Naslov (eng)

Spatio-temporal interpolation of climate elements using geostatistics and machine learning: doctoral dissertation

Autor

Sekulić, Aleksandar M., 1991-, 67908361

Doprinosi

Kilibarda, Milan, 1983-, 20083559
Bajat, Branislav, 1963-, 12617575
Luković, Jelena, 1979-, 5440103
Pejović, Milutin, 1983-, 67911177
Nikolić, Mladen, 1981-, 13483111

Opis (srp)

Gridovani podaci dnevnih klimatskih elemenata visoke rezolucije predstavljaju znacajan izvor in- ˇ formacija koje se koriste kao ulazni podaci za analize u klimatologiji, meteorologiji, poljoprivredi, hidrologiji, ekologiji i ostalim istraziva ˇ ckim oblastima i disciplinama. Prostorno-vremenske inter- ˇ polacione metode cesto se koriste za kreiranje gridovanih dnevnih klimatskih elemenata. Glob- ˇ alni model prostorno-vremenskog regresionog kriginga za srednje dnevne temperature iznad povrsi ˇ Zemlje je pojednostavljen koristeci samo geometrijski temperaturni trend, digitalni model terena ´ i topografski indeks vlaznosti (bez MODIS LST snimaka) kao prediktore i kalibrisan za podru ˇ cje ˇ Hrvatske koristeci podatke iz 2008 godine u ovoj disertaciji. Na osnovu prostorne kros-validacije, ´ tacnost kalibrisanog modela iznosi R ˇ 2=97.8% i RMSE=1.2 ◦C, sto je pobolj ˇ sanje od 3.4% i 0.7 ˇ ◦C u odnosu na globalni model. Prilagodeni model srednjih dnevnih temperatura nadmasuje ostale ve ˇ c´ razvijene modele za podrucje Hrvatske u pogledu ta ˇ cnosti i ima sli ˇ cnu ili ve ˇ cu ta ´ cnost u odnosu na ˇ modele za druga lokalna podrucja ili dr ˇ zave. Rezultati pokazuju da se globalni model prostorno- ˇ vremenskog regresionog kriginga moze prilagoditi lokalnim podru ˇ cjima koriste ˇ ci mre ´ zu nacional- ˇ nih meteoroloskih stanica i tako proizvesti gridovane podatke srednjih dnevnih temperatura ve ˇ ce ´ tacnosti sa prostornom rezolucijom od 1 km. Kalibrisani model za podru ˇ cje Hrvatske jo ˇ s uvek ima ˇ manju tacnost u planinskim predelima, ˇ sto ga ˇ cini pogodnim za primenu u poljoprivrednim po- ˇ drucjima koja su na ni ˇ zim nadmorskim visinama. ˇ Algoritmi masinskog u ˇ cenja kombinovani sa inovativnim prostornim prediktorima predstavljaju ˇ novi oblik modela za prostornu ili prostorno-vremensku interpolaciju, koji mogu da se koriste i za interpolaciju klimatskih elemenata. U ovoj disertaciji je predstavljena i testirana inovativna Random Forest Spatial Interpolation (RFSI) metodologija za prostornu ili prostorno-vremensku interpolaciju. RFSI metodologija je bazirana na Random Forest algoritmu masinskog u ˇ cenja koji koristi inovativne ˇ prostorne prediktore: opazanja na ˇ n najblizih lokacija i rastojanja do njih. RFSI metodologija je ˇ primenjena i testirana na tri studije slucaja. U prvoj sinteti ˇ ckoj studiji, koja predstavlja simulirani ˇ set podataka, tacnost RFSI metodologije je pore ˇ dena sa tacno ˇ sˇcu obi ´ cnog kriging-a, ˇ Random Forest for spatial prediction (RFsp) metode, metode inverznih distanci (eng. inverse distance weighting), najblizeg suseda (eng. ˇ nearest neighbour) i mapiranja povrsi trenda (eng. ˇ trend surface mapping). U ovom slucaju, RFSI je pokazao ve ˇ cu ta ´ cnost u pore ˇ denju sa metodama najblizeg suseda i mapiranja ˇ povrsi trenda i sli ˇ cnu ta ˇ cnost kao RFsp i metoda inverznih distanci. Obi ˇ cni kriging je o ˇ cekivano dao ˇ bolje rezultate od RFSI metodologije iz razloga sto je simulirani set podataka kreiran geostatisti ˇ ckom ˇ simulacijom i samim tim obicni kriging predstavlja optimalnu metodu interpolacije u ovom slu ˇ caju. ˇ U ostale dve studije slucaja, koje se odnose na dnevne koli ˇ cine padavina za podru ˇ cje Katalonije ˇ za 2016–2018 period i srednje dnevne temperature za podrucje Hrvatske za 2008 godinu, ta ˇ cnost ˇ ix UNIVERZITET U BEOGRADU Gradevinski fakultet Odsek za geodeziju i geoinformatikuRFSI metodologije je poredena sa tacno ˇ sˇcu prostorno-vremenskog regresionog kriginga, metode in- ´ verznih distanci, standardnom Random Forest i RFsp metodom koristeci ugnje ´ zdenu prostornu kros- ˇ validaciju. RFSI metodologija je pokazala najbolje rezultate u ovim studijama. RFSI metodologija se preporucuje za interpolaciju slo ˇ zenih parametara zbog osobine ˇ Random Forest algoritma da moze da ˇ modelira nelinearne veze izmedu prediktora i modeliranog parametra. RFSI metodologija se takode moze koristiti za prostornu ili prostorno-vremensku interpolaciju bilo kog drugog parametra ˇ zivotne ˇ sredine. Koristeci RFSI metodologiju za prostorno-vremensku interpolaciju, kreiran je MeteoSerbia1km ´ set podataka koji predstavlja prvi set gridovanih dnevnih klimatskih elemenata (maksimalne, minimalne i srednje temperature, atmosferskog pritiska na nivou mora i kolicine padavina) sa pros- ˇ tornom rezolucijom od 1 km za podrucje Srbije za period 2000–2019. Agregacijom dnevnih gri- ˇ dovanih podataka dodatno su kreirani gridovani podaci mesecnih i godi ˇ snjih proseka (ukupne ˇ kolicine za padavine) i gridovani podaci dnevnih, mese ˇ cnih i godi ˇ snjih dugoro ˇ cnih proseka kli- ˇ matskih elemenata. Tacnost dnevnih MeteoSerbia1km gridovanih podaka je ocenjena pomo ˇ cu ´ ugnjezdene prostorne kros-validacije. Ta ˇ cnost dnevnih temperatura i atmosferskog pritiska na ˇ nivou mora je visoka, dok je tacnost dnevnih padavina o ˇ cekivano ne ˇ sto manja zbog slo ˇ zenosti samih ˇ padavina. Dnevni MeteoSerbia1km gridovani podaci su takode poredeni sa E-OBS setom dnevnih gridovanih podataka sa prostornom rezolucijom od 10 km i pokazuju visok stepen korelacije, osim za padavine. RFSI metodolgija je automatizovana i implementirana u okviru R paketa meteo, kroz cetiri ˇ nove R funkcije za procese kreiranja, predikcije, kalibrisanja i kros-validacije RFSI modela.

