Siyabonga ngokuvakashela i-Nature.com. Inguqulo yesiphequluli oyisebenzisayo inokusekelwa okulinganiselwe kwe-CSS. Ukuze uthole ulwazi olungcono kakhulu, sincoma ukuthi usebenzise isiphequluli esibuyekeziwe (noma uvale imodi yokuhambisana ku-Internet Explorer). Okwamanje, ukuqinisekisa ukwesekwa okuqhubekayo, sizobonisa isayithi ngaphandle kwezitayela kanye ne-JavaScript.
Ukungcola komhlabathi kuyinkinga enkulu ebangelwa imisebenzi yabantu. Ukusatshalaliswa kwendawo kwezinto ezingaba yingozi (ama-PTE) kuyahlukahluka ezindaweni eziningi zasemadolobheni nasezindaweni eziseduze nedolobha. Ngakho-ke, kunzima ukubikezela ngokwendawo okuqukethwe ama-PTE enhlabathini enjalo. Amasampula angu-115 atholakale kuFrydek Mistek eCzech Republic. Ukuhlushwa kwe-Calcium (Ca), i-magnesium (Mg), i-potassium (K) kanye ne-nickel (Ni) kunqunywe kusetshenziswa i-inductively combined plasma emission spectrometry. I-response variable yi-Ni kanti izibikezeli yi-Ca, Mg, kanye ne-K. I-correlation matrix phakathi kwe-response variable kanye ne-predictor variable ikhombisa ukuhlangana okwanelisayo phakathi kwezinto. Imiphumela yokubikezela ikhombisile ukuthi i-Support Vector Machine Regression (SVMR) yenze kahle, yize iphutha layo lesikwele elilinganiselwe le-root mean (RMSE) (235.974 mg/kg) kanye ne-mean absolute error (MAE) (166.946 mg/kg) laliphezulu kunezinye izindlela ezisetshenzisiwe. Amamodeli axubile e-Empirical Bayesian Kriging-Multiple Linear Regression (EBK-MLR) asebenza. Kubi kakhulu, njengoba kufakazelwa ama-coefficients okunquma angaphansi kuka-0.1. Imodeli ye-Empirical Bayesian Kriging-Support Vector Machine Regression (EBK-SVMR) yayiyimodeli engcono kakhulu, enamanani aphansi e-RMSE (95.479 mg/kg) kanye ne-MAE (77.368 mg/kg) kanye ne-coefficient ephezulu yokunquma (R2 = 0.637). Umphumela wobuchwepheshe bokumodela be-EBK-SVMR ubonakala kusetshenziswa imephu yokuzihlela. Ama-neurons ahlanganisiwe endizeni yengxenye yemodeli ye-hybrid CakMg-EBK-SVMR abonisa amaphethini amaningi emibala abikezela amazinga e-Ni enhlabathini yasemadolobheni kanye neyasemadolobheni. Imiphumela ikhombisa ukuthi ukuhlanganisa i-EBK kanye ne-SVMR kuyindlela ephumelelayo yokubikezela amazinga e-Ni enhlabathini yasemadolobheni kanye neyasemadolobheni.
I-Nickel (Ni) ibhekwa njengesakhi esincane sezitshalo ngoba inegalelo ekufakweni kwe-nitrogen emoyeni (N) kanye ne-urea metabolism, kokubili okudingekayo ukuze imbewu ikhule. Ngaphezu kokufaka kwayo isandla ekukhuleni kwembewu, i-Ni ingasebenza njengesivimbeli se-fungus kanye ne-bacteria futhi ikhuthaze ukuthuthukiswa kwezitshalo. Ukuntuleka kwe-nickel enhlabathini kuvumela isitshalo ukuthi siyimunce, okuholela ku-chlorosis yamaqabunga. Isibonelo, ubhontshisi kanye nobhontshisi oluhlaza kudinga ukusetshenziswa komanyolo osuselwa ku-nickel ukuze kulungiswe ukulungiswa kwe-nitrogen2. Ukusetshenziswa okuqhubekayo komanyolo osuselwa ku-nickel ukuze kucebise inhlabathi futhi kwandise ikhono lemidumba lokulungisa i-nitrogen enhlabathini kwandisa njalo ukugcwala kwe-nickel enhlabathini. Nakuba i-nickel iyisakhi esincane sezitshalo, ukudla kwayo ngokweqile enhlabathini kungenza umonakalo omkhulu kunokuhle. Ubuthi be-nickel enhlabathini bunciphisa i-pH yenhlabathi futhi buvimbele ukumuncwa kwensimbi njengesakhi esibalulekile ekukhuleni kwezitshalo1. Ngokusho kukaLiu3, i-Ni itholakale iyinto ye-17 ebalulekile edingekayo ekuthuthukisweni nasekukhuleni kwezitshalo. Ngaphezu kwendima ye-nickel ekuthuthukisweni nasekukhuleni kwezitshalo, abantu bayayidinga ngezinhlelo ezahlukene. Ukukhiqiza i-electroplating, ukukhiqizwa kwe-nickel esuselwa ku-nickel ama-alloy, kanye nokukhiqizwa kwamadivayisi okuthungela kanye nama-spark plug embonini yezimoto konke kudinga ukusetshenziswa kwe-nickel emikhakheni eyahlukene yezimboni.Ngaphezu kwalokho, ama-alloy asekelwe ku-nickel kanye nezinto ezifakwe ngogesi ziye zasetshenziswa kabanzi ezintweni zasekhishini, izesekeli ze-ballroom, izimpahla zemboni yokudla, ugesi, ucingo kanye nekhebula, ama-jet turbine, izifakelo zokuhlinzwa, izindwangu, kanye nokwakhiwa kwemikhumbi5.Amazinga acebile nge-Ni enhlabathini (okungukuthi, inhlabathi engaphezulu) ahlotshaniswa nemithombo yabantu kanye nemvelo, kodwa ngokuyinhloko, i-Ni iwumthombo wemvelo kunokuba ibe umthombo wabantu4,6.Imithombo yemvelo ye-nickel ihlanganisa ukuqhuma kwezintaba-mlilo, izitshalo, imililo yamahlathi, kanye nezinqubo ze-geological; noma kunjalo, imithombo yabantu ihlanganisa amabhethri e-nickel/cadmium embonini yensimbi, i-electroplating, i-arc welding, i-diesel kanye namafutha kaphethiloli, kanye nokukhishwa komoya okuvela ekushisweni kwamalahle kanye nokushiswa kwemfucuza kanye nodaka Ukuqongelela kwe-nickel7,8.Ngokusho kukaFreedman noHutchinson9 noManiwa et al. 10, imithombo eyinhloko yokungcoliswa komhlabathi ongaphezulu endaweni eseduze neseduze ikakhulukazi izincibilikisi kanye nezimayini ezisekelwe ku-nickel-copper. Inhlabathi ephezulu ezungeze ifektri yokuhlanza i-nickel-copper yaseSudbury eCanada yayinamazinga aphezulu okungcoliswa kwe-nickel ku-26,000 mg/kg11. Ngokuphambene nalokho, ukungcoliswa okuvela ekukhiqizweni kwe-nickel eRussia kuholele ekuqongeleleni okuphezulu kwe-nickel enhlabathini yaseNorway11. Ngokusho kuka-Alms et al. 12, inani le-nickel ekhishwa yi-HNO3 endaweni yokulima ephezulu esifundeni (ukukhiqizwa kwe-nickel eRussia) lalisukela ku-6.25 kuya ku-136.88 mg/kg, okuhambisana nesilinganiso esingu-30.43 mg/kg kanye nokuhlushwa okuyisisekelo okungu-25 mg/kg. Ngokusho kwe-kabata 11, ukusetshenziswa komanyolo we-phosphorus enhlabathini yezolimo enhlabathini yasemadolobheni noma eseduze nedolobha ngezikhathi zezitshalo ezilandelanayo kungangenisa noma kungcolise inhlabathi. Imiphumela engaba khona ye-nickel kubantu ingaholela kumdlavuza nge-mutagenesis, umonakalo we-chromosome, ukukhiqizwa kwe-Z-DNA, ukulungiswa kwe-DNA excision okuvinjiwe, noma izinqubo ze-epigenetic13. Ekuhlolweni kwezilwane, i-nickel itholakale inamandla okubangela izinhlobo ezahlukene zezimila, futhi ama-nickel complexes e-carcinogenic angase abhebhethekise izimila ezinjalo.
