Beira Lake, located in Colombo, Sri Lanka, has suffered severe anthropogenic impacts, with previous restoration attempts failing due to a limited understanding of pollutant dynamics. Aiming to fill this gap, a comprehensive study was conducted during dry and wet seasons to assess the spatiotemporal water pollution of Beira Lake, employing key physicochemical parameters, numerical indices, and remote sensing analysis. The water pollution index (WPI) results categorize Beira Lake as highly polluted, with WPI values ranging from 2.38 ± 0.92 in the wet season to 2.53 ± 1.32 in the dry season. Comparatively higher COD levels recorded in the Beira Lake network, especially for Gangarama Lake show significant pollution levels during both the dry and wet seasons, e.g., the highest COD levels, at 306.40 mg/L, were observed during the wet season. The Trophic State Index (TSI) results indicate eutrophic and hypereutrophic conditions in Beira Lake, which are particularly pronounced during the wet season. The heavy metal pollution index (HPI) results suggest elevated heavy metal concentrations in Beira Lake, especially in the wet season. Combined with field investigation results, a remote sensing data analysis between 2016 and 2023 reveals significant improvements in water transparency, suggesting positive effects of recent management interventions. Parameters demanding attention include COD, nitrate, and total phosphate levels due to their consistent exceedance of permissible limits. The PCA results of indices correlations between wet and dry seasons offer valuable insights into the complex dynamics of Beira Lake’s water quality. The study makes recommendations for restoring Beira Lake, including stringent pollution controls, regular dredging, green infrastructure implementation, implementing new rules and regulations, and community engagement.

Understanding the spatiotemporal characteristics of drought and its impacts on vegetation is a timely prerequisite to ensuring agricultural, environmental, and socioeconomic sustainability in Sri Lanka. We investigated the drought characteristics (duration, severity, frequency, and intensity) from 1990 to 2020 by using the Standardized Precipitation Evapotranspiration Index (SPEI) at various timescales and the cumulative and lagged effects on vegetation between 2000 and 2020 across the climatic zones of Sri Lanka (Dry, Wet, and Intermediate). SPEI indexes at 1-, 3-, 6-, 12-, and 24-month scales were used to analyze the drought characteristics. Frequent droughts (~13%) were common in all zones, with a concentration in the Dry zone during the last decade. Drought occurrences mostly ranged from moderate to severe in all zones, with extreme events more common in the Dry zone. This research used SPEI and the Standardized Normalized Difference Vegetation Index (SNDVI) at 0 to 24-month scales to analyze the cumulative and lagged effects of drought on vegetation. Cumulated drought effects and vegetation had maximum correlation coefficient values concentrated in the −0.41–0.98 range in Sri Lanka. Cumulated drought effects affected 40% of Dry and 16% of Intermediate zone vegetation within 1–4 months. The maximum correlation between the lagged drought effect and vegetation SNDVI showed coefficient values from −0.31–0.94 across all zones, and the high correlation areas were primarily distributed in Dry and Intermediate zones. Over 60% of the Dry and Intermediate zones had a lagged drought impact within 0 to 1 month, while 52% of the Wet zone experienced it over 11 months. The resulting dominant shorter timescale responses indicate a higher sensitivity of vegetation to drought in Sri Lanka. The findings of this study provide important insights into possible spatiotemporal changes of droughts and their possible impact on vegetation across climate zones.

Semantic segmentation and height estimation in remote sensing imagery are two pivotal tasks for scene understanding, and they are highly interrelated. Although deep learning methods have achieved remarkable progress in these tasks in recent years, several challenges remain. Recent studies have shown that multi-task learning methods can enhance the complementarity of task-related features, thus maximizing the prediction accuracy of multiple tasks at a low computational cost. However, due to factors such as complex semantic categories and the inconsistent spatial scales of remotely sensed images, existing multi-task learning methods often fail to achieve better results on these two tasks. To address this issue, we propose CTME-Net, a novel architecture termed the Cross-Task Mutual Enhancement Network, designed to jointly perform height estimation and semantic segmentation tasks on remote sensing imagery. Firstly, to generate discriminative initial features for each task branch and activate dedicated pathways for cross-task feature disentanglement, we design a universal initial feature embedding module for each downstream task. Secondly, to address the impact of redundancy in general features during global–local fusion, we develop an Adaptive Task-specific Feature Distillation Module that enhances the model’s ability to acquire task-specific features. Finally, we propose a task feature interaction module to optimize features across tasks through mutual optimization, maximizing task-specific feature expression. We conduct extensive experiments on the ISPRS Vaihingen and Potsdam datasets to validate the effectiveness of our approach. The results demonstrate that our proposed method outperforms existing methods in both height estimation and semantic segmentation.

Generative adversarial networks (GANs) have become an important approach for translating remote sensing images across sensing modalities, acquisition domains, and degraded observation conditions. This paper reviews 111 studies published between 2018 and 2026 and analyzes the field from three translation perspectives: intra-modal or domain translation, cross-modal translation, and image restoration or completion. The review also examines the use of translated imagery in change detection, object detection, semantic segmentation, and data augmentation. Existing methods are compared in terms of supervision strategy, data pairing, network architecture, feature representation, loss-function design, datasets, and evaluation practice. The literature shows a strong concentration on SAR-to-optical translation, accompanied by growing use of multiscale feature extraction, attention mechanisms, contrastive learning, cross-modal alignment, and pretrained representations. However, translation quality remains closely related to dataset characteristics, sensor configuration, preprocessing, and evaluation protocol. Pixel-level, structural, perceptual, distributional, sensor-related, and downstream metrics provide complementary evidence of translation performance. Major challenges include the non-unique nature of cross-modal mappings, preservation of geometric and sensor-related information, dependence on auxiliary observations in restoration tasks, and limited evidence for cross-domain generalization. Future research should develop task- and sensor-aware evaluation and strengthen cross-region and cross-sensor validation.