Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence

FACTA UNIVERSITATIS Series: Automatic Control and Robotics Vol. 24, No 2, 2025, pp. 147 - 166 https://doi.org/10.22190/FUACR250903011L © 2025 by University of Niš, Serbia | Creative Commons License: CC BY-NC-ND Regular Paper OPTIMISING PROCESS AUTOMATION OF GEOSPATIAL DATA PIPELINES BY ARTIFICIAL INTELLIGENCE UDC (004.8:(911+681.518)) Polina Lemenkova1,2 1National University of Science and Technology, Moscow Institute of Steel and Alloys (MISIS), College of Computer Sciences, Department of Automated Control Systems, Russia 2Russian Academy of Sciences, Institute of Geography, Cartography and Remote Sensing Department, Russia ORCID iD: Polina Lemenkova https://orcid.org/0000-0002-5759-1089 Abstract. This study aims at using advanced GeoAI tools for monitoring landscapes in Italy using machine learning (ML) methods for remote sensing (RS) data processing. Changes in land cover types were identified using AI-processed satellite images. The methodology is based on the four ML algorithms of Python library Scikit-Learn embedded in the GRASS GIS: SupportVectorMachine (SVM), Decision Tree Classifier (DTC), RandomForest (RF) and Multilayer Perceptron Classifier (MLPC) of Artificial Neural Network (ANN). The multispectral satellite Landsat imagery was processed and analysed for changes in categories. The workflow of image processing includes classification for automatic detection of land categories. The presented maps demonstrated spatio-temporal vegetation dynamics and changes in land cover types detected using times series of the RS data. The topology of patches was detected by ML considering differences among spectral reflectance of pixels. ML algorithms recognised. Streamlined workflow through integration of RS and ML algorithms for model training, prediction and classification in GRASS GIS environment. This study has shown the advantages of AI methods for automation of RS data processing. Key words: GIS, remote sensing, image processing, data analysis, machine learning. 1. INTRODUCTION Nowadays, business companies, universities and enterprises employ a variety of Artificial Intelligence (AI) techniques in geospatial solutions to enhance decision-making processes and support cartographic analysis [1–5]. The most popular AI branch is Machine Learning (ML) analytical tools which leverages powerful and robust workflow for geospatial data analysis. Received September 3, 2025 / Accepted November 13, 2025 Corresponding author: Polina Lemenkova National University of Science and Technology, Moscow Institute of Steel and Alloys (MISIS), College of Computer Sciences, Department of Automated Control Systems, Leninskiy Prospekt 6/7, Moscow 119049, Russia E-mail: [email protected] 148 P. LEMENKOVA The use of ML in image analysis ensures quick, multifaceted processing of large amounts of geospatial data to support decision-making and GeoAI intelligence solutions. This is possible through automated workflow in satellite image segmentation, image analysis and classification for extracting structured information from data. [6-7]. For example, the use of AI supports environmental monitoring, land planning, cartographic plotting, topographic LiDAR-based modelling and Remote Sensing (RS) data processing [8-10]. AI enables performing such fast queries and ensures representation in cartographic data analytics that makes fast multidimensional analysis of spatial datasets possible. In order to analyze and illustrate the potential security facets, this study develops an engineering procedure that designs the conceptual model matching to updated geospatial applications. This procedure is based on the model-driven architecture for creating secure AI-supported workflow where the conceptual model of GIS analysis is enhanced through novel AI implementations in geospatial technologies, Fig. 1. Fig. 1 Conceptual scheme of analytical geospatial data processing system. Source: author Data analytics and management strategy pose a challenge to Geographic Information System (GIS) technologies for meeting the big Earth data era. This urgency is also well recognised in research and governmental organizations. Accordingly, it has been incorporated by the environmental planning systems, and analytical enterprises that employ GIS. The use of AI and ML in geospatial data processing ensures the following advantages: i. Coordination and control of data-driven research on environmental risk assessment; ii. Quality control and updating of the data for operative monitoring; iii. Carrying out geospatial data management and reporting; iv. Establishing effective links between GIS datasets and processing centers; v. Resolving possible issues in GIS through metadata analytics vi. Employing algorithms of Python's ML libraries for data automation vii. Image analysis using Deep Learning Frameworks: TensorFlow, PyTorch, Keras viii. Managing data generated in large projects and research centers Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence 149 Data analytics, when employed the algorithms of AI and ML, ensures the compliance of research activities with the fundamental principles of research integrity in computer science (CS) and data science (DS) at national and international levels. Therefore, special attention is paid to AI-supported data analysis as an essential tool for data processing and automation. 