The growing demand for land to accommodate renewable energy infrastructure has intensified competition with biodiversity conservation, agriculture, and ecosystem services. In Portugal, electricity system decarbonisation relies heavily on utility-scale solar energy (USSE) facilities, yet the spatial extent of land transformation associated with photovoltaic development has not been systematically assessed. This study provides an assessment of the land occupancy of USSE facilities and associated land use and land cover (LULC) changes in continental Portugal over the past two decades, as well as their spatial relationship with areas designated for land and nature conservation. A geospatial database of USSE installations (≥1 MW) was developed through the integration of multiple data sources using geographic information systems (GIS). The geometric consistency of spatial features was ensured through harmonisation and validation procedures involving GIS-based corrections supported by Sentinel-2 satellite imagery. Spatial overlay analyses were conducted with multitemporal LULC datasets and with land-use planning constraints, including areas classified for nature conservation, ecological reserves, and agricultural reserves. The results indicate that USSE deployment has been predominantly located in the southern regions of Portugal, although the location of planned projects indicates a northward shift. The implementation of USSE facilities has been mainly associated with LULC changes in forest land, agricultural areas, pastures and shrubland. Spatial overlaps were observed with areas classified within the national ecological and agricultural reserves. These patterns may be indicative of growing land-use conflicts, but the extent to which these developments align with land-use planning objectives and conservation priorities requires further examination.

Abstract The delineation of the Wildland-Urban Interface (WUI) is fundamental for wildfire risk management, yet it is highly sensitive to the underlying building data used. Global building datasets offer unprecedented coverage but may introduce biases by including different types of built-up structures, potentially leading to an overestimation of exposure and a misallocation of critical resources. This study aims to map the WUI and assess wildfire exposure in mainland Portugal by comparing the efficacy of different building datasets: the official residential database (BGE21), Microsoft Building Footprints (MSB24), and the World Settlement Footprint (WSF19). We employed a point-based mapping methodology (100-m radius, > 6.17 buildings/km 2 ) to classify WUI into Intermix and Interface zones. Our analysis revealed that the choice of dataset drastically alters WUI estimates. MSB24, which includes all structure types, identified 67% more buildings than the residential-focused BGE21, resulting in a 73% larger WUI area. Spatial agreement was low, with only 46% of the total WUI area being identified by all three datasets, falling to just 29% for the more vulnerable Intermix zones. While MSB24 showed high recall (0.97 Intermix , 0.99 Interface ), its precision was low (0.43 Intermix , 0.67 Interface ), confirming a significant overestimation of critical zones. Analysis of wildfire perimeters (2000–2023) showed that burned area within the WUI was disproportionately higher in years with smaller total fire extent (e.g., 2006, 2018) rather than in megafire years (e.g., 2003, 2017). We conclude that while global datasets like MSB24 are valuable for emergency response due to their high coverage, their use for preventive planning and resource allocation may cause inefficiencies. We recommend that local authorities prioritize validated residential data, like BGE21, for strategic wildfire prevention and mitigation planning in Portugal to ensure resources are targeted efficiently.