Ukuhlolwa kokungcoliswa kwenhlabathi kuye kwachuma ezikhathini zamuva ngenxa yezinkinga eziningi ezihlobene nempilo ezivela ebudlelwaneni bomhlabathi nezitshalo, ubudlelwano bezinto eziphilayo zomhlabathi nomhlabathi, ukuwohloka kwemvelo, kanye nokuhlolwa komthelela wendawo. Kuze kube manje, ukubikezela indawo kwezinto ezingaba yingozi (ama-PTE) njenge-Ni enhlabathini kube nzima futhi kudla isikhathi kusetshenziswa izindlela zendabuko. Ukufika kwemephu yomhlabathi yedijithali (i-DSM) kanye nempumelelo yayo yamanje15 kuye kwathuthukisa kakhulu imephu yomhlabathi ebikezelayo (i-PSM). Ngokusho kukaMinasny noMcBratney16, imephu yomhlabathi ebikezelayo (i-DSM) iye yabonakala iyisigaba esivelele sesayensi yomhlabathi. ULagacherie noMcBratney, 2006 bachaza i-DSM ngokuthi "ukudalwa nokugcwaliswa kwezinhlelo zolwazi lomhlabathi wendawo ngokusebenzisa izindlela zokubuka ezisendaweni kanye nezaselabhorethri kanye nezinhlelo zokucabanga komhlabathi zendawo nezingezona ezendawo". UMcBratney et al. 17 zichaza ukuthi i-DSM noma i-PSM yanamuhla iyindlela ephumelela kakhulu yokubikezela noma ukumaka ukusatshalaliswa kwendawo kwama-PTE, izinhlobo zenhlabathi kanye nezakhiwo zenhlabathi. I-Geostatistics kanye ne-Machine Learning Algorithms (MLA) amasu okulingisa e-DSM adala amamephu edijithali ngosizo lwamakhompyutha esebenzisa idatha ebalulekile nencane.
I-Deutsch18 kanye ne-Olea19 bachaza i-geostatistics ngokuthi “iqoqo lamasu ezinombolo abhekene nokumelwa kwezimfanelo zendawo, ikakhulukazi besebenzisa amamodeli e-stochastic, njengokuthi ukuhlaziywa kochungechunge lwesikhathi kuveza kanjani idatha yesikhathi.” Ngokuyinhloko, i-geostatistics ihilela ukuhlolwa kwe-variograms, okuvumela i-Quantify futhi ichaze ukuncika kwamanani endawo kusuka kusethi ngayinye yedatha20. UGumiaux et al. 20 baqhubeka babonisa ukuthi ukuhlolwa kwe-variograms ku-geostatistics kusekelwe ezimisweni ezintathu, okuhlanganisa (a) ukubala isikali sokuhlobana kwedatha, (b) ukuhlonza nokubala i-anisotropy ekungafani kwesethi yedatha kanye (c) ngaphezu kwalokho Ngaphezu kokucabangela iphutha elingokwemvelo ledatha yokulinganisa ehlukaniswe nemiphumela yendawo, imiphumela yendawo nayo iyalinganiselwa. Ngokwakhela kule mibono, amasu amaningi okuhlanganisa asetshenziswa ku-geostatistics, okuhlanganisa i-kriging ejwayelekile, i-co-kriging, i-kriging evamile, i-kriging ye-Bayesian empirical, indlela elula ye-kriging kanye namanye amasu okuhlanganisa aziwayo ukumaka noma ukubikezela i-PTE, izici zenhlabathi, kanye nezinhlobo zenhlabathi.
Ama-Algorithm Okufunda Komshini (i-MLA) ayindlela entsha kakhulu esebenzisa amakilasi edatha amakhulu angewona aqondile, aqhutshwa ama-algorithms asetshenziswa kakhulu ekumbeni idatha, ukuhlonza amaphethini kudatha, futhi asetshenziswa ngokuphindaphindiwe ekuhlukaniseni emikhakheni yesayensi efana nesayensi yomhlabathi kanye nemisebenzi yokubuyisa. Amaphepha amaningi ocwaningo athembele kumamodeli e-MLA ukubikezela i-PTE enhlabathini, njengoTan et al. 22 (amahlathi angahleliwe okulinganisa insimbi esindayo enhlabathini yezolimo), uSakizadeh et al. 23 (ukumodela kusetshenziswa imishini yokusekela i-vector kanye namanethiwekhi e-neural artificial) ukungcoliswa komhlabathi). Ngaphezu kwalokho, uVega et al. 24 (IKHADI lokulingisa ukugcinwa kwensimbi esindayo kanye nokumuncwa enhlabathini) uSun et al. 25 (ukusetshenziswa kwe-cubist ukusatshalaliswa kwe-Cd enhlabathini) kanye namanye ama-algorithms afana nomakhelwane oseduze we-k, ukuhlehla okukhushuliwe okujwayelekile, kanye nokuhlehla okukhushuliwe Izihlahla nazo zisebenzise i-MLA ukubikezela i-PTE enhlabathini.
Ukusetshenziswa kwama-algorithm e-DSM ekubikezeleni noma ekumepheni kubhekene nezinselele eziningana. Abalobi abaningi bakholelwa ukuthi i-MLA ingcono kune-geostatistics kanye nokuphikisana nalokho. Nakuba enye ingcono kunenye, ukuhlanganiswa kwalokhu okubili kuthuthukisa izinga lokunemba kokumapha noma ukubikezela ku-DSM15. UWoodcock noGopal26 Finke27; uPontius noCheuk28 noGrunwald29 baphawula ngokushiyeka kanye namaphutha athile ekumepheni komhlaba okubikezelwe. Ososayensi bomhlabathi bazame amasu ahlukahlukene ukuze bathuthukise ukusebenza kahle, ukunemba, kanye nokubikezela kokumapha nokubikezela kwe-DSM. Ukuhlanganiswa kokungaqiniseki nokuqinisekisa kungenye yezici eziningi ezahlukene ezihlanganiswe ku-DSM ukuze kuthuthukiswe ukusebenza kahle futhi kuncishiswe amaphutha. Kodwa-ke, u-Agyeman et al. 15 bachaza ukuthi ukuziphatha kokuqinisekisa kanye nokungaqiniseki okwethulwe ngokudalwa kwemephu nokubikezela kufanele kuqinisekiswe ngokuzimela ukuze kuthuthukiswe ikhwalithi yemephu. Ukulinganiselwa kwe-DSM kungenxa yekhwalithi yomhlabathi ehlakazekile ngokwendawo, okubandakanya ingxenye yokungaqiniseki; noma kunjalo, ukuntuleka kokuqiniseka ku-DSM kungavela emithonjeni eminingi yephutha, okungukuthi iphutha le-covariate, iphutha lemodeli, iphutha lendawo, kanye nephutha lokuhlaziya 31. Ukulingisa ukungalungi okubangelwa yi-MLA kanye nezinqubo ze-geostatistical kuhlotshaniswa nokuntuleka kokuqonda, ekugcineni okuholela ekwenzeni lula kakhulu inqubo yangempela32. Kungakhathaliseki ukuthi hlobo luni lokulingisa, ukungalungi kungabangelwa amapharamitha okulingisa, izibikezelo zemodeli yezibalo, noma ukuhunyushwa kwe-interpolation33. Muva nje, kuvele umkhuba omusha we-DSM okhuthaza ukuhlanganiswa kwe-geostatistics kanye ne-MLA ekumepheni nasekubikezeleni. Ososayensi abaningana bomhlabathi nababhali, njengoSergeev et al. 34; Subbotina et al. 35; Tarasov et al. 36 kanye noTarasov et al. 37 basebenzise ikhwalithi enembile ye-geostatistics kanye nokufunda komshini ukukhiqiza amamodeli ahlanganisiwe athuthukisa ukusebenza kahle kokubikezela kanye nokumapha. ikhwalithi. Amanye alawa mamodeli e-hybrid noma ahlanganisiwe e-algorithm yi-Artificial Neural Network Kriging (ANN-RK), i-Multilayer Perceptron Residual Kriging (MLP-RK), i-Generalized Regression Neural Network Residual Kriging (GR-NNRK)36, i-Artificial Neural Network Kriging-Multilayer Perceptron (ANN-K-MLP)37 kanye ne-Co-Kriging kanye ne-Gaussian Process Regression38.
Ngokusho kukaSergeev et al., ukuhlanganisa amasu ahlukahlukene okulingisa kunamandla okususa amaphutha nokwandisa ukusebenza kahle kwemodeli ye-hybrid ephumayo kunokuthuthukisa imodeli yayo eyodwa. Kulesi simo, leli phepha elisha lithi kuyadingeka ukusebenzisa i-algorithm ehlanganisiwe ye-geostatistics kanye ne-MLA ukudala amamodeli e-hybrid afanele ukubikezela ukucebisa kwe-Ni ezindaweni zasemadolobheni nasezindaweni eziseduze nedolobha. Lolu cwaningo luzoncika ku-Empirical Bayesian Kriging (EBK) njengemodeli eyisisekelo futhi luyixube namamodeli e-Support Vector Machine (SVM) kanye ne-Multiple Linear Regression (MLR). Ukuxutshwa kwe-EBK nanoma iyiphi i-MLA akwaziwa. Amamodeli amaningi axubile abonwayo ayinhlanganisela ye-normal, residual, regression kriging, kanye ne-MLA. I-EBK iyindlela yokuhlanganisa i-geostatistical esebenzisa inqubo ye-spatially stochastic etholakala endaweni njengensimu engahleliwe engahleliwe/engahleliwe enemingcele yendawo echazwe phezu kwensimu, okuvumela ukuhlukahluka kwendawo39. I-EBK isetshenziswe ezifundweni ezahlukahlukene, okuhlanganisa ukuhlaziya ukusatshalaliswa kwe-organic carbon enhlabathini yasepulazini40, ukuhlola ukungcola kwenhlabathi41 kanye nokumaka inhlabathi. izakhiwo42.
