Tinokutendai nekushanyira Nature.com. Shanduro yebrowser yauri kushandisa ine rutsigiro rushoma rweCSS. Kuti uwane ruzivo rwakanaka, tinokurudzira kuti ushandise browser yakagadziridzwa (kana kudzima compatibility mode muInternet Explorer). Panguva ino, kuti tive nechokwadi chekuti rutsigiro runoramba ruchienderera mberi, ticharatidza saiti iyi isina masitayera uye JavaScript.
Kusvibiswa kwevhu idambudziko guru rinokonzerwa nemabasa evanhu. Kupararira kwenzvimbo yezvinhu zvinogona kuva nechepfu (PTEs) kunosiyana munzvimbo zhinji dzemaguta nedziri pedyo nemaguta. Nokudaro, zvakaoma kufanotaura zviri mukati mePTEs muvhu rakadaro. Samples 115 dzakawanikwa kubva kuFrydek Mistek muCzech Republic. Calcium (Ca), magnesium (Mg), potassium (K) uye nickel (Ni) concentrations zvakaonekwa uchishandisa inductively coupled plasma emission spectrometry. Response variable ndiNi uye predictors ndiCa, Mg, naK. Correlation matrix pakati pe response variable ne predictor variable inoratidza hukama hunogutsa pakati pezvinhu. Mhedzisiro yekufanotaura yakaratidza kuti Support Vector Machine Regression (SVMR) yakashanda zvakanaka, kunyangwe inofungidzirwa root mean square error (RMSE) (235.974 mg/kg) uye mean absolute error (MAE) (166.946 mg/kg) yaive yakakwira kupfuura dzimwe nzira dzakashandiswa. Mixed models yeEmpirical Bayesian Kriging-Multiple Linear Regression (EBK-MLR) perform. Zvakaipa, sezvinoratidzwa nema coefficients ekugadzika pasi pe 0.1. Modhi yeEmpirical Bayesian Kriging-Support Vector Machine Regression (EBK-SVMR) ndiyo yaive modhi yakanakisa, ine RMSE yakaderera (95.479 mg/kg) uye MAE (77.368 mg/kg) uye coefficient yakakwira yekugadzika (R2 = 0.637). Kubuda kwehunyanzvi hweEBK-SVMR kunoonekwa uchishandisa mepu inozvironga. Ma neuron akaunganidzwa mudenderedzwa rechikamu chehybrid model CakMg-EBK-SVMR anoratidza mapatani emavara akawanda anofanotaura kuwanda kweNi muvhu remaguta neremunharaunda. Mhedzisiro yacho inoratidza kuti kusanganisa EBK neSVMR inzira inoshanda yekufanotaura kuwanda kweNi muvhu remaguta nerenharaunda.
Nickel (Ni) inoonekwa sechinhu chinovaka muviri chezvirimwa nekuti inobatsira mukugadzirisa nitrogen (N) uye urea metabolism, zvese zvinodiwa kuti mbeu dzimere. Pamusoro pekubatsira kwayo mukumera kwembeu, Ni inogona kushanda semushonga wefungus uye bacteria uye inokurudzira kukura kwezvirimwa. Kushaikwa kwenickel muvhu kunoita kuti chirimwa chikwanise kuinwa, zvichikonzera chlorosis yemashizha. Semuenzaniso, nyemba nebhinzi dzinoda kushandiswa kwefetereza inobva kunickel kuti iwedzere nitrogen fixation 2. Kuramba uchishandisa fetereza inobva kunickel kuti iwedzere kupfuma kwevhu uye kuwedzera kugona kwembeu kugadzirisa nitrogen muvhu kunowedzera huwandu hwenickel muvhu. Kunyangwe nickel iri chinhu chinovaka muviri chezvirimwa, kudya kwayo zvakanyanya muvhu kunogona kukuvadza kupfuura kubatsira. Chepfu yenickel muvhu inoderedza pH yevhu uye inodzivisa kutora iron sechinovaka muviri chakakosha pakukura kwezvirimwa 1. Sekureva kwaLiu3, Ni yakawanikwa iri chinhu chechi17 chakakosha chinodiwa pakukura kwezvirimwa. Pamusoro pebasa renickel mukukura kwezvirimwa, vanhu vanoida kuti ishandiswe zvakasiyana-siyana. Kugadzira electroplating, kugadzirwa kwenickel-based ma alloys, uye kugadzirwa kwemidziyo yekupisa uye ma spark plugs muindasitiri yemotokari zvese zvinoda kushandiswa kwe nickel muzvikamu zvakasiyana zvemaindasitiri. Pamusoro pezvo, ma alloys akavakirwa pa nickel nezvinhu zvakagadzirwa ne electroplated zvakashandiswa zvakanyanya mumidziyo yekicheni, zvishandiso zve ballroom, zvinhu zveindasitiri yechikafu, magetsi, waya netambo, ma jet turbines, ma surgical implants, machira, uye kuvaka ngarava5. Mazinga eNi-rich muvhu (kureva, ivhu repamusoro) anonzi anokonzerwa ne anthropogenic uye zvakasikwa, asi zvikuru, Ni itsime rechisikigo kwete anthropogenic4,6. Manyuko echisikigo e nickel anosanganisira kuputika kwemakomo, zvinomera, moto wemasango, uye maitiro e geological; zvisinei, manyuko e anthropogenic anosanganisira mabhatiri e nickel/cadmium muindasitiri yesimbi, electroplating, arc welding, diesel nemafuta emafuta, uye kuburitswa kwemhepo kubva mukupisa marasha uye tsvina netsvina Nickel accumulation7,8. Sekureva kwaFreedman naHutchinson9 naManiwa et al. 10, nzvimbo huru dzinokonzera kusvibiswa kwevhu repamusoro munzvimbo dziri pedyo nepedyo dzinosanganisira simbi dzesimbi dzinoshandiswa mu nickel-copper nemigodhi. Ivhu repamusoro rakatenderedza nzvimbo yekucheneserwa kwe nickel-copper muSudbury muCanada raiva nehuwandu hwakanyanya hwekusvibiswa kwe nickel pa26,000 mg/kg11. Kusiyana neizvi, kusvibiswa kunokonzerwa nekugadzirwa kwe nickel muRussia kwakakonzera kuwanda kwe nickel muvhu reNorway11. Sekureva kwaAlms et al. 12, huwandu hweHNO3-extractable nickel munzvimbo inorimwa zvakanyanya munharaunda (kuburitswa kwenickel muRussia) hwaive kubva pa6.25 kusvika 136.88 mg/kg, zvichienderana neavhareji ye30.43 mg/kg uye huwandu hwekutanga hwe25 mg/kg. Sekureva kwa kabata 11, kushandiswa kwefosforasi fetereza muvhu rekurima muvhu remaguta kana rekunze kwemaguta munguva dzezvirimwa zvinotevedzana kunogona kupinza kana kusvibisa ivhu. Mhedzisiro inogona kuitika yenickel muvanhu inogona kutungamira kugomarara kuburikidza nekuchinja kwemasero, kukuvara kwechromosome, kugadzirwa kweZ-DNA, kugadziriswa kweDNA yakavharika, kana maitiro e epigenetic13. Mukuyedza kwemhuka, nickel yakawanikwa ine mukana wekukonzera mhando dzakasiyana dzemamota, uye carcinogenic nickel complexes inogona kuwedzera mamota akadaro.