2. RESEARCH FORMULATION The RS data provide an excellent source of information for Earth observation. Such data are widely used and their applications are discussed in the relevant literature on environmental monitoring and landscape mapping [11-13]. Understanding and monitoring landscape dynamics is critical for addressing the impacts of climate change, biodiversity loss, and socio-economic transitions [14,15]. The essential role of RS in landscape analysis consists in the complex structure of landscapes that are shaped by interactions between natural processes and human activities [16,17]. The question of how to efficiently, accurately and automatically detect and recognize those changes is essential for RS data analysis through object detection using cartographic visualization. The key role in extracting information from the RS data for thematic mapping plays the methods of image processing. Advanced methods enable to perform analysis of raster scenes through diverse techniques, including, among others, classification, segmentation, spectral enhancing, contrasting, to mention a few of them. Interpretation of the extracted information ensures the identification of the objects visible on the Earth’s surface. Many landscapes have complex structure that requires advanced methods of image processing. One of the examples is the Central Apennines, a biodiversity hotspot [18]. Its diverse ecosystems provide critical services, including water regulation [19,20], carbon sequestration [21], and habitat provision [22], and also support human livelihoods through forestry and agriculture. Recently, these landscapes have undergone changes, driven by land abandonment [23], forest encroachment [24] and climate-related processes, including droughts and disturbance [25]. Landscape changes are related to the dynamics of land use. For instance, traditional agriculture and silviculture practices declined throughout the 20th century due to rural depopulation [26]. This affected agricultural and agroforestry sectors in Apennines and caused changes in landscape mosaic, as in other mountainous regions [27]. The widespread land abandonment resulted in forest expansion and habitat homogenization. Over the 1936-2018 period, for instance, forest cover increased from 29% (1.673.266 ha) to 38% (2.214.671 ha) in Central Italy [28], while this area has been characterized as one of the land use change hotspots of Europe [29-31]. While some of these processes have positive effects, such as ecosystem restoration and increased forest cover, others pose significant challenges, including biodiversity loss [32-33], soil erosion [34], and landscape diversity [35]. The inherited properties of the Earth’s land surface includes complex parameters such as texture, curvature of patches, context of disposition of several objects representing the scene, shape, form and structure of objects. Increased variability of images with different properties and resolution requires advanced methods of their processing compared to the traditional approaches of image analysis that use pixel-based classification algorithms. 150 P. LEMENKOVA 3. STATE-OF-THE-ART Computer vision provides valuable insights into the cumulative effects of land cover changes [36]. AI significantly facilitates the workflow of image processing through process optimization: the increase of the speed and precision of data processing. For example, comparing several images, it is possible to detect dynamics in forest biodiversity and ecosystem structure. Furthermore, the development of automated approaches for RS data analysis offers broader implications for monitoring landscape changes in other regions facing similar challenges. Using the ensemble algorithms presented by Scikit-Learn library of Python, ML provides one of such cartographic approaches due to its effectiveness and applicability. Specifically for RS data processing, ML has been successfully applied to existing problem of classification due to its effectiveness, high level of automation and robustness. Although traditional classification methods using in GIS can perform image processing, it has a generally higher time-consuming performance. Recently, the use of ML methods that emerged and evolved rapidly has been in the centre of attention for RS data processing [37-40]. ML has become an available approach in cartography along with the advances in programming, and these methods are increasingly used in various tasks of environmental monitoring and mapping. The popularity of such methods is explained by the improved speed of data processing and improved quality of image analysis. The use of these tools ensures automation of satellite data processing, supports operative monitoring using time series and has an important effect on accuracy and effectiveness of data processing for cartographic visualization. High level of automation in data processing achieved by ML enables to rapidly identify objects on the images using computer vision techniques. Examples of the existing advanced methods of Geospatial AI include algorithms of ML, DL and the convolutional neural networks (CNNs). They are used for classification, image segmentation and spatial analysis using the following tools: ▪ Pyramid scene parsing network (PSPNet) [41], ▪ DeepLabV3 (PyTorch) [42], Unet [43], open source TensorFlow and Keras [44], ▪ Segment Anything Model (SAM) Meta-AI [45], ▪ Python’s Scikit-Learn library with several embedded algorithms [46], ▪ TensorBoard toolkit in ArcGIS (API for Python v. 1.8.3) by arcis.learn module. Such methods enable to perform the image analysis workflow, including semantic segmentation, feature extraction, pattern recognition and supervised classification. For environmental monitoring, the application of AI presents high perspectives due to automated framework to detection and robust attribution of land cover changes [47–50]. The ML methods are based on diverse approaches and algorithms. For example, one of the application of AI in image analysis is semantic segmentation. Semantic segmentation enables modelling of the images and feature extraction on the scenes for pattern recognition in the environmentally constrained regions. Among others, semantic segmentation contributes to image enhancement (noise reduction) and information extraction: panoptic, instance and semantic segmentation, image translation and classification for LULC change detection on sequential time series of images. In view of these limitations, the use of the novel AI methods for RS data processing and programming algorithms for classification approaches has recently drawn much attention. Among them, Python algorithms present a powerful instrument for image processing that have been investigated earlier. The AI algorithms for the mapping of crop Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence 151 fields using Landsat imagery were reported in existing papers. Examples of AI in geospatial analysis include Deep Learning (DL) [51-52], Convolutional Neural Networks (CNN) [53] and ML [54] techniques. These employ the robust algorithms of classification, such as Decision Tree Classifier (DTC) [55-56], Support Vector Machine (SVM) [57-58], Artificial Neural Network (ANN) [59-60], Random Forest (RF) [61-63], Naive Bayes [64], Gradient boosting [65-67], to mention a few of them. Application of such methods enhances the precision of object recognition on the image through advanced level of pixel analysis. Advanced studies proposed an integrated application of the ML algorithms based on the multi-source data combining LiDAR and Landsat where connected components in the RS data are particularly well suited for complex forest canopy analysis using data on cover and height. ML applications in RS data processing include crop mapping for food and agricultural research, hazard risk assessment using fire mapping, urban sprawl and environmental monitoring. 4. OBJECTIVES AND GOALS This research strives to tackle research questions relevant to geospatial data analytics, providing results that are applicable within and beyond the applications of AI into remote sensing data processing. Here, we consider the three types of AI algorithms – DTC, SVM and ANN by multilayer perceptron classification (MLPC) – for improving high performance analytic tool in remote sensing. The study has the following scientific objectives: 1. To assess how AI applications in Geographic Resources Analysis Support System (GRASS) GIS affect and improve the efficiency of geo-data analytics, 2. To highlight areas of maximum improvements in geospatial query analysis and to quantify the percentage of land cover changes using satellite image analysis and AI; 3. To evaluate and address the growing importance of leveraging databases for AI-based query analysis in response to increasing geospatial data analytics demands (analytical review and state of the art); 4. To identify the multiple shifts in geospatial data management strategy using AIsupported GIS and develop narratives, to interpret the updates in data analytics 5. To promote the advancements in RS data processing which has substantially improved through AI application; 6. To perform the comprehensive and systematic analysis of AI applications, combining GIS with AI and databases in big data environments. The specific objective is to apply methods for image processing for geo-information extraction. To do this, this work presents the use of ML to recognize landscape patches using classification by GRASS GIS modules. 5. NOVELTY AND MOTIVATION The importance and novelty of the presented article consists in the integration of GIS and AI for monitoring landscapes. Landscapes represent the land surface where environmental processes interplay to provide habitat and resources for life. The cumulative effects from climate change, biophysical and anthropogenic processes affect landscapes and cause land cover changes, decline of vegetation and ecosystem degradation. Such issues are visible on spaceborne RS data and can be detected for operative environmental monitoring using 152 P. LEMENKOVA GeoAI. Coupled technological innovation that integrates technical methods and environmental analysis lies in addressing a critical absence and need in the current applications of the CS in geoinformatics. The lack of research contribution to the improved GIS functionalities requires development of up-to-date applications of AI to spatial data analytics. Here, we demonstrate the improved workflow pipeline of image analysis through the combinations of several AI capabilities in GRASS GIS as an advanced environmental toolset. To this end, the GRASS GIS software is employed as a methodological tool for image analysis aimed at information extraction from the scikit-learn library. GRASS GIS was selected due to its technical functionality and multiple modules which ensure the feasibility of image analysis in the implementation of the ML for time series analysis. With AI, geospatial data may be effectively converted into valuable analytical geoinformation. The advantages of using AI-enhanced GIS consist in the effective employment of information for decision making for environmental solutions (land planning, monitoring and cadastral assessment). To this end, we analysed the data organization in GIS, derived from data warehouses of USGS and relational sources optimized for AI in environmental research. In this way, the presented research task integrates the Computer Science (CS), Geoinformatics and RS domains. This study contributes to the reports on the use of AI-supported solutions of GRASS GIS analytics. Such analysis consists in the improvements of the GIS technologies and practices that recent research do not sufficiently cover. By creating new AI-enhanced GIS workflow on image analysis, this study offers a practical update of the functionality of RS data processing in GRASS GIS models. 6. DATA AND METHODS 6.1. Datasets The geospatial data management strives at fulfillment of open science data principles embraces an open-access FAIR (Findable, Accessible, Interoperable, Reusable) principle. When it comes to environmental data collection, archiving, and disclosure, it is important to account for the concerns of data open availability and the use of public data obtained from Earth observation centers. Various studies on land cover mapping are based on satellite data: SPOT, WorldView, Landsat [68-69]. This work employs a series of eight Landsat-8-9 OLI/TIRS images: four images in spring period 2018, 2019, 2022 and 2023, and four images in the autumn period in the same years. The choice of these data is explained by reputable history, free availability and quality (30 m resolution for multi-spectral bands). Landsat sensors have geometric features performing acquisitions at constant acquisition angles. The data were collected from the United States Geological Survey (USGS) EarthExplorer repository as a time series. The quality of data makes it possible to exploit the synergy of the multi-temporal time series. The data contained seven multispectral Landsat bands for each image (Aerosol, Blue, Green, Red, NIR, SWIR-1, SWIR-2). Collectively, the dataset presents the short-time series of the satellite images and demonstrate landscape dynamics over the period 2018-2023. 