Naslov
Toward to development of the new national climate monitoring products

Autori
Perčec Tadić, Melita

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, stručni

Izvornik
Geostatistics for Environmental Applications – geoENV 2016. Book of Abstracts. / Soares, Amílcar ; Ribeiro, Manuel ; Quintão Maria João ; Pereira, Maria João - Lisabon : Instituto Superior Técnico, Universidade de Lisboa, 2016, 9-9

ISBN
978-989-98342-7-9

Skup
Geostatistics for Environmental Applications – geoENV 2016.

Mjesto i datum
Lisabon, Portugal, 06.07.2016. - 08.07.2016

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Nije recenziran

Ključne riječi
climate service, monitoring products

Sažetak
Climate services provide climate information to the public and professionals in different sectors. Those sectors can range from health sector where information about prolonged extreme climate event like heat wave or cold is important, or the information about the prolonged drought is important in energy sector for power consumption planning in near or more distant future. National meteorological services provide climate products based on data from the national weather network, while different methods or background information is used to provide the estimates for the locations where there are no measurements. A range of spatial or spatial-temporal interpolation techniques (splines, regression, kriging, neural networks and machine learning techniques) have been used for the interpolation of meteorological data to produce maps. Applications are mostly on the monthly or annual time scale because climate fields are much easier to model than weather situations since they are more spatially homogenous. Also, it has been shown that most of the techniques used for the interpolation of the climatic variables perform better if auxiliary grids are used. The predictors can be morphological like elevation from the DEM, exposition, distance from the coast-line, but also time dependent remote sensed images like land surface temperature or satellite estimated precipitation rate. Benefit of using auxiliary information is available in several almost identical interpolation techniques called kriging with external drift, universal kriging , or regression kriging hence we will explore their performance on the monthly temperature and precipitation and with some new predictors in the R computational framework. The advantages of machine learning will be examined in this particular case with the randomForest R package. The methods are compared through cross-validation.

Izvorni jezik
Engleski

Znanstvena područja
Fizika, Geologija

POVEZANOST RADA