Pregled bibliografske jedinice broj: 1198973
Agroecological and pedological modeling of cropland suitability for soybean cultivation by integrating satellite remote sensing data and machine learning
Agroecological and pedological modeling of cropland suitability for soybean cultivation by integrating satellite remote sensing data and machine learning, 2022., doktorska disertacija, Osijek
CROSBI ID: 1198973 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Agroecological and pedological modeling of
cropland suitability for soybean cultivation by
integrating satellite remote sensing data and
machine learning
Autori
Radočaj, Dorijan
Vrsta, podvrsta i kategorija rada
Ocjenski radovi, doktorska disertacija
Mjesto
Osijek
Datum
21.03
Godina
2022
Stranica
92
Mentor
Jurišić, Mladen ; Antonić, Oleg
Ključne riječi
geographic information system ; vegetation index ; biophysical variables ; automation ; Sentinel-2
Sažetak
Increased global food demand pressured farmers into producing higher crop yields to keep track of the population growth and increased life standard. The standard method of calculating cropland suitability in previous studies was GIS-based multicriteria analysis, most often combined with the Analytic Hierarchy Process (AHP). While this is a flexible and widely accepted procedure, it has major fundamental disadvantages in a cropland suitability determination, including lack of accuracy assessment, high human subjectivity and computational inefficiency in case of complex input data. To improve these disadvantages, a reliable, objective and computationally efficient procedure for determining the cropland suitability, consisting of: 1) a computationally efficient validation method of cropland suitability using global satellite missions in high (Sentinel-2) and medium spatial resolution (PROBA-V) ; 2) an automatization method of spatial modeling of abiotic criteria for the example of soil texture according to a globally accepted standard ; 3) suitability prediction method based on machine learning algorithms and globally available spatial data, which allows high reliability of prediction with reduced user subjectivity compared to GIS-based multicriteria analysis. The proposed methods add to an important paradigm shift in a cropland suitability determination, which focuses on the objective, automated and computationally efficient prediction method, as well as ensuring reliable validation data from open data sources on a global scale.
Izvorni jezik
Engleski
Znanstvena područja
Biologija, Poljoprivreda (agronomija), Interdisciplinarne biotehničke znanosti