Kuongororwa kwekusvibiswa kwevhu kwakabudirira munguva pfupi yapfuura nekuda kwematambudziko akasiyana-siyana ane chekuita nehutano anobva muhukama hwevhu nemiti, hukama hwevhu nehupenyu hwevhu, kuora kwezvakatipoteredza, uye kuongororwa kwemhedzisiro yezvakatipoteredza. Kusvika pari zvino, kufanotaura kwenzvimbo yezvinhu zvinogona kuva nechepfu (PTEs) zvakaita seNi muvhu kwave kwakaoma uye kuchitora nguva uchishandisa nzira dzechinyakare. Kuuya kwedhijitari yekupenda ivhu (DSM) uye kubudirira kwayo pari zvino15 kwakavandudza zvikuru kufanotaura kwekupenda ivhu (PSM). Sekureva kwaMinasny naMcBratney16, kufanotaura kwekupenda ivhu (DSM) kwaratidza kuva chikamu chikuru chesainzi yevhu.Lagacherie naMcBratney, 2006 vanotsanangura DSM se "kugadzira nekuzadza masisitimu eruzivo rwevhu kuburikidza nekushandisa nzira dzekutarisa dziri munzvimbo nenzvimbo uye murabhoritari uye masisitimu ekufungidzira ivhu munzvimbo nenzvimbo".McBratney et al. 17 inoratidza kuti DSM kana PSM yemazuva ano ndiyo nzira inoshanda zvikuru yekufanotaura kana kugadzira mapa ekugoverwa kwePTE, mhando dzevhu uye hunhu hwevhu. Geostatistics uye Machine Learning Algorithms (MLA) inzira dzeDSM dzekugadzira mamepu dzinogadzira mamepu edhijitari nerubatsiro rwemakombiyuta achishandisa data rakakosha uye diki.
Deutsch18 naOlea19 vanotsanangura geostatistics se "muunganidzwa wematekiniki enhamba anobata nekumiririrwa kwehunhu hwenzvimbo, kunyanya vachishandisa mamodheru estochastic, akadai sekuti kuongorora nguva kunoratidza sei data renguva." Kunyanya, geostatistics inosanganisira kuongororwa kwevariograms, izvo zvinobvumira Quantify uye kutsanangura kutsamira kwenzvimbo kubva kudata rega rega20.Gumiaux et al. 20 vanoratidza zvakare kuti kuongororwa kwevariograms mu geostatistics kwakavakirwa pamisimboti mitatu, inosanganisira (a) kuverenga chiyero chekubatana kwedata, (b) kuziva uye kuverenga anisotropy mukusiyana kwedata uye (c) kuwedzera kune Kuwedzera pakufunga nezvekukanganisa kwedata rekuyera rakaparadzaniswa nemhedzisiro yemuno, mhedzisiro yenzvimbo inoongororwawo. Zvichivakwa pane izvi pfungwa, matekiniki mazhinji einterpolation anoshandiswa mu geostatistics, kusanganisira general kriging, co-kriging, ordinary kriging, empirical Bayesian kriging, simple kriging method nedzimwe nzira dzinozivikanwa dzeinterpolation kuti dzigadzire kana kufanotaura PTE, hunhu hwevhu, uye mhando dzevhu.
Magadzirirwo eMachine Learning (MLA) inzira itsva inoshandisa makirasi makuru edata asiri emutsara, anotsigirwa nemaalgorithms anoshandiswa zvakanyanya mukuchera data, kuona mapatani mudata, uye anoshandiswa kakawanda mukuisa muzvikamu mune zvesainzi senge sainzi yevhu uye mabasa ekudzoka. Mapepa mazhinji ekutsvagisa anovimba nemamodeli eMLA kufanotaura PTE muvhu, senge Tan et al. 22 (masango asina kurongeka ekuongorora simbi inorema muvhu rekurima), Sakizadeh et al. 23 (kutevedzera uchishandisa michina yekutsigira vector uye network dze neural dzekugadzira) kusvibiswa kwevhu ). Pamusoro pezvo, Vega et al. 24 (CART yekutevedzera kuchengetedza simbi inorema uye kunyudzwa muvhu) Sun et al. 25 (kushandiswa kwe cubist ndiko kugoverwa kweCd muvhu) uye mamwe maalgorithms akadai se k-nearest neighbor, generalized boosted regression, uye boosted regression Trees dzakashandisawo MLA kufanotaura PTE muvhu.
Kushandiswa kwema algorithms eDSM mukufanotaura kana kumepu kunosangana nematambudziko akati wandei. Vanyori vazhinji vanotenda kuti MLA iri nani pane geostatistics uye zvinopesana. Kunyangwe imwe iri nani pane imwe, kusanganiswa kwezviviri kunovandudza mwero wekururama kwemepu kana kufanotaura muDSM15. Woodcock naGopal26 Finke27; Pontius naCheuk28 naGrunwald29 vanotaura nezvekusakwana uye zvimwe zvikanganiso mukufanotaura kwemepu yevhu. Masayendisiti evhu akaedza nzira dzakasiyana-siyana dzekuvandudza kushanda, kururama, uye kufanotaura kweDSM mepu nekufanotaura. Kusanganiswa kwekusava nechokwadi nekusimbisa ndeimwe yezvinhu zvakasiyana zvakabatanidzwa muDSM kuti zviwedzere kushanda uye kuderedza zvikanganiso. Zvisinei, Agyeman et al. 15 vanotsanangura kuti maitiro ekusimbisa uye kusava nechokwadi kunounzwa nekugadzirwa kwemepu nekufanotaura kunofanirwa kusimbiswa zvakazvimiririra kuti kunatsiridze mhando yemepu. Miganho yeDSM inokonzerwa nehunhu hwevhu hwakapararira munzvimbo, hunosanganisira chikamu chekusava nechokwadi; zvisinei, kushaikwa kwechokwadi muDSM kunogona kubva pazvikonzero zvakawanda zvekukanganisa, zvinoti covariate error, model error, location error, uye analytical Error 31. Kuenzanisa kusakarurama kunokonzerwa muMLA uye maitiro e geostatistical kunobatanidzwa nekushaikwa kwekunzwisisa, pakupedzisira zvichitungamira mukurerutsa zvakanyanya maitiro chaiwo32. Pasinei nerudzi rwemuenzaniso, kusakarurama kunogona kuverengerwa ku model parameters, mathematical model prediction, kana interpolation33. Munguva pfupi yapfuura, maitiro matsva eDSM akabuda anokurudzira kubatanidzwa kwe geostatistics ne MLA mu mepu nekufanotaura. Nyanzvi dzakawanda dzesainzi yevhu nevanyori, vakaita saSergeev et al. 34; Subbotina et al. 35; Tarasov et al. 36 naTarasov et al. 37 vakashandisa mhando chaiyo ye geostatistics ne machine learning kugadzira ma hybrid models anovandudza kushanda kwekufanotaura nekupa mapu. mhando. Mamwe emamodheru aya ehybrid kana akabatanidzwa ndeaya: Artificial Neural Network Kriging (ANN-RK), Multilayer Perceptron Residual Kriging (MLP-RK), Generalized Regression Neural Network Residual Kriging (GR-NNRK)36, Artificial Neural Network Kriging-Multilayer Perceptron (ANN-K-MLP)37 uye Co-Kriging neGaussian Process Regression38.
Sekureva kwaSergeev nevamwe vake, kusanganisa matekiniki akasiyana-siyana ekutevedzera kune mukana wekubvisa zvikanganiso uye kuwedzera kushanda kwemuenzaniso wehybrid unobva wavapo pane kugadzira muenzaniso wayo mumwe chete. Mumamiriro ezvinhu aya, bepa idzva iri rinoti zvakakosha kushandisa algorithm yakabatana ye geostatistics ne MLA kugadzira ma optimal hybrid models kufanotaura kuwedzera kweNi munzvimbo dzemaguta nedziri pedyo nemaguta. Chidzidzo ichi chichavimba ne Empirical Bayesian Kriging (EBK) se base model uye kuisanganisa ne Support Vector Machine (SVM) uye Multiple Linear Regression (MLR) models. Hybridization yeEBK nechero MLA haizivikanwe. Ma multiple mixed models anoonekwa i combinations ye ordinary, residual, regression kriging, uye MLA.EBK inzira ye geostatistical interpolation inoshandisa spatially stochastic process iyo inowanikwa se non-stationary/stationary random field ine defined localization parameters pamusoro pemunda, zvichibvumira spatial variation39.EBK yakashandiswa muzvidzidzo zvakasiyana-siyana, kusanganisira kuongorora kugoverwa kwe organic carbon muvhu repurazi40, kuongorora kusvibiswa kwevhu41 uye kugadzira mapping ivhu. zvivakwa42.