6.2. Methodology The methodology explores the AI applications to multidimensional data processing in RS and GIS for efficient support of fast mapping, automated querying and complex calculations Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence 153 of geo-data (time series analysis of the satellite images). AI is a novel development in CS implemented in the environmental analysis, such as risk assessment and mapping natural hazards. The data processing aims at handling data pools to coordinate the information flows through research analytics and updates of actual data obtained for operative monitoring. Here, we used several ML algorithms embedded in GRASS GIS to test their functionality and speed of data processing. This set of several algorithms – SVM, RF, MLPC and DTC, – is central to our research goal, and aims at selecting the optimal algorithm that works best. By comparing these methods we ensure the productive pipeline workflow of image analysis, Fig. 2. Fig. 2 Workflow of the methodology: steps of satellite image processing by Artificial Intelligence (AI) algorithms. Diagram source: author The preliminary results were obtained and reported [70], where the performance of each algorithm was evaluated for the calendar year: that is, the images were classified by four algorithms and compared for 2019, 2022, and 2023. To continue that work, here we complete the inverse research question, and compare the classified images for each algorithm separately within the time series of the calendar years. All the scenes were evaluated for two different seasons: spring and autumn. Such algorithms as SVM, RF and others organise, implement and perform the classification of objects detectable on the images. Here we improve the GIS functionality in the image analysis through AI with the aim of identification of hidden relationships between environmental elements (vegetation and soil). The updated relational database presents a set of images imported into the GRASS GIS from the open repository of USGS. The workflow of image processing was optimised through the algorithms of pattern recognition. The indexing of pixels enabled their assignment into land cover classes tailored for environmental data analytics. The major commands of scripts of the GRASS GIS used for image analysis are included and commented below in the listings. 154 P. LEMENKOVA Listing 1: importing the image subset with 7 Landsat bands and displaying the raster map grass r.import input=/Users/polinalemenkova/grassdata/Italy/LC08_L2SP_191030_ 20180420_20200901_02_T1_SR_B1.TIF output=L_2018_01 extent=region resolution=region # repeated for all raster scenes to be imported in the GRASS GIS input=/Users/polinalemenkova/grassdata/Italy/LC08_L2SP_191030_ 20180420_20200901_02_T1_SR_B7.TIF output=L_2018_07 extent=region resolution=region g.list rast The clustering and classification enabled to group the data by 'i.group' module and set the computational region to match the scene, as in Listing 2. Listing 2 g.region raster=L_2018_01 -p i.group group=L_2018 subgroup=res_30m \ input=L_2018_01,L_2018_02,L_2018_03,L_2018_04,L_2018_05,L_2 018_06,L_2018_07 At the next step, we generated signature file which was used as a training datasets and report using k-means clustering algorithm, Listing 3. Listing 3 i.cluster group=L_2018 subgroup=res_30m \ signaturefile=cluster_L_2018 \ classes=10 reportfile=rep_clust_L_2018.txt --overwrite # Classification by i.maxlik module i.maxlik group=L_2018 subgroup=res_30m \ signaturefile=cluster_L_2018 \ output=L_2018_clusters reject=L_2018_cluster_reject Cartographic visualization was performed using mapping functionalities of GRASS GIS, Listing 4. Here, we adjusted the style of vector and raster objects, placed the color legend and added necessary annotations. Listing 4 r.colors L_2018_clusters color=roygbiv g.region raster=L_2018_01 -p d.mon wx0 d.vect isolines color='100:93:134' width=0 d.rast L_2018_clusters d.grid -g size=00:30:00 color=grey width=0.1 fontsize=16 text_color=grey d.legend raster=L_2018_clusters title="Clusters 2018" title_fontsize=19 font="Helvetica" fontsize=17 bgcolor=white border_color=white d.out.file output=Italy_2018 format=jpg --overwrite Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence 155 Mapping rejection probability was performed for analysis of accuracy using the estimated chi-square test in the GRASS GIS. It shows the probability that pixels are correctly identified and assigned to the corrected categories, Listing 5. Listing 5 d.mon wx1 g.region raster=L_2018_clusters -p r.colors L_2018_cluster_reject color=soilmoisture -e d.rast L_2018_cluster_reject d.grid -g size=00:30:00 color=grey width=0.1 fontsize=16 text_color=grey d.legend raster=L_2018_cluster_reject title="2018" title_fontsize=19 font="Helvetica" fontsize=17 bgcolor=white border_color=white d.out.file output=Italy_2018_reject format=jpg The ML block of image processing was performed using specially designed module of GRASS GIS ‘r.learn.train’. It enabled generating training pixels from the land cover classification, Listing 6. These training pixels were used to perform a classification on the target Landsat image. The example below is given for the algorithm of RF. Listing 6 r.random input=L_2018_clusters seed=100 \ npoints=1000 raster=training_pixels r.learn.train group=L_2018 training_map=training_pixels \ model_name=RandomForestClassifier \ n_estimators=500 save_model=rf_model.gz --overwrite The prediction performance was evaluated using ‘r.learn.predict’ module, Listing 7. here, we also checked raster categories applied to the classification output. Listing 7 r.learn.predict