Background: Sustainable vineyard management and planning require reliable methods for identification and monitoring. This systematic review synthesises and appraises the literature on automatic vineyard identification using remote sensing (RS), from classical techniques to artificial intelligence (AI), describing the state of the art, patterns, challenges, and gaps. Methods: Guided by PRISMA and informed by selected SWiM reporting items, we conducted a systematic search across multiple databases, gathering all relevant records up to 13 July 2025, and included 108 sources, of which 80 empirical studies contributed to the synthesis. The risk of bias was assessed by adapting the principles of PROBAST-AI and QUADAS-2 to the agricultural context, covering data representativeness, sensors/pre-processing, ground-truth, validation, and portability; its application also guided the selection and organisation of the synthesis. Results: The analysis was narrative and structured by scale and application objective (regional, parcel, row, and plant). The most common tasks were classification (28%), detection (26%), and segmentation (24%), with multitask pipelines being frequent. We observe a clear transition from pixel-based approaches using satellite imagery to methodologies that integrate very-high-resolution UAV imagery, 3D reconstruction, and Deep Learning (DL). UAVs dominate row and plant-level analyses, whereas Sentinel-2 has become the main tool for multitemporal regional monitoring. DL models, such as CNNs and Vision Transformers (ViTs), tend to deliver superior performance in canopy segmentation and parcel classification. The assessment identified model validation as the weakest methodological domain across studies. Discussion: The main limitations lie in weak spatiotemporal portability of models and high computational costs, aggravated by reliance on large volumes of annotated data. Promising directions include multisensory fusion (UAV + satellite) and the integration of 3D information into DL pipelines, which increase robustness and operational applicability. These advances are enabling high-value, specialised objectives such as mapping in complex terrain, detecting abandoned vineyards, and identifying missing plants.

Urban areas across Europe are undergoing rapid morphological transformations driven by densification, redevelopment, and infrastructure expansion. Monitoring these urban changes requires operational, harmonized, and reproducible approaches grounded in Earth Observation. This study presents a Copernicus use case demonstrating how the High-Resolution Layer Imperviousness Change (2015–2018) and Urban Atlas datasets can be integrated with the Guidos Toolbox (GTB) to quantify structural urban change across six metropolitan areas (Milan, Sofia, Riga, Warsaw, Viseu, Santander). Morphological Spatial Pattern Analysis (MSPA) and entropy-based indicators were applied to characterize land take, fragmentation, compaction, and internal reorganization of impervious surfaces. The combined framework captured both configurational morphology and spatial disorder, revealing divergent development patterns: pronounced heterogeneity and fragmentation in Sofia, stabilization or compact growth in Milan, Warsaw, and Santander, controlled densification in Riga, and localized intensification without outward expansion in Viseu. All analyses rely on openly accessible Copernicus data and open-source tools, ensuring full reproducibility and transferability. Outputs were disseminated through a FAIR-compliant geoportal developed within a Copernicus FPCUP project, supporting transparency and reuse. The findings underscore the value of Copernicus services for operational urban monitoring and provide a scalable methodology to support European land-use policies, including the Zero Net Land Take 2050 target and the EU Soil Strategy.

National Forest Inventories (NFIs) provide valuable land cover (LC) information but often lack spatial continuity and an adequate update frequency. Satellite-based remote sensing offers a viable alternative, employing machine learning to extract thematic data. State-of-the-art methods such as convolutional neural networks rely on fully pixel-level annotated images, which are difficult to obtain. Although reference LC datasets have been widely used to derive annotations, NFIs consist of point-based data, providing only sparse annotations. Weakly supervised and self-supervised learning approaches help address this issue by reducing dependence on fully annotated images and leveraging unlabeled data. However, their potential for large-scale LC mapping needs further investigation. This study explored the use of NFI data with deep learning and weakly supervised and self-supervised methods. Using Sentinel-2 images and the Portuguese NFI, which covers other LC types beyond forest, as sparse labels, we performed weakly supervised semantic segmentation with a convolutional neural network to create an updated and spatially continuous national LC map. Additionally, we investigated the potential of self-supervised learning by pretraining a masked autoencoder on 65,000 Sentinel-2 image chips and then fine-tuning the model with NFI-derived sparse labels. The weakly supervised baseline achieved a validation accuracy of 69.60%, surpassing Random Forest (67.90%). The self-supervised model achieved 71.29%, performing on par with the baseline using half the training data. The results demonstrated that integrating both learning approaches enabled successful countrywide LC mapping with limited training data.