Ngakolunye uhlangothi, i-Self-Organizing Graph (SeOM) iyi-algorithm yokufunda esetshenziswe ezihlokweni ezahlukene ezifana noLi et al. 43, Wang et al. 44, Hossain Bhuiyan et al. 45 kanye noKebonye et al. 46. Thola izimfanelo zendawo kanye nokuqoqwa kwezinto. UWang et al. 44 uchaza ukuthi i-SeOM iyindlela yokufunda enamandla eyaziwa ngekhono layo lokuqoqa nokucabanga ngezinkinga ezingezona eziqondile. Ngokungafani nezinye izindlela zokuqaphela amaphethini ezifana nokuhlaziywa kwezingxenye eziyinhloko, ukuqoqa okungacacile, ukuqoqa okuhlelekile, kanye nokwenza izinqumo zemigomo eminingi, i-SeOM ingcono ekuhleleni nasekuboneni amaphethini e-PTE. Ngokusho kukaWang et al. 44, i-SeOM ingaqoqa ngokwendawo ukusatshalaliswa kwama-neurons ahlobene futhi inikeze umbono wedatha onesinqumo esiphezulu. I-SeOM izobona ngeso lengqondo idatha yokubikezela ye-Ni ukuthola imodeli engcono kakhulu yokuchaza imiphumela yokuhumusha okuqondile.
Leli phepha lihlose ukukhiqiza imodeli yokumapha eqinile enokunemba okuhle kakhulu kokubikezela okuqukethwe yi-nickel enhlabathini yasemadolobheni kanye neyasemadolobheni. Sicabanga ukuthi ukuthembeka kwemodeli exubile kuncike kakhulu ethonyeni lamanye amamodeli anamathele kumodeli eyisisekelo. Siyazivuma izinselelo ezibhekene ne-DSM, futhi ngenkathi lezi zinselelo zixazululwa ngezindlela eziningi, inhlanganisela yentuthuko kumamodeli we-geostatistics kanye ne-MLA ibonakala ikhuphuka kancane kancane; ngakho-ke, sizozama ukuphendula imibuzo yocwaningo engase iveze amamodeli axubile. Kodwa-ke, imodeli inembe kangakanani ekubikezeleni isici esiqondiwe? Futhi, liyini izinga lokuhlola ukusebenza kahle ngokusekelwe ekuqinisekisweni nasekuhlolweni kokunemba? Ngakho-ke, imigomo ethile yalolu cwaningo kwakuwukudala (a) imodeli yokuxuba ehlanganisiwe ye-SVMR noma i-MLR kusetshenziswa i-EBK njengemodeli eyisisekelo, (b) ukuqhathanisa amamodeli atholakele (c) ukuphakamisa imodeli yokuxuba engcono kakhulu yokubikezela amazinga e-Ni enhlabathini yasemadolobheni noma yasemadolobheni, kanye (d) ukusetshenziswa kwe-SeOM ukudala imephu enesinqumo esiphezulu yokwehluka kwendawo ye-nickel.
Lolu cwaningo lwenziwa eCzech Republic, ikakhulukazi esifundeni saseFrydek Mistek esifundeni saseMoravia-Silesian (bheka uMfanekiso 1). Indawo yendawo yocwaningo ilukhuni kakhulu futhi ikakhulukazi iyingxenye yesifunda saseMoravia-Silesian Beskidy, esiyingxenye yomngcele ongaphandle wezintaba zaseCarpathian. Indawo yocwaningo itholakala phakathi kuka-49° 41′ 0′ N no-18° 20′ 0′ E, kanti ukuphakama kuphakathi kuka-225 no-327 m; Nokho, uhlelo lokuhlukanisa iKoppen lwesimo sezulu sesifunda lulinganiswa ngokuthi i-Cfb = isimo sezulu esifudumele saselwandle, Kukhona imvula eningi ngisho nasezinyangeni ezomile. Amazinga okushisa ayahlukahluka kancane unyaka wonke phakathi kuka-−5 °C no-24 °C, akuvamile ukuba ngaphansi kuka-−14 °C noma ngaphezulu kuka-30 °C, kuyilapho isilinganiso semvula yonyaka siphakathi kuka-685 no-752 mm47. Indawo elinganisiwe yokuhlola yonke indawo ingamakhilomitha-skwele angu-1,208, kanye no-39.38% wendawo etshaliwe kanye no-49.36% wokumbozwa kwehlathi. Ngakolunye uhlangothi, indawo esetshenziswe kulolu cwaningo ingamakhilomitha-skwele angu-889.8. E-Ostrava nasezindaweni ezizungezile, imboni yensimbi kanye nemisebenzi yensimbi iyasebenza kakhulu. Izimboni zensimbi, imboni yensimbi lapho i-nickel isetshenziswa khona ezinsimbini ezingagqwali (isb. ukumelana nokugqwala komoya) kanye nezinsimbi ze-alloy (i-nickel yandisa amandla e-alloy ngenkathi igcina ukuguquguquka kwayo okuhle nokuqina), kanye nezolimo ezijulile njengokusetshenziswa komanyolo we-phosphate kanye nokukhiqizwa kwemfuyo kuyimithombo engaba khona yocwaningo lwe-nickel ku isifunda (isb., ukwengeza i-nickel ezimvini ukwandisa amazinga okukhula ezimvini nasezinkomeni ezingondliwe kahle). Okunye ukusetshenziswa kwe-nickel ezimbonini ezindaweni zocwaningo kufaka phakathi ukusetshenziswa kwayo ekufakweni kwe-electroplating, okuhlanganisa izinqubo zokumboza i-nickel kanye ne-nickel engenawo ugesi. Izakhiwo zomhlabathi zihlukaniswa kalula ngombala wenhlabathi, isakhiwo, kanye nokuqukethwe kwe-carbonate. Ukuthungwa komhlabathi kuphakathi kuya kokuncane, okutholakala ezintweni eziyinhloko. Ziyi-colluvial, alluvial noma aeolian ngokwemvelo. Ezinye izindawo zomhlabathi zibonakala zinamabala ebusweni nangaphansi komhlabathi, ngokuvamile zinokhonkolo kanye nokumhlophe. Kodwa-ke, ama-cambisol nama-stagnosol yizinhlobo zenhlabathi ezivame kakhulu esifundeni48. Njengoba ukuphakama kusuka ku-455.1 kuya ku-493.5 m, ama-cambisol abusa iCzech Republic49.
Imephu yendawo yokufunda [Imephu yendawo yokufunda yadalwa kusetshenziswa i-ArcGIS Desktop (ESRI, Inc, inguqulo 10.7, i-URL: https://desktop.arcgis.com).]
Kutholakale amasampula enhlabathi engaphezulu angu-115 enhlabathini yasemadolobheni kanye neyasemadolobheni esifundeni saseFrydek Mistek. Iphethini yesampula esetshenzisiwe kwakuyigridi evamile enamasampula enhlabathi ahlukaniswe ngamakhilomitha angu-2 × 2, kanti inhlabathi ephezulu yalinganiswa ngokujula okungu-0 kuya ku-20 cm kusetshenziswa idivayisi ye-GPS ephethwe ngesandla (Leica Zeno 5 GPS). Amasampula apakishwa ezikhwameni ze-Ziploc, alebula kahle, bese ethunyelwa elabhorethri. Amasampula omiswe emoyeni ukuze akhiqize amasampula acoliwe, acoliwe ngohlelo lomshini (i-Fritsch disc mill), futhi ahlungwe (usayizi wesihlungo ongu-2 mm). Beka igremu eli-1 lamasampula enhlabathi omisiwe, ahlungwe futhi ahlungwe emabhodleleni e-teflon alebula ngokucacile. Esitsheni ngasinye se-Teflon, khipha u-7 ml we-35% HCl kanye no-3 ml we-65% HNO3 (usebenzisa i-dispenser ezenzakalelayo - eyodwa ye-asidi ngayinye), mboza kancane bese uvumela amasampula ukuthi ame ubusuku bonke ukuze kuvele impendulo (uhlelo lwe-aqua regia). Beka i-supernatant epuletini lensimbi elishisayo (izinga lokushisa: 100 W kanye no-160 °C) amahora ama-2 ukuze kube lula inqubo yokugaya amasampula, bese upholisa. Dlulisa i-supernatant ebhodleleni elingu-50 ml bese uyixuba ibe ngu-50 ml ngamanzi axubile. Ngemuva kwalokho, hlunga i-supernatant exubile ibe yipayipi le-PVC elingu-50 ml ngamanzi axubile. Ngaphezu kwalokho, i-1 ml yesisombululo sokuxuba yaxubaniswa ne-9 ml yamanzi axubile futhi yahlungwa ibe yipayipi elingu-12 ml elilungiselelwe ukuhlushwa kwe-PTE. Ukuhlushwa kwama-PTE (As, Cd, Cr, Cu, Mn, Ni, Pb, Zn, Ca, Mg, K) kunqunywe yi-ICP-OES (Inductively Coupled Plasma Optical Emission Spectroscopy) (Thermo Fisher Scientific, USA) ngokwezindlela ezijwayelekile kanye nesivumelwano. Qinisekisa izinqubo zokuqinisekisa ikhwalithi nokulawula (QA/QC) (SRM NIST 2711a Montana II Soil). Ama-PTE anemikhawulo yokuthola engaphansi kwesigamu akhishwe kulolu cwaningo. Umkhawulo wokuthola we-PTE okusetshenziswe kulolu cwaningo kwakungu-0.0004.(wena).Ngaphezu kwalokho, inqubo yokulawula ikhwalithi kanye nokuqinisekisa ikhwalithi yokuhlaziywa ngakunye kuqinisekiswa ngokuhlaziya amazinga okubhekisela.Ukuqinisekisa ukuthi amaphutha ancishisiwe, kwenziwa ukuhlaziywa okuphindwe kabili.
I-Empirical Bayesian Kriging (EBK) ingenye yezindlela eziningi zokuhlanganisa i-geostatistical ezisetshenziswa ekubumbeni emikhakheni eyahlukene njengesayensi yomhlabathi. Ngokungafani nezinye izindlela zokuhlanganisa i-kriging, i-EBK ihlukile ezindleleni zendabuko zokuhlanganisa i-kriging ngokucabangela iphutha elilinganiswe yimodeli ye-semivariogram. Ku-interpolation ye-EBK, amamodeli amaningana e-semivariogram abalwa ngesikhathi sokuhlanganisa, kunokuba kube yi-semivariogram eyodwa. Amasu okuhlanganisa avula indlela yokungaqiniseki kanye nokuhlela okuhlobene nalokhu kuhlelwa kwe-semivariogram eyakha ingxenye eyinkimbinkimbi kakhulu yendlela eyanele yokuhlanganisa i-kriging. Inqubo yokuhlanganisa i-EBK ilandela izindlela ezintathu eziphakanyiswe yi-Krivoruchko50, (a) imodeli ilinganisa i-semivariogram kusuka kusethi yedatha yokufaka (b) inani elisha elibikezelwe lendawo ngayinye yedatha yokufaka ngokusekelwe ku-semivariogram ekhiqizwe kanye (c) imodeli yokugcina ye-A ibalwa kusuka kusethi yedatha elingisiwe. Umthetho we-equation ye-Bayesian unikezwa njenge-posterior
Lapho i-\(Prob\left(A\right)\) imelela ithuba elingaphansi, \(Prob\left(B\right)\) linganakwa ezimweni eziningi, \(Prob (B,A)\ ).Ukubalwa kwe-semivariogram kusekelwe emthethweni kaBayes, okhombisa ukuthambekela kwamasethi edatha okubonwa angadalwa kusuka kuma-semivariogram. Inani le-semivariogram libe selinqunywa kusetshenziswa umthetho kaBayes, ochaza ukuthi kungenzeka kangakanani ukudala isethi yedatha yokubonwa kusuka ku-semivariogram.
Umshini we-vector wokusekela uyi-algorithm yokufunda komshini ekhiqiza i-hyperplane ehlukanisa kahle ukuhlukanisa amakilasi afanayo kodwa angazimele ngomugqa. I-Vapnik51 idale i-algorithm yokuhlukanisa ngenhloso, kodwa muva nje isetshenziswe ukuxazulula izinkinga eziqondiswe ekubuyiseleni emuva. Ngokusho kukaLi et al.52, i-SVM ingenye yezindlela ezinhle kakhulu zokuhlukanisa futhi isetshenziswe emikhakheni eyahlukene. Ingxenye yokubuyisela emuva ye-SVM (Support Vector Machine Regression - SVMR) isetshenziswe kulokhu kuhlaziywa. UCherkassky noMulier53 baqale i-SVMR njengokubuyela emuva okusekelwe ku-kernel, ukubalwa kwayo okwenziwe kusetshenziswa imodeli yokubuyela emuva eqondile enemisebenzi yendawo yamazwe amaningi. UJohn et al54 babika ukuthi ukumodela kwe-SVMR kusebenzisa ukubuyela emuva komugqa oqondile, okudala ubudlelwano obungewona omugqa futhi kuvumela imisebenzi yendawo. Ngokusho kukaVohland et al. 55, i-epsilon (ε)-SVMR isebenzisa isethi yedatha eqeqeshiwe ukuthola imodeli yokumelwa njengomsebenzi ongazweli i-epsilon osetshenziswa ukumaka idatha ngokuzimela nge-epsilon bias engcono kakhulu kusukela ekuqeqeshweni kwedatha ehlobene. Iphutha lebanga elisethwe ngaphambilini alinakwa kusukela enanini langempela, futhi uma iphutha likhulu kune-ε(ε), izakhiwo zomhlabathi ziyalikhokhela. Imodeli iphinde yehlise ubunzima bedatha yokuqeqesha ibe yisethi ebanzi yamavektha okusekela. Isibalo esiphakanyiswe yi-Vapnik51 siboniswe ngezansi.
lapho u-b emele umkhawulo we-scalar, \(K\left({x}_{,}{ x}_{k}\right)\) umele umsebenzi we-kernel, \(\alpha\) umele i-Lagrange multiplier, u-N umele isethi yedatha yezinombolo, \({x}_{k}\) umele ukufaka idatha, kanye no-\(y\) ukukhishwa kwedatha. Enye yama-kernel ayisihluthulelo asetshenziswayo umsebenzi we-SVMR, okuwumsebenzi we-Gaussian radial basis (RBF). I-RBF kernel isetshenziswa ukunquma imodeli ye-SVMR efanele, okubaluleke kakhulu ukuthola i-penalty set factor C ecashile kakhulu kanye ne-kernel parameter gamma (γ) yedatha yokuqeqesha ye-PTE. Okokuqala, sihlole isethi yokuqeqesha bese sihlola ukusebenza kwemodeli kusethi yokuqinisekisa. Ipharamitha yokuqondisa esetshenzisiwe yi-sigma kanti inani lendlela yi-svmRadial.
Imodeli yokubuyela emuva eqondile eningi (i-MLR) iyimodeli yokubuyela emuva emele ubudlelwano phakathi kwe-response variable kanye nenani le-predictor variables ngokusebenzisa amapharamitha ahlanganisiwe aqondile abalwe kusetshenziswa indlela ye-least squares. Ku-MLR, imodeli ye-least squares iwumsebenzi wokubikezela wezakhiwo zomhlabathi ngemuva kokukhethwa kwe-definitives echazayo. Kubalulekile ukusebenzisa impendulo ukusungula ubudlelwano obuqondile kusetshenziswa i-definitives echazayo. I-PTE isetshenziswe njenge-response variable ukusungula ubudlelwano obuqondile ne-definitives echazayo. I-MLR equation iyi
lapho u-y kuyi-response variable, \(a\) kuyi-intercept, u-n yinombolo yezibikezeli, \({b}_{1}\) kuyi-regression engaphelele yama-coefficients, \({x}_{ i}\) imele i-predictor noma i-explanatory variable, kanye ne-\({\varepsilon }_{i}\) imele iphutha kumodeli, eyaziwa nangokuthi i-residual.
Amamodeli axubile atholakale ngokuhlanganisa i-EBK ne-SVMR kanye ne-MLR. Lokhu kwenziwa ngokukhipha amanani abikezelwe kusukela ekuhlanganisweni kwe-EBK. Amanani abikezelwe atholakale ku-Ca, K, kanye ne-Mg ehlanganisiwe atholakala ngenqubo yokuhlanganisa ukuthola iziguquguquko ezintsha, njenge-CaK, CaMg, kanye ne-KMg. Izakhi i-Ca, K kanye ne-Mg bese zihlanganiswa ukuthola iguquguquko lesine, i-CaKMg. Sekukonke, iziguquguquko ezitholiwe yi-Ca, K, Mg, CaK, CaMg, KMg kanye ne-CaKMg. Lezi ziguquguquko zaba izibikezelo zethu, ezisiza ukubikezela ukugcwala kwe-nickel enhlabathini yasemadolobheni kanye neyasemadolobheni. I-algorithm ye-SVMR yenziwe kuma-predictors ukuthola imodeli exubile ye-Empirical Bayesian Kriging-Support Vector Machine (EBK_SVM). Ngokufanayo, iziguquguquko nazo zihanjiswa ngepayipi nge-algorithm ye-MLR ukuthola imodeli exubile ye-Empirical Bayesian Kriging-Multiple Linear Regression (EBK_MLR). Ngokuvamile, iziguquguquko i-Ca, K, I-Mg, i-CaK, i-CaMg, i-KMg, ne-CaKMg zisetshenziswa njengezibikezeli zokuqukethwe kwe-Ni enhlabathini yasemadolobheni kanye neyasemadolobheni. Imodeli eyamukelekayo kakhulu etholakele (i-EBK_SVM noma i-EBK_MLR) izobe isibonakala kusetshenziswa igrafu ezihlelayo. Ukuhamba komsebenzi kwalolu cwaningo kuboniswe kuMfanekiso 2.
