned area maps, a moderate spatial resolution with high temporal resolution remote sensing data is considered as the best alternative. Currently, the most widely used remote sensing data are Advanced Very High Resolution Radiometer (AVHRR) images and Moderate-Resolution Imaging Spectroradiometer (MODIS) images. Although the AVHRR provides continuous observations for burned area analyses, some studies have identified several sources of potential errors in burned area discrimination from this sensor, mainly due to its radiometric instability, cloud obscuration, and transmission problems. Most of these problems have been notably reduced in the MODIS sensor, which offers greater spectral, spatial, and radiometric resolution than the AVHRR. This study proposes an algorithm to map areas burned by forest fire using MODIS time series data in Heilongjiang Valley, China. The algorithm is divided into two steps: First, the “core” pixels were extracted to represent the most possible burned pixels based on a comparison of the temporal change of the Global Environmental Monitoring Index (GEMI), the Burned Area Index (BAI), and the MODIS active fire products between pre- and post-fire spatial patterns. Second, a 15-km distance was set to extract the entire burned area near the “core” pixels. These more relaxed conditions were used to identify the fire pixels for reducing the omission error as much as possible. The algorithm comprehensively considered the thermal characteristics and the spectral change between pre- and post-fire spatial patterns, which were represented by the MODIS fire products and the spectral index, respectively. Heilongjiang province in China was selected as the typical study area to validate the accuracy of the algorithm. The results showed that with the use of the MODIS fire products, the accuracy of the algorithm was improved, with an overall accuracy of 71% and a highest accuracy of 84%. Consequently, the algorithm used in this study produced a long-time series of burned area maps of the study area from 2000 to 2011 with a relatively high accuracy. According to the burned area maps, the study area has been seriously affected by fire disasters on average of 0.53 million ha of burned land each year. The most affected years were 2003 and 2008 with burned areas exceeding 1 million ha. The least affected year was 2010 with a burned area of just 0.18 million ha. The relatively large disparity between the maximum and minimum values of the areas burned by forest fire indicates that there is a fluctuation in the severity of disaster during the studied period. burned area; MODIS; GEMI; BAI 中國(guó)科學(xué)院戰(zhàn)略性先導(dǎo)科技專項(xiàng)(XDA05090310) 2013- 12- 31; 日期:2014- 11- 03 10.5846/stxb201312313076 *通訊作者Corresponding author.E-mail: zhangshuwen@neigae.ac.cn 楊偉, 張樹(shù)文, 姜曉麗.基于MODIS時(shí)序數(shù)據(jù)的黑龍江流域火燒跡地提取.生態(tài)學(xué)報(bào),2015,35(17):5866- 5873. Yang W, Zhang S W, Jiang X L.Burned area mapping for Heilongjiang basin based on MODIS time series data.Acta Ecologica Sinica,2015,35(17):5866- 5873.2 結(jié)果與分析
3 結(jié)論與討論