The current land cover (LC) mapping paradigm relies on automatic satellite imagery classification, predominantly through supervised methods, which depend on training data to calibrate classification algorithms. Hence, training data have a critical influence on classification accuracy. Although research on specific aspects of training data in the LC classification context exists, a study that organizes and synthetizes the multiplicity of aspects and findings of these researches is needed. In this article, we review the training data used for LC classification of satellite imagery. A protocol of identification and selection of relevant documents was followed, resulting in 114 peer-reviewed studies included. Main research topics were identified and documents were characterized according to their contribution to each topic, which allowed uncovering subtopics and categories and synthetizing the main findings regarding different aspects of the training dataset. The analysis found four research topics, namely construction of the training dataset, sample quality, sampling design and advanced learning techniques. Subtopics included sample collection method, sample cleaning procedures, sample size, sampling method, class balance and distribution, among others. A summary of the main findings and approaches provided an overview of the research in this area, which may serve as a starting point for new LC mapping initiatives.

Recent advances in satellite data availability, computing storage and processing power introduced a new land cover monitoring paradigm, settled on a continuous and timely identification of changes. The Continuous Change Detection and Classification (CCDC) algorithm has emerged as a powerful tool for continuous monitoring, being noteworthy for its ability to process high temporal frequency satellite data with components of seasonality, trend and break. Studies using CCDC were mostly limited to Landsat data, which offer lower spatial and temporal resolution in comparison to Sentinel-2 data. Therefore, our study aims to explore the potential of CCDC with Sentinel-2 data. For that purpose, an extensive reference dataset was developed for change detection accuracy assessment, comprising 290 sites of 200 m radius in a disturbance prone region in Central Portugal, ensuring an adequate representation of areas of vegetation loss. We focused on two specific forest species from this region, eucalyptus and maritime pine. Change date was determined through interpretation of orthophotos and satellite time series. We explored determinant aspects to CCDC performance, namely cloud and cloud shadow masking, algorithm parameterization, use of distinct vegetation indices and detection timeliness. Optimal accuracy was achieved with s2cloudless masking, lambda of 200, chi-square of 0.999, minYears of 1 and the Normalized Difference Vegetation Index. We computed the time lag vs omission error curve, showing comparable results (omission error rate close to 20 % was obtained with a time lag from 30 to 40 days) to methods designed to achieve near-real-time detection. Detections were spatially coherent, with patches of vegetation loss detected only with minor errors, mostly located in polygon borders. Disturbances in the first months resulted in poor model fitting, which undermined detection performance in some cases. Overall, results demonstrated how CCDC and Sentinel-2 data can be used to successfully monitor vegetation loss in a timely manner, especially as the satellite's time series grows.

The Wildland-Urban Interface (WUI), where vegetation and built-up structures intermingle, encompasses a variety of territorial elements that interact spatially, being variable both in space and time. Mapping the WUI at finer scales is paramount to assess wildfire exposure and define tailored mitigation strategies. Our aim was to develop a semi-automated method to map the WUI at municipal level, leveraging recent advances in data and technology. We tested the procedure in four municipalities of mainland Portugal with different fire history, biophysical conditions, and sociodemographic contexts. We considered WUI as either intermix or interface. Our approach integrates both building location data and high-resolution vegetation maps, to calculate the density of buildings and forest cover proportion within different circular moving window sizes. Within each radius, we evaluated the total area and spatial distribution of the WUI types, as well as the number of buildings within WUI and within the fire perimeters recorded between the years 2000 and 2022 and analysed the differences between municipalities. We then compared the mapped WUI with previous WUI mappings for mainland Portugal, to identify common spots and potential spatial divergences. We found that the area mapped as WUI within all four municipalities ranged from about 400 km2 to 1135 km2 depending on the radius size. A distinct distribution for each type of WUI was observed as the radius size increased: the intermix WUI showed a tendency to increase, and the interface WUI increased only between the radius of 100 and 200 m, decreasing gradually in subsequent radii. Between 39.4% and 45.5% of the nearly 200,000 buildings in the study areas were within WUI, depending on radius size and a total of 5436 buildings were within the historic fire perimeter. Although the comparison with other maps showed fair agreement, due to differences in data and methodology, common areas mapped as WUI were found, which suggests that these areas should receive greater attention from decision-makers regarding fire management strategies, since their classification as WUI remains consistent across different methodologies.

