@CONFERENCE{MTMT:34799801,
	title = {Evaluating the efficacy of multitemporal TLS and UAS surveys for quantifying wind erosion magnitudes of sand dune topography},
	url = {https://m2.mtmt.hu/api/publication/34799801},
	author = {Bertalan, László and Négyesi, Gábor and Szabó, Gergely and Túri, Zoltán and Szabó, Szilárd},
	booktitle = {EGU General Assembly 2024 : abstracts},
	unique-id = {34799801},
	year = {2024},
	orcid-numbers = {Bertalan, László/0000-0002-5963-2710; Szabó, Szilárd/0000-0002-2670-7384}
}

@article{MTMT:34742919,
	title = {Predictive machine learning for gully susceptibility modeling with geo-environmental covariates: main drivers, model performance, and computational efficiency},
	url = {https://m2.mtmt.hu/api/publication/34742919},
	author = {Phinzi, Kwanele and Szabó, Szilárd},
	doi = {10.1007/s11069-024-06481-9},
	journal-iso = {NAT HAZARDS},
	journal = {NATURAL HAZARDS},
	unique-id = {34742919},
	issn = {0921-030X},
	abstract = {Currently, machine learning (ML) based gully susceptibility prediction is a rapidly expanding research area. However, when assessing the predictive performance of ML models, previous research frequently overlooked the critical component of computational efficiency in favor of accuracy. This study aimed to evaluate and compare the predictive performance of six commonly used algorithms in gully susceptibility modeling. Artificial neural networks (ANN), partial least squares, regularized discriminant analysis, random forest (RF), stochastic gradient boosting, and support vector machine (SVM) were applied. The comparison was conducted under three scenarios of input feature set sizes: small (six features), medium (twelve features), and large (sixteen features). Results indicated that SVM was the most efficient algorithm with a medium-sized feature set, outperforming other algorithms across all overall accuracy (OA) metrics (OA = 0.898, F 1-score = 0.897) and required a relatively short computation time (< 1 min). Conversely, ensemble-based algorithms, mainly RF, required a larger feature set to reach optimal accuracy and were computationally demanding, taking about 15 min to compute. ANN also showed sensitivity to the number of input features, but unlike RF, its accuracy consistently decreased with larger feature sets. Among geo-environmental covariates, NDVI, followed by elevation, TWI, population density, SPI, and LULC, were critical for gully susceptibility modeling. Therefore, using SVM and involving these covariates in gully susceptibility modeling in similar environmental settings is strongly suggested to ensure higher accuracy and minimal computation time.},
	keywords = {machine learning; Computational efficiency; Predictive modeling; gully erosion; Geo-environmental predictors},
	year = {2024},
	eissn = {1573-0840},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384}
}

@article{MTMT:34741333,
	title = {Classification Assessment Tool: A program to measure the uncertainty of classification models in terms of class-level metrics},
	url = {https://m2.mtmt.hu/api/publication/34741333},
	author = {Szabó, Szilárd and Holb, Imre and Abriha-Molnár, Vanda Éva and Szatmári, Gábor and Singh, Sudhir Kumar and Abriha, Dávid},
	doi = {10.1016/j.asoc.2024.111468},
	journal-iso = {APPL SOFT COMPUT},
	journal = {APPLIED SOFT COMPUTING},
	volume = {155},
	unique-id = {34741333},
	issn = {1568-4946},
	abstract = {Accuracy assessments are important steps of classifications and get higher relevance with the soar of machine 
		and deep learning techniques. We provided a method for quick model evaluations with several options: calculate 
		the class level accuracy metrics for as many models and classes as needed; calculate model stability using random 
		subsets of the testing data. The outputs are single calculations, summaries of the repetitions, and/or all accuracy 
		results per repetitions. Using the application, we demonstrated the possibilities of the function and analyzed the 
		accuracies of three experiments. We found that some popular metrics, the binary Overall Accuracy, Sensitivity, 
		Precision, and Specificity, as well as ROC curve, can provide false results when the true negative cases dominate. 
		F1-score, Intersection over Union and the Matthews correlation coefficient were reliable in all experiments. 
		Medians and interquartile ranges (IQR) of the repeated sampling from the testing dataset showed that IQR were 
		small when a model was almost perfect or completely unacceptable; thus, IQR reflected the model stability, 
		reproducibility. We found that there were no general, statistically justified relationship with the median and IQR, 
		furthermore, correlations of accuracy metrics varied by experiments, too. Accordingly, a multi-metric evaluation 
		is suggested instead of a single metric.},
	keywords = {REPETITION; Python; model evaluation; Model stability},
	year = {2024},
	eissn = {1872-9681},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384; Szatmári, Gábor/0000-0003-3201-598X}
}