group=L_2018 \ load_model=rf_model.gz output=rf_classification -- overwrite r.category rf_classification The scripts presented above illustrated the practical steps of the workflow used to explore the computational resource requirements of GRASS GIS. The use of scripts facilitates image processing through automation achieved by the the AI-based algorithms. The classification techniques were derived from the Scikit-Learn. As intuitive classifiers, the supervised learning approaches imply the probabilistic approach and normal distribution of the data and straightforward implementation. This approach evaluates the likelihood of the pixels to belong to the target land cover class. During the process of image partition, the pixels of the image are assigned to the target classes. Afterwards, the algorithms verify the correctness of this assignment. This is based on the decision rule that evaluates the suitability of pixels into categories according to threshold values. The Python's Scikit-learn library for ML tasks was employed in this work. ScikitLearn is integrated in GRASS GIS through special modules to leverage within its environment. Such integration supports supervised classification and regression on rasters using various algorithms. Three modules of Scikit-Learn were used in this work: 162 P. LEMENKOVA 3. Complex issues of raster data processing and analytical problems related to spatial analysis were tackled through AI-enhanced data analysis. This creates a multi-faceted understanding of the engines and algorithms shaping pattern recognition tasks and trends in geospatial data organization and queries, by using advanced integrated approaches of data science and GIS. This paper contributed to the development of the advanced solutions of data processing in GeoAI. This interdisciplinary, data-driven study demonstrated the improvements in GIS implementation in analytical environments to support the upgrade and product development in RS data processing. The widespread use of the AI can be illustrated by such products as the Python libraries: Keras and Tensorflow. Moreover, AI functionality is included in selected GIS: GRASS GIS and the Analysis Services of ArcGIS suite. Here, we demonstrated the use of AI in GIS, which illustrates its importance for image analysis, as also proved by constantly increasing number of applications and reports in environmental mapping [71-77]. Future studies can explore the AI applications to multidimensional data processing in GIS for efficient supporting of fast geospatial querying and complex calculations of large RS data volumes. For such tasks, ML and AI are robust tools, implemented in environmental monitoring through integrating geoinformatics with CS. Such approaches are challenging since they enable to develop novel methods of information extraction. The transfer of data between the existing pool of spatial databases (collection of satellite images, such as Landsat or Sentinel) and the analytical centers is aimed at information extraction and generating new knowledge. With these regards, further works can continue data analysis to extract information from the satellite images using Python tools for RS data analysis. As shown in this study, GRASS GIS excellently integrates the GeoAI through the addon of scikit-learn. Further research can benefit from these methods which employ the advanced ML algorithms of RS data analysis and contribute to the development of GeoAI. Acknowledgement: The author thanks the anonymous reviewers for their support and comments. REFERENCES [1] Z. Wang, G. Mountrakis, N. J. Pastick, “Deep learning approaches on Landsat observations: a review and meta-analysis”, International Journal of Remote Sensing, vol. 46, no. 15, p. 5601–5627, 2025. https://doi.org/10.1080/01431161.2025.2520050 [2] P. Lemenkova, “Artificial Intelligence for Computational Remote Sensing: Quantifying Patterns of Land Cover Types around Cheetham Wetlands, Port Phillip Bay, Australia”, J. Mar. Sci. Eng., 12(8), 1279, 2024d. https://doi.org/10.3390/jmse12081279 [3] X. Zhang, Y. Zhou, J. Luo, “Deep learning for processing and analysis of remote sensing big data: a technical review”, Big Earth Data, vol. 6, no. 4, p. 527–560, 2021. https://doi.org/10.1080/20964471.2021.1964879 [4] P. Lemenkova, “Deep Learning Methods of Satellite Image Processing for Monitoring of Flood Dynamics in the Ganges Delta, Bangladesh”, Water, vol. 16, no. 8, p. 1141, 2024c. https://doi.org/10.3390/w16081141 [5] G. Richardson, A. Knudby, M. A. Crowley, M. Sawada, W. Chen, “Machine learning approaches to Landsat change detection analysis”, Canadian Journal of Remote Sensing, vol. 51, no. 1, 2025. https://doi.org/10.1080/07038992.2024.2448169 [6] J. Babbar and N. Rathee, "Satellite Image Analysis: A Review," 2019 IEEE International Conference on Electrical, Computer and Communication Technologies (ICECCT), Coimbatore, India, 2019, pp. 1-6, https://doi.org/10.1109/ICECCT.2019.8869481 [7] P. Lemenkova, “Exploitation d'images satellitaires Landsat de la région du Cap (Afrique du Sud) pour le calcul et la cartographie d'indices de végétation à l'aide du logiciel GRASS GIS”, Physio-Géo, vol. 20, no. 1, p. 113-129, 2024b. https://doi.org/10.4000/11pyj Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence 163 [8] F. Chouikhi, A. Ben Abbes, I. Riadh Farah, “Supervised desertification classification using Siamese Variational Autoencoder”, International Journal of Image and Data Fusion, vol. 16, no. 1, 2025. https://doi.org/10.1080/19479832.2025.2476544 [9] Z. Sun, L. Di, H. Fang, “Using long short-term memory recurrent neural network in land cover classification on Landsat and Cropland data layer time series”, International Journal of Remote Sensing, vol. 40, no. 2, p. 593–614, 2019. https://doi.org/10.1080/01431161.2018.1516313 [10] P. Lemenkova, “Approche cartographique par le SIG GRASS pour l'analyse de la structure du paysage au Libéria, Afrique de l'Ouest”, Dynamiques environnementales, vol. 53, p. 1-36, 2024a. https://doi.org/ 10.4000/12n0l [11] M. Klaučo, B. Gregorová, U. Stankov, V. Marković, P. Lemenkova, “Determination of ecological significance based on geostatistical assessment: a case study from the Slovak Natura 2000 protected area”, Open Geosciences, vol. 5, no. 1, p. 28-42, 2013. https://doi.org/10.2478/s13533-012-0120-0 [12] S. Xiong, S. Du, X. Zhang, S. Ouyang, W. Cui, “Fusing Landsat-7, Landsat-8 and Sentinel-2 surface reflectance to generate dense time series images with 10m spatial resolution”, International Journal of Remote Sensing, vol. 43, no. 5, p. 1630–1654, 2022. https://doi.org/10.1080/01431161.2022.2047240 [13] A. Banskota, N. Kayastha, M. J. Falkowski, M. A. Wulder, R. E. Froese, J. C. White, “Forest Monitoring Using Landsat Time Series Data: A Review”, Canadian Journal of Remote Sensing, vol. 40, no. 5, p. 362–384, 2014. https://doi.org/10.1080/07038992.2014.987376 [14] P. Lindh, P. Lemenkova, “Optimized Workflow Framework in Construction Projects to Control the Environmental Properties of Soil”, Algorithms, vol. 16(6), p. 303, 2023. https://doi.org/10.3390/a16060303 [15] P. Lemenkova, “Image Segmentation of the Sudd Wetlands in South Sudan for Environmental Analytics by GRASS GIS Scripts”, Analytics, vol. 2, no. 3, p. 745-780, 2023. https://doi.org/10.3390/analytics2030040 [16] T. Kim, Y. Yun, S. Park, J. Oh, Y. Han, “Change detection over the Aral Sea using relative radiometric normalization based on deep learning”, Remote Sensing Letters, vol. 14, no. 8, p. 821–832, 2023. https://doi.org/10.1080/2150704X.2023.2242589 [17] X. Yang, “Use of Archival Landsat Imagery to Monitor Urban Spatial Growth”, Urban Remote Sensing, X. Yang (Ed.). https://doi.org/10.1002/9780470979563.ch2 [18] V. M., Granito, D. Lunghini, “Biodiversity of macrofungi in the beech forests and calcareous grasslands of the Simbruini Mountains Regional Park (central Apennines, Italy)”, Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, vol. 145, no. 2, p. 381–396, 2011. https://doi.org/10.1080/11263504.2011.563523 [19] P. Farabollini, P. De Pari, M. E. Discenza, M. Minnillo, C. Carabella, G. Paglia, E. Miccadei, Geomorphological evidence of debris flows and landslides in the Pescara del Tronto area (Sibillini Mts, Marche Region, Central Italy). Journal of Maps, vol. 17, no. 3, p. 90–99, 2020. https://doi.org/10.1080/17445647.2020.1827055 [20] P. Lemenkova, “Using R packages 'tmap', 'raster' and 'ggmap' for cartographic visualization: An example of dem-based terrain modelling of Italy, Apennine Peninsula”, Zbornik radova - Geografski fakultet Univerziteta u Beogradu, vol. 68, p. 99-116, 2020. https://doi.org/10.5937/zrgfub2068099L [21] M. Marchetti, L. Sallustio, M. Ottaviano, A. Barbati, P. Corona, R. Tognetti, L. Zavattero, G. Capotorti, “Carbon sequestration by forests in the National Parks of Italy”, Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, 146(4), 1001–1011, 2012. https://doi.org/10.1080/11263504.2012. 738715 [22] C. Catonica, “A new species and a new record of Festuca (Poaceae) from the Gran Sasso of Italy (central Apennines)”, Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, vol. 135, no. 3, p. 271–283, 2001. https://doi.org/10.1080/11263500112331350910 [23] S. Landi, F. Binazzi, R. Papini, G. d’Errico, S. Del Duca, A. Fabiani, S. Mocali, “Effect of Long-Term Set-Aside Management System on Soil Health by Evaluation of Biodiversity Indicators”, Diversity, vol. 17, no. 4, p. 240, 2025. https://doi.org/10.3390/d17040240 [24] C. Blasi, S. Burrascano, E. Del Vico, R. Di Pietro, M. Iocchi, L. Rosati, “Cynosurion cristati grasslands in the central Apennines (Tyrrhenian sector): A phytosociological survey in the Lepini and Prenestini mountains”, Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, vol. 143, no. 1, p. 69–77, 2009. https://doi.org/10.1080/11263500903498192 [25] B. Petriccione, “Short-term changes in key plant communities of Central Apennines (Italy)”, Acta Botanica Gallica, vol. 152, no. 4, p. 545–561, 2005. https://doi.org/10.1080/12538078.2005.10515513 [26] E. Biondi, M. Allegrezza, “Syntaxonomy of Pinus nigra s.l. communities in the Erico-Pinetea class and their distribution in the central Apennines and Balkan province”, Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, vol. 154, no. 2, p. 248–258, 2019. https://doi.org/10.1080/11263504.2019. 1701128 [27] G. Filibeck, L. Cancellieri, F. Bartolucci, U. Becker, F. Conti, S. Maestri, M. Mürz, E. Schommer, M. G. Sperandii, T. Becker, “Festuca valesiaca Schleich. ex Gaudin newly discovered in the Central Apennines 164 P. LEMENKOVA (Italy): a further example of steppe relict in the Abruzzo “dry valleys.” Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, vol. 154, no. 5, p. 593–600, 2019. https://doi.org/10.1080/11263504.2019.1651784 [28] Ferrara, M. Biró, L. Malatesta, Z. Molnár, S. Mugnoz, F. M. Tardella, A. Catorci, “Land-use modifications and ecological implications over the past 160 years in the central Apennine mountains”, Landscape Research, vol. 46, no. 7, p. 932–944, 2021. https://doi.org/10.1080/01426397.2021.1922997 [29] F. Lombardi, M. Klopcic, P. Di Martino, R. Tognetti, G. Chirici, A. Boncina, M. Marchetti, “Comparison of forest stand structure and management of silver fir–European beech forests in the Central Apennines, Italy and in the Dinaric Mountains, Slovenia”, Plant Biosystems - An International Journal Dealing with all Aspects of Plant Biology, vol. 146, no. 1, p. 114–123, 2012. https://doi.org/10.1080/11263504.2011.623190 [30] P. Lemenkova, “On forest hydrology in South Tyrol: Processes, modelling and assessment”, Agriculture and Forestry, vol. 71, no. 2, 2025. https://doi.org/10.17707/AgricultForest.71.2.01 [31] P. Lemenkova, “Python for environmental modelling of mixed coniferous forests dominated by Norway spruce (Picea abies [L.] Karst.) and Swiss stone pine (Pinus cembra L.),” Topola, no. 215, pp. 5-16, 2025, https://doi.org/10.5937/topola2515005L [32] G. Mei, S. Pesaresi, G. Corti, S. Cocco, C. Colpi, and F. Taffetani, Changes in vascular plant species composition, top-soil and seed-bank along coppice rotation in an Ostrya carpinifolia forest. Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, 154(2), 259–268, 2019. https://doi.org/10.1080/11263504.2019.1609618 [33] P. Lemenkova, “Gathering predictors of biodiversity change and reconstructing land cover history in Central Apennines using machine learning and remote sensing data”, Journal of Anatolian Geography, vol. 2, no. 1, p. 36--47. https://doi.org/10.5281/zenodo.15767237 [34] Catorci, G. Ottaviani, I. V. Kosić, S. Cesaretti, Effect of spatial and temporal patterns of stress and disturbance intensities in a sub-Mediterranean grassland. Plant Biosystems - An International Journal Dealing with All Aspects of Plant Biology, 146(2), 352–367, 2011. https://doi.org/10.1080/11263504.2011.623192 [35] P. Lemenkova, “Contribution of epiphytes and fog to patterns of atmospheric fluxes in mountainous forests (Picea abies L. and Pinus cembra L.)”, Acta herbologica, vol. 34, no. 1, pp. 7-21, 2025, https://doi.org/10.5937/ 34ah-57901 [36] O. S. Ahmed, S. E. Franklin, M. A. Wulder, “Interpretation of forest disturbance using a time series of Landsat imagery and canopy structure from airborne lidar”, Canadian Journal of Remote Sensing, vol. 39, no. 6, p. 521–542, 2014. https://doi.org/10.5589/m14-004 [37] P. Lemenkova, “Automation of image processing through ML algorithms of GRASS GIS using embedded Scikit-Learn library of Python”, Examples and Counterexamples, vol. 7, no. 100180, 2025. https://doi.org/10.1016/j.exco.2025.100180 [38] D. He, Q. Shi, X. Liu, Y. Zhong, G. Xia, L. Zhang, “Generating annual high resolution land cover products for 28 metropolises in China based on a deep super-resolution mapping network using Landsat imagery”, GIScience & Remote Sensing, vol. 59, no. 1, p. 2036–2067, 2022. https://doi.org/10.1080/15481603.2022. 2142727 [39] P. Lemenkova, “Statistical analysis of the Mariana Trench geomorphology using R programming language”, Geodesy and Cartography, vol. 45, no. 2, p. 57-84, 2019. https://doi.org/10.3846/gac.2019.3785 [40] D. H. Jr. Knepper, “Mapping Hydrothermal Alteration With Landsat Thematic Mapper Data”, Remote Sensing in Exploration Geology: Golden, Colorado to Washington, D.C., June 30–July 8, 1989. https://doi.org/10.1002/9781118669877.ch3 [41] Boonpook, W., Tan, Y., Nardkulpat, A., Torsri, K., Torteeka, P., Kamsing, P., Sawangwit, U., Pena, J., Jainaen, M.: Deep learning semantic segmentation for land use and land cover types using Landsat 8 imagery. ISPRS International Journal of Geo-Information 12(1) 2023. https://doi.org/10.3390/ijgi12010014 [42] Shi, S., Wu, J., Yao, K., Meng, Q.: Deep learning-based contrail segmentation in thermal infrared satellite cloud images via frequency-domain enhancement. Remote Sensing 17(18), 2025. https://doi.org/10.3390/rs17183145 [43] Jiao, L., Huo, L., Hu, C., Tang, P., Zhang, Z.: Refined unet v4: End-to-end patch-wise network for cloud and shadow segmentation with bilateral grid. Remote Sensing 14(2), 2022. https://doi.org/10.3390/rs14020358 [44] Parente, L., Taquary, E., Silva, A.P., Souza, C., Ferreira, L.: Next generation mapping: Combining deep learning, cloud computing, and big remote sensing data. Remote Sensing 11(23), 2019. https://doi.org/10.3390/rs11232881 [45] Sun, J., Yan, S., Alexandridis, T., Yao, X., Zhou, H., Gao, B., Huang, J., Yang, J., Li, Y.: Enhancing crop mapping through automated sample generation based on segment anything model with mediumresolution satellite imagery. Remote Sensing 16(9), 2024. https://doi.org/10.3390/rs16091505 [46] Lemenkova, P.: Deep learning methods of satellite image processing for monitoring of flood dynamics in the ganges delta, bangladesh. Water 16(8), 2024. https://doi.org/10.3390/w16081141 Optimising Process Automation of Geospatial Data Pipelines by Artificial Intelligence 165 [47] Manthira Moorthi, S., Misra, I., Kaur, R., Darji, N.P., Ramakrishnan, R.: Kernel based learning approach for satellite image classification using support vector machine. In: 2011 IEEE Recent Advances in Intelligent Computational Systems, pp. 107–110 (2011). https://doi.org/10.1109/RAICS.2011.6069282 [48] Nagaraj, R., Arulvadivelan, V., Gouthamkumar, K., Dharshen, K., Kumar, L.S.: Surface water mapping and volume estimation of Lake Victoria using machine learning algorithms. In: 2023 International Conference on Signal Processing, Computation, Electronics, Power and Telecommunication (IConSCEPT), pp. 1–6 (2023). https://doi.org/10.1109/IConSCEPT57958.2023.10170600 [49] Lemenkova, P.: Sentinel-2 for high resolution mapping of slope-based vegetation indices using machine learning by Saga GIS. Transylvanian Review of Systematical and Ecological Research 22(3), 17–34 (2020) https://doi.org/10.2478/trser-2020-0015 [50] Tan, C., Sun, Y., Li, G.-F., Jiang, G.-Z., Kong, J.-Y., Tao, B.: Research on image segmentation based on support vector machine. In: 2018 International Conference on Machine Learning and Cybernetics (ICMLC), 2, 650–655 (2018). https://doi.org/10.1109/ICMLC.2018.8526926 [51] Cao, X. (2024). Meta-Heuristic Hyperparametric Optimization Model BO-GCN-LSTM for Spatial InteractionOriented Prediction Based on Deep Learning. 2024 5th International Conference on Intelligent Computing and Human-Computer Interaction (ICHCI), 65–68. doi:10.1109/ICHCI63580.2024.10808144 [52] Khurram, S., Pour, A. B., Bagheri, M., Helmy Ariffin, E., Akhir, M. F., & Bahri Hamzah, S. (2025). Satellite-Based Multi-Decadal Shoreline Change Detection by Integrating Deep Learning with DSAS: Eastern and Southern Coastal Regions of Peninsular Malaysia. Remote Sensing, 17(19), 3334. https://doi.org/10.3390/rs17193334 [53] Zheng, Z., Yu, Y., Yang, X., Yuan, X., & Hou, Z. (2025). A Novel Framework for Long-Term Forest Disturbance Monitoring: Synergizing the LandTrendr Algorithm with CNN in Northeast China. Remote Sensing, 17(21), 3521. https://doi.org/10.3390/rs17213521 [54] Rahaman, M., Southworth, J., Wen, Y., & Keellings, D. (2025). Assessing Model Trade-Offs in Agricultural Remote Sensing: A Review of Machine Learning and Deep Learning Approaches Using Almond Crop Mapping. Remote Sensing, 17(15), 2670. [55] Yu, L., Xie, H., Xu, Y., Li, Q., Jiang, Y., Tao, H., & Aihemaiti, M. (2024). Identification and Monitoring of Irrigated Areas in Arid Areas Based on Sentinel-2 Time-Series Data and a Machine Learning Algorithm. Agriculture, 14(10), 1693. https://doi.org/10.3390/agriculture14101693 [56] Lemenkova, P. (2025). Machine Learning Methods of Remote Sensing Data Processing for Mapping Salt Pan Crust Dynamics in Sebkha de Ndrhamcha, Maurhttps://doi.org/10.3390/rs17152670itania. Artificial Satellites, 60(2). 