Agricultural drought (AGD) prediction is critical for agricultural water conservation in India, where ∼60% of the population depends on farming. This study presents an explainable data-driven framework to predict AGD at sub-seasonal-to-seasonal (S2S) scales (1–3-month lead) using a Stacked Long Short-Term Memory (SLSTM) model at 0.25° grid resolution across croplands in India. The framework integrates local antecedent land-atmosphere variables and non-local climate drivers, represented by Sea Surface Temperature (SST) anomalies, with Soil Moisture Percentile (SMP) as the drought indicator. A novel extreme-aware custom loss function combining weighted mean squared error and a time-shift penalty is used to train SLSTM model specifically to capture drought events. The model is trained from June 1981–May 2015 and tested from June 2015–May 2022. SLSTM model outperformed benchmark deep learning models (XGBoost and LSTM) and the SEAS5 dynamical forecast system in predicting major AGD events, including the 2015–16 event. At a 1-month lead time, SLSTM model demonstrated strong performance, with percentage correct and probability of detection between 0.6–1 and false alarm rate between 0–0.3, particularly in Northern, Southern, and Eastern India. Although performance declined with increasing lead time, the model remains conservative while minimizing overprediction. Predicted drought area correlates reasonably well with that of observations across all three lead times. We further analyzed shifts in grid-wise predictor importance between local and non-local features in response to changes in lead time at seasonal and climatological scales. This research supports the development of early warning systems for AGD risk management in India.

The Himalayas, a critical component of the global water cycle and a hotspot for biodiversity, have been increasingly affected by severe wildfires. These fires disrupt eco-hydrological processes, posing threats to ecosystems, air quality, and human health. Despite their growing prevalence, the dynamics of forest fires in this region, particularly the role of meteorological factors, remain underexplored. Using the Weather Research and Forecasting Model-Fire (WRF-Fire), we simulated forest fires in the Western Himalayas during May 2018 to investigate the relationship between weather conditions and fire behaviour. We conducted four experimental simulations: a control (Fire_ctr), increased temperature by 2 K (Fire_Ti), reduced relative humidity by 15

India is the second-largest contributor to global greening. Despite significant increase in green cover, satellite-based MODIS estimates reveal no significant increase in vegetation productivity, with major forested areas showing declines in net primary productivity (NPP), due to warming. A major limitation of MODIS vegetation productivity dataset is that it does not consider the improved biochemical rate of plant photosynthesis under higher atmospheric CO _2 levels, known as CO _2 fertilization (CFE). In this study, we incorporate the direct effect of CFE into satellite-based MODIS vegetation productivity estimates and reassess NPP trends across India from 2001 to 2024. Our analysis confirms that the NPP trend values nearly doubled after accounting for the direct effect of CFE, significantly reducing the percentage of grids exhibiting a negative trend from 21.21% to 8.86% across total vegetated land and from 51.81% to 25.23% across forested land, compared to the trends observed without CFE. Notably, most of the grids showing negative trends transitioned into statistically non-significant categories, either negative or positive. These grids are primarily located along the Western Ghats and southern peninsular India, where vegetation productivity showed only a modest increase in response to the direct effect of CFE. This limited response is likely due to concurrent warming in these regions, which dampens the benefits of CFE. In contrast, northwestern India, where atmospheric moisture stress and temperature have declined, exhibited a stronger increase in vegetation productivity after accounting for CFE — enhancing already positive trends. Overall, our findings suggest that incorporating the direct effect of CFE significantly improves the estimation of vegetation productivity trends across India. However, regional climate patterns have a strong influence on the magnitude of the increase in vegetation productivity due to the direct effect of CFE. This study enhances understanding of climate-vegetation interactions under increasing atmospheric CO _2 concentration and climate change.

Abstract In a warming climate, elevated atmospheric CO₂ concentrations are altering ecological and hydrological systems globally. As the second-largest contributor to global greening, driven largely by agricultural intensification and with the country's rapid economic growth, India’s vegetation dynamics have significant implications for the global water and carbon cycles and their interactions. India, as a major contributor to global greening, offers key insights into the complex interactions among climate variability, land-use shifts, and anthropogenic activities. However, large uncertainties persist in India’s water cycle, particularly in understanding monsoon precipitation processes, sparsely monitored in-situ evapotranspiration, and soil moisture, which propagate into uncertainties in global model products and predictions over the region and fundamentally limit reliable projections of vegetation productivity and carbon dynamics. Vegetation, in turn, provides feedback to the water cycle through both biophysical and physiological processes, emphasizing the need to quantify two-way land–atmosphere interactions. Understanding and reducing these inherent uncertainties are necessary for projecting the future of India’s ecosystems under ongoing global climate change. This review synthesizes current understanding of terrestrial carbon–water interactions in India, exploring regional carbon budgets, atmospheric controls on vegetation productivity, the profound effects of land management, ecosystem feedback via evapotranspiration and land surface change. We highlight emerging state-of-the-art methods, such as causal inference and machine learning, which offer powerful tools for accurately diagnosing and predicting these coupled processes. Despite these advancements, significant limitations persist: long-term field observations remain sparse, plant physiological processes are underrepresented in current land-surface models, and human interventions like irrigation are often poorly parameterized. To address these critical gaps, we propose an integrated, multiscale research framework that cohesively connects field data, remote sensing, and land-surface modeling. Strengthening this integrated understanding is essential for improving projections of ecosystem change and hydroclimatic extremes in India under projected global warming scenarios, supporting climate resilience and sustainable development.

Emerging evidence suggests that climate-induced changes in vegetation profoundly affect the water cycle and hydrological processes. However, state-of-the-art hydrological models typically overlook this dynamic aspect, relying instead on static vegetation parameters when analysing the hydrological impacts of climate change. Although land surface models within earth system models account for dynamic vegetation, their computational demands render them impractical for integration into hydrological frameworks. In the current study focusing on India, we demonstrate the pivotal role of vegetation variability in influencing evapotranspiration, particularly during the latter half of the Indian summer monsoon season and the subsequent post-monsoon period characterized by high vegetation activity. Incorporating spatial and temporal vegetation variability into the widely used Variable Infiltration Capacity (VIC) model revealed an 18% increase in total annual evapotranspiration compared to conventional simulations. To efficiently integrate varying vegetation dynamics into the VIC framework, we developed a grid-scale Machine Learning model based