Kune rumwe rutivi, Self-Organizing Graph (SeOM) inzira yekudzidza yakashandiswa muzvinyorwa zvakasiyana-siyana zvakaita saLi et al. 43, Wang et al. 44, Hossain Bhuiyan et al. 45 naKebonye et al. 46. Sarudza hunhu hwenzvimbo uye kuunganidzwa kwezvinhu. Wang et al. 44 vanotsanangura kuti SeOM inzira yekudzidza ine simba inozivikanwa nekukwanisa kwayo kuronga nekufungidzira matambudziko asiri emutsara. Kusiyana nedzimwe nzira dzekuziva mapatani dzakadai sekuongorora zvikamu zvikuru, kusanganisa zvinhu zvisina kusimba, kusanganisa zvinhu zvine hunhu, uye kugadzira sarudzo dzezviyero zvakawanda, SeOM iri nani pakuronga nekuona mapatani ePTE. Sekureva kwaWang et al. 44, SeOM inogona kuunganidza kugoverwa kwema neuron ane hukama uye kupa ruzivo rwepamusoro-soro. SeOM ichaona data rekufanotaura reNi kuti iwane modhi yakanakisa yekuratidza mhedzisiro yekududzirwa kwakananga.
Chinyorwa ichi chine chinangwa chekugadzira modhi yakasimba yemepu ine kururama kwakanyanya kwekufanotaura huwandu hwenickel muvhu remaguta neremunharaunda. Tinofungidzira kuti kuvimbika kwemodhi yakasanganiswa kunonyanya kutsamira papesvedzero yemamwe mamodhi akabatana nemodhi yekutanga. Tinobvuma matambudziko ari kutarisana neDSM, uye nepo matambudziko aya ari kugadziriswa pamativi akawanda, kusanganiswa kwekufambira mberi mu geostatistics uye mamodhi eMLA kunoita sekunge kuri kuwedzera; saka, tichaedza kupindura mibvunzo yekutsvagisa inogona kuburitsa mamodhi akasanganiswa. Zvisinei, modhi yacho yakarurama sei pakufanotaura chinhu chakanangana nacho? Zvakare, ndeipi nhanho yekuongorora kushanda zvakanaka zvichibva pakusimbiswa uye kuongororwa kwekururama? Saka, zvinangwa chaizvo zvechidzidzo ichi zvaive (a) kugadzira modhi yakasanganiswa yeSVMR kana MLR uchishandisa EBK semodhi yekutanga, (b) kuenzanisa mamodhi akabuda (c) kupa modhi yakanakisa yekufanotaura kuwanda kweNi muvhu remaguta kana renharaunda, uye (d) kushandiswa kweSeOM kugadzira mepu ine resolution yepamusoro yekuchinja kwenzvimbo yenickel.
Chidzidzo ichi chiri kuitwa muCzech Republic, kunyanya mudunhu reFrydek Mistek mudunhu reMoravia-Silesian (ona Mufananidzo 1). Nzvimbo yenzvimbo yekuongorora yakaoma zvikuru uye inonyanya kuve chikamu chedunhu reMoravia-Silesian Beskidy, iro riri chikamu chemucheto wekunze kweMakomo eCarpathian. Nzvimbo yekuongorora iri pakati pe49° 41′ 0′ N uye 18° 20′ 0′ E, uye kukwirira kuri pakati pe225 ne327 m; Zvisinei, Koppen classification system yemamiriro ekunze enzvimbo iyi inonzi Cfb = mamiriro ekunze egungwa ane mwero, Kune mvura yakawanda inonaya kunyangwe mumwedzi yakaoma. Kupisa kunosiyana zvishoma mugore rose pakati pe −5 °C ne 24 °C, kashoma kudonha pasi pe −14 °C kana kupfuura 30 °C, nepo avhareji yemvura inonaya pagore iri pakati pe 685 ne 752 mm47. Nzvimbo inofungidzirwa yekuongorora nzvimbo yese i 1,208 square kilometers, ne 39.38% yenyika yakarimwa uye 49.36% yenzvimbo yakafukidzwa nemasango. Kune rumwe rutivi, nzvimbo yakashandiswa muchidzidzo ichi i 889.8 square kilometers. Mukati nekwakapoteredza Ostrava, indasitiri yesimbi nemabasa esimbi zvinoshanda zvikuru. Zvigayo zvesimbi, indasitiri yesimbi uko nickel inoshandiswa musimbi dzisina ngura (semuenzaniso kudzivirira ngura yemhepo) nesimbi dze alloy (nickel inowedzera simba re alloy uku ichichengetedza ductility yayo yakanaka uye kusimba), uye kurima kwakasimba senge phosphate fertilizer application uye kugadzirwa kwezvipfuyo zvinogona kuwanikwa mukutsvaga nickel mu Dunhu iri (semuenzaniso, kuwedzera nickel kumakwayana kuti kuwedzere kukura kwemakwayana nemombe dzisina kudya zvakakwana). Mamwe mashandisirwo emaindasitiri enickel munzvimbo dzekutsvagisa anosanganisira kushandiswa kwayo mukugadzira electroplating, kusanganisira electroplating nickel uye electroless nickel plating processes. Hunhu hwevhu hunogona kusiyanisa zviri nyore neruvara rwevhu, chimiro, uye carbonate. Maumbirwo evhu ari pakati nepakati kusvika patete, anobva kune zvinhu zvemubereki. Iwo ari macolluvial, alluvial kana aeolian muzvisikwa. Dzimwe nzvimbo dzevhu dzinoita sedzakachekwa pamusoro nepasi pevhu, kazhinji nekongiri uye bleaching. Zvisinei, cambisols ne stagnosols ndiwo marudzi evhu anonyanya kuwanikwa munharaunda iyi48. Nekukwirira kuri pakati pe455.1 kusvika 493.5 m, cambisols inotonga Czech Republic49.
Mepu yenzvimbo yekudzidza [Mepu yenzvimbo yekudzidza yakagadzirwa uchishandisa ArcGIS Desktop (ESRI, Inc, vhezheni 10.7, URL: https://desktop.arcgis.com).]
Sampuli dzevhu repamusoro dzinosvika 115 dzakatorwa kubva muvhu remumaguta neremumaguta mudunhu reFrydek Mistek. Muenzaniso wemuenzaniso wakashandiswa waive wegrid renguva dzose rine sampuro dzevhu dzakaparadzaniswa ne2 × 2 km, uye ivhu repamusoro rakayerwa pakadzika kwe0 kusvika 20 cm uchishandisa mudziyo weGPS unobatwa nemaoko (Leica Zeno 5 GPS). Sampuli dzinoiswa mumabhegi eZiploc, dzakanyorwa zvakanaka, uye dzinotumirwa kurabhoritari. Sampuli dzakaomeswa nemhepo kuti dzigadzire sampuro dzakapwanywa, dzakapwanywa nesystem yemakanika (Fritsch disc mill), uye dzakaseferwa (saizi yesipo 2 mm). Isa 1 gram yesampuru dzevhu dzakaomeswa, dzakasanganiswa uye dzakaseferwa mumabhodhoro eteflon ane mavara akajeka. Mumudziyo wega wega weTeflon, buritsa 7 ml ye35% HCl uye 3 ml ye65% HNO3 (uchishandisa automatic dispenser - imwe yeasidhi yega yega), vhara zvishoma uye rega sampuro dzimire husiku hwese kuti dziite reaction (aqua regia program). Isa supernatant paplate yesimbi inopisa (tembiricha: 100 W uye 160 °C) kwemaawa maviri kuti zviite kuti sampuli dzigayiwe, wozodzitonhodza. Isa supernatant muflask ye50 ml woisanganisa kusvika 50 ml nemvura yakabviswa ion. Mushure meizvozvo, sefa supernatant yakaderedzwa muchubhu yePVC ye50 ml nemvura yakabviswa ion. Pamusoro pezvo, 1 ml yemushonga wekusanganiswa yakasanganiswa ne9 ml yemvura yakabviswa ion ndokuseferwa muchubhu ye12 ml yakagadzirirwa PTE pseudo-concentration. Kuwanda kwePTEs (As, Cd, Cr, Cu, Mn, Ni, Pb, Zn, Ca, Mg, K) kwakatsanangurwa neICP-OES (Inductively Coupled Plasma Optical Emission Spectroscopy) (Thermo Fisher Scientific, USA) zvichienderana nenzira dzakajairwa uye chibvumirano. Ita shuwa kuti Quality Assurance and Control (QA/QC) maitiro (SRM NIST 2711a Montana II Soil). PTEs dzine miganhu yekuona iri pasi pehafu hadzina kubviswa muchidzidzo ichi. Muganho wekuona wePTE Zvakashandiswa muchidzidzo ichi zvaive 0.0004.(iwe). Pamusoro pezvo, maitiro ekudzora hunhu uye ekusimbisa hunhu hweongororo yega yega anosimbiswa nekuongorora zviyero zvekutarisa. Kuti zvive nechokwadi chekuti zvikanganiso zvaderedzwa, ongororo mbiri dzakaitwa.