Ukusebenzisa i-SeOM sekuyindlela ethandwayo yokuhlela, ukuhlola, nokubikezela idatha emkhakheni wezezimali, ezempilo, embonini, izibalo, isayensi yomhlabathi, nokuningi. I-SeOM idalwe kusetshenziswa amanethiwekhi e-neural okwenziwa kanye nezindlela zokufunda ezingaqondiswanga zokuhlela, ukuhlola, kanye nokubikezela. Kulolu cwaningo, i-SeOM yasetshenziswa ukubona ngeso lengqondo amazinga e-Ni ngokusekelwe kumodeli engcono kakhulu yokubikezela i-Ni enhlabathini yasemadolobheni kanye neyasemadolobheni. Idatha ecutshungulwa ekuhlolweni kwe-SeOM isetshenziswa njengezinguquko ze-vector zokufaka ze-n43,56. UMelssen et al. 57 ichaza ukuxhumana kwevektha yokufaka kunethiwekhi ye-neural ngokusebenzisa ungqimba olulodwa lokufaka kuvektha yokukhipha enevektha yesisindo esisodwa. Umphumela okhiqizwa yi-SeOM imephu enezinhlangothi ezimbili equkethe ama-neurons noma ama-node ahlukene ahlanganiswe kumamephu e-topological ayisithupha, ayindilinga, noma ayisikwele ngokuya ngokusondela kwawo. Uma kuqhathaniswa osayizi bemephu ngokusekelwe ku-metric, i-quantization error (QE) kanye ne-topographic error (TE), imodeli ye-SeOM ene-0.086 kanye ne-0.904, ngokulandelana, iyakhethwa, okuyiyunithi yemephu engu-55 (5 × 11). Isakhiwo se-neuron sinqunywa ngokuya ngenani lama-node ku-empiric equation.
Inani ledatha elisetshenziswe kulolu cwaningo lingamasampula angu-115. Indlela engahleliwe isetshenziswe ukuhlukanisa idatha ibe yidatha yokuhlola (25% yokuqinisekisa) kanye namasethi edatha okuqeqesha (75% yokulinganisa). Isethi yedatha yokuqeqesha isetshenziselwa ukukhiqiza imodeli yokubuyela emuva (ukulinganisa), kanti isethi yedatha yokuhlola isetshenziselwa ukuqinisekisa ikhono lokuhlanganisa58. Lokhu kwenzelwa ukuhlola ukufaneleka kwamamodeli ahlukahlukene okubikezela okuqukethwe kwe-nickel enhlabathini. Wonke amamodeli asetshenzisiwe adlule enkambisweni yokuqinisekisa ephindwe kayishumi, ephindaphindwa izikhathi ezinhlanu. Ama-variable akhiqizwe yi-EBK interpolation asetshenziswa njengezibikezeli noma ama-variable achazayo ukubikezela i-target variable (PTE). Ukumodela kuphathwa ku-RStudio kusetshenziswa i-packages library(Kohonen), library(caret), library(modelr), library(“e1071″), library(“plyr”), library(“caTools”), library(“prospectr”) kanye nama-library(“Metrics”).
Kusetshenziswe amapharamitha ahlukahlukene okuqinisekisa ukuthola imodeli engcono kakhulu efanelekile yokubikezela amazinga e-nickel enhlabathini kanye nokuhlola ukunemba kwemodeli kanye nokuqinisekiswa kwayo. Amamodeli e-hybridization ahlolwe kusetshenziswa i-mean absolute error (MAE), i-root mean square error (RMSE), kanye ne-R-squared noma i-coefficient determination (R2). I-R2 ichaza ukungafani kwezilinganiso empendulweni, emelelwa yimodeli yokubuyela emuva. I-RMSE kanye nobukhulu bokungafani ezilinganisweni ezizimele zichaza amandla okubikezela emodeli, kuyilapho i-MAE inquma inani langempela lenani. Inani le-R2 kumele libe phezulu ukuze kuhlolwe imodeli yengxube engcono kakhulu kusetshenziswa amapharamitha okuqinisekisa, lapho inani lisondela ku-1, ukunemba kuyanda. Ngokusho kukaLi et al. 59, inani le-R2 criterion elingu-0.75 noma ngaphezulu libhekwa njengesibikezeli esihle; kusukela ku-0.5 kuya ku-0.75 ukusebenza kwemodeli okwamukelekayo, kanti ngaphansi kuka-0.5 ukusebenza kwemodeli okungamukeleki. Lapho ukhetha imodeli esebenzisa izindlela zokuhlola ze-RMSE kanye ne-MAE validation criteria, amanani aphansi atholiwe anele futhi abhekwa njengokukhetha okungcono kakhulu. I-equation elandelayo ichaza indlela yokuqinisekisa.
lapho u-n emele usayizi wenani elibonwe\({Y}_{i}\) emele impendulo elinganisiwe, kanti \({\widehat{Y}}_{i}\) naye umele inani lempendulo elibikezelwe, ngakho-ke, kokubonwa kokuqala kuka-i.
Izincazelo zezibalo zeziguquguquko zokubikezela kanye nezimpendulo zethulwe kuThebula 1, ezibonisa isilinganiso, ukuphambuka okujwayelekile (SD), i-coefficient of variation (CV), ubuncane, ubukhulu, i-kurtosis, kanye nokugoba. Amanani amancane kanye naphezulu ezinto alandelana ngokulandelana okwehlayo kwe-Mg < Ca < K < Ni kanye ne-Ca < Mg < K < Ni, ngokulandelana. Ukuhlushwa kwe-response variable (Ni) okuthathwe isampula endaweni yocwaningo kusuke ku-4.86 kuya ku-42.39 mg/kg. Ukuqhathanisa i-Ni nesilinganiso somhlaba (29 mg/kg) kanye nesilinganiso saseYurophu (37 mg/kg) kubonise ukuthi isilinganiso sejiyometri esibalwe sonke sendawo yocwaningo besingaphakathi kobubanzi obubekezelelekayo. Noma kunjalo, njengoba kuboniswe yi-Kabata-Pendias11, ukuqhathanisa ukuhlushwa kwe-nickel (Ni) okujwayelekile ocwaningweni lwamanje nenhlabathi yezolimo eSweden kukhombisa ukuthi ukuhlushwa kwe-nickel okujwayelekile kwamanje kuphakeme. Ngokufanayo, ukuhlushwa kwe-Frydek Mistek enhlabathini yasemadolobheni kanye neyasemadolobheni ocwaningweni lwamanje (Ni 16.15 mg/kg) kwakuphakeme kunesivunyelwe. umkhawulo wama-60 (10.2 mg/kg) we-Ni enhlabathini yasemadolobheni yasePoland ebikwe nguRóżański et al.Ngaphezu kwalokho, uBretzel noCalderisi61 baqophe amazinga aphansi kakhulu e-Ni (1.78 mg/kg) enhlabathini yasemadolobheni eTuscany uma kuqhathaniswa nocwaningo lwamanje. UJim62 uphinde wathola amazinga aphansi e-nickel (12.34 mg/kg) enhlabathini yasemadolobheni yaseHong Kong, aphansi kune-nickel yamanje kulolu cwaningo. UBirke et al63 babike amazinga aphakathi e-Ni angu-17.6 mg/kg endaweni yakudala yezimayini neyasemadolobheni eSaxony-Anhalt, eJalimane, okwakungu-1.45 mg/kg ephakeme kune-Ni evamile endaweni (16.15 mg/kg).Ucwaningo lwamanje.Okuqukethwe kwe-nickel okweqile enhlabathini kwezinye izindawo zasemadolobheni nasezindaweni zasemadolobheni endaweni yocwaningo kungase kubangelwe kakhulu embonini yensimbi nensimbi kanye nemboni yensimbi.Lokhu kuhambisana nocwaningo lukaKhodadoust et al. 64 ukuthi imboni yensimbi kanye nokusebenza kwensimbi kuyimithombo eyinhloko yokungcoliswa kwe-nickel enhlabathini. Kodwa-ke, izibikezelo nazo zazisukela ku-538.70 mg/kg kuya ku-69,161.80 mg/kg ye-Ca, 497.51 mg/kg kuya ku-3535.68 mg/kg ye-K, kanye no-685.68 mg/kg kuya ku-5970.05 mg/kg ye-Mg.Jakovljevic et al. Abangu-65 bahlole okuqukethwe yi-Mg kanye ne-K okuphelele kwenhlabathi enkabeni yeSerbia. Bathole ukuthi ukugcwala okuphelele (410 mg/kg kanye no-400 mg/kg, ngokulandelana) kwakuphansi kunokugcwala kwe-Mg kanye ne-K kocwaningo lwamanje. Okungahlukaniseki, empumalanga yePoland, u-Orzechowski noSmolczynski66 bahlole okuqukethwe okuphelele kwe-Ca, Mg kanye ne-K futhi babonise ukugcwala okumaphakathi kwe-Ca (1100 mg/kg), i-Mg (590 mg/kg) kanye ne-K (810 mg/kg). Okuqukethwe enhlabathini ephezulu kuphansi kunento eyodwa kulolu cwaningo. Ucwaningo lwamuva nje lukaPongrac et al. 67 lubonise ukuthi okuqukethwe okuphelele kwe-Ca okuhlaziywe enhlabathini ezintathu ezahlukene eScotland, e-UK (inhlabathi yaseMylnefield, inhlabathi yaseBalruddery kanye nenhlabathi yaseHartwood) kubonise okuqukethwe okuphezulu kwe-Ca kulolu cwaningo.
