基于神经网络模型的富硒特色农作物产区预测——以河北省张北县为例  

Prediction of selenium-rich characteristic crops production area based on neural network model:a case study of Zhangbei County,Hebei province

作　　者：徐丹虹 刘继红 张素荣 刘宏伟 白耀楠 李状 XU Danhong;LIU Jihong;ZHANG Surong;LIU Hongwei;BAI Yaonan;LI Zhuang(Tianjin Center,China Geological Survey(North China Center of Geoscience Innovation),Tianjin 300170,China;Xiongan Urban Geological Research Center,China Geological Survey,Tianjin 300170,China;Tianjin Key Laboratory of Coast Geological Processes and Environmental Safety,Tianjin 300170,China)

机构地区：[1]中国地质调查局天津地质调查中心(华北地质科技创新中心),天津300170 [2]中国地质调查局雄安城市地质研究中心,天津300170 [3]天津市海岸带地质过程与环境安全重点实验室,天津300170

出　　处：《华北地质》2024年第2期28-37,共10页North China Geology

基　　金：中国地质调查项目:“京津冀协同发展区暨雄安新区资源环境承载能力监测评价”(DD20221727);“京津冀鲁耕地区土地质量地球化学调查”(DD20160325);“华北地区自然资源动态监测与风险评估”(DD20230101)。

摘　　要：【研究目的】近年来,富硒产业迅速发展,但因土壤硒含量与农作物籽实硒含量并没有显著的正相关关系,依据土壤硒含量进行富硒农产品种植区划具有一定的不确定性。因此,如何通过建立农产品富硒模型预测富硒农产品产出区,进行合理规划是急需解决的科学问题。【研究方法】本文以河北省张北县农作物、土壤为研究对象,在测试土壤pH、有机碳、SiO_(2)、Al_(2)O_(3)、CaO、MgO、Fe_(2)O_(3)、K_(2)O、Na_(2)O、Se等指标和搜集1/25万多目标区域地球化学数据的基础上,研究了农作物吸收硒的影响因素,建立了基于神经网络的农作物硒含量预测模型,进行富硒特色农作物产出区的预测。【研究结果】张北县表层土壤中仅有2.31%的点位达到了富硒土壤标准,主要分布在三号乡北部地区。满足富硒标准的莜麦有6件,富硒率为16.67%,亚麻有7件,富硒率为21.86%。亚麻、莜麦对硒的富集能力与土壤中的有机碳、Al_(2)O_(3)、Fe_(2)O_(3)含量呈负相关关系,与pH值呈正相关关系。据此,通过机器学习构建了亚麻、莜麦籽实硒含量预测模型,结果较为可靠。富硒亚麻、莜麦的可能产区主要分布在张北、小二台、台路沟、油娄沟、郝家营等乡镇,产出区面积分别为2486 km^(2)和2406 km^(2)。【结论】通过农作物富硒预测模型圈定的富硒作物产区面积远大于富硒土壤面积,对张北县发展富硒农业具有十分重要的意义。This paper is the result of geochemistry.[Objective]In recent years,selenium enrichment industry has developed rapidly,but because there is no significant positive correlation between selenium content in soil and selenium content in crop seeds,the planting regionalization of selenium-enriched agricultural products based on soil selenium content has certain uncertainties.Therefore,how to establish the selenium-rich model of agricultural products to predict the production area of selenium-rich agricultural products and make reasonable planning is an urgent scientific problem.[Methods]The crops and soils were collected from Zhangbei county of Hebei province.And pH,organic carbon,silica,aluminum trioxide,calcium oxide,magnesium oxide,ferric oxide,potassium oxide,sodium oxide,selenium,arsenic,mercury,cadmium,lead,copper,zinc,chromium,and nickel were measured.The influencing factors of selenium uptake in crops were studied,and the prediction model of crop selenium content based on neural network was established,and the yield area of selenium rich crops was predicted,base on the results of measurement and the data of 1/250000 multiple geochemistry collected.[Results]There were only 2.31 percent of topsoil sites reached the standard of selenium-enriched,and they mainly distributed in the northern area of Sanhao Town.There are 6 naked oats which meet the standard of selenium-enriched,and the selenium-enriched rate is 16.67 percent.While there are 7 flaxseeds which meet the standard of selenium-enriched,and the selenium-enriched rate is 21.86 percent.There was a negative relationship between crops'bioconcentration factor for selenium with organic carbon,ferric oxide and aluminum trioxide,a positive relationship with pH in soil.The prediction model of selenium content in flax and naked oat seed was established by machine learning,and the results were reliable.The potential production areas of flax and naked oats were mainly distributed in Zhangbei,Xiaoertai,Tailugou,Youlouou and Haojiaying Town,etc,and the output areas were

关 键 词：硒 张北县 土壤 农作物 预测模型 评价 

分 类 号：S153.6[农业科学—土壤学]