基于聚類算法與多智能體的細(xì)碎化耕地空間優(yōu)化研究

關(guān)鍵詞：耕地細(xì)碎化；聚類算法；多智能體；耕地空間優(yōu)化；新北區(qū) 中圖分類號：F301.21;TP311.13;F323.211文獻(xiàn)標(biāo)志碼：A文章編號：2095-6819（2025）051196-10doi：10.13254/jare2024.0438

Spatial optimization of fragmented farmland based on clustering algorithm and multi-agent system

HUANG Bingyuan 1^{1，2，3，?} ， HUANG Qiuhac ^{1，2，3，5*} ，YANG Lihong1，2，3，4， CHEN Zhenjie1，23 （1.ScholofGeographyandOceanSciene，NanjingUniversityNanjing103，China;.KeyLaboratoryoftheCoastalZone ExploitationndProteoistryofatualsouesng，Cia;3.KeybatoryfoandtelitRoesin Applications，inistryfaturalesources，anjingo3China;4JangsurovincialKeyLaboratoryofGeogaphicIfoatio ScienceandTechnologyanjngO3，China;5.JiangsuCenterforColaborativeIovationiGeograpicalInformationRsouce Development and Application，Nanjing 21oo23，China）

Abstract：Toaievelargescaleutlatinofculivatedlandresoucestroughmropareladjustments，addressingthhallgsf fragmentedcultivatedlandgoveane，suingfoodsurityndpromotigsaiableiculturalevelopent，tisdd landscapepatteindicesandacomprehensiveweightedoverlaymethodtoevaluate tefragmentationandarablelandresouesuitability intheXinbeiDistrictofChangzhuCity.eKmedoidscusteringalgoritwasusedtocategoretypesoffarlandfragmentatnda multi-agentmodelwasdevelopedtooptimizethespatiallayoutofthecultivatedland.TheresultsshowedthatusingtheK-medods clusteringalgiatedlandsaiditooigosdlampepedndateddfrdt proportion of non-contiguous and regular cultivated land was 53.68% ，suggesting severe cultivated land fragmentation. The overall level of landresourcesuitabilityforcultivationinthestudyareawasrelativelyhigh，withtheproportionofhighlyandrelativelyhighlysuitable plotsbeing 14.03% and 44.83% ，respectively.The spatial distribution showedapatern of higher in the north and west and lower in the southndeastTultgntoelielyidpatialagetfgmnedasmprovgovalblad suitabilityiooossedadoutiniol optimizationresults，butrequiredlargerscalefarmlandadjustments，withareasofhgherrablelandresourcesuitabilitytakingonore responsibilityforlandrestoration.Thefarmlandspatialoptimizationmodeldevelopedinthisstudydemonstratesstrongoptiization effectiveness and provides a scientific basis for addressing regional farmland fragmentation.

Keywords：farmlandfragmentation;clusteringalgorithm;multi-agentsystem;farmland spatialoptimization;XinbeiDistrict

耕地細(xì)碎化是自然資源稟賦與人類活動(dòng)相互作用的復(fù)雜結(jié)果，耕地被分割成大小不一、空間分散的破碎地塊，是我國耕地資源利用的典型特征[1-2]。耕地細(xì)碎化會(huì)增加農(nóng)業(yè)生產(chǎn)成本3，降低農(nóng)田使用效率，阻礙農(nóng)業(yè)機(jī)械化發(fā)展，甚至導(dǎo)致農(nóng)藥和化肥的過度使用等，對區(qū)域糧食安全和農(nóng)業(yè)可持續(xù)發(fā)展造成威脅，已成為限制我國耕地資源高效利用的重要因素。長三角地區(qū)經(jīng)歷了快速的工業(yè)化和城市化，未來農(nóng)業(yè)用地?cái)U(kuò)展?jié)摿κ钟邢?。因此，提高耕地的集約化利用水平是未來保障區(qū)域糧食安全的關(guān)鍵措施。如何通過優(yōu)化耕地空間布局降低耕地細(xì)碎度、提高耕地利用效率、促進(jìn)耕地規(guī)模化利用已成為新時(shí)期推進(jìn)農(nóng)業(yè)現(xiàn)代化發(fā)展亟需解決的問題之一。

耕地細(xì)碎化評價(jià)作為有效發(fā)現(xiàn)和科學(xué)解決問題的重要過程，基本形成了以社會(huì)經(jīng)濟(jì)指數(shù)法景觀格局指數(shù)法9-10和多維綜合評價(jià)法為主的研究方法。目前學(xué)界多采用景觀格局指數(shù)法從規(guī)模、形態(tài)、分布等方面選取指標(biāo)對耕地細(xì)碎化程度進(jìn)行定量化和多尺度的評價(jià)。耕地細(xì)碎化治理研究普遍基于評價(jià)結(jié)果從整治分區(qū)、驅(qū)動(dòng)因素、治理措施等方面展開。整治分區(qū)方法如四象限分析法[12]、三維魔方模型[]和聚類算法等，其中聚類算法常用的有 K-means 聚類[13]二步聚類及K-means聚類[4]等。驅(qū)動(dòng)因素主要包括自然條件、政策制度、社會(huì)經(jīng)濟(jì)等[15]，常用的計(jì)算方法有雙向固定效應(yīng)模型[、地理加權(quán)回歸模型]隨機(jī)森林模型和偏相關(guān)分析方法[18等；……