Empirical Bayesian Kriging (EBK) ndeimwe yenzira dzakawanda dzekubatanidza manhamba dzinoshandiswa mukuita modhi munzvimbo dzakasiyana siyana dzakadai sesainzi yevhu. Kusiyana nedzimwe nzira dzekubatanidza manhamba, EBK yakasiyana nenzira dzechinyakare dzekuita kriging nekufunga nezvechikanganiso chakafungidzirwa ne semivariogram model. Mu EBK interpolation, mamodheru akati wandei e semivariogram anoverengerwa panguva yekubatanidza, pane semivariogram imwe chete. Nzira dzekubatanidza dzinoita kuti pave nekusava nechokwadi uye mapurogiramu ane chekuita nechirongwa ichi che semivariogram chinoumba chikamu chakaoma kwazvo chenzira yakakwana yekufungidzira. Maitiro ekubatanidza manhamba eEBK anotevera zviyero zvitatu zvakakurudzirwa naKrivoruchko50, (a) modhi inofungidzira semivariogram kubva mudhata rekupinda (b) kukosha kutsva kwakafanotaurwa kwenzvimbo yega yega yedhata rekupinda zvichibva pane semivariogram yakagadzirwa uye (c) modhi yekupedzisira yeA inoverengerwa kubva mudhata rekufananidzira. Mutemo weBayesian equation unopiwa seposterior
Apo \(Prob\left(A\right)\) inomiririra mukana wepamberi, \(Prob\left(B\right)\) wepakati unoregeredzwa muzviitiko zvakawanda, \(Prob (B,A)\ ) .Kuverengera kwesemivariogram kwakavakirwa pamutemo waBayes, unoratidza mukana wedata rekucherechedza rinogona kugadzirwa kubva kusemivariograms. Kukosha kwesemivariogram kunozoonekwa uchishandisa mutemo waBayes, unotaura kuti zvingangoita sei kugadzira data rekucherechedza kubva kusemivariogram.
Muchina wekutsigira vector ialgorithm yekudzidza muchina inogadzira hyperplane inoparadzanisa zvakanaka kuti isiyanise makirasi akafanana asi asiri akazvimiririra. Vapnik51 yakagadzira algorithm yekusarudza chinangwa, asi ichangobva kushandiswa kugadzirisa matambudziko akatarisana nekudzoka. Sekureva kwaLi et al.52, SVM ndeimwe yenzira dzakanakisisa dzekusarudza uye yakashandiswa muminda yakasiyana-siyana. Chikamu chekudzoka kweSVM (Support Vector Machine Regression - SVMR) chakashandiswa mukuongorora uku. Cherkassky naMulier53 vakatanga SVMR senzira yekudzoka yakavakirwa pakernel, iyo yakaverengerwa uchishandisa nzira yekudzoka kwemutsara ine mabasa enzvimbo akawanda. John et al54 vanotaura kuti SVMR modeling inoshandisa hyperplane linear regression, iyo inogadzira hukama husina kurongeka uye inobvumira mabasa enzvimbo. Sekureva kwaVohland et al. 55, epsilon (ε)-SVMR inoshandisa data rakadzidziswa kuti iwane modhi yekumiririra sebasa risinganzwisisike repsilon rinoshandiswa kuronga data rakazvimiririra ne epsilon bias yakanakisa kubva pakudzidziswa pane data rakabatana. Kukanganisa kwedaro rakatarwa kunoregeredzwa kubva pamutengo chaiwo, uye kana kukanganisa kuri kukuru kupfuura ε(ε), hunhu hwevhu hunoritsiva. Modhi iyi inoderedzawo kuoma kwedata rekudzidzisa kuita subset yakakura yekutsigira vectors. Equation yakakurudzirwa naVapnik51 inoratidzwa pazasi.
apo b inomiririra scalar threshold, \(K\left({x}_{,}{ x}_{k}\right)\) inomiririra kernel function, \(\alpha\) inomiririra Lagrange multiplier, N inomiririra nhamba yedata, \({x}_{k}\) inomiririra data input, uye \(y\) idata output. Imwe yema key kernels anoshandiswa iSVMR operation, inova Gaussian radial basis function (RBF). RBF kernel inoshandiswa kuona optimal SVMR model, iyo yakakosha kuti uwane penalty set factor C uye kernel parameter gamma (γ) yePTE training data. Kutanga, takaongorora training set uye tobva taedza model performance pa validation set. Direction parameter inoshandiswa i sigma uye method value i svmRadial.
Modhi yeregression ine mutsara wakawanda (MLR) imodhi yeregression inomiririra hukama huripo pakati pe response variable nenhamba ye predictor variables kuburikidza nekushandisa linear pooled parameters dzakaverengerwa uchishandisa nzira ye least squares. MuMLR, modhi ye least squares ibasa rekufanotaura kwehunhu hwevhu mushure mekusarudza ma explanatory variables. Zvakakosha kushandisa response kuti ugadzire linear relationship uchishandisa explanatory variables.PTE yakashandiswa se response variable kuti ugadzire linear relationship ne explanatory variables. Iyo MLR equation ndiyo
apo y iri mhinduro inochinja, \(a\) iri intercept, n inhamba yezvinofanotaura, \({b}_{1}\) iri kudzoreredzwa kwechikamu chema coefficients, \({x}_{ i}\) inomiririra chiratidzo chinofanotaura kana chinotsanangudza, uye \({\varepsilon }_{i}\) inomiririra kukanganisa mumuenzaniso, inozivikanwawo se residual.
Mamodheru akasanganiswa akawanikwa nekubatanidza EBK neSVMR neMLR. Izvi zvinoitwa nekubvisa mavalue akafanotaurwa kubva mukubatanidzwa kweEBK. Mavalue akafanotaurwa akawanikwa kubva kuCa, K, neMg akabatanidzwa anowanikwa kuburikidza nemaitiro ekubatanidza kuti pawane mavalue matsva, akadai seCaK, CaMg, neKMg. Zvinhu Ca, K neMg zvinozobatanidzwa kuti pave nechinhu chechina, CaKMg. Pakazara, mavalue akawanikwa ndiCa, K, Mg, CaK, CaMg, KMg neCaKMg. Aya mavalue akava mavalue edu, zvichibatsira kufanotaura kuwanda kwe nickel muvhu remaguta nerekunze kwemaguta. Iyo SVMR algorithm yakaitwa pazviratidzi kuti pave nemuenzaniso wakasanganiswa weEmpirical Bayesian Kriging-Support Vector Machine (EBK_SVM). Saizvozvo, mavalue anofambiswawo kuburikidza neMLR algorithm kuti pave nemuenzaniso wakasanganiswa weEmpirical Bayesian Kriging-Multiple Linear Regression (EBK_MLR). Kazhinji, mavalue Ca, K, Mg, CaK, CaMg, KMg, uye CaKMg zvinoshandiswa sezviyero zvekufanotaura huwandu hweNi muvhu remaguta neremumaguta. Muenzaniso unogamuchirwa zvikuru wakawanikwa (EBK_SVM kana EBK_MLR) ucharatidzwa uchishandisa girafu rinozvironga. Mafambiro ebasa rechidzidzo ichi anoratidzwa muMufananidzo 2.
