Naslov
Inverse kriging - a new “old” method for improvement of spatial data quality

Autori
Hackenberger Kutuzović, Domagoj ; Ćaleta, Bruno ; Đerđ, Tamara ; Hackenberger Kutuzović, Branimir

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

Izvornik
Environmental Assessments and the European Green Deal '22 - Book of Abstracts / Antonić, O. ; Mikulić, N. ; Celinšćak, M. - Zagreb : Croatian Association of Experts in Nature and Environmental Protection, Zagreb, Croatia, 2022, 122-122

Skup
European and regional conference „Environmental assessments and the European Green Deal '22“

Mjesto i datum
Vodice, Hrvatska, 14.09.2022. - 17.09.2022

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Domaća recenzija

Ključne riječi
sub-sampling ; UAVs ; satellite data ; GAN

Sažetak
Kriging is the common name for a group of methods used to estimate values between sampling sites (interpolation) and prediction outside of the extent of the sampling area (extrapolation) based on a relatively small amount of data. Back in 2002, the so-called reverse kriging method was published. This method was supposed to obtain higher quality data based on a certain sub-sampling in the area modeled by kriging. Unfortunately, the existing reverse kriging method only partially satisfies the need for an artificial increase in spatial data resolution and has rarely been applied so far. Although similar algorithms are used in image sharpening, they do not serve to obtain finer detail in space because these algorithms are based on tracking common extremes of pixel values. Reverse kriging is especially important in the fine structuring of spatial data related to climate change, such as temperatures, greenhouse gas concentrations, etc. It is crucial in increasing the resolution of data that is difficult or rare to measure. However, the most efficient application of reverse kriging is by using a combination of low resolution (satellite) and high resolution (drones) remote sensing data. To increase the efficiency of reverse kriging, we used algorithms that include Generative Adversarial Networks (GAN). In this presentation, we present that this algorithm can be used to optimize subsampling and obtain the highest possible resolution by the given remote sensing parameters.

Izvorni jezik
Engleski

Znanstvena područja
Interdisciplinarne prirodne znanosti

POVEZANOST RADA