CSIR Fourth Paradigm Institute

Landslides are among the most destructive geohazards, often triggered by extreme rainfall and intensified by fragile geological settings and human activities. On July 30, 2024, a catastrophic rainfall-induced landslide struck Vythiri Taluka in Wayanad district, Kerala, India, causing fatalities and widespread destruction. This study employs Persistent Scatterer Interferometry (PSI) using Sentinel-1 SAR data along with GNSS observations to investigate pre-event deformation and slope instability. PSI analysis of 33,253 scatterer points revealed cumulative Line of Sight (LoS) displacements up to 65 mm, with abnormal variations of ~ 20 mm detected on 20th July, ten days prior to failure. Average displacement rates of 2.0 mm/year and localized rates up to 5.0 mm/year near the crown highlight significant precursory deformation. GNSS-derived E-W compressional and N-S extensional strain confirmed localized active fault systems in the region. Integration of rainfall records demonstrated extreme precipitation (> 360 mm in 24 h) as a critical triggering factor, acting in concert with fractured lithologies and tectonic activity to destabilize slopes. The findings suggest that a weak zone formed well before the event, with deformation serving as a precursor to failure. This study underscores the importance of continuous monitoring with PSI-InSAR for landslide hazard mitigation.

Because rainfall forecasting has a direct impact on crop yield, resource management, and overall farm productivity, it is essential to agricultural planning. Rainfall patterns are influenced by a number of internal and external factors, such as variations in temperature, humidity, atmospheric pressure, and monsoon variability. Rainfall forecasting accuracy is crucial for maximizing irrigation schedules, preserving soil moisture, and guaranteeing sustainable farming methods. Rainfall patterns that are unpredictable result in crop failures, inefficient use of resources, and financial losses for farmers. Using past weather data and climate parameters, this review combines cutting-edge machine learning and deep learning models to increase the accuracy of rainfall predictions. These models enhance forecasting accuracy by analyzing historical weather patterns and environmental variables, thereby increasing readiness for severe weather events. For farmers, policymakers, and agricultural researchers to develop data-driven strategies for mitigating climate-related risks, accurate rainfall forecasts provide vital insights. Using Karnataka’s main crops as a case study, this study expands on forecasted rainfall data by estimating crop yield. Accurate yield forecasts are crucial for both food security and economic stability, as climate variations have a significant impact on agricultural productivity. The effectiveness of DL models in predicting agricultural output is analyzed through a review of published literature and forecasting analyses presented between 2020 and 2025, focusing on trends in methodologies, results, and research gaps rather than providing new empirical rainfall or agricultural yield estimates. This review helps improve food security, optimize resource use, and make informed decisions by leveraging advancements in climate modeling and agricultural analytics. The results provide a thorough framework for flexible farming practices, guaranteeing sustainable crop yields in the face of changing environmental.

Urban flooding affects millions of people in Indian major cities during past decades and is expected to be severe in near future under the warming climate scenario. Rapid urbanization and demography changes in the Bengaluru city in the past has led to significant hydrological and environmental stress contributing to the severity of urban flooding. The whole Bengaluru city (Urban as well as Rural) consists of four watershed valley and here, we have chosen Koramangala-Challaghatta (KC) valley as our study area which is known for its low-lying topography and frequent flooding events. In this study, we selected four Extreme Rainfall Events (EREs), occurred on 07 September 2015, 1 June 2016, 15 August 2017, and 24 September 2018, for simulation using the Storm Water Management Model (SWMM). The digital elevation model (DEM) is examined to understand the topography of high and low-lying areas to assess the flood impacts of EREs in various regions. Based on the frequency distribution, the 99th percentile of daily rainfall intensity (39.5 mm/day) is considered as the threshold for defining EREs in the study region, and long-term rainfall data (1998–2024) revealed a clear increasing trend in frequency of EREs. Results indicate that rainfall exceeding 30 mm/day leads to flooding in more than half of the area in KC Valley, with flash floods occurring when rainfall surpasses 55 mm/day. The most flood-prone locations are those near historical natural lakes, where urban encroachment has diminished natural flood detention capacity. Evaluation using multiple statistical metrics demonstrates reliable performance of SWMM in flood simulation, particularly for high-intensity rainfall events. The bias in simulated stormwater discharge ranges from + 4 to −4 m³/s, indicating useful skill of the model. The reliable methodological framework to tackle hydrometeorological challenges presented in this study can be handy tools for designing integrated flood management strategies for urban cities.