Indian Institute of Remote Sensing

Abstract. Accurate classification of Land Use and Land Cover (LULC) from satellite imagery is vital for environmental monitoring, sustainable urban development, and resource management. With the increasing availability of multi-spectral data from Earth observation missions such as Sentinel-2, deep learning provides powerful solutions for automating LULC classification. In this study, we present a lightweight Convolutional Neural Network (CNN) architecture tailored for 4-band satellite imagery. Unlike conventional approaches that rely solely on RGB inputs, our model incorporates Red, Green, Blue, and Near-Infrared (NIR) bands to capture a broader range of surface and vegetation characteristics. The architecture combines stacked convolutional blocks with batch normalization, pooling layers, and dropout regularization, ensuring both strong accuracy and efficient computation. Training was further enhanced through data augmentation strategies such as rotation, flipping, and zooming. Using the EuroSAT dataset (27,000 images across 10 classes), the model achieved a test accuracy of 96% and a macro-averaged F1-score of 0.96, with excellent performance in challenging categories such as Residential, SeaLake, and Forest. The compact design of the model makes it highly suitable for deployment in time-sensitive or resource-limited scenarios, including monitoring of city growth, assessing agricultural productivity, and supporting rapid response to environmental hazards.

<title>Abstract</title> This study examined long-term trends in precipitation and temperature in the Gomti River Basin of northern India from 1980 to 2023, using gridded climate data. Mann-Kendall tests and Sen's slope estimators were applied to analyse trends at annual, seasonal and monthly scales. The results showed a statistically significant decreasing trend in annual precipitation (-2.52 mm/year) and monsoon rainfall (-0.169 mm/year). At the district level, Ayodhya, Sitapur, and Unnao exhibited significant declines in annual rainfall. Pre-monsoon rainfall increased slightly in some of the central districts. For temperature, a marginally increasing trend was observed in the annual mean temperature (0.008°C/year), with the most pronounced warming during the monsoon season (0.014°C/year). Monthly analysis revealed significant precipitation decreases in June-July and increases in October-November, indicating a shift in seasonal rainfall patterns. The findings highlight changes in the basin's hydroclimatic regime, with implications for water resources and agricultural practices. This analysis provides insights to support climate adaptation planning in the region.

Landslides are among the most hazardous geomorphological processes, especially in mountainous terrains where they threaten lives, disrupt infrastructure, and degrade ecosystems. The Chamoli district in Uttarakhand, India, with its rugged topography, erratic rainfall, and anthropogenic disturbances, is highly susceptible to such events. This research focuses on generating a detailed landslide susceptibility map for Chamoli using four machine learning algorithms Naïve Bayes (NB), K-Nearest Neighbors (KNN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). A total of sixteen causative factors were selected and processed using Geographic Information System (GIS) techniques. To avoid multicollinearity and ensure the reliability of input variables, statistical validation was performed. The landslide inventory, consisting of 778 past events, was split into 70% training and 30% testing datasets for model development and evaluation. The models were assessed using statistical indicators such as sensitivity, specificity, precision, accuracy, F1-score, Matthews Correlation Coefficient (MCC), and the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC). Among the models, XGBoost showed the highest performance (AUC = 0.95), followed by RF (0.83), KNN (0.79), and NB (0.78). The susceptibility analysis revealed that XGBoost categorized 17.53% of the area as highly vulnerable, outperforming RF (14.08%), KNN (14.00%), and NB (4.55%). The enhanced accuracy of XGBoost and RF stems from their ensemble learning approach, which effectively captures nonlinear relationships and mitigates overfitting. The resulting susceptibility maps are crucial tools for risk management, infrastructure planning, and sustainable development in hazard-prone Himalayan regions. This study reinforces the importance of machine learning in natural hazard assessment and suggests that future efforts should incorporate real-time data, finer-resolution remote sensing, and deep learning methods to further refine prediction accuracy.

Indian historic cities serve as cultural anchors and are vital to heritage tourism, yet their unregulated urban expansion has become a major concern. Long-term monitoring of built-up area growth is crucial for informed and sustainable urban governance. However, the absence of satellite data before 1975 limits the ability to track historical urbanization trends. To bridge this temporal data gap and enhance the accuracy of future urban growth predictions, this study develops a semi-automated methodology that integrates georeferenced and vectorised historical maps with remote sensing data. Focusing on the historic cities of Varanasi and Hyderabad, the study reconstructs two centuries of built-up area growth. Varanasi exhibited an average annual built-up growth rate of approximately 3.35%. A discernible north-westward shift in the urban centroid was observed, with buffer analysis around the Kashi Vishwanath Temple indicating intensified urbanization within the 5–10 km and &gt;20 km zones. Hyderabad showed an average annual built-up growth rate of about 3.04%. The city’s centroid exhibited a northward drift until 1995, followed by a south-eastward shift, aligning with the growth of the IT corridor and associated infrastructure in that region. Buffer analysis further revealed that urbanization in Hyderabad has been more prominent beyond the 20 km radius, underscoring peripheral expansion driven by economic clustering. This study demonstrates the efficacy of combining historical cartographic archives with satellite imagery for reconstructing long-term urban dynamics. The proposed methodology not only enhances the temporal depth of urban change analysis but also provides actionable insights for planners and policymakers to promote resilient, culturally sensitive urban development strategies.

Earthquakes have always been a permanent threat to humanity. Monitoring the earthquake precursors using space-based methods can be a new introduction to earthquake studies. The promising analysis of the earthquake precursors such as Ionospheric Perturbations is based on the detection of Ionospheric electron content. Within approximately 3 weeks before an event, negative TEC anomalies have been witnessed. A careful examination of solar and geomagnetic perturbations has been done to ensure that TEC fluctuation in the ionosphere over the investigated time period can be attributed solely to the seismic activity within the earth. Major tectonic plate boundaries, 2-D maps of global ionospheric anomalies, and Earthquake Preparation Zone have all been considered in order to evaluate the TEC fluctuations in the spatial domain during the time leading up to the earthquake. Statistical ARIMA model is used in order to give the backup for the data availability which matches with the data derived from the IGS. 2-D zone-time series LLT maps show how anomaly is spreading in the nearest fault position and EPZ.