Találati lista | Tudóstér

001-es BibID:	BIBFORM127416
Első szerző:	Phinzi, Kwanele
Cím:	Improving the spatial prediction of topsoil properties in a typical grazing area using multi-season PlanetScope spectral covariates and data mining techniques / Kwanele Phinzi, László Bertalan, Gashaw Gismu Chakilu, Szilárd Szabó
Dátum:	2025
ISSN:	1865-0473 1865-0481
Megjegyzések:	Understanding the spatial distribution of topsoil properties in grassland ecosystems is essential for improving soil ecosystem services, quality, and erosion resilience. The availability of free, high-resolution satellite imagery and advanced data mining techniques offers new opportunities for efficient soil property assessment. This study aimed to evaluate the potential utility of multi-season PlanetScope imagery to predict soil organic carbon (SOC), pH, and calcium carbonate (CaCO3). Using random sampling, 121 topsoil samples (0-30 cm depth) were collected with an auger across grasslands, bare soil, and eroded areas within a typical grazing land use. Three data mining techniques: random forest (RF), extreme gradient boosting (XGB), and support vector machines (SVM), were applied and evaluated using a 10-fold cross-validation. The results indicated that multi-season spectral covariates considerably improved the accuracy of the target soil properties compared to single-season imagery. SVM was the most effective algorithm for predicting SOC, achieving a root mean square error (RMSE) of 0.52%, mean absolute error (MAE) of 0.24%, and R² of 0.92. RF was the best-performing algorithm for predicting soil pH (RMSE=0.22, MAE=0.17, and R² = 0.97) and CaCO3 (RMSE=0.55%, MAE=0.42%, and R² = 0.96). While XGB failed to capture the variability in soil pH, the other models generated interpretable maps that accurately represented the distribution of soil properties across different land cover categories. The green-red vegetation index (GRVI) was the most critical covariate for predicting SOC, while elevation and the topographic wetness index (TWI) were key predictors for soil pH and CaCO3, respectively. This study underscores the potential of multi-season PlanetScope imagery for accurately predicting soil properties and recommends conducting similar studies in diverse geographical settings to validate these findings and develop more generalizable models.
Tárgyszavak:	Természettudományok Földtudományok idegen nyelvű folyóiratközlemény külföldi lapban folyóiratcikk Digital soil mapping Machine learning Multispectral Remote sensing Soil property prediction
Megjelenés:	Earth Science Informatics. - 18 : 2 (2025), p. 222. -
További szerzők:	Bertalan László (1989-) (geográfus) Chakilu, Gashaw Gismu Szabó Szilárd (1974-) (geográfus)
Internet cím:	Szerző által megadott URL DOI Intézményi repozitóriumban (DEA) tárolt változat
Borító:
Saját polcon:

001-es BibID:	BIBFORM115323
035-os BibID:	(WoS)001086503900002 (Scopus)85174061140
Első szerző:	Phinzi, Kwanele
Cím:	Understanding the role of training sample size in the uncertainty of high-resolution LULC mapping using random forest / Kwanele Phinzi, Njoya Silas Ngetar, Quoc Bao Pham, Gashaw Gismu Chakilu, Szilárd Szabó
Dátum:	2023
ISSN:	1865-0473 1865-0481
Megjegyzések:	High-resolution sensors onboard satellites are generally reputed for rapidly producing land-use/land-cover (LULC) maps with improved spatial detail. However, such maps are subject to uncertainties due to several factors, including the training sample size. We investigated the effects of different training sample sizes (from 1000 to 12,000 pixels) on LULC classification accuracy using the random forest (RF) classifier. Then, we analyzed classification uncertainties by determining the median and the interquartile range (IQR) of the overall accuracy (OA) values through repeated k-fold cross-validation. Results showed that increasing training pixels significantly improved OA while minimizing model uncertainty. Specifically, larger training samples, ranging from 9000 to 12,000 pixels, exhibited narrower IQRs than smaller samples (1000-2000 pixels). Furthermore, there was a significant variation (Chi2 = 85.073; df = 11; p < 0.001) and a significant trend (J-T = 4641, p < 0.001) in OA values across various training sample sizes. Although larger training samples generally yielded high accuracies, this trend was not always consistent, as the lowest accuracy did not necessarily correspond to the smallest training sample. Nevertheless, models using 9000-11,000 pixels were effective (OA > 96%) and provided an accurate visual representation of LULC. Our findings emphasize the importance of selecting an appropriate training sample size to reduce uncertainties in high-resolution LULC classification.
Tárgyszavak:	Természettudományok Földtudományok idegen nyelvű folyóiratközlemény külföldi lapban folyóiratcikk High-resolution sensor LULC Training sample size Random forest Classification uncertainty
Megjelenés:	Earth Science Informatics. - 16 : 4 (2023), p. 3667-3677. -
További szerzők:	Ngetar Njoya Silas Pham, Quoc Bao Chakilu, Gashaw Gismu Szabó Szilárd (1974-) (geográfus)
Pályázati támogatás:	K138079 Egyéb K138503 Egyéb
Internet cím:	Szerző által megadott URL DOI Intézményi repozitóriumban (DEA) tárolt változat
Borító:
Saját polcon: