ABSTRACT. One-quarter of the world's urban population resides in informal settlements, a figure projected to grow in coming years. These settlements house approximately 1 billion people who often face precarious living conditions, including substandard housing, insecure land tenure, and limited access to essential services like clean water, electricity, and sanitation. While improving these conditions remains an urgent global priority, rapid urban growth has led to outdated and unreliable data on these areas, hindering effective urban planning and service delivery.

To address this challenge, the United Nations Innovation Technology Accelerator for Cities (UNITAC Hamburg) developed the Building and Establishment Automated Mapper (BEAM). This machine learning tool uses aerial imagery to map informal structures and was tested in partnership with the Human Settlement Unit of eThekwini Municipality in South Africa. This research examines the challenges of mapping informal settlements through the lens of BEAM's development and implementation in eThekwini.

ABSTRACT. The growth of deprived urban areas (DUA), often associated with slums or informal settlements, is one of the consequences of rapid urbanization. Earth Observation (EO) data provides valuable information for mapping and monitoring such urban areas to assess the Sustainable Development Goal (SDG) indicator 11.1.1. Previous studies show that building density is one of the most informative morphometric variables to map DUA. However, building density when available are often mono-temporal and lack information about the exact date it was assessed. To address this gap, we present a deep learning-based approach that integrates building density regression from EO data to guide the learning process of a pixel-wise classification network. Our methodology optimizes the combined loss function of a dual-output semantic segmentation model, balancing classification and regression tasks, using Sentinel-1 and Sentinel-2 as input. This balance improves the accuracy of building density predictions, which, in turn, enhances the detection of DUAs and model interpretability. We evaluated our approach in Salvador (Brazil) and Nairobi (Kenya), achieving improvements of 9.75\% and 0.60\%, respectively, compared to previous studies.

ABSTRACT. Over one billion people globally live in slums, informal settlements, and other deprived areas. However, maps of deprived areas are often unavailable or over-simplistic, distinguishing only between slums and formal areas. Recent research advocates for a multidimensional approach to better account for the complexity of deprivation. Previous studies have mapped the unplanned urbanization domain of deprivation using morphometrics derived from building footprint data. This study explores leveraging Earth observation data for scalability and regular updates by using high-resolution satellite imagery and deep learning to map indicators for morphological informality in Nairobi, Kenya. The proposed model combines a ResNet backbone with three classification heads to map the two indicators: irregular settlement layout (ISL) and small, dense structures (SDS), alongside building presence. The model was trained on automatically generated reference data using building footprint morphometrics and clustering, and its outputs were validated through community-sourced annotations obtained via participatory action research. The results demonstrate the potential of high-resolution satellite imagery for mapping ISL (F1 80.90) and SDS (F1 78.73). Nonetheless, further research is required on the geographic transferability of the proposed method.

ABSTRACT. Artificial Intelligence (AI) and Earth Observation (EO) are revolutionizing global efforts to achieve the United Nations’ (UN) Sustainable Development Goals (SDGs). EO provides essential multiscale data for monitoring natural ecosystems, resources, and infrastructure, enabling cost-effective and evidence-based decision-making. This paper introduces SmartSDGTunisia, a framework designed to address Tunisia’s specific challenges in achieving SDGs 1, 2, 6, 11, 13, and 15. By integrating AI and EO technologies, the framework aims to enhance monitoring, evaluation, and progress toward sustainable development goals.

ABSTRACT. This study assesses the impact of climate change on the probability of desertification in Africa. Using CMIP6 climate models and a comparative analysis of projections based on SSP245 and SSP585 scenarios. We classified regions at risk of desertification using machine learning models such as Random Forest (RF) and performed spatial analysis using Google Earth Engine (GEE). With an accuracy of 0.8991, the GFDL-ESM4 model combined with RF shows a remarkable overall accuracy. The study predicts how desertification will change under the SSP245 and SSP585 scenarios, with an increase in the probability of desertification, particularly in the Sahel and sub-Saharan Africa. Both scenarios increase the number of critical and vulnerable areas, particularly in semi-arid regions, highlighting the importance of sustainable land management and emission reductions to limit the impacts of climate change.3- Experiment and Results.

