Cities worldwide are attempting to be labelled as smart, but truly classifying as such remains a great challenge. This study aims to use artificial intelligence (AI) to classify the performance of smart cities and identify the factors linked to their smartness. Based on residents’ perceptions of urban structures and technological applications, this study included 200 cities globally. For 147 cities, we gathered the perceptions of 120 residents per city through a survey of 39 questions covering two main pillars: ‘Structures’, referring to the existing infrastructure of the city, and the ‘Technology’ pillar that describes the technological provisions and services available to the inhabitants. These pillars were evaluated across five key areas: health and safety, mobility, activities, opportunities, and governance. For the remaining 53 cities, scores were derived by analyzing pertinent data collected from various online resources. Multiple machine learning algorithms, including Random Forest, Artificial Neural Network, Support Vector Machine, and Gradient Boost, were tested and compared in order to select the best one. The results showed that Random Forest and the Artificial Neural Network are the best trained models that achieved the highest levels of accuracy. This study provides a robust framework for using machine learning to identify and assess smart cities, offering valuable insights for future research and urban planning.

The present study focuses on the validation and comparison of eight different physical models for soil salinity mapping in an arid landscape using two independent Landsat-Operational Land Imager (OLI) datasets: simulated and image data. The examined and compared models were previously developed for different semi-arid and arid geographic regions around the world, i.e., Latino-America, the Middle East, North and East Africa and Asia. These models integrate different spectral bands and unlike mathematical functions in their conceptualization. To achieve the objectives of the study, four main steps were completed. For simulated data, a field survey was organized, and 100 soil samples were collected with various degrees of salinity levels. The bidirectional reflectance factor was measured above each soil sample in a goniometric laboratory using an analytical spectral device (ASD) FieldSpec-4 Hi-Res spectroradiometer. These measurements were resampled and convolved in the solar-reflective bands of the Operational Land Imager (OLI) sensor using a radiative transfer code and the relative spectral response profiles characterizing the filters of the OLI sensor. Then, they were converted in terms of the considered models. Moreover, the OLI image acquired simultaneously with the field survey was radiometrically preprocessed, and the models were implemented to derive soil salinity maps. The laboratory analyses were performed to derive electrical conductivity (EC_-Lab) from each soil sample for validation and comparison purposes. These steps were undertaken between predicted salinity (EC_-Predicted) and the measured ground truth (EC_-Lab) in the same way for simulated and image data using regression analysis (p ˂ 0.05), coefficient of determination (R²), and root mean square error (RMSE). Moreover, the derived maps were visually interpreted and validated by comparison with observations from the field visit, ancillary data (soil, geology, geomorphology and water table maps) and soil laboratory analyses. Regardless of data sources (simulated or image) or the validation mode, the results obtained show that the predictive models based on visible- and near-infrared (VNIR) bands and vegetation indices are inadequate for soil salinity prediction in an arid landscape due to serious signals confusion between the salt crust and soil optical properties in these spectral bands. The statistical tests revealed insignificant fits (R² ≤ 0.41) with very high prediction errors (RMSE ≥ 0.65), while the model based on the second-order polynomial function and integrating the shortwave infrared (SWIR) bands provided the results of best fit, with the field observations (EC_-Lab), yielding an R² of 0.97 and a low overall RMSE of 0.13. These findings were corroborated by visual interpretation of derived maps and their validation by comparison with the ground truthing.

Since the 1990s, Morocco’s agriculture has been characterized by the co-existence and transformation of both modern and traditional smallholder systems. In the Atlas Mountains, the effects of rural–urban transformation have led to intensified irrigated agriculture in some agricultural areas, while others were abandoned. To better understand these effects, this study aimed at (1) analyzing the land use and land cover (LULC) changes, (2) assessing the structure and dynamics of vegetation, and (3) comparing a Support Vector Machine (SVM) classification approach with a seasonal rules-based approach. We, therefore, employed a semi-automatic supervised classification of LULC using Landsat data from the 1990s to the 2020s to distinguish between Open Canopy Vegetation, Bareland, Forest, and Water. Overall accuracies achieved ranged from 88% to 90% in 1990, 2000, 2010, and 2020. SVM results indicated the share of Bareland as >80% of the landscape in all periods. With the seasonal rules-based approach, 10% less Bareland was detected than with the SVM approach. Our findings indicate the limitation of detecting vegetation reflectance in semi-arid mountainous regions such as that prevailing in Morocco using a single machine learning method.

Coastal urbanization is increasingly constrained by legacy land-use patterns and escalating climate risks, yet long-term morphological trajectories remain poorly quantified due to the absence of multispectral data in pre-satellite archives. This study introduces a scalable deep learning pipeline that bridges a century-scale domain gap through an attention-enhanced Pix2Pix colorization stage and a few-shot U-Net++ segmentation stage, enabling automated reconstruction of urban expansion from panchromatic historical aerial imagery (1920–1971) and digital aerial photographs (1997) to contemporary very-high-resolution satellite data (2024) in Les Sables-d’Olonne, France. The novelty of the approach lies in coupling generative colorization with epoch-specific fine-tuning to overcome radiometric and annotation bottlenecks that have historically prevented quantitative urban reconstruction from pre-satellite archives. The colorization stage achieved high spectral fidelity (PSNR 35.21 dB, SSIM 0.9762), and segmentation performed strongly on modern imagery (mIoU 0.9789). While the segmentation model performed strongly on modern imagery, direct transfer to historical data exhibited substantial domain shift due to radiometric discrepancies. Few-shot adaptation on year-specific calibration sets recovered reliable building footprints (mIoU 0.53–0.65) across the full timeline. Multi-scalar analysis of the reconstructed footprints revealed constrained anisotropic expansion: early saturation of the coastal historic core, followed by rapid inland peri-urbanization post-1971 driven by geographic barriers. This spatiotemporal shift has entrenched spatial lock-in, placing recent development in retro-littoral zones that are vulnerable to submersion and characterized by severe vegetation loss. The framework unlocks previously inaccessible historical archives for quantitative urban monitoring, providing critical insights into legacy effects of unconstrained growth and informing resilient coastal planning under climate change.

Coastal urbanization is increasingly constrained by legacy land-use patterns and escalating climate risks, yet long-term morphological trajectories remain poorly quantified due to the absence of multispectral data in pre-satellite archives. This study presents a scalable deep learning pipeline that bridges this century-scale domain gap, enabling automated reconstruction of urban expansion from panchromatic historical aerial imagery (1920–1971) and digital aerial photographs (1997) to contemporary very-high-resolution satellite data (2024) in Les Sables-d’Olonne, France. Spectral restoration was performed using an attention-enhanced Pix2Pix generative adversarial network with hybrid inference, achieving high fidelity (PSNR 35.21 dB, SSIM 0.9762). Semantic segmentation was conducted with U-Net++, yielding strong performance on modern data (mIoU 0.9789). However, direct transfer to historical periods suffered from severe domain shift due to radiometric variations.To overcome this limitation without extensive manual annotation, few-shot adaptation was applied on year-specific calibration sets, producing reliable building footprints (mIoU 0.53–0.65) despite degradation. Multi-scalar analysis of the reconstructed footprints revealed constrained anisotropic expansion: early saturation of the coastal historic core, followed by rapid inland peri-urbanization post-1971 driven by geographic barriers. This spatiotemporal shift has entrenched spatial lock-in, placing recent development in retro-littoral zones vulnerable to submersion and characterized by severe vegetation loss. This framework unlocks previously inaccessible historical archives for quantitative urban monitoring, providing critical insights into legacy effects of unconstrained growth and informing resilient coastal planning under climate change.