Kushandisa SeOM kwave chishandiso chinozivikanwa chekuronga, kuongorora, uye kufanotaura data muchikamu chezvemari, hutano, indasitiri, nhamba, sainzi yevhu, nezvimwewo. SeOM yakagadzirwa uchishandisa network dze neural dzekugadzira uye nzira dzekudzidza dzisina kutarisirwa dzekuronga, kuongorora, uye kufanotaura. Muchidzidzo ichi, SeOM yakashandiswa kufungidzira kuwanda kweNi zvichibva pane muenzaniso wakanakisa wekufanotaura Ni muvhu remaguta nerekunze kwemaguta. Data rakagadziriswa muongororo yeSeOM rinoshandiswa se n input-dimensional vector variables43,56.Melssen et al. 57 inotsanangura kubatana kwevector yekupinda mu neural network kuburikidza ne single input layer kune output vector ine single weight vector. Output inogadzirwa neSeOM i mepu ine mativi maviri ine ma neuron akasiyana kana ma nodes akarukwa kuita hexagonal, denderedzwa, kana square topological maps zvichienderana nekuswedera kwavo. Kuenzanisa saizi dzemepu zvichibva pametric, quantization error (QE) uye topographic error (TE), modhi yeSeOM ine 0.086 na0.904, zvichiteerana, inova unit yemepu 55 (5 × 11). Maumbirwo e neuron anotsanangurwa zvichienderana nehuwandu hwema nodes mu empirical equation.
Huwandu hwedata rakashandiswa muchidzidzo ichi i115 samples. Nzira isina kurongeka yakashandiswa kupatsanura data kuita data rekuyedza (25% yekusimbisa) uye seti dzedata rekudzidzisa (75% yekuenzanisa). Data rekudzidzisa rinoshandiswa kugadzira regression model (calibration), uye data rekuyedza rinoshandiswa kusimbisa kugona kwe generalization58. Izvi zvakaitwa kuti vaongorore kukodzera kwemamodheru akasiyana-siyana ekufanotaura huwandu hwe nickel muvhu. Mamodheru ese akashandiswa akapfuura ne ten-fold cross-validation process, yakadzokororwa kashanu. Mavariables anogadzirwa neEBK interpolation anoshandiswa se predictors kana explanatory variables kufanotaura target variable (PTE). Modelling inobatwa muRStudio uchishandisa packages library(Kohonen), library(caret), library(modelr), library(“e1071″), library(“plyr”), library(“caTools”), library(“prospectr”) uye library(“Metrics”).
Maparamita akasiyana-siyana ekusimbisa akashandiswa kuona modhi yakanakisa yakakodzera kufanotaura kuwanda kwe nickel muvhu uye kuongorora kururama kwemodhi uye kusimbiswa kwayo. Mamodhi eHybridization akaongororwa achishandisa mean absolute error (MAE), root mean square error (RMSE), uye R-squared kana coefficient determination (R2). R2 inotsanangura kusiyana kwehuwandu mumhinduro, inomiririrwa ne regression model.RMSE uye variance magnitude mu independent measures zvinotsanangura simba rekufanotaura remodhi, nepo MAE ichisarudza chaiyo quantitative value. R2 value inofanira kunge yakakwira kuti iongorore modhi yakanakisa yemusanganiswa uchishandisa ma paramita ekusimbisa, kana kukosha kuri pedyo ne1, kururama kwacho kwakakwira. Sekureva kwaLi et al. 59, R2 criterion value ye0.75 kana kupfuura inoonekwa se predictor yakanaka; kubva pa0.5 kusvika pa0.75 inogamuchirwa modhi performance, uye pasi pe0.5 inogamuchirwa modhi performance. Pakusarudza modhi uchishandisa RMSE neMAE validation criteria evaluation methods, ma values ​​akaderera akawanikwa akakwana uye akaonekwa sesarudzo yakanakisa. Equation inotevera inotsanangura nzira yekusimbisa.
apo n inomiririra saizi yehuwandu hwakaonekwa\({Y}_{i}\) inomiririra mhinduro yakayerwa, uye \({\widehat{Y}}_{i}\) inomiririrawo kukosha kwemhinduro kwakafanotaurwa, saka, kwekutarisa kwekutanga kwe i.
Tsananguro dzezviverengero zvezviratidziro uye mhinduro dzinoratidzwa muTafura 1, zvichiratidza avhareji, standard deviation (SD), coefficient of variation (CV), minimum, maximum, kurtosis, uye skewness. Zviyero zvepasi uye maximum zvezvinhu zviri muhurongwa hwekuderera kweMg < Ca < K < Ni uye Ca < Mg < K < Ni, zvichiteerana. Kuwanda kwemhinduro (Ni) kwakatorwa kubva munzvimbo yekudzidza kwakabva pa4.86 kusvika 42.39 mg/kg. Kuenzanisa kweNi neavhareji yepasi rose (29 mg/kg) uye avhareji yeEurope (37 mg/kg) kwakaratidza kuti avhareji yejiometri yakaverengerwa yenzvimbo yekudzidza yaive mukati mehuwandu hunoshivirira. Zvisinei, sezvakaratidzwa neKabata-Pendias11, kuenzanisa avhareji yehuwandu hwenickel (Ni) muchidzidzo chazvino nevhu rekurima muSweden kunoratidza kuti avhareji yehuwandu hwenickel yazvino yakakwira. Saizvozvowo, avhareji yehuwandu hweFrydek Mistek muvhu remaguta nerekunze kwemaguta muchidzidzo chazvino (Ni 16.15 mg/kg) yaive yakakwira kupfuura inobvumidzwa. muganho we60 (10.2 mg/kg) weNi muvhu remaguta rePoland wakataurwa naRóżański et al. Uyezve, Bretzel naCalderisi61 vakanyora huwandu hwakaderera hweNi (1.78 mg/kg) muvhu remaguta muTuscany zvichienzaniswa neongororo yazvino. Jim62 akawanawo huwandu hwakaderera hweNickel (12.34 mg/kg) muvhu remaguta reHong Kong, hwakaderera pane huwandu hweNickel hwazvino muchidzidzo ichi. Birke et al63 vakashuma huwandu hweNi ye17.6 mg/kg munzvimbo yekare yemigodhi neindasitiri yemaguta muSaxony-Anhalt, Germany, hwaive hwakakwira ne1.45 mg/kg kupfuura huwandu hweNi munzvimbo iyi (16.15 mg/kg). Tsvagiridzo yazvino. Huwandu hweNickel hwakawanda muvhu mune mamwe maguta nenzvimbo dzemumaguta munzvimbo iyi inogona kunge yakakonzerwa neindasitiri yesimbi nesimbi neindasitiri yesimbi. Izvi zvinoenderana neongororo yaKhodadoust et al. 64 kuti indasitiri yesimbi nekugadzira simbi ndizvo zvinonyanya kukonzera kusvibiswa kwe nickel muvhu. Zvisinei, zvaifungidzirwawo zvakabva pa538.70 mg/kg kusvika 69,161.80 mg/kg yeCa, 497.51 mg/kg kusvika 3535.68 mg/kg yeK, uye 685.68 mg/kg kusvika 5970.05 mg/kg yeMg.Jakovljevic et al. 65 vakaongorora huwandu hwese hweMg neK muvhu riri pakati peSerbia. Vakaona kuti huwandu hwese (410 mg/kg uye 400 mg/kg, zvichiteerana) hwakanga hwakaderera pane huwandu hweMg neK hwechidzidzo chazvino. Zvisingazivikanwe, kumabvazuva kwePoland, Orzechowski naSmolczynski66 vakaongorora huwandu hwese hweCa, Mg neK uye vakaratidza huwandu hweCa (1100 mg/kg), Mg (590 mg/kg) uye K (810 mg/kg). Hunhu hwepamusoro hwevhu hwakaderera pane chinhu chimwe chete muchidzidzo ichi. Chidzidzo chakaitwa naPongrac et al. 67 chakaratidza kuti huwandu hwese hweCa hwakaongororwa muvhu nhatu dzakasiyana muScotland, UK (ivhu reMylnefield, ivhu reBalruddery nevhu reHartwood) hwakaratidza huwandu hweCa hwakakwira muchidzidzo ichi.