@article{MTMT:34476689,
	title = {Impact of the El Niño on Fire Dynamics on the African Continent},
	url = {https://m2.mtmt.hu/api/publication/34476689},
	author = {de Oliveira-Júnior, José Francisco and Mendes, David and Szabó, Szilárd and Singh, Sudhir Kumar and Jamjareegulgarn, Punyawi and Cardoso, Kelvy Rosalvo Alencar and Bertalan, László and da Silva, Marcos Vinicius and da Rosa Ferraz Jardim, Alexandre Maniçoba and da Silva, Jhon Lennon Bezerra and Lyra, Gustavo Bastos and Abreu, Marcel Carvalho and Filho, Washington Luiz Félix Correia and de Sousa, Amaury and de Barros Santiago, Dimas and da Silva Santos, Iwldson Guilherme and Maksudovna, Vafaeva Khristina},
	doi = {10.1007/s41748-023-00363-z},
	journal-iso = {EARTH SYST ENVIRON},
	journal = {EARTH SYSTEMS AND ENVIRONMENT},
	volume = {-},
	unique-id = {34476689},
	issn = {2509-9426},
	abstract = {Several studies investigated the occurrence of fires in Africa with numerical modeling or applied statistics; however, only a few studies focused on the influence of El Niño on the fire risk using a coupled model. The study aimed to assess the influence of El Niño on wildfire dynamics in Africa using the SPEEDY-HYCOM model. El Niño events in the Eastern Tropical Pacific were classified via sea surface temperature (SST) anomaly based on a predefined climatology between 1961 and 2020 for the entire time series of SST, obtaining linear anomalies. The time series of the SST anomalies was created for the region between 5° N and 5° S and 110° W and 170° W. The events were defined in three consecutive 3-month periods as weak, moderate, and strong El Niño conditions. The Meteorological Fire Danger Index (MFDI) was applied to detect fire hazards. The MFDI simulated by the SPEEDY-HYCOM model for three El Niño categories across different lagged months revealed relevant distinctions among the categories. In the case of ‘Weak’, the maximum variability of fire risk observed at time lags (0, -3, -6, and -9 months) was primarily in Congo, Gabon, and Madagascar. The ‘Moderate’ pattern had similar characteristics to ‘Weak’ except for the lag-6 months and its occurrence in the equatorial zone of Africa. ‘Strong’ showed a remarkable impact in East Africa, resulting in high fire risk, regardless of time lags. Precipitation and evaporation simulations (SPEEDY-HYCOM) indicated that El Niño categories in Africa need particular attention in the central, southern, and southeastern regions emphasizing the significance of lag-0 and lag-6 (evaporation) as well as lag-0, lag-6, and lag-9 (precipitation). The SPEEDY-HYCOM coupled model in conjunction with the MFDI was efficient in assessing climate variabilities in Africa during El Niño events. This model allows the analysis and prediction of wildfire risks based on El Niño events, providing crucial information for wildfire management and prevention. Its simulations uncover significant variations in risks among different El Niño categories and lagged months, contributing to the understanding and mitigation of this environmental challenge.},
	keywords = {fire risk; Coupled modelling; El Nino categories; SPEEDY-HYCOM model},
	year = {2024},
	eissn = {2509-9434},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384; Bertalan, László/0000-0002-5963-2710}
}

@article{MTMT:34405801,
	title = {Yield and cost–benefit analyses for apple scab sanitation practices in integrated and organic apple management systems},
	url = {https://m2.mtmt.hu/api/publication/34405801},
	author = {Antal, Gabriella and Szabó, Szilárd and Szarvas, Péter and Holb, Imre},
	doi = {10.1002/ppp3.10460},
	journal-iso = {Plants, People, Planet},
	journal = {PLANTS, PEOPLE, PLANET},
	volume = {6},
	unique-id = {34405801},
	issn = {2572-2611},
	abstract = {Societal Impact statement
		Reduced fungicide use lowers environmental pollution and enables safer food production. The usage of fungicides in apple orchards can be reduced through the application of sanitation practices which decrease the inoculum sources of apple scab disease on fallen leaves. This study found two non-chemical sanitation practices, namely the collection of fallen leaves (CFL) and CFL combined with straw mulch in tree rows, were beneficial. These two practices are not only biologically and environmentally valuable, as they reduce disease levels and can replace chemical fungicides, but they are also economically efficient options for integrated and organic orchards compared to non-sanitized ones.
		