ABSTRACT. Urban trees have proven an effective adaptation and mitigation strategy against the background of ongoing urbanization and climate change. However, area-wide data on trees in cities is still largely missing. Furthermore, ecosystem services provided by urban trees have only been quantified for individual trees within the urban forest to date. This study presents an integrative approach based on remote sensing for detection and characterization of urban trees, which serves as input for the modeling of ecosystem services using the process-based tree growth model “CityTree”. In addition, the remote sensing-based parameter setting is compared to in-situ measurements to assess the plausibility of the ecosystem services estimation. The results show that tree dimensions from remote sensing generally agree well with in-situ measurements. Furthermore, the classification of tree genera based on multi-temporal very-high resolution (VHR) satellite data achieved good accuracies around 70%. Finally, this study demonstrates the effectiveness of remote sensing-based individual tree parameters for the estimation of ecosystem services, while the comparison with in-situ parameters confirmed the plausibility of the results.

ABSTRACT. The role of green spaces in sustaining a good quality of life for urban dwellers has been widely appreciated. Earth observation has reached a next level in terms of deep time-series and unprecedented processing means, in order to monitor the actual quality of vegetation, in green spaces, for large regions. We exploit satellite big data to estimate NDVI and then convert it to vegetation percentage, for the largest urban areas of Greece, using the pixel dichotomy method, for the past two decades. Interesting trends appear in the results, both in terms of the overall change of vegetation and in the spatial distribution of greening and browning areas.

ABSTRACT. Megacities, as epicenters of population growth and economic activity, contribute significantly to atmospheric emissions. Despite numerous studies either examining urban growth or air pollution trends in megacities, a systematic analysis that integrates both quantities within the context of economic development remains lacking. This study addresses this gap by leveraging Earth Observation (EO) data to investigate trends in settlement growth and tropospheric NO2 pollution across 38 megacities from 1996 to 2015. Using the World Bank’s income classifications, it is hypothesized whether the observed dynamics follows the Environmental Kuznets Curve (EKC), which posits a non-linear relationship between economic growth and environmental quality. The findings reveal an average annual NO2 increase of 5.06 ± 0.83% and settlement growth of 2.87%, with substantial variation across income groups. High-income cities show declining NO2 trends, attributed to advanced emission controls and stringent environmental policies, yet still exhibit high per capita pollution levels. Conversely, upper-middle-income cities experience rapid NO2 increases driven by industrial activity and urban sprawl. Settlement growth dynamics were observed to correlate with economic development phases, with upper-middle-income cities leading in urban expansion. The results support the theoretical framework of the EKC hypothesis, highlighting a non-linear interplay between urbanization, economic development and environmental quality. These insights into socio-economic transitions underscore the necessity of differentiated strategies, including investments in green technology, stricter emission standards, and enhanced international collaboration, to foster sustainable urbanization.

ABSTRACT. Urban Heat Islands (UHI) represent a significant challenge in cities, increasing heat stress and impacting public health. Therefore, it is important to promote mitigation measurements, such as planting trees. This study investigates the relationship between land surface temperature (LST) and tree canopy cover across different urban land use classes in the city of Munich, Germany. Therefore, we extracted mean LST and percentage of tree canopy cover from multi-sensor earth observation (EO)-data. We related these data to five land use categories: Residential, Industrial, Recreational, Traffic, and Mixed. We used violin plots and linear regression models to analyze the relationships. Our analyses revealed a consistent negative relationship between tree canopy cover and LST across all land use categories, with coefficients ranging from -0.038 °C/%-of tree cover in Recreational areas to -0.083 °C/%-of tree cover in Mixed areas. These findings highlight the importance of urban trees to reduce the LST, particularly in areas affected by high temperatures with limited vegetation.

ABSTRACT. Recent widespread marine mucilage formations in the Mediterranean, particularly in Türkiye and Italy, have raised significant environmental concerns. Monitoring and analyzing these events are crucial for the protection of marine ecosystems. Recently, unmixing-based approaches on hyperspectral satellite data have been proposed for marine mucilage monitoring, offering not only promising performance, but also key advantages such as not requiring labeled training data or carefully selected thresholds. Additionally, accounting for spectral variability has been shown to result in increased monitoring performance. Building on the previous studies, this paper presents marine mucilage monitoring using unmixing with the Perturbed Linear Mixing Model (PLMM). The recent mucilage event off the coast of Italy is investigated, on hyperspectral data acquired by the Earth Surface Mineral Dust Source Investigation (EMIT) mission. Experimental results demonstrate the approach's effectiveness in mucilage monitoring, and have the potential to provide new insights.

ABSTRACT. Image classification is a pivotal task in remote sensing applications, essential for extracting valuable insights across various sectors such as environmental monitoring and urban planning. However, the process is fraught with challenges, including the selection of suitable classification algorithms and the interpretation of satellite imagery. Therefore, this study addressed these challenges by comparing the performance of different image classification algorithms, namely; Support Vector Machine (SVM), Maximum Likelihood Classifier (MLC), Random Forest (RF), and K-Nearest Neighbor (KNN), for land use/land cover (LULC) classification in the Ibadan metropolis. Utilizing Sentinel 2A and Landsat 8 satellite data, the research evaluates the algorithms' accuracy and effectiveness in classifying water, built-up areas, bare land, and vegetation for the years 2015, 2019, and 2023. AgroR package on R 4.4.1 was used to perform the Two-way Analysis of Variance (ANOVA) and the Duncan Multiple Range Test (DMRT) was used to assess the significant difference between the algorithms and the satellite sensor. The results highlight significant changes in land cover dynamics, with urbanization driving the expansion of built-up areas at the expense of vegetation. Statistical analysis of the overall accuracies revealed that there were no significant differences among the classification algorithms, with MLC having the highest mean value of 0.891, followed closely by SVM (0.885), KNN (0.877), and RF (0.866). The study underscores the importance of comparative analysis in selecting the most suitable algorithm for accurate land cover mapping, providing valuable insights for urban planners and policymakers to address the challenges of rapid urbanization and land cover changes in developing cities like Ibadan.

ABSTRACT. This work demonstrates how legacy archives of daylight satellites can be utilized to inform urban studies with long-spanning panel data on building stock evolution. We generate an 11-year panel data set with 5-meter resolution of Beijing. We predict building footprint and height with a U-Net taking numerous scenes from the RapidEye archive. We introduce a rigorous post-processing pipeline to address prediction variation over time with a change point detection algorithm, Pruned Exact Linear Time (PELT). The method is straightforward and does not require any longitudinal labels that are hard to obtain.

ABSTRACT. The need for high-resolution remote sensing images is growing rapidly, driven by applications in urban planning, environmental monitoring, and resource management. This study introduces a novel framework for urban remote sensing image super-resolution by integrating Denoising Diffusion Probabilistic Models (DDPM) with Generative Adversarial Networks (GANs). The hybrid model leverages DDPM's robust denoising and feature reconstruction capabilities alongside GANs' ability to enhance perceptual fidelity and spatial detail. Evaluated on benchmark datasets (NWPU-RESISC45, UCMerced Land Use, and Sfax aerial imagery), the model achieves a PSNR of 31.60 dB, SSIM of 0.81, and LPIPS of 0.31, outperforming existing state-of-the-art methods. With an inference time of 2.70 seconds per image, the framework is computationally efficient and well-suited for large-scale urban applications.

ABSTRACT. Geological maps play a pivotal role in urban planning providing essential information about surface and subsurface conditions to ensure safe and sustainable infrastructure development. This paper examines the use of machine learning and deep learning techniques in geological mapping for urban environments, focusing on both regional analysis and detailed assessments. For regional mapping, medium-resolution data like Landsat, ASTER, and Sentinel-2 were processed with classical ML algorithms, such as Support Vector Machines (SVM), to classify lithological units and detect geological features, offering valuable insights during the preliminary phases of urban planning. In contrast, detailed geological mapping requires high-resolution datasets. Advanced DL models, including convolutional neural networks, recurrent neural networks, and Graph Convolutional Networks, address challenges like data complexity, noise, and class imbalance. A case study conducted on the northern and western margins of Tunis highlights the practical application of various techniques. Lithological unit classification was performed using SVM. The choice of SVM was based on a comparative analysis with multiple supervised classifiers. The data utilized included ASTER VNIR-SWIR and SRTM DEM data, and the results demonstrated an overall accuracy of 78.44%. Despite its strengths, the study encountered challenges related to imbalanced datasets and temporal variability in the remote sensing data. Potential solutions, such as synthetic data generation via Generative Adversarial Networks and weighted loss functions, are proposed to improve model performance.

ABSTRACT. Urbanization transforms rural areas into urban spaces, leading to notable social, economic, and environmental changes. This process is often associated with human activity increase, transport, and energy consumption, resulting in higher carbon monoxide (CO) emissions from vehicular traffic and industrial activities. Additionally, urban development often reduces green spaces due to infrastructure and construction projects, aggravating city air pollution. In this context, the work presented in this article conducts a temporal analysis of the relationship between vegetation indices, such as NDVI, SAVI, and MSAVI, with the atmospheric pollutant carbon monoxide (CO), and land occupation across different regions of Algeria. The study is based on five years of data, covering the four seasons of each year, using Sentinel-5P data to assess carbon monoxide emissions and Sentinel-2 data to determine the three vegetation indices NDVI, SAVI and MSAVI. The results of the temporal analysis revealed some interesting elements. Over the first years considered in the study, an inverse correlation is observed between vegetation indices and carbon monoxide, confirming that the presence of vegetation directly influences the concentration of carbon monoxide in the air. However, the latest annual data series show a decrease in the estimated concentration of carbon monoxide (CO), independently of the increase or decrease in vegetation indices. This may be explained by the introduction of new measures to reduce carbon emissions.

ABSTRACT. Supervised learning is by far the dominant train- ing paradigm in the context of semantic segmentation models. However, it assumes implicitly that the deployment (or target) data follow the same distribution as the training data. And unfortunately this assumption is often violated in practice, leading to the infamous domain shift, coupled with degraded segmentation performances. Domain adaptation is a common approach for addressing it, and involves adapting a model to the target domain’s data. Domain generalization on the other hand, assumes no access to target domain data, and aims to exploit the source (or training) domains’ data, in such a way so as to achieve maximum performance with the unseen target data. This paper focuses on domain generalization for the semantic segmentation of optical remote sensing images, through an adversarial strategy, and contrary to the state-of- the-art, investigates the concurrent use of feature maps from multiple layers instead of only the bottleneck features. Through a proper selection of loss weights, we encourage the model to learn progressively more domain invariant features across the model’s layers. The proposed approach is validated with the FLAIR dataset, where it achieves superior performance w.r.t. state-of- the-art studies.

ABSTRACT. In modern cities, Urban Heat Islands (UHI) present a major challenge for sustainable development and environmental management. This phenomenon, characterized by a significant temperature differences between urban and rural areas, negatively impacts public health, energy consumption, and biodiversity. In response to these challenges, public authorities and researchers are increasingly interested in utilizing advanced technologies to better understand and mitigate the effects of UHIs. In Algeria, the integration of climatic and environmental data into innovative tools, such as Webmapping systems, plays a crucial role in monitoring and managing these heat islands. In this work, an interactive platform is developed, with the aim to allow for a detailed, real-time visualization of urban temperatures and climate dynamics, enabling informed decision-making by local authorities, urban planners, and researchers. Webmapping, as a tool of territorial intelligence, helps analyze and plan strategies for mitigating UHI effects, contributing to a more resilient urban development in the face of climate challenges.

ABSTRACT. In this paper, we introduce a novel method for fusing hyperspectral and multispectral images to generate an unobservable hyperspectral image with high spatial and spectral resolution by combining the strengths of both. This work is based on an extended version of the Linear Mixing Model (LMM) and incorporates a spectral library approach to address spectral variability (SV). The proposed method consists of four stages. The first stage involves constructing the spectral library using an unmixing approach designed to handle spectral variability. The second stage employs a spectral library pruning strategy. The third stage extracts spatial information from the multispectral image, and finally, the fourth stage generates the fused product. The proposed approach is tested on synthetic data, the obtained results demonstrates clearly the effectiveness of the introduced method by providing fusion product with higher spectral and spatial fidelities.

ABSTRACT. The rapid growth of urban areas and human activi ties makes automatic change detection in remote sensing essential for tasks such as monitoring urban growth, environmental changes, and disasters. Deep learning methods have shown great promise for these tasks, yet they often lack transparency, making it challenging to build trust in their results. This study addresses model interpretability in change detection by combining the predictive accuracy of deep learning with explainable artificial intelligence (XAI) techniques. A Siamese U-Net model, paired with Grad-CAM and Grad-CAM++ visualizations, is utilized to produce intuitive visual explanations of the model’s predictions. Validated three change detection datasets, namely CDD, LEVIR CD, and SyntheWorld, which contain high resolution optical satellite images from various urban and natural environments, the presented approach enhances the interpretability of deep learning-based change detection by providing insights into how the model classifies each pixel as ”change” or ”no change”, and aims to improve confidence in its outputs.

ABSTRACT. The objective of this paper is to analyse the impacts of ICT integration in the energy sector and in the achievement of sustainable development goals in the European Economic Area countries. We investigate the evidence for a modified environmental KUZNETS curve (EKC) for a group of 13 countries (Austria, Belgium, Denmark, Finland, Germany, Greece, Ireland, Italy, Lithuania, Luxembourg, Norway, Portugal, Spain), The Panel-VAR model and the Granger causality are adopted to explore the dynamic causal relationships between information and communication technologies (ICT), renewable energy (RE) and fossil energy (FE) consumption, economic growth (GDP) and adjusted net savings (ANS). The results show that most countries in the EEA are increasingly aware that the energy transition is a means of combating climate change and preserving the environment for future generations. The findings also emphasize that new technologies improve energy efficiency and reduce the costs of renewable energy, which supports the energy transition and enhances well-being and sustainability, particularly in urban areas where energy demands and environmental challenges are concentrated. Policies in these countries should encourage the use of electricity from renewable energy by providing the necessary credits and by adopting new technologies, strategies, and regulations that promote clean energy efficiency in both production and consumption, with a particular focus on urban sustainability and smart city initiatives.

ABSTRACT. Satellite data have been widely used to detect and monitor flooding events, especially in rural areas. The integration of data acquired by different sensors, operating at different wavelengths within the electromagnetic spectrum and using different technologies, can further improve the capabilities of providing more frequent information, which is crucial for the effective management of this dynamic phenomena. In any case, suitable and robust methodologies for the analysis of these data must be used to reduce false alarms. In this work, we presented an approach based on the combination of microwave data acquired by the C-band SAR aboard Sentinel 1 satellites with the optical data collected by the MultiSpectral Instrument on board Sentinel 2 platforms and the Operational Land Imager on Landsat 8 and 9. Different methodologies will be tested within the Google Earth Engine (GEE) cloud computing system, where all historical data are available and accessible, as well as other datasets/tools useful for developing the proposed methodologies. A selected number of events that occurred worldwide, trying to focus on urban areas, were utilized as test cases, and the results were compared with flood maps made available by the Copernicus Emergency Monitoring Service system to assess their accuracy. The performance achieved is discussed in this study.