Nekuda kwekusiyana kwehuwandu hwezvinhu zvakayerwa, kugoverwa kwedata kwezvinhu kunoratidza kusarongeka kwakasiyana. Kusarongeka uye kurtosis yezvinhu zvakabva pa1.53 kusvika 7.24 uye 2.49 kusvika 54.16, zvichiteerana. Zvinhu zvese zvakaverengerwa zvine kusarongeka uye kurtosis mazinga ari pamusoro pe +1, zvichiratidza kuti kugoverwa kwedata hakuna kurongeka, kwakakombama munzira chaiyo uye kwakasvika pakakwirira. MaCV anofungidzirwa ezvinhu anoratidzawo kuti K, Mg, naNi zvinoratidza kusarongeka kuri pakati nepakati, nepo Ca ine kusarongeka kwakanyanya. MaCV eK, Ni naMg anotsanangura kugoverwa kwawo kwakafanana. Uyezve, kugoverwa kweCa hakuna kufanana uye masosi ekunze anogona kukanganisa huwandu hwayo hwekuwedzerwa.
Kubatana kwezvinofanotaura zvinhu nezvinopindura kwakaratidza hukama hunogutsa pakati pezvinhu (ona Mufananidzo 3). Kubatana kwakaratidza kuti CaK yakaratidza hukama huri pakati nepakati ne r value = 0.53, sezvakaita CaNi. Kunyangwe Ca na K zvichiratidza hukama huri pakati nepakati, vaongorori vakaita saKingston et al. 68 naSanto69 zvinoratidza kuti huwandu hwazvo muvhu hwakaenzana. Zvisinei, Ca neMg zvinopesana neK, asi CaK inoenderana zvakanaka. Izvi zvinogona kunge zvichikonzerwa nekushandiswa kwemafetiraiza akadai sepotassium carbonate, iyo iri pamusoro pe56% mu potassium. Potassium yaive yakabatana zvine mwero nemagnesium (KM r = 0.63). Muindasitiri yemafetiraiza, zvinhu zviviri izvi zvine hukama hwepedyo nekuti potassium magnesium sulfate, potassium magnesium nitrate, uye potash zvinoshandiswa muvhu kuti zviwedzere kushomeka kwazvo. Nickel ine hukama hwepakati neCa, K neMg ine r values ​​​​= 0.52, 0.63 na 0.55, zvichiteerana. Hukama hunosanganisira calcium, magnesium, uye PTE dzakadai senickel hwakaoma, asi zvakadaro, magnesium inodzivirira kunyudzwa kwecalcium, calcium inoderedza mhedzisiro yemagnesium yakawandisa, uye magnesium necalcium zvese zvinoderedza mhedzisiro yechepfu yenickel muvhu.
Matrix yekubatana kwezvinhu zvinoratidza hukama huripo pakati pezvinofanotaura nemhinduro (Cherechedza: mufananidzo uyu unosanganisira scatterplot pakati pezvinhu, mazinga ekukosha akavakirwa pa p < 0,001).
Mufananidzo 4 unoratidza kugoverwa kwezvinhu munzvimbo. Sekureva kwaBurgos et al70, kushandiswa kwekugoverwa kwenzvimbo inzira inoshandiswa kuyera nekuratidza nzvimbo dzinopisa munzvimbo dzakasvibiswa. Kuwedzerwa kweCa muMufananidzo 4 kunogona kuonekwa kuchamhembe kwakadziva kumadokero kwemepu yekugoverwa kwenzvimbo. Mufananidzo unoratidza nzvimbo dzinopisa dzeCa dziri pakati nepakati kusvika pakakwirira. Kuwedzerwa kwecalcium kuchamhembe kwakadziva kumadokero kwemepu kungangodaro kuri nekuda kwekushandiswa kwequicklime (calcium oxide) kuderedza acidity yevhu uye kushandiswa kwayo mumagayo esimbi sealkaline oxygen mukuita simbi. Kune rumwe rutivi, vamwe varimi vanosarudza kushandisa calcium hydroxide muvhu rine acidic kuti vabvise pH, iyo inowedzerawo calcium muvhu71. Potassium inoratidzawo nzvimbo dzinopisa kuchamhembe kwakadziva kumadokero nekumabvazuva kwemepu. Northwest inzvimbo huru yekurima, uye pateni yepotassium iri pakati nepakati kusvika yakakwira inogona kunge iri nekuda kwekushandiswa kweNPK nepotash. Izvi zvinoenderana nezvimwe zvidzidzo, zvakaita seMadaras naLipavský72, Madaras et al.73, Pulkrabová et al.74, Asare et al. al.75, avo vakaona kuti kugadzikana kwevhu uye kurapwa neKCl neNPK zvakakonzera huwandu hwakawanda hweK muvhu. Kuwedzerwa kweSpatial Potassium kuchamhembe kwakadziva kumadokero kwemepu yekugovera kunogona kunge kuri nekuda kwekushandiswa kwefetereza ine potassium yakadai sepotassium chloride, potassium sulfate, potassium nitrate, potash, uye potash kuwedzera huwandu hwepotassium muvhu risina kunaka.Zádorová et al. 76 naTlustoš et al. 77 yakatsanangura kuti kushandiswa kwefetereza ine K kwakawedzera huwandu hweK muvhu uye kwaizowedzera zvakanyanya huwandu hwezvinovaka muviri muvhu kwenguva refu, kunyanya K neMg zvichiratidza nzvimbo inopisa muvhu. Nzvimbo dzine mwero dziri kuchamhembe kwakadziva kumadokero kwemepu uye kumaodzanyemba kwakadziva kumabvazuva kwemepu. Kugadzika kweColloidal muvhu kunoderedza huwandu hwemagnesium muvhu. Kushaikwa kwayo muvhu kunoita kuti zvirimwa zviratidze yellow intervein chlorosis. Fetereza ine Magnesium, yakadai sepotassium magnesium sulfate, magnesium sulfate, neKieserite, inorapa kushomeka (zvirimwa zvinoita sepepuru, tsvuku, kana brown, zvichiratidza kushomeka kwemagnesium) muvhu rine pH yakajairika6. Kuunganidzwa kwenickel pamusoro pevhu remaguta nepamativi eguta kunogona kunge kuri nekuda kwemabasa evanhu akadai sekurima uye kukosha kwenickel mukugadzirwa kwesimbi isina ngura78.
Kugoverwa kwezvinhu munzvimbo [mepu yekugoverwa kwenzvimbo yakagadzirwa uchishandisa ArcGIS Desktop (ESRI, Inc, Version 10.7, URL: https://desktop.arcgis.com).]
Mhedzisiro yemuenzaniso wekushanda kwezvinhu zvakashandiswa muchidzidzo ichi inoratidzwa muTafura 2. Kune rumwe rutivi, RMSE neMAE zveNi zvese zviri pedyo ne zero (0.86 RMSE, -0.08 MAE). Kune rumwe rutivi, ese ari maviri RMSE neMAE values ​​​​dzeK zvinogamuchirwa. Mhedzisiro yeRMSE neMAE yaive yakakura kune calcium ne magnesium. Mhedzisiro yeCa neK MAE neRMSE yakakura nekuda kwedatasets dzakasiyana. RMSE neMAE zvechidzidzo ichi zvichishandisa EBK kufanotaura Ni zvakawanikwa zviri nani pane mhedzisiro yaJohn et al. 54 vachishandisa synergistic kriging kufanotaura kuwanda kweS muvhu vachishandisa data rakaunganidzwa rakafanana. Zvakabuda zveEBK zvatakadzidza zvinoenderana nezvaFabijaczyk et al. 41, Yan et al. 79, Beguin et al. 80, Adhikary et al. 81 naJohn et al. 82, kunyanya K naNi.