Ngenxa yokugxila okuhlukile okulinganisiwe kwezinto ezithathwe isampula, ukusatshalaliswa kwesethi yedatha yezinto kubonisa ukugoba okuhlukile. Ukugoba kanye ne-kurtosis yezinto kuqale ku-1.53 ​​kuya ku-7.24 kanye no-2.49 kuya ku-54.16, ngokulandelana. Zonke izinto ezibaliwe zinamazinga okugoba kanye ne-kurtosis angaphezu kuka-+1, okubonisa ukuthi ukusatshalaliswa kwedatha akujwayelekile, kugoba ngendlela efanele futhi kufinyelele esicongweni. Ama-CV alinganisiwe ezinto abonisa nokuthi i-K, i-Mg, ne-Ni abonisa ukugoba okuphakathi, kuyilapho i-Ca inokuguquguquka okuphezulu kakhulu. Ama-CV e-K, i-Ni ne-Mg achaza ukusatshalaliswa kwawo okufanayo. Ngaphezu kwalokho, ukusatshalaliswa kwe-Ca akufani futhi imithombo yangaphandle ingathinta izinga layo lokucebisa.
Ukuhlangana kweziguquguquko zokubikezela nezinto zokuphendula kubonise ukuhlangana okwanelisayo phakathi kwezinto (bheka uMfanekiso 3). Ukuhlangana kubonise ukuthi i-CaK ibonise ukuhlangana okumaphakathi nenani lika-r = 0.53, njengoba kwenza i-CaNi. Nakuba i-Ca ne-K zibonisa ukuhlangana okulinganiselwe komunye nomunye, abacwaningi abanjengoKingston et al. I-68 ne-Santo69 zisikisela ukuthi amazinga azo enhlabathini ayalingana ngokuphambene. Kodwa-ke, i-Ca ne-Mg ziphikisana ne-K, kodwa i-CaK ihlobene kahle. Lokhu kungase kube ngenxa yokusetshenziswa kwamanyolo afana ne-potassium carbonate, aphezulu ngo-56% ku-potassium. I-Potassium yayihlobene ngokulinganiselwe ne-magnesium (KM r = 0.63). Embonini yamanyolo, lezi zakhi ezimbili zihlobene eduze ngoba i-potassium magnesium sulfate, i-potassium magnesium nitrate, kanye ne-potassium kusetshenziswa enhlabathini ukuze kwandiswe amazinga azo okuntuleka. I-Nickel ihlobene ngokulinganiselwe ne-Ca, K kanye ne-Mg ngamanani e-r = 0.52, 0.63 kanye no-0.55, ngokulandelana. Ubudlelwano obuhilela i-calcium, i-magnesium, kanye nama-PTE afana ne-nickel buyinkimbinkimbi, kodwa noma kunjalo, i-magnesium ivimbela ukumuncwa kwe-calcium, i-calcium inciphisa imiphumela ye-magnesium eningi, futhi i-magnesium ne-calcium zombili zinciphisa imiphumela enobuthi ye-nickel enhlabathini.
I-matrix yokuxhumana kwezinto ezibonisa ubudlelwano phakathi kwezibikezeli nezimpendulo (Qaphela: lesi sibalo sihlanganisa i-scatterplot phakathi kwezinto, amazinga okubaluleka asekelwe ku-p <0,001).
Isithombe 4 sibonisa ukusatshalaliswa kwendawo kwezinto. Ngokusho kukaBurgos et al70, ukusetshenziswa kokusatshalaliswa kwendawo kuyindlela esetshenziswa ukulinganisa nokugqamisa izindawo ezishisayo ezindaweni ezingcolile. Amazinga okunotha kwe-Ca ku-Fig. 4 angabonakala engxenyeni esenyakatho-ntshonalanga yemephu yokusatshalaliswa kwendawo. Lesi sibalo sibonisa izindawo ezishisayo zokunotha kwe-Ca eziphakathi nendawo kuya phezulu. Ukunotha kwe-calcium enyakatho-ntshonalanga yemephu kungenzeka kungenxa yokusetshenziswa kwe-quicklime (i-calcium oxide) ukunciphisa i-acidity yomhlabathi kanye nokusetshenziswa kwayo ezigayweni zensimbi njenge-alkaline oxygen enkambisweni yokwenza insimbi. Ngakolunye uhlangothi, abanye abalimi bakhetha ukusebenzisa i-calcium hydroxide enhlabathini ene-acidic ukuze kuncishiswe i-pH, okwandisa nokuqukethwe kwe-calcium enhlabathini71. I-Potassium iphinde ibonise izindawo ezishisayo enyakatho-ntshonalanga nasempumalanga yemephu. INyakatho-ntshonalanga iyindawo enkulu yezolimo, futhi iphethini ye-potassium ephakathi nendawo kuya phezulu ingase ibangelwe ukusetshenziswa kwe-NPK kanye ne-potassium. Lokhu kuhambisana nezinye izifundo, njengeMadaras neLipavský72, Madaras et al.73, Pulkrabová et al.74, Asare et al.75, abaqaphele ukuthi ukuzinza kwenhlabathi kanye nokwelashwa nge-KCl kanye ne-NPK kuholele ekuqukweni okuphezulu kwe-K enhlabathini. Ukucebisa nge-Spatial Potassium enyakatho-ntshonalanga yemephu yokusatshalaliswa kungenzeka kungenxa yokusetshenziswa kwamanyolo asekelwe ku-potassium njenge-potassium chloride, i-potassium sulfate, i-potassium nitrate, i-potassium, kanye ne-potassium ukwandisa okuqukethwe kwe-potassium enhlabathini empofu.Zádorová et al. 76 kanye noTlustoš et al. 77 ichaze ukuthi ukusetshenziswa komanyolo osekelwe ku-K kwandisa okuqukethwe kwe-K enhlabathini futhi kuzokwandisa kakhulu okuqukethwe kwezakhamzimba enhlabathini ngokuhamba kwesikhathi, ikakhulukazi i-K ne-Mg ezibonisa indawo eshisayo enhlabathini. Izindawo ezishisayo eziphakathi nendawo enyakatho-ntshonalanga yemephu naseningizimu-mpumalanga yemephu. Ukuqina kwe-Colloidal enhlabathini kunciphisa ukugcwala kwe-magnesium enhlabathini. Ukuntuleka kwayo enhlabathini kubangela izitshalo ukuthi zibonise i-chlorosis ephuzi ye-intervein. Umanyolo osekelwe ku-magnesium, njenge-potassium magnesium sulfate, i-magnesium sulfate, ne-Kieserite, welapha ukuntuleka (izitshalo zibonakala zinsomi, zibomvu, noma zinsundu, okubonisa ukuntuleka kwe-magnesium) enhlabathini ene-pH evamile6. Ukuqongelela kwe-nickel ezindaweni zomhlabathi zasemadolobheni nasezindaweni zasemadolobheni kungase kube ngenxa yemisebenzi yabantu njengezolimo kanye nokubaluleka kwe-nickel ekukhiqizweni kwensimbi engagqwali78.
Ukusatshalaliswa kwendawo kwezinto [imephu yokusatshalaliswa kwendawo yadalwa kusetshenziswa i-ArcGIS Desktop (ESRI, Inc, Version 10.7, URL: https://desktop.arcgis.com).]