The current land cover (LC) mapping paradigm relies on automatic satellite imagery classification, predominantly through supervised methods, which depend on training data to calibrate classification algorithms. Hence, training data have a critical influence on classification accuracy. Although research on specific aspects of training data in the LC classification context exists, a study that organizes and synthetizes the multiplicity of aspects and findings of these researches is needed. In this article, we review the training data used for LC classification of satellite imagery. A protocol of identification and selection of relevant documents was followed, resulting in 114 peer-reviewed studies included. Main research topics were identified and documents were characterized according to their contribution to each topic, which allowed uncovering subtopics and categories and synthetizing the main findings regarding different aspects of the training dataset. The analysis found four research topics, namely construction of the training dataset, sample quality, sampling design and advanced learning techniques. Subtopics included sample collection method, sample cleaning procedures, sample size, sampling method, class balance and distribution, among others. A summary of the main findings and approaches provided an overview of the research in this area, which may serve as a starting point for new LC mapping initiatives.

The free availability of Sentinel-1 and 2 imageries enables the production of high resolution (10 m) global Land Use Land Cover (LULC) maps by a wide range of institutions, which often make them publicly available. This raises several issues: Which map should be used for each type of application? How accurate are these maps? What is the level of agreement between them? This motivated us to assess the thematic accuracy of six LULC maps for continental Portugal with 10 m spatial resolution with reference dates between 2017 and 2020, using the same method and the same reference database, in a bid to make the results comparable. The overall accuracy and the per class user’s and producer’s accuracy are compared with the ones reported by the map producers, at the national, European, or global level, according to their availability. The nomenclatures of the several maps were then analyzed and compared to generate a harmonized nomenclature to which all maps were converted into. The harmonized products were compared directly with a visual analysis and the proportion of regions equally classified was computed, as well as the area assigned per product to each class. The accuracy of these harmonized maps was also assessed considering the previously used reference database. The results show that there are significant differences in the overall accuracy of the original products, varying between 42% and 72%. The differences between the user’s and producer’s accuracy per class are very large for all maps. When comparing the obtained results with the ones reported by the map producers for Portugal, Europe or globally (depending on what is available) the results obtained in this study have lower accuracy metrics values for all maps. The comparison of the harmonized maps shows that they agree in 83% of the study area, but there are differences in terms of detail and area of the classes, mainly for the class “Built up” and “Bare land”.

The Portuguese Land Cover Monitoring System SMOS (Sistema de Monitorização da Ocupação do Solo) is an initiative conceived and developed by Direção-Geral do Território (DGT) with the aim of continuously produce reference and thematic cartographic products on land use land cover for continental Portugal. SMOS is a collaborative and multifunctional system that involves the Public Administration, the national scientific and technological system, the private sector and the citizen, guided by user needs and open data policy. SMOS represents a paradigm shift in cartography production and uses the latest developments in space technologies and Artificial Intelligence to create maps with more detail, quality and promptness. Here, SMOS is presented, including the initial steps leading to the implementation of the system, its organization and governance, cartographic products and viewers, and foreseen developments.

Land Use and Land Cover (LULC) datasets are widely used across disciplines, with many users demanding more and better information. Understanding the uncertainties and errors associated to the main LULC datasets is a required step to facilitate their correct use, as well as to identify what could be improved in the future production of these products. CORINE Land Cover is probably the most well-known and used LULC dataset in Europe, especially valuable for the rich time-series that it provides. Despite being produced through a change mapping first approach, which tries to avoid technical errors and uncertainties in the temporal analysis of LULC changes, the Copernicus Land Monitoring Service distributes status layers of CORINE (CSL), which are not valid for change analysis because of their associated errors and uncertainties. The CORINE layers of changes (CHA) remove a lot of these issues, but do not meet the needs of many users. In Portugal, the national authority in charge of producing CORINE, the DGT, has implemented a backdating approach to produce consistent CSL layers that allow change analysis with low levels of uncertainty. Throughout this paper, we evaluate the changes that can be analyzed through all available CORINE layers in Portugal: Copernicus CSL layers; the national DGT CSL layers; and CHA layers. To this end, we aim to assess what type of changes can be studied through each type of layer, their associated sources of uncertainty and the relevance and utility of the Portuguese backdating approach to produce a consistent time-series of LULC maps. The results prove how the Portuguese CORINE layers distributed by the Copernicus Land Monitoring Service contain important sources of uncertainty, which however have been removed through the national backdating methodology. This methodology can be therefore exported for the production of CORINE in other European countries.

<p>Wildfires occur unevenly in the territory, driven by different biophysical and social factors. Understanding the spatial and temporal distribution of wildfires can help identifying common characteristics and/or dissimilarities between regions. In this research, we use specific fire metrics, from historical fire data, to explore the possibility to identify groups of municipalities based on their pyrosimilarities. We apply a clustering model based on the method k means to identify and compare groups of municipalities (defining pyroregions) of mainland Portugal (n=277), using fire data from the last 22 years (between 2000 and 2021). The fire metrics used were: (a) cumulative percentage of total burned area, (b) cumulative percentage of burned area in the summer months, (c) mean annual number of fires and (d) GINI index applied for burned area over time. We used tools available in Geographic Information Systems (ArcGIS Pro) linked with python programming, to apply the cluster method and map the results. Our preliminary results divided the mainland in 5 clusters. CL1 (n=66) is seen in the west coast and is characterised by a burned area concentrated in a few years (high Gini index), but the fires occur mainly outside the summer months; CL2 (n=50) cover municipalities in the northeast and is characterised by a high mean number of fires dispersed over the years (low GINI index); CL3 (n=26) located in the central Portugal has a high percentage of cumulative burned area throughout the years, but with low number of fires, concentrated in time; CL4 (n=63) covers the municipalities in the southwest and south and shows a low mean number of fires but  these occur mainly in the summer season and, CL5 (n=55) appears throughout the country, but is more concentrated in the west and is characterised by intermediate values in all analysed metrics. The extreme wildfires that occurred in 2017 in Portugal influence the clustering; for example, CL1 occurs on the west central coast, where the consolidated maritime pine forest has burned in October 2017, outside the summer months. The next steps of this analysis are: (i) apply other clustering methods to compare with these clusters identified with k means and their characteristics and (ii) analyse the explanatory variables that influence these fire patterns.</p> <p> </p> <p>This work was funded by FCT, I.P.: BB and AG in the scope of PhD projects [2022.12095.BD], [2020.07651.BD], SO under the contract ‘2020.03873.CEECIND′, Centre for Geographical Studies—University of Lisbon and FCT under Grant number [UIDB/00295/2020 + UIDP/ 00295/2020].</p>

This work presents a comparison between a global and a national land cover map, namely the ESA WorldCover 2020 (WC20) and the Portuguese use/land cover map (Carta de Uso e Ocupação do Solo 2018) (COS18). Such a comparison is relevant given the current amount of publicly available LULC products (either national or global) where such comparative studies enable a better understanding regarding different sets of LULC information and their production, focus and characteristics, especially when comparing authoritative maps built by national mapping agencies and global land cover focused products. Moreover, this comparison is also aimed at complementing the global validation report released with the WC20 product, which focused on global and continental level accuracy assessments, with no additional information for specific countries. The maps were compared by following a framework composed by four steps: (1) class nomenclature harmonization, (2) computing cross-tabulation matrices between WC20 and the Portuguese map, (3) determining the area occupied by each harmonized class in each data source, and (4) visual comparison between the maps to illustrate their differences focusing on Portuguese landscape details. Some of the differences were due to the different minimum mapping unit ofCOS18 and WC20, different nomenclatures and focuses on either land use or land cover. Overall, the results show that while WC20 detail is able to distinguish small occurrences of artificial surfaces and grasslands within an urban environment, WC20 is often not able to distinguish sparse/individual trees from the neighboring cover, which is a common occurrence in the Portuguese landscape. While selecting a map, users should be aware that differences between maps can have a range of causes, such as scale, temporal reference, nomenclature and errors.