Kushanda kwenzira dzakasiyana-siyana dzekufanotaura huwandu hwenickel muvhu remaguta nerekunze kwemaguta kwakaongororwa uchishandisa mashandiro emamodheru (Tafura 3). Kusimbiswa kwemuenzaniso uye kuongororwa kwekururama kwakasimbisa kuti Ca_Mg_K predictor pamwe chete neEBK SVMR model zvakaburitsa kushanda kwakanakisa. Calibration model Ca_Mg_K-EBK_SVMR model R2, root mean square error (RMSE) uye mean absolute error (MAE) zvaive 0.637 (R2), 95.479 mg/kg (RMSE) uye 77.368 mg/kg (MAE) Ca_Mg_K-SVMR yaive 0.663 (R2), 235.974 mg/kg (RMSE) uye 166.946 mg/kg (MAE). Zvisinei, R2 values ​​​​dzakanaka dzakawanikwa dzeCa_Mg_K-SVMR (0.663 mg/kg R2) uye Ca_Mg-EBK_SVMR (0.643 = R2); mhedzisiro yavo yeRMSE neMAE yaive yakakwira kupfuura yeCa_Mg_K-EBK_SVMR (R2 0.637) (ona Tafura 3). Pamusoro pezvo, RMSE neMAE zveCa_Mg-EBK_SVMR (RMSE = 1664.64 uye MAE = 1031.49) modhi ndeye 17.5 ne13.4, zvichiteerana, izvo zvakakura kupfuura zveCa_Mg_K-EBK_SVMR. Saizvozvowo, RMSE neMAE zveCa_Mg-K SVMR (RMSE = 235.974 uye MAE = 166.946) modhi yakakura ne2.5 ne2.2 kupfuura zveCa_Mg_K-EBK_SVMR RMSE neMAE, zvichiteerana. Mhedzisiro yeRMSE yakaverengerwa inoratidza kuti data rakanyatsobatanidzwa sei nemutsara wezvakanakisa. RSME neMAE zvepamusoro zvakaonekwa. Sekureva kwa Kebonye et al. 46 naJohn et al. 54, kana RMSE neMAE zviri pedyo kusvika pazero, mhedzisiro yacho inowedzera kunaka. SVMR neEBK_SVMR dzine RSME neMAE values ​​dzakakwira. Zvakaonekwa kuti fungidziro dzeRSME dzaive dzakakwira nguva dzose kupfuura MAE values, zvichiratidza kuvapo kwe outliers. Sekureva kwaLegates naMcCabe83, kuti RMSE inodarika avhareji absolute error (MAE) inokurudzirwa sechiratidzo chekuvapo kwe outliers. Izvi zvinoreva kuti dataset yakawanda, MAE neRMSE values ​​dzinokwira. Kururama kwekuongorora kwe cross-validation yeCa_Mg_K-EBK_SVMR mixed model yekufanotaura huwandu hweNi muvhu remaguta nemaguta kwaive 63.70%. Sekureva kwaLi et al. 59, iyi level yekururama imodhi inogamuchirwa performance rate. Mhedzisiro iripo inoenzaniswa neongororo yapfuura yakaitwa naTarasov et al. 36 vane modhi yakasanganiswa yakagadzira MLPRK (Multilayer Perceptron Residual Kriging), ine chekuita neEBK_SVMR accuracy evaluation index yakataurwa muchidzidzo chazvino, RMSE (210) uye The MAE (167.5) yaive yakakwira kupfuura zvatakawana muchidzidzo chazvino (RMSE 95.479, MAE 77.368). Zvisinei, pakuenzanisa R2 yechidzidzo chazvino (0.637) neyaTarasov et al. 36 (0.544), zviri pachena kuti coefficient of determination (R2) yakakwira mumuenzaniso uyu wakasanganiswa. Muganho wekukanganisa (RMSE neMAE) (EBK SVMR) wemuenzaniso wakasanganiswa wakaderera kaviri. Saizvozvowo, Sergeev et al.34 vakanyora 0.28 (R2) yemuenzaniso wakasanganiswa wakagadzirwa (Multilayer Perceptron Residual Kriging), nepo Ni muchidzidzo chazvino vakanyora 0.637 (R2). Mwero wekufanotaura kwemuenzaniso uyu (EBK SVMR) i63.7%, nepo kufanotaura kwechokwadi kwakawanikwa naSergeev et al. 34 kuri 28%. Mepu yekupedzisira (Mufananidzo 5) yakagadzirwa uchishandisa muenzaniso weEBK_SVMR uye Ca_Mg_K senzira yekufanotaura inoratidza kufanotaura kwenzvimbo dzinopisa uye pakati nepakati kusvika nickel munzvimbo yese yekudzidza. Izvi zvinoreva kuti huwandu hwenickel munzvimbo yekudzidza hunonyanya kuva pakati nepakati, nehuwandu hwakawanda munzvimbo dzakatarwa.
Mepu yekupedzisira yekufanotaura inomiririrwa uchishandisa modhi yehybrid EBK_SVMR uye uchishandisa Ca_Mg_K senzira yekufanotaura. [Mepu yekugovera nzvimbo yakagadzirwa uchishandisa RStudio (vhezheni 1.4.1717: https://www.rstudio.com/).]
Zvinoratidzwa muMufananidzo 6 zvinoratidza huwandu hwePTE sechikamu chemuviri chine ma neuron ega ega. Hapana chimwe chezvikamu zvemuviri chakaratidza patani yemavara akafanana sezvakaratidzwa. Zvisinei, nhamba yakakodzera yema neuron pamepu yakadhirowewa i55. SeOM inogadzirwa uchishandisa mavara akasiyana-siyana, uye kana mapatani emavara akafanana, hunhu hwemasamples hwacho hunoenzana zvakanyanya. Zvichienderana nechiyero chavo chemavara chaiwo, zvinhu zvega zvega (Ca, K, uye Mg) zvakaratidza mapatani emavara akafanana nema neuron ega ega epamusoro uye akawanda ma neuron akaderera. Saka, CaK neCaMg zvakafanana nema neuron epamusoro-soro uye mapatani emavara akaderera kusvika pakati nepakati. Mamodheru ese ari maviri anofanotaura huwandu hweNi muvhu nekuratidza mavara epakati kusvika akakwirira akadai setsvuku, orenji neyero. Modheru yeKMg inoratidza mapatani akawanda emavara akakwirira zvichibva pahuwandu hwakarurama uye mavara emavara akaderera kusvika pakati nepakati. Pachiyero chakarurama chemavara kubva pakaderera kusvika kumusoro, patani yekugoverwa kwezvikamu zvemodheru yakaratidza patani yemavara akakwirira inoratidza huwandu hunogona kuwanikwa hwe nickel muvhu (ona Mufananidzo 4). Chikamu chechikamu chemodheru yeCakMg chinoratidza patani yemavara akasiyana-siyana kubva pakaderera kusvika pakakwirira zvichienderana neruvara rwakarurama. chiyero.Uyezve, kufanotaura kwemuenzaniso wehuwandu hwenickel (CakMg) kwakafanana nekugoverwa kwenzvimbo kwenickel kunoratidzwa muMufananidzo 5. Magirafu ese ari maviri anoratidza huwandu hwakakwira, hwepakati nepashoma hwehuwandu hwenickel muvhu remaguta nerekunze kwemaguta.Mufananidzo 7 unoratidza nzira yecontour muboka re k-means pamepu, rakakamurwa kuita mapoka matatu zvichibva pakukosha kwakafanotaurwa mumuenzaniso wega wega. Nzira yecontour inomiririra huwandu hwakakodzera hwemapoka.Pamasampuli evhu 115 akaunganidzwa, chikamu 1 chakawana akawanda masampuli evhu, 74. Chikamu 2 chakagamuchira 33 samples, nepo chikwata 3 chakagamuchira 8 samples. Musanganiswa we planar predictor wezvikamu zvinomwe wakarerutswa kuti ubvumire kududzirwa kwakarurama kwemapoka.Nekuda kwemaitiro akawanda evanhu uye echisikigo anokanganisa kuumbwa kwevhu, zvakaoma kuva nemapatani emapoka akasiyana-siyana mumepu yeSeOM yakagoverwa78.
Kuburitswa kwechikamu chimwe nechimwe cheEmpirical Bayesian Kriging Support Vector Machine (EBK_SVM_SeOM) variable.[Mamepu eSeOM akagadzirwa achishandisa RStudio (vhezheni 1.4.1717: https://www.rstudio.com/).]
Zvikamu zvakasiyana zvekupatsanura ma cluster [Mapu eSeOM akagadzirwa achishandisa RStudio (vhezheni 1.4.1717: https://www.rstudio.com/).]
Chidzidzo chiripo chinoratidza zvakajeka matekiniki ekugadzira nickel muvhu remaguta neremumaguta. Chidzidzo ichi chakaedza matekiniki akasiyana ekugadzira, kusanganisa zvinhu netekiniki dzekugadzira, kuti vawane nzira yakanakisisa yekufanotaura huwandu hwenickel muvhu. Maitiro eSeOM compositional planar spatial etekiniki yemuenzaniso akaratidza patani yemavara akawanda kubva pasi kusvika pakakwirira pachikero chemavara chakarurama, zvichiratidza huwandu hweNi muvhu. Zvisinei, mepu yekugovera nzvimbo inosimbisa kugoverwa kwenzvimbo kwezvikamu zvinoratidzwa neEBK_SVMR (ona Mufananidzo 5). Mhedzisiro yacho inoratidza kuti support vector machine regression model (Ca Mg K-SVMR) inofanotaura huwandu hweNi muvhu semodhi imwe chete, asi ma parameter ekuongorora ekusimbisa uye accuracy anoratidza zvikanganiso zvakanyanya maererano neRMSE neMAE. Kune rumwe rutivi, tekiniki yekugadzira yakashandiswa neEBK_MLR model hainawo mhedzisiro nekuda kwekukosha kwakaderera kwe coefficient of determination (R2). Mhedzisiro yakanaka yakawanikwa uchishandisa EBK SVMR nezvinhu zvakasanganiswa (CaKMg) neRMSE neMAE errors yakaderera ne accuracy ye63.7%. Zvinobuda kuti kusanganisa EBK Algorithm ine algorithm yekudzidza kwemuchina inogona kugadzira algorithm yakasanganiswa inogona kufanotaura huwandu hwePTE muvhu. Zvakabuda zvinoratidza kuti kushandisa Ca Mg K senzira dzekufanotaura huwandu hweNi munzvimbo yekudzidza kunogona kuvandudza kufanotaura kweNi muvhu. Izvi zvinoreva kuti kushandiswa nguva dzose kwefetereza inobva ku nickel uye kusvibiswa kwevhu mumaindasitiri neindasitiri yesimbi kune katsika kekuwedzera huwandu hwenickel muvhu. Chidzidzo ichi chakaratidza kuti EBK model inogona kuderedza huwandu hwekukanganisa uye kuvandudza kururama kwemuenzaniso wekupararira kwevhu muvhu remaguta kana rekunze kwemaguta. Kazhinji, tinokurudzira kushandisa EBK-SVMR model kuongorora nekufanotaura PTE muvhu; pamusoro pezvo, tinokurudzira kushandisa EBK kusanganisa nemaalgorithms akasiyana-siyana ekudzidza kwemuchina. Hunhu hweNi hwakafanotaurwa uchishandisa zvinhu se covariates; zvisinei, kushandisa covariates dzakawanda kwaizovandudza zvikuru mashandiro emuenzaniso, izvo zvinogona kutorwa semuganho webasa riripo. Chimwe chiganho chechidzidzo ichi ndechekuti huwandu hwedatasets i115. Saka, kana data rakawanda rikapihwa, mashandiro enzira yakakurudzirwa ye optimized hybridization anogona kuvandudzwa.
PlantProbs.net.Nickel muzvirimwa nevhu https://plantprobs.net/plant/nutrientImbalances/sodium.html (Yakasvikwa musi wa28 Kubvumbi 2021).
Kasprzak, KS Nickel inofambira mberi mukudzidzisa kwezvakatipoteredza kwemazuva ano. surroundings.toxicology.11, 145–183 (1987).
Cempel, M. & Nikel, G. Nickel: Ongororo yezvainobva uye chepfu yezvakatipoteredza. Polish J. Environment.Stud.15, 375–382 (2006).
Freedman, B. & Hutchinson, TC Zvinosvibisa zvinobva mumhepo uye kuungana muvhu nemiti pedyo nechinhu chinonyungudutsa nickel-copper muSudbury, Ontario, Canada.can.J. Bot.58(1), 108-132.https://doi.org/10.1139/b80-014 (1980).
Manyiwa, T. nevamwe. Simbi dzinorema dziri muvhu, zvirimwa uye njodzi dzine chekuita nemhuka dzinodya mombe dziri pedyo nemugodhi wemhangura-nickel weSelebi-Phikwe muBotswana. surroundings.Geochemistry.Health https://doi.org/10.1007/s10653-021-00918-x (2021).
Cabata-Pendias.Kabata-Pendias A. 2011. Zvinhu zvinotevedzana muvhu uye… – Google Scholar https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Kabata-Pendias+A.+2011.+Trace+ Elements+in+soils+and+plants.+4th+ed.+New+York+%28NY%29%3A+CRC+Press&btnG= (Yakasvikwa musi wa24 Mbudzi 2020).
Almås, A., Singh, B., Agriculture, TS-NJ ya & 1995, isina kutsanangurwa. Mhedzisiro yeindasitiri yenickel yeRussia pahuwandu hwesimbi inorema muvhu rekurima nehuswa muSoer-Varanger, Norway.agris.fao.org.
Nielsen, GD nevamwe. Kunyudzwa kwenickel uye kuchengetwa kwayo mumvura yekunwa zvine chekuita nekudya uye kunzwisiswa kwenickel.toxicology.application.Pharmacodynamics.154, 67–75 (1999).
Costa, M. & Klein, CB Nickel carcinogenesis, mutation, epigenetics kana sarudzo.zvakakomberedza. Maonero Ehutano.107, 2 (1999).
Ajman, PC; Ajado, SK; Borůvka, L.; Bini, JKM; Sarkody, VYO; Cobonye, ​​​​NM; Kuongororwa kwemaitiro ezvinhu zvinogona kuva nechepfu: ongororo yezvinyorwa. Environmental Geochemistry and Health. Springer Science & Business Media BV 2020. https://doi.org/10.1007/s10653-020-00742-9.
Minasny, B. & McBratney, AB Kugadzira Ivhu reDigital: Nhoroondo Pfupi uye Zvimwe Zvidzidzo. Geoderma 264, 301–311.https://doi.org/10.1016/j.geoderma.2015.07.017 (2016).
McBratney, AB, Mendonça Santos, ML & Minasny, B. Pamapu evhu redhijitari.Geoderma 117(1-2), 3-52.https://doi.org/10.1016/S0016-7061(03)00223-4 (2003).
Kugadzira Modeling yeDeutsch.CV Geostatistical Reservoir,… – Google Scholar https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=CV+Deutsch%2C+2002%2C+Geostatistical+Reservoir+Modeling%2C +Oxford+University+Press%2C+376+pages.+&btnG= (Yakasvikwa musi wa28 Kubvumbi 2021).