Imiphumela yenkomba yokusebenza kwemodeli yezinto ezisetshenziswe kulolu cwaningo iboniswe kuThebula 2. Ngakolunye uhlangothi, i-RMSE kanye ne-MAE ye-Ni zombili ziseduze no-zero (0.86 RMSE, -0.08 MAE). Ngakolunye uhlangothi, womabili amanani e-RMSE kanye ne-MAE e-K ayamukeleka. Imiphumela ye-RMSE kanye ne-MAE yayinkulu nge-calcium kanye ne-magnesium. Imiphumela ye-Ca kanye ne-K MAE kanye ne-RMSE mikhulu ngenxa yamasethi edatha ahlukene. I-RMSE kanye ne-MAE yalolu cwaningo esebenzisa i-EBK ukubikezela i-Ni kutholakale ukuthi ingcono kunemiphumela kaJohn et al. 54 besebenzisa i-synergistic kriging ukubikezela amazinga e-S enhlabathini besebenzisa idatha efanayo eqoqwe. Imiphumela ye-EBK esiyifundile ihlobene naleyo kaFabijaczyk et al. 41, Yan et al. 79, Beguin et al. 80, Adhikary et al. 81 kanye noJohn et al. 82, ikakhulukazi i-K kanye ne-Ni.
Ukusebenza kwezindlela ngazinye zokubikezela okuqukethwe yi-nickel enhlabathini yasemadolobheni kanye neyasemadolobheni kuhlolwe kusetshenziswa ukusebenza kwamamodeli (Ithebula 3). Ukuqinisekiswa kwemodeli kanye nokuhlolwa kokunemba kuqinisekisile ukuthi isibikezeli se-Ca_Mg_K esihlanganiswe nemodeli ye-EBK SVMR siveze ukusebenza okuhle kakhulu. Imodeli yokulinganisa i-Ca_Mg_K-EBK_SVMR imodeli R2, iphutha lesikwele le-root mean (RMSE) kanye nephutha eliphelele eliphakathi (MAE) kwakungu-0.637 (R2), 95.479 mg/kg (RMSE) kanye no-77.368 mg/kg (MAE) Ca_Mg_K-SVMR kwakungu-0.663 (R2), 235.974 mg/kg (RMSE) kanye no-166.946 mg/kg (MAE). Noma kunjalo, amanani amahle e-R2 atholakale ku-Ca_Mg_K-SVMR (0.663 mg/kg R2) kanye ne-Ca_Mg-EBK_SVMR (0.643 = R2); imiphumela yabo ye-RMSE kanye ne-MAE yayiphakeme kunaleyo ye-Ca_Mg_K-EBK_SVMR (R2 0.637) (bheka Ithebula 3). Ngaphezu kwalokho, i-RMSE kanye ne-MAE yemodeli ye-Ca_Mg-EBK_SVMR (RMSE = 1664.64 kanye ne-MAE = 1031.49) ingu-17.5 kanye no-13.4, ngokulandelana, emikhulu kunaleyo ye-Ca_Mg_K-EBK_SVMR. Ngokufanayo, i-RMSE kanye ne-MAE yemodeli ye-Ca_Mg-K SVMR (RMSE = 235.974 kanye ne-MAE = 166.946) inkulu ngo-2.5 ​​kanye no-2.2 kunaleyo ye-Ca_Mg_K-EBK_SVMR RMSE kanye ne-MAE, ngokulandelana. Imiphumela ye-RMSE ebaliwe ikhombisa ukuthi isethi yedatha igxile kangakanani emgqeni wokulingana okungcono kakhulu. I-RSME ephezulu kanye ne-MAE yabonwa. Ngokusho UKebonye et al. 46 kanye noJohn et al. 54, lapho i-RMSE kanye ne-MAE zisondela ku-zero, imiphumela iba ngcono. I-SVMR kanye ne-EBK_SVMR zinamanani aphezulu e-RSME kanye ne-MAE alinganisiwe. Kwabonwa ukuthi izilinganiso ze-RSME beziphakeme njalo kunezilinganiso ze-MAE, okubonisa ukuba khona kwezinto ezingaphandle. Ngokusho kukaLegates noMcCabe83, izinga lapho i-RMSE idlula khona i-mean absolute error (MAE) linconywa njengesibonakaliso sokuba khona kwezinto ezingaphandle. Lokhu kusho ukuthi lapho isethi yedatha ingafani kakhulu, kulapho amanani e-MAE kanye ne-RMSE aphezulu khona. Ukunemba kokuhlolwa kokuqinisekiswa kwemodeli exubile ye-Ca_Mg_K-EBK_SVMR yokubikezela okuqukethwe kwe-Ni ezindaweni zasemadolobheni kanye nasezindaweni zasemadolobheni kwakungama-63.70%. Ngokusho kukaLi et al. 59, leli zinga lokunemba liyisilinganiso sokusebenza kwemodeli esamukelekayo. Imiphumela yamanje iqhathaniswa nocwaningo lwangaphambilini lukaTarasov et al. 36 imodeli yabo ehlanganisiwe eyadala i-MLPRK (Multilayer Perceptron Residual Kriging), ehlobene nenkomba yokuhlola ukunemba kwe-EBK_SVMR ebikwe ocwaningweni lwamanje, i-RMSE (210) kanye ne-MAE (167.5) yayiphezulu kunemiphumela yethu ocwaningweni lwamanje (RMSE 95.479, MAE 77.368). Kodwa-ke, uma kuqhathaniswa i-R2 yocwaningo lwamanje (0.637) nekaTarasov et al. 36 (0.544), kusobala ukuthi i-coefficient of determination (R2) iphakeme kule modeli exubile. Umkhawulo wephutha (RMSE kanye ne-MAE) (EBK SVMR) yemodeli exubile uphansi kabili. Ngokufanayo, uSergeev et al.34 babhale u-0.28 (R2) wemodeli ehlanganisiwe ethuthukisiwe (i-Multilayer Perceptron Residual Kriging), kanti u-Ni ocwaningweni lwamanje babhale u-0.637 (R2). Izinga lokunemba kokubikezela kwale modeli (EBK SVMR) lingu-63.7%, kanti ukunemba kokubikezela okutholwe nguSergeev et al. 34 kungu-28%. Imephu yokugcina (Umfanekiso 5) edalwe kusetshenziswa imodeli ye-EBK_SVMR kanye ne-Ca_Mg_K njengesibikezeli ikhombisa izibikezelo zezindawo ezishisayo futhi eziphakathi nendawo kuya endaweni yonke yokufunda. Lokhu kusho ukuthi ukuhlushwa kwe-nickel endaweni yokufunda kuphakathi kakhulu, kanye nokuhlushwa okuphezulu kwezinye izindawo ezithile.
Imephu yokugcina yokubikezela imelelwa kusetshenziswa imodeli ye-hybrid EBK_SVMR futhi kusetshenziswa i-Ca_Mg_K njengesibikezeli.[Imephu yokusatshalaliswa kwendawo yadalwa kusetshenziswa i-RStudio (inguqulo 1.4.1717: https://www.rstudio.com/).]
Okwethulwe kuMfanekiso 6 ukuhlushwa kwe-PTE njengendilinga yokwakheka equkethe ama-neurons ngamanye. Akukho neyodwa yendilinga yezingxenye ebonise iphethini yombala efanayo njengoba kuboniswe. Kodwa-ke, inani elifanele lama-neurons ngemephu edwetshiwe lingu-55. I-SeOM ikhiqizwa kusetshenziswa imibala ehlukahlukene, futhi uma amaphethini ombala afana kakhulu, kulapho izakhiwo zamasampula zifana khona. Ngokwesilinganiso sazo sombala esiqondile, izakhi ngazinye (i-Ca, i-K, ne-Mg) zibonise amaphethini ombala afanayo nama-neurons aphezulu kanye nama-neurons amaningi aphansi. Ngakho-ke, i-CaK ne-CaMg zabelana ngokufana okuthile nama-neurons asezingeni eliphezulu kakhulu kanye namaphethini ombala aphansi kuya kwaphakathi. Womabili amamodeli abikezela ukuhlushwa kwe-Ni enhlabathini ngokubonisa imibala ephakathi kuya phezulu yemibala efana nobomvu, i-orenji kanye nophuzi. Imodeli ye-KMg ibonisa amaphethini amaningi ombala aphezulu ngokusekelwe ezilinganisweni eziqondile kanye nama-patches ombala aphansi kuya kwaphakathi. Esikalini esiqondile sombala kusukela phansi kuya phezulu, iphethini yokusabalalisa ehleliwe yezingxenye zemodeli ibonise iphethini yombala ephezulu ekhombisa ukuhlushwa okungenzeka kwe-nickel enhlabathini (bheka uMfanekiso 4). Indilinga yendilinga yemodeli ye-CakMg ikhombisa iphethini yombala ehlukahlukene kusukela phansi kuya phezulu ngokusho kombala onembile. isikali.Ngaphezu kwalokho, ukubikezela komodeli kokuqukethwe kwe-nickel (i-CakMg) kufana nokusatshalaliswa kwendawo kwe-nickel okuboniswe kuMfanekiso 5.Womabili amagrafu abonisa isilinganiso esiphezulu, esiphakathi nesiphansi sokugcwala kwe-nickel enhlabathini yasemadolobheni kanye neyasemadolobheni.Isithombe 7 sibonisa indlela ye-contour ekuqoqweni kwe-k-means kumephu, ihlukaniswe ngamaqoqo amathathu ngokusekelwe enanini elibikezelwe kumodeli ngayinye.Indlela ye-contour imelela inani elifanele lamaqoqo.Kumasampula enhlabathi angu-115 aqoqwe, isigaba 1 sithole amasampula enhlabathi amaningi, angu-74.I-Cluster 2 ithole amasampula angu-33, kanti i-cluster 3 ithole amasampula angu-8.Inhlanganisela yokubikezela ye-planar enezingxenye eziyisikhombisa yenziwe lula ukuvumela ukuchazwa kweqembu okulungile.Ngenxa yezinqubo eziningi ze-anthropogenic kanye nezemvelo ezithinta ukwakheka kwenhlabathi, kunzima ukuba namaphethini eqembu ahlukaniswe kahle kumephu ye-SeOM esatshalalisiwe78.