Change detection in highly dynamic areas, such as vegetated areas, is necessary to improve the thematic accuracy of national land use/cover map production. Sentinel-1 data have proved its potential for change detection. However, its suitability in mountainous and sparse forested areas is not well known. Here, we propose a straightforward clear-cut detection method, using monthly backscatter and coherence composites in both polarizations. The monthly clear-cut masks are estimated using a multivariate alteration detection (MAD) algorithm, using the previous and following composite, and a threshold value corresponding to the 98th percentile. The method was applied to a test site in northern Portugal, mainly characterized by the presence of dense eucalyptus forest, in a mountainous area. A Sentinel-1 time series from February to October 2018 was considered. Results showed that the monthly clear-cut estimation is not only highly influenced by rainfall events, with F1-score values less than 0.11, for the rainy season, but also by terrain-induced geometric distortions and foliage characteristics of the eucalyptus stands. The highest accuracy metrics were obtained for June, with an F1-score value of 0.45, which was still considered unfavorable. Inaccurate monthly estimations are also affected by clear-cut events that occur over more than one month, with the sum of all masks producing results with F1- scores less than 0.41. When applied to another region, located southwest of the former site and near the coastline, but characterized by smoother slopes and scattered and sparse forested stands, the method retrieved equivalent recall, but increased precision of around 0.78 (false alarms reduction), resulting in higher F1-score values (0.56).

Land use/land cover (LULC) change detection and classification in maps based on automated data processing are becoming increasingly sophisticated in Earth Observation (EO).There is a growing number of annual maps available, with diverse but related production structures consisting primarily of classification and postclassification phases, the latter of which deals with inaccuracies of the first.The methodology production of the "Carta de Ocupação do Solo conjuntural" (COSc), a thematic land cover map of continental Portugal produced by the Directorate-General for Territory (DGT) mostly based on Sentinel-2 images classification, includes a semi-automatic phase of correction that combines expert knowledge and ancillary data in if-thenelse rules validated by photointerpretation.Although this approach reduces misclassifications from an initial Random Forest (RF) prediction map, improving consistency between years and compliance with ecological succession, requires a lot of time-consuming semi-automatic procedures.This work evaluates the relevance of exploring an additional set of variables for automatic classification over disturbance-prone areas.A multitemporal dataset with 124 variables was analysed using data dimensionality reduction techniques, resulting in the identification of 35 major explanatory indicators, which were then used as inputs for RF classification with cross-validation.The estimated importance of the explanatory variables shows that composites of spectral bands, which are already included in the current COSc workflow, in conjunction with the inclusion of additional data namely, historical land cover information and change detection coefficients, from the Continuous Change Detection and Classification (CCDC) algorithm, are relevant for predicting land cover classes after disturbance.Since map updating is a more challenging task for disturbed pixels, we focused our analysis on locations where COSc indicated potential land cover change.Nonetheless, the overall classification accuracy for our experiments was 72.34 % which is similar to the accuracy of COSc for this region of Portugal.The findings suggest new variables that could improve future COSc maps.

The current land cover mapping paradigm relies on automatic classification of satellite images, with supervised methods being the most used, implying training data to have a crucial role. Aspects such as training sample size and quality should be carefully considered. This paper proposes assessing the use of a detailed class nomenclature to reinforce class diversity in the training sample. A Random Forest (RF) classification of Sentinel-2 multi-temporal data was conducted. Additionally, the effect of sample size and class distribution were evaluated. The results indicate that the use of a detailed nomenclature provided better results in terms of classification accuracy. With respect to sample distribution, adopting class sizes proportional to their occurrence in a reference land cover map exhibited superior performance in comparison to an equal size approach. The effect of sample size on classification performance was limited, as previous studies with RF suggested.