ABSTRACT. The Democratic Republic of the Congo is one of the poorest countries in the world despite its wealth of natural resources. Mining, whether formal or informal, is the country's most important economic sector. Against this background, we analyze in this study whether and, if so, to what extent mining regions have led to particularly dynamic settlement dynamics between 2016 and 2023. As a primary source, we used the growth of settlements near mines mentioned in the book "Cobalt Red: How the Blood of the Congo Powers Our Lives" in relation to growth rates at national level. We found that within the region of the cobalt belt in the south of the Democratic Republic of Congo, the increase in buildings which strongly correlates with the number of residents living in those places increased by over 130% for cities near mining activities and over 50% for the region of the cobalt belt. In comparison we measure only 20 to 30% for the entire country (2016 to 2023). This indicates that mining has a large impact on the local environment, economy, and society.

ABSTRACT. The urban fact is diluted, mutated following the interventions of development actors and to the detriment of Public Utility Easements. The use of mapping and remote sensing is essential for the detection and understanding of the trajectory of urban development. Urbanization within the city of Jendouba was triggered during the colonial period while taking a linear form according to the land communication axes, in particular the railway network. The typology of housing and changes in land use are at the origin of the development of the urban fact.

ABSTRACT. Wastewater treatment plants (WWTPs) account for roughly 1–2% of global greenhouse gas emissions. During the treatment process, gases such as carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O) are emitted into the atmosphere. This issue becomes particularly concerning for WWTPs situated in urban areas, where high population densities result in large volumes of wastewater that must be treated daily. The increased load can overwhelm treatment capacities, leading to inefficiencies in pollutant removal. This study highlights the enhanced usefulness of the Sentinel-2 (S2) data for mapping wastewater in urban WWTPs and monitoring their discharge into surrounding water bodies. Specifically, the proposed method relies on adapted algorithms designed to address the reflectance spectrum, considering atmospheric interactions. In comparison to the matched filtering results, derived from the classification of the same image using the extracted reflectance spectrum, the method has demonstrated its effectiveness in monitoring and analyzing the spatial distribution and flow of wastewater. This approach is a crucial step in understanding and addressing environmental risks, especially in densely populated urban areas where WWTPs are vital for wastewater management and ecosystem protection.

ABSTRACT. Change detection (CD) is a critical task in remote sensing, especially for monitoring dynamic urban environments. In this work, we present a novel, self-supervised CD approach that leverages a two-player architecture to improve the reliability and accuracy of change predictions. The 2Player framework consists of two cooperating models: Player 1, a change detection model, and Player 2, a supporting autoencoder trained for reconstruction. The two models work in tandem: after reconstructing the second image from the first one, the second player highlights areas of potential change where reconstruction is difficult. On the other hand, the first player focuses on predicting changes based on these indications and shares its change maps with the autoencoder, enabling it to disregard these areas during reconstruction. This mutual guidance creates a robust self-supervised mechanism that we test with a FC-Siam-Diff model for the first player and a standard UNet for the second. We evaluate the 2Player framework on a refined version of the HRSCD dataset, where labels have been enhanced using IGN’s BD TOPO® model, resulting in a cleaner dataset of 10,000 images. Experimental results show that our approach achieves improved performances in detecting changes, outperforming traditional unsupervised baselines.

ABSTRACT. The widespread availability of satellite imagery data has enabled advancements in Land Use/Land Cover (LULC) andUrban Fabric (UF) mapping through deep learning. However, maintaining up-to-date urban land cover maps is challenged by the high cost and operational constraints of continuous field data collection. This study explores the feasibility of updating urban LULC maps using SITS-based semantic segmentation model strained on historical data, specifically examining a transfer scenario where a model trained on 2015 data is applied to 2020 imagery. We benchmark the performance of two convolution-based architectures (Unet and Unet3D), plus a recent spatio-temporal transformer-based approach (TSViT) and a proposed variant, named TSViT+SW, which incorporates a shifted window attention scheme. Experimental evaluations covering the urban area of Lyon, France, reveal that the proposed TSViT+SW model achieves the best results among transferred models, minimizing performance degradation compared to the ideal in-year training scenario. This work offers insights into the potential and limitations of using historical data to update urban land cover in the absence of fresh labeled data.