		Summary
		Severe fungicide use can be reduced by applications of sanitation practices in order to reduce scab incidence, yield and fruit quality losses in apple orchards. In a 5-year study, we aimed to investigate the effect of sanitation practices on biological and cost–benefit parameters in two sustainable apple management systems, and to find significant correlations among the parameters.
		We investigated the effect of five sanitation treatments (lime sulphur, leaf collection, mulching, lime sulphur + leaf collection, leaf collection + mulching) on four biological (scab incidence, fruit parameters: total yield, yield class I and II) and seven cost–benefit (three cost types, three annual revenue types, income surplus/deficit) parameters in integrated and organic apple orchards. Correlation, linear regression and principal component analyses (PCA) were performed to find correlations among biological and cost–benefit parameters.
		Results showed that fruit scab incidence was 3.4–8.1 times higher, while total yield was 1.4–1.8 times lower in the organic management system than in the integrated one. The treatment of leaf collection and/or leaf collection + mulching showed higher total cost (180.3 and 675.2 EUR ha−1) but lower scab incidence (5.3 and 27.3%; 4.8 and 26.7%, integrated and organic, respectively) and higher yield with greater total revenues (10,235 and 10,329 EUR ha−1; 8,136 and 8,230 EUR ha−1, integrated and organic, respectively) and income surpluses (851 and 451 EUR ha−1; 897 and 496 EUR ha−1, integrated and organic, respectively) compared to non-sanitized control treatments in most cases. Other sanitation treatments provided fewer biological and/or no financial benefits. Results from correlation and linear regression analyses indicated strong relationships among the factors of total yield vs surplus, class I vs surplus, and fruit scab vs class II) in both management systems. Further relationships were detected among almost all parameters in the PCA.
		Overall, our study demonstrated that two non-chemical sanitation treatments could not only reduce scab incidence and increase fruit yield, but could show positive cost–benefit outcomes in both management systems.},
	keywords = {sanitation; Yield loss; Apple scab; cost-benefit analyses; integrated and organic fruit production; lime sulphur; leaf collection; mulch cover},
	year = {2024},
	eissn = {2572-2611},
	pages = {470-489},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384}
}

@article{MTMT:34396433,
	title = {Assessment of multi-source satellite products using hydrological modelling approach},
	url = {https://m2.mtmt.hu/api/publication/34396433},
	author = {Mahanta, Aiswarya Rani and Rawat, Kishan Singh and Kumar, Nirmal and Szabó, Szilárd and Srivastava, Prashant K. and Singh, Sudhir Kumar},
	doi = {10.1016/j.pce.2023.103507},
	journal-iso = {PHYS CHEM EARTH (2002-)},
	journal = {PHYSICS AND CHEMISTRY OF THE EARTH (2002-)},
	volume = {133},
	unique-id = {34396433},
	issn = {1474-7065},
	abstract = {Multi-source satellite products performance evaluation for varied geographical
		locations aids in quantification of hydrological variables and is useful in the strategy
		making and conservation of the hydrological resources available in a basin. The work
		was focused on assessing utility of multi-source satellite datasets to obtain the
		estimation of hydrologic variables and provide solution for areas that are poorly
		gauged or un-gauged. Assessment of the multi-source-satellite products was
		performed for the poorly gauged river basin with the help of SWAT concerning the
		Palar River basin, India. We analysed time series at the monthly, seasonal, and annual
		scales to quantify surface runoff, water yield, ET, & PET at the calibration site and for
		the entire basin for the period 2003 to 2021, depending on the common period of
		availability of all the data sets. SWAT model estimated highest monthly water yield
		during November–December, with annual water yield being maximum (220 mm in
		2010) and average (99 mm), which can be used to understand water resources for
		irrigation, drinking aspects, and net storage. Average monthly surface runoff patterns
		were similar for SWAT, TerraClimate, and FLDAS. The FLDAS and SWAT simulated
		surface runoff show a resemblance in pattern and magnitude for the monthly and
		annual time series of the average basin scenario. The monthly PET obtained from
		SWAT and ERA-5 show a similar pattern for the entire basin and at the calibration
		site. The ET derived from satellite observation has over-predicted the model output at
		both the calibration site and entire basin},
	keywords = {MODIS; Water yield; SWAT; MERRA-2; FLDAS; ERA-5},
	year = {2024},
	eissn = {1873-5193},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384}
}