Unmanned aerial vehicles (UAVs) are platforms that have been increasingly used over the last decade to collect data for forest insect pest and disease (FIPD) monitoring. These machines provide flexibility, cost efficiency, and a high temporal and spatial resolution of remotely sensed data. The purpose of this review is to summarize recent contributions and to identify knowledge gaps in UAV remote sensing for FIPD monitoring. A systematic review was performed using the preferred reporting items for systematic reviews and meta-analysis (PRISMA) protocol. We reviewed the full text of 49 studies published between 2015 and 2021. The parameters examined were the taxonomic characteristics, the type of UAV and sensor, data collection and pre-processing, processing and analytical methods, and software used. We found that the number of papers on this topic has increased in recent years, with most being studies located in China and Europe. The main FIPDs studied were pine wilt disease (PWD) and bark beetles (BB) using UAV multirotor architectures. Among the sensor types, multispectral and red–green–blue (RGB) bands were preferred for the monitoring tasks. Regarding the analytical methods, random forest (RF) and deep learning (DL) classifiers were the most frequently applied in UAV imagery processing. This paper discusses the advantages and limitations associated with the use of UAVs and the processing methods for FIPDs, and research gaps and challenges are presented.

Monitoring land-use patterns and its trends provides useful information for impact evaluation and policy design. The latest in-depth studies of land-use dynamics for continental Portugal are outdated, and have not examined how municipalities may be classified into a typology of observed dynamics or considered the trajectory profiles of land-use transitions. This paper presents a comprehensive analysis of the spatiotemporal dynamics of land-use in continental Portugal from 1995 to 2018. Our multi-scalar approach used land-use maps in geographic information systems with the following objectives: (i) quantify variations of land-use classes, (ii) assess the transitions between uses, and (iii) derive a municipal typology of land-use dynamics. The methodology employed involved calculating statistical indicators of land-use classes, transition matrices between uses and combinatorial analysis for the most common trajectory-profiles. For the typology, a principal component analysis was used for dimensionality reduction and the respective components were classified by testing several clustering techniques. Results showed that the land-use transitions were not homogeneous in space or time, leading to the growth of territorial asymmetries. Forest (Δ5%), water bodies (Δ28%) and artificial surfaces (Δ35%) had a greater expansion, as opposed to agricultural areas, which had the biggest decline (Δ-8%). Despite the decline of agricultural activities, olive-grove expansion (Δ7%) was a relevant dynamic, and in the case of forests, the increment of eucalyptus (Δ34%) replaced native species such as the maritime pine (Δ-20%). A land-use-dynamics typology was estimated, dividing continental Portugal into 11 clusters, which is informative for sectoral policies and spatial planning, as zonings in need of interventions tailored to their specificities. The findings are a contribution to the study of land-use dynamics in continental Portugal, presenting various challenges for sustainable land uses with regard to the urban system, forest management, food production, soil preservation, and ecosystem protection.

Portugal is building a land cover monitoring system to deliver land cover products annually for its mainland territory. This paper presents the methodology developed to produce a prototype relative to 2018 as the first land cover map of the future annual map series (COSsim). A total of thirteen land cover classes are represented, including the most important tree species in Portugal. The mapping approach developed includes two levels of spatial stratification based on landscape dynamics. Strata are analysed independently at the higher level, while nested sublevels can share data and procedures. Multiple stages of analysis are implemented in which subsequent stages improve the outputs of precedent stages. The goal is to adjust mapping to the local landscape and tackle specific problems or divide complex mapping tasks in several parts. Supervised classification of Sentinel-2 time series and post-classification analysis with expert knowledge were performed throughout four stages. The overall accuracy of the map is estimated at 81.3% (±2.1) at the 95% confidence level. Higher thematic accuracy was achieved in southern Portugal, and expert knowledge significantly improved the quality of the map.