ABSTRACT. Urban green spaces play a crucial role in enhancing the sustainability of cities, making it a necessity for their measurement methods to be precise and reliable. Mapping of urban green has been standardized with multiple data sources, such as the Normalized Difference Vegetation Index (NDVI) derived from satellite imagery, and the Green View Index (GVI), derived from street-level imagery. Both are popular, complementary methods for quantifying urban green. This case study, in Athens, Greece, aims to investigate the strength of their relationship for different resolutions of NDVI. In addition, as NDVI is sensitive to height variations of the imaged area, this relationship was also quantified in terms of the effect of building heights. Results indicate a strong relationship between the two indices, but possibly not linear. Incorporating building heights into appropriate models is a method that could support a more holistic approach.

ABSTRACT. Urban trees play an important role in the city's thermal comfort, carbon storage capacity, rainwater infiltration, capture of some pollutants or to enhance biodiversity. The classification of tree species in urban areas is essential for monitoring their health and distribution, particularly in the context of increasingly frequent heatwaves and global warming. In this study, we present a deep learning-based approach to classify the 20 most representative urban tree species in Strasbourg using Satellite Image Time Series (SITS) from Sentinel-2 and PlanetScope. We implement and compare three state-of-the-art models used for Time Series Classification (TSC) — InceptionTime, H-InceptionTime, and LITE — augmented with sensor fusion (features fusion) to exploit the complementary spectral and temporal information from both sensors. Our results show that the sensor fusion approach significantly improves classification accuracy, with the best scores for H-InceptionTime model with an accuracy of 69%. Also, the error can be explained by the specific spatial organisation of the trees in urban areas, where dense and heterogenous tree distribution inside parks is more prone to errors than aligned ones.

ABSTRACT. Intensification of the urban heat island (UHI) effect due to rapid urbanization poses significant health risks to urban residents during heatwaves. Therefore, quantitative assessment of cooling strategies has thus become essential. This study employed Weather Research and Forecasting (WRF) model to simulate air temperatures directly related to human perception. To improve simulation accuracy, remote sensing data were integrated into the WRF framework to dynamically update land cover and urban canopy parameters. Subsequently, hourly meteorological data were utilized to validate the simulation accuracy. Based on the optimal simulation case LCZBH, we evaluated six cooling strategies about green roofs and cool roofs. The cooling effectiveness was analyzed during different time periods and across various Local Climate Zones (LCZs). The results showed that (1) replacing MODIS with LCZ data and incorporating localized building height data reduced the simulation errors of air temperature in WRF; (2) Temperature differences across LCZ types in the baseline were mainly observed in the afternoon and nighttime, with relatively higher temperatures in LCZ2, LCZ4, and LCZ10, and lower temperatures in LCZ9; (3) Increasing roof reflectivity was more effective for urban cooling than enhancing green roof coverage; (4) For the same cooling strategy, higher air temperatures corresponded to greater cooling potential. Specifically, cooling potential was higher during daytime than nighttime, and greater in high-temperature zones like LCZ2 and LCZ10 compared to low-temperature zones (e.g., LCZ9). The accuracy of WRF-simulated temperatures can be improved by integrating remote sensing data. The findings contributed to a better understanding of the spatial and temporal variations in urban air temperatures and cooling potential and supported the selection and implementation of cooling measures by governments effectively.

ABSTRACT. In many developing countries, obtaining accurate and high-resolution GDP estimates is crucial for guiding economic policy and challenging due to limited data availability. This study introduces a novel deep learning fusion model to estimate the aggregated values of the Gross Domestic Product (GDP) in Brazil on a high spatial resolution grid of 1 km2. This is done by a combination of remote sensing data, specifically optical imagery and nighttime light emissions. The model processes these data streams separately before fusing them for GDP prediction. This approach allows for the extraction of both physical and socioeconomic features relevant to economic activity, providing valuable insights for economic planning and policy making. Our fusion model achieved high r² values of up to 0.75 and was trained and tested in 29 Brazilian cities, demonstrating its effectiveness and scalability for large-scale urban economic estimations.