@article{MTMT:34148822,
	title = {Assessment of the environmental kuznets curve within EU-27: Steps toward environmental sustainability (1990–2019)},
	url = {https://m2.mtmt.hu/api/publication/34148822},
	author = {Mohammed, Safwan and Gill, Abid Rashid and Ghosal, Kaushik and Al-Dalahmeh, Main and Alsafadi, Karam and Szabó, Szilárd and Oláh, Judit and Alkerdi, Ali and Ocwa, Akasairi and Harsányi, Endre},
	doi = {10.1016/j.ese.2023.100312},
	journal-iso = {Environmental Science and Ecotechnology},
	journal = {Environmental Science and Ecotechnology},
	volume = {18},
	unique-id = {34148822},
	issn = {2666-4984},
	year = {2024},
	eissn = {2666-4984},
	orcid-numbers = {Gill, Abid Rashid/0000-0003-1961-5139; Ghosal, Kaushik/0000-0003-3235-0278; Alsafadi, Karam/0000-0001-8925-7918; Szabó, Szilárd/0000-0002-2670-7384; Oláh, Judit/0000-0003-2247-1711}
}

@article{MTMT:34092358,
	title = {Deep learning‐based training data augmentation combined with post‐classification improves the classification accuracy for dominant and scattered invasive forest tree species},
	url = {https://m2.mtmt.hu/api/publication/34092358},
	author = {Likó, Szilárd Balázs and Holb, Imre and Oláh, Viktor and Burai, Péter and Szabó, Szilárd},
	doi = {10.1002/rse2.365},
	journal-iso = {REMOTE SENS ECOL CON},
	journal = {REMOTE SENSING IN ECOLOGY AND CONSERVATION},
	unique-id = {34092358},
	abstract = {Species composition of forests is a very important component from the point of view of nature conservation and forestry. We aimed to identify 10 tree species in a hilly forest stand using a hyperspectral aerial image with a particular focus on two invasive species, namely Ailanthus tree and black locust. Deep learning‐based training data augmentation (TDA) and post‐classification techniques were tested with Random Forest and Support Vector Machine (SVM) classifiers. SVM had better performance with 81.6% overall accuracy (OA). TDA increased the OA to 82.5% and post‐classification with segmentation improved the total accuracy to 86.2%. The class‐level performance was more convincing: the invasive Ailanthus trees were identified with 40% higher producer's and user's accuracies (PA and UA) to 70% related to the common technique (using a training dataset and classifying the trees). The PA and UA did not change in the case of the other invasive species, black locust. Accordingly, this new method identifies well Ailanthus, a sparsely distributed species in the area; while it was less efficient with black locust that dominates larger patches in the stand. The combination of the two ancillary steps of hyperspectral image classification proved to be reasonable and can support forest management planning and nature conservation in the future.},
	keywords = {Black locust; Multiresolution segmentation; Convolutional-Neural-Network; RANDOM-FOREST; Support-vector-machine; ailanthus},
	year = {2024},
	eissn = {2056-3485},
	orcid-numbers = {Oláh, Viktor/0000-0001-5410-5914; Szabó, Szilárd/0000-0002-2670-7384}
}

@article{MTMT:34220831,
	title = {Recognizing our editorial colleagues M Duane Nellis and Prashant Srivastava},
	url = {https://m2.mtmt.hu/api/publication/34220831},
	author = {Lulla, Kamlesh and Rundquist, Brad and Szabó, Szilárd},
	doi = {10.1080/10106049.2023.2252709},
	journal-iso = {GEOCAR INT},
	journal = {GEOCARTO INTERNATIONAL},
	volume = {38},
	unique-id = {34220831},
	issn = {1010-6049},
	year = {2023},
	eissn = {1752-0762},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384}
}

@article{MTMT:34194422,
	title = {Understanding the role of training sample size in the uncertainty of high-resolution LULC mapping using random forest},
	url = {https://m2.mtmt.hu/api/publication/34194422},
	author = {Phinzi, Kwanele and Ngetar, Njoya Silas and Pham, Quoc Bao and Chakilu, Gashaw Gismu and Szabó, Szilárd},
	doi = {10.1007/s12145-023-01117-1},
	journal-iso = {EARTH SCI INF},
	journal = {EARTH SCIENCE INFORMATICS},
	volume = {16},
	unique-id = {34194422},
	issn = {1865-0473},
	abstract = {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 (Chi 2  = 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.},
	keywords = {training sample size; classification uncertainty analysis; High-resolution sensor},
	year = {2023},
	eissn = {1865-0481},
	pages = {1-11},
	orcid-numbers = {Szabó, Szilárd/0000-0002-2670-7384}
}