Opis (eng)

High resolution daily maps for climate elements are a valuable source of information and serve as an input for climatology, meteorology, agriculture, hydrology, ecology, and many other research areas and disciplines. Spatio-temporal interpolation methods are o‰en used for creation of daily maps for climate elements. In this research, already existing spatio-temporal geostatistical interpolation methods and newly developed spatio-temporal interpolation methods based on machine learning algorithms are applied to and evaluated on climate element case studies. A spatio-temporal regression kriging model for global land areas for mean daily temperature is simpli€ed by using only a geometric temperature trend, digital elevation model, and topographic wetness index (without MODIS LST) as covariates and adapted for Croatian territories for the year 2008 in this dissertation. ‘e leave-one-out and 5-fold cross-validation show that the accuracy of the model a‰er adaptation is 97.8% in R2 and 1.2 ◦C in RMSE, which is an improvement of 3.4% in R2 and 0.7 ◦C in RMSE. ‘e adapted daily mean temperature model also outperforms previously developed models for Croatia and shows similar or be‹er accuracy in comparison with models for other local areas. ‘e results show that the spatio-temporal regression kriging model for global land areas can be adapted to local areas using a national weather station network, thus providing more accurate daily mean temperature maps at a 1 km spatial resolution. ‘e proposed adapted geostatistical model for Croatia still provides larger prediction errors in mountainous regions making it convenient for application in agricultural areas that are at lower altitudes. A di‚erent approach to spatial or spatio-temporal interpolation of climate elements is to use machine learning algorithms together with spatial covariates. A novel Random Forest Spatial Interpolation (RFSI) methodology for spatial or spatio-temporal interpolation is proposed and evaluated in this dissertation. ‘e RFSI methodology is based on the Random Forest algorithm that uses innovative spatial predictors: observations at n nearest locations and distances to them. ‘e RFSI methodology is applied and evaluated in three case studies. In the €rst, a synthetic (simulated) case study, the accuracy of RFSI is compared with the accuracy of ordinary kriging, Random Forest for spatial prediction (RFsp), inverse distance weighting, nearest neighbour, and trend surface mapping interpolation methods. In this case study, RFSI outperforms nearest neighbour and trend surface mapping and has similar accuracy as RFsp and inverse distance weighting. RFSI is outperformed by ordinary kriging because this case study is created by geostatistical simulation and consequentially ordinary kriging is an optimal interpolation method in this case. In the following two real-world case studies, a daily precipitation for Catalonia for the 2016–2018 period and a daily mean temperature for Croatia for the year 2008, the accuracy of RFSI is compared with the accuracy of spatio-temporal regression kriging, inverse distance weighting, standard Random Forest and RFsp using a nested vii UNIVERSITY OF BELGRADE Faculty of Civil Engineering Department of Geodesy and Geoinformaticsk-fold leave-location-out cross-validation and RFSI outperformed all of them. RFSI is recommended for the interpolation of complex variables due to Random Forest’s ability to model non-linear relations between covariates and target variables. RFSI can be used for spatial or spatio-temporal interpolation of any environmental variable. Next, a MeteoSerbia1km dataset — a €rst gridded dataset for daily climate elements (maximum, minimum, and mean temperature, mean sea level pressure, and total precipitation) at a 1 km spatial resolution for Serbian territories for the 2000–2019 period — is created using RFSI methodology for spatio-temporal interpolation. Additionally, monthly and annual summaries and daily, monthly, and annual long term means maps of the climate elements are generated by aggregating the daily MeteoSerbia1km maps. ‘e nested 5-fold leave-location-out cross-validation is used to access the accuracy of the MeteoSerbia1km daily dataset. ‘e accuracy is high for daily temperature variables and sea level pressure and lower for daily precipitation which was expected due to its complexity. MeteoSerbia1km daily maps are further compared with the 10-km E-OBS daily maps and show high correlation with them except for daily precipitation. ‘e automation of the RFSI methodology is implemented within the R package meteo, in the form of four new R functions for creation, prediction, tuning, and cross-validation processes of RFSI model.

Opis (eng)

Geodesy - Modelling and management in geodesy / Geodezija - Modeliranje i menadzment u geodeziji Datum odbrane: 09.04.2021.

Jezik

engleski

Datum

2021

Licenca

Ovo delo je licencirano pod uslovima licence
Creative Commons CC BY-NC-ND 2.0 AT - Creative Commons Autorstvo - Nekomercijalno - Bez prerada 2.0 Austria License.

http://creativecommons.org/licenses/by-nc-nd/2.0/at/legalcode

Predmet

OSNO - Opšta sistematizacija naučnih oblasti, Geodezija

prostorno-vremenska interpolacija, masinsko u ˇ cenje, random forest, RFSI, dnevne ˇtemperature, dnevne padavine, MeteoSerbia1km, R, meteo

OSNO - Opšta sistematizacija naučnih oblasti, Geodezija

spatio-temporal interpolation, kriging, machine learning, random forest, RFSI, dailytemperature, daily precipitation, MeteoSerbia1km, R, meteo