Ukukhishwa kwendiza yengxenye yi-Empirical Bayesian Kriging Support Vector Machine (EBK_SVM_SeOM) variable ngayinye.[Amamephu e-SeOM adalwe kusetshenziswa i-RStudio (inguqulo 1.4.1717: https://www.rstudio.com/).]
Izingxenye ezahlukene zokuhlukaniswa kwamaqoqo [Amamephu e-SeOM adalwe kusetshenziswa i-RStudio (inguqulo 1.4.1717: https://www.rstudio.com/).]
Ucwaningo lwamanje lubonisa ngokucacile amasu okulingisa amazinga e-nickel enhlabathini yasemadolobheni kanye neyasemadolobheni. Ucwaningo luhlole amasu ahlukene okulingisa, luhlanganisa izakhi namasu okulingisa, ukuze kutholakale indlela engcono kakhulu yokubikezela amazinga e-nickel enhlabathini. Izici zesikhala se-SeOM compositional planar technique zohlelo lokulingisa zibonise iphethini yombala ophezulu kusukela phansi kuya phezulu esikalini sombala esinembile, okubonisa amazinga e-Ni enhlabathini. Kodwa-ke, imephu yokusabalalisa indawo iqinisekisa ukusatshalaliswa kwendawo kwezingxenye eziboniswe yi-EBK_SVMR (bheka Umfanekiso 5). Imiphumela ikhombisa ukuthi imodeli yokusekela ye-vector machine regression (Ca Mg K-SVMR) ibikezela amazinga e-Ni enhlabathini njengemodeli eyodwa, kodwa imingcele yokuhlola yokuqinisekisa nokunemba ikhombisa amaphutha aphezulu kakhulu ngokwe-RMSE kanye ne-MAE. Ngakolunye uhlangothi, indlela yokulingisa esetshenziswa nemodeli ye-EBK_MLR nayo inephutha ngenxa yenani eliphansi le-coefficient of determination (R2). Imiphumela emihle itholakale kusetshenziswa i-EBK SVMR kanye nezinto ezihlanganisiwe (CaKMg) ngamaphutha aphansi e-RMSE kanye ne-MAE ngokunemba okungu-63.7%. Kuvela ukuthi ukuhlanganisa i-EBK I-algorithm ene-algorithm yokufunda komshini ingakhiqiza i-algorithm ehlanganisiwe engabikezela ukuhlushwa kwama-PTE enhlabathini. Imiphumela ikhombisa ukuthi ukusebenzisa i-Ca Mg K njengezibikezeli ukubikezela ukuhlushwa kwe-Ni endaweni yokucwaninga kungathuthukisa ukubikezela kwe-Ni enhlabathini. Lokhu kusho ukuthi ukusetshenziswa okuqhubekayo komanyolo osuselwa ku-nickel kanye nokungcola kwezimboni kwenhlabathi yimboni yensimbi kuthambekele ekwandiseni ukuhlushwa kwe-nickel enhlabathini. Lolu cwaningo luveze ukuthi imodeli ye-EBK inganciphisa izinga lamaphutha futhi ithuthukise ukunemba kwemodeli yokusatshalaliswa kwendawo yenhlabathi enhlabathini yasemadolobheni noma eseduze nedolobha. Ngokuvamile, siphakamisa ukusebenzisa imodeli ye-EBK-SVMR ukuhlola nokubikezela i-PTE enhlabathini; ngaphezu kwalokho, siphakamisa ukusebenzisa i-EBK ukuze ixube ngama-algorithm ahlukahlukene okufunda komshini. Ukuhlushwa kwe-Ni kwabikezelwa kusetshenziswa izakhi njenge-covariates; noma kunjalo, ukusebenzisa ama-covariates amaningi kuzothuthukisa kakhulu ukusebenza kwemodeli, okungabhekwa njengomkhawulo womsebenzi wamanje. Omunye umkhawulo walolu cwaningo ukuthi inani lamasethi edatha lingu-115. Ngakho-ke, uma kunikezwa idatha eyengeziwe, ukusebenza kwendlela ephakanyisiwe ye-hybridization ethuthukisiwe kungathuthukiswa.
PlantProbs.net. I-Nickel Ezitshalweni Nasemhlabathini https://plantprobs.net/plant/nutrientImbalances/sodium.html (Kufinyelelwe ngo-28 Ephreli 2021).
UKasprzak, KS Nickel uthuthuka ku-toxicology yesimanje yezemvelo. surroundings.toxicology.11, 145–183 (1987).
I-Cempel, M. & Nikel, G. Nickel: Ukubuyekezwa kwemithombo yayo kanye nobuthi bemvelo. I-Polish J. Environment.Stud.15, 375–382 (2006).
Freedman, B. & Hutchinson, TC Ukungena kokungcola okuvela emkhathini kanye nokuqongelela enhlabathini nasezitshalweni eduze kwesincibilikisi se-nickel-copper eSudbury, e-Ontario, eCanada.can.J. Bot.58(1), 108-132.https://doi.org/10.1139/b80-014 (1980).
UManyiwa, T. nabanye. Izinsimbi ezisindayo enhlabathini, ezitshalweni kanye nezingozi ezihlobene nezilwane ezidla imfuyo eduze kwemayini yethusi ne-nickel yaseSelebi-Phikwe eBotswana. surroundings.Geochemistry.Health https://doi.org/10.1007/s10653-021-00918-x (2021).
I-Cabata-Pendias. I-Kabata-Pendias A. 2011. Izakhi zokulandelela enhlabathini kanye… – I-Google Scholar https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=I-Kabata-Pendias+A.+2011.+Trace+ Elements+in+soils+and+plants.+4th+ed.+New+York+%28NY%29%3A+CRC+Press&btnG= (Kufinyelelwe ngo-24 Nov 2020).
U-Almås, A., Singh, B., Ezolimo, TS-NJ ka-& 1995, engachazwanga. Imiphumela yemboni ye-nickel yaseRussia ekugxilweni kwensimbi esindayo enhlabathini yezolimo kanye notshani eSoer-Varanger, eNorway.agris.fao.org.
UNielsen, GD nabanye. Ukumuncwa kwe-nickel kanye nokugcinwa kwayo emanzini okuphuza kuhlobene nokudla kanye nokuzwela kwe-nickel.toxicology.application.Pharmacodynamics.154, 67–75 (1999).
UCosta, M. & Klein, CB Nickel carcinogenesis, ukuguqulwa kwezakhi zofuzo, i-epigenetics noma ukukhetha.okuzungezile.Umbono Wezempilo.107, 2 (1999).
U-Ajman, PC; U-Ajado, SK; UBorůvka, L.; UBini, JKM; USarkody, VYO; UCobonye, ​​​​NM; Ukuhlaziywa kwethrendi yezinto ezingaba yingozi: isibuyekezo se-bibliometric. I-Environmental Geochemistry and Health. I-Springer Science & Business Media BV 2020. https://doi.org/10.1007/s10653-020-00742-9.
UMinasny, B. kanye noMcBratney, AB Ukumapha Inhlabathi Yedijithali: Umlando Omfushane kanye Nezifundo Ezithile.Geoderma 264, 301–311.https://doi.org/10.1016/j.geoderma.2015.07.017 (2016).
McBratney, AB, Mendonça Santos, ML & Minasny, B. Kumephu yenhlabathi yedijithali.Geoderma 117(1-2), 3-52.https://doi.org/10.1016/S0016-7061(03)00223-4 (2003).
Ukumodela kwe-Deutsch.CV Geostatistical Reservoir,… – I-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= (Kufinyelelwe ngo-28 Ephreli 2021).