ABSTRACT. This work shows the current approaches to recognize and locate urban land-uses and understand the complex urban environment for sustainable urban systems using machine learning methods and the most physical scattering mechanism-based polarimetric decomposition models. This paper intends to identify the urban area from polarimetric Synthetic aperture radar (PolSAR) data using the CNN model.This model has shown its effectiveness in classification. For this purpose, we propose to use a feature vector that contains the elements of the 6 real elements extracted from the coherency matrix and the three physical descriptors (entropy, alpha angle and anisotropy) from the eigenvalue/eingenvector decomposition. The used elements are roll invariants, which make them not sensitive to the orientation angle, which can create ambiguity between oriented buildings and vegetation. To test the performance of the proposed method, ALOS-2/PALSAR-2L-band over the San-Francisco, USA were used. The results show that man-made targets could be effectively discriminated from natural scenes using the proposed method.

ABSTRACT. Urban residents are especially vulnerable to rising temperatures due to the urban heat island (UHI) effect. Yet, there is limited understanding of if and how they express heat-related concerns and how these responses are influenced by the urban tissue. In this study, we integrated social media data, on-site weather records, and remote sensing data to explore the relationship between heat-related social media content, urban temperatures, and urban morphology. We collected geolocated Twitter data posted from Mexico City during 2022. In addition, we used weather station air temperature data, and land surface temperature (LST) estimated from MODIS imagery. To represent the urban morphology of the city, we used the local climate zones (LCZ) classification scheme. Our results reveal an exponential increase in the frequency of heat-related Tweets as air temperatures rise. Furthermore, the proportion of heat-related Tweets is higher in LCZ with urban morphologies that exhibit elevated LST. These findings show that social media serves as a medium to communicate concerns about urban heat, with these expressions varying according to the urban morphology and its thermal characteristics.

ABSTRACT. The recurring phenomenon of marine mucilage in the Mediterranean region highlights the urgent need for effective environmental monitoring of marine ecosystems under stress. This study leverages Sentinel-2 multispectral imagery and incorporates a wide range of spectral indices to develop a CNN-based classification model for marine mucilage detection and mapping. The model was trained and validated on data of the Sea of Marmara, acquired in 2021, and tested on data from the Adriatic Sea, acquired in August 10, 2024, to evaluate its generalizability across geographically distinct regions. A SHAP-based explainability analysis was employed to assess the contributions of individual spectral indices to model predictions. The results highlight the robustness of key indices in mucilage detection and reveal new insights into their applicability across diverse environmental contexts, providing a foundation for broader monitoring applications in the Mediterranean.

ABSTRACT. Recently, there has been growing interest in achieving the Sustainable Development Goals (SDGs) due to the increasing global population and the observed changes in Earth's natural environment. In this study, we aim to contribute to the achievement of three Sustainable Development Goals (SDGs), namely SDG2, SDG13, and SDG15 by providing accurate SM retrieval. The selection of an accurate SM retrieval method for agricultural semi-arid regions is challenging. Therefore, the main goal of this study is to evaluate different Machine Learning (ML) algorithms to identify the most powerful algorithm capable of estimating SM conditions in those areas based on the combined use of radar and optical data. The different ML algorithms are implemented using a synthetic database obtained by a Calibrated Water Cloud Model (CWCM). The statistical evaluations are carried out using five corn fields in Huamantla, Central Mexico. The results demonstrate that the Random Forest Regressor (RFR) achieves the lowest error values, thus outperforming the other models. In order to improve the synthetic dataset and forecast accuracy, we consider the integration of surface scattering models with the WCM in the next step of this research.

ABSTRACT. This paper explores the use of synthetic datasets to improve aerial image segmentation, addressing the need for large and diverse data for model training. Current benchmarks often lack real-world conditions, such as high-altitude and nadir perspectives. To overcome this, we propose a controlled data generation approach using the CARLA simulator to generate aerial images of different towns under different weather and time of day conditions, with dynamic traffic elements. We compare our dataset with existing real and synthetic datasets, and evaluate model performance by training the DeepLabV3+ neural network on our dataset and testing on real data. The results show that incorporating synthetic data yields performance comparable to training on real data alone, highlighting its complementary value.