37-69. https://doi.org/10.2478/arsa-2025-0003 [57] Tangthaikwan, K., Keeratipranon, N., & Agsornintara, A. (2017). Multiclass support vector machine for classification spatial data from satellite image. 2017 9th International Conference on Knowledge and Smart Technology (KST), 111–115. doi:10.1109/KST.2017.7886107 [58] Lemenkova, P. (2024). Support Vector Machine Algorithm for Mapping Land Cover Dynamics in Senegal, West Africa, Using Earth Observation Data. Earth, 5(3), 420–462. doi:10.3390/earth5030024 [59] Jiang, X., Lin, M., & Zhao, J. (2011). Woodland Cover Change Assessment Using Decision Trees, Support Vector Machines and Artificial Neural Networks Classification Algorithms. 2011 Fourth International Conference on Intelligent Computation Technology and Automation, 2, 312–315. doi:10.1109/ICICTA.2011.363 [60] Anurogo, W., Tarigan, A. P. A., Seftyarizki, D., Prihantarto, W. J., Woo, J., dos Santos Catarino, L., Arora, A. S., Gohaud, E., Meller, B., & Schuetze, T. (2025). Cellular Automata–Artificial Neural Network Approach to Dynamically Model Past and Future Surface Temperature Changes: A Case of a Rapidly Urbanizing Island Area, Indonesia. Land, 14(8), 1656. https://doi.org/10.3390/land14081656 [61] Lemenkova, P. (2024c). Random Forest Classifier Algorithm of Geographic Resources Analysis Support System Geographic Information System for Satellite Image Processing: Case Study of Bight of Sofala, Mozambique. Coasts, 4, 127–149. doi:10.3390/coasts4010008 [62] Echevarría-Rubio, J. M., Martínez-Flores, G., & Morales-Pérez, R. A. (2025). Machine and Deep Learning Framework for Sargassum Detection and Fractional Cover Estimation Using Multi-Sensor Satellite Imagery. Data, 10(11), 177. https://doi.org/10.3390/data10110177 [63] Lemenkova, P. (2025). Machine learning algorithms for optimization of image classification in spatially constrained regions: A case of Eritrea, East Africa. Engineering Today, 4(2), 13-27. https://doi.org/10. 5937/engtoday2500008L [64] Sharma, N., Singh, S., & Kaur, K. (2025). Optimised DNN-Based Agricultural Land Mapping Using Sentinel-2 and Landsat-8 with Google Earth Engine. Land, 14(8), 1578. https://doi.org/10.3390/land14081578 [65] Lemenkova, P. (2025). Reclassification Scheme for Image Analysis in GRASS GIS Using Gradient Boosting Algorithm: A Case of Djibouti, East Africa. Journal of Imaging, 11(8), 249. https://doi.org/10.3390/jimaging11080249 166 P. LEMENKOVA [66] Bhaga, T.D., Dube, T., Shekede, M.D., Shoko, C.: Investigating the effectiveness of Landsat-8 OLI and Sentinel-2 MSI satellite data in monitoring the effects of drought on surface water resources in the western cape province, South Africa. Remote Sensing Applications: Society and Environment, 32, 101037 (2023) https://doi.org/10.1016/j.rsase.2023.101037 [67] Feyisa, G.L., Meilby, H., Fensholt, R., Proud, S.R.: Automated water extraction index: A new technique for surface water mapping using Landsat imagery. Remote Sensing of Environment, 140, 23–35 (2014) https://doi.org/10.1016/j.rse.2013.08.029 [68] Zhao, Q., Yu, L., Du, Z., Peng, D., Hao, P., Zhang, Y., & Gong, P. (2022). An Overview of the Applications of Earth Observation Satellite Data: Impacts and Future Trends. Remote Sensing, 14(8), 1863. https://doi.org/10.3390/rs14081863 [69] Czapiewski, S., & Szumińska, D. (2022). An Overview of Remote Sensing Data Applications in Peatland Research Based on Works from the Period 2010–2021. Land, 11(1), 24. https://doi.org/10.3390/land11010024 [70] Lemenkova, P. (2025). Machine Learning Algorithms of Remote Sensing Data Processing for Mapping Changes in Land Cover Types over Central Apennines, Italy. Journal of Imaging, 11(5), 153. https://doi.org/10.3390/jimaging11050153 [71] Rabie, A. B., Elhag, M., & Subyani, A. (2025). Remote Sensing, GIS, and Machine Learning in Water Resources Management for Arid Agricultural Regions: A Review. Water, 17(21), 3125. https://doi.org/10.3390/ w17213125 [72] Nikolić, R. R. (2025). Deforestation in the Special Nature Reserve Gornje Podunavlje: Insights from PlanetScope, Sentinel-2 and Landsat-8 remote sensing data. Tehnika, 80(3), 263-272. https://doi.org/10.5937/ tehnika2503263N [73] Čurlin, M. T., Drobnjak, S. M., Potić, I. M., Đorđević, D. R., & Bakrač, S. T. (2024). Application of machine learning methods in the classification of satellite images. Tehnika, 79(2), 152-158. https://doi.org/10.5937/tehnika2402152C [74] Choi, J. W., Hidayat, M. S., Cho, S. B., Hwang, W.-H., Lee, H., Cho, B.-K., Kim, M. S., Baek, I., & Kim, G. (2025). Recent Trends in Machine Learning, Deep Learning, Ensemble Learning, and Explainable Artificial Intelligence Techniques for Evaluating Crop Yields Under Abnormal Climate Conditions. Plants, 14(18), 2841. https://doi.org/10.3390/plants14182841 [75] Meyer, A., Milani, G., Lack, N., Schönholzer, M., & Jordan, D. (12 2025). Machine learning-based multimodal data fusion for the Swiss land use statistics. European Journal of Remote Sensing, 58(1), 2508420. doi:10.1080/22797254.2025.2508420 [76] Buthelezi, M. N. M., Lottering, R. T., Peerbhay, K. Y., & Mutanga, O. (10 2024). Predicting land use and land cover change dynamics in the eThekwini Municipality: a machine learning approach with Landsat imagery. Journal of Spatial Science, 69(4), 1241–1263. doi:10.1080/14498596.2024.2378362 [77] Miao, X., Heaton, J. S., Zheng, S., Charlet, D. A., & Liu, H. (03 2012). Applying tree-based ensemble algorithms to the classification of ecological zones using multi-temporal multi-source remote-sensing data. International Journal of Remote Sensing, 33(6), 1823–1849. doi:10.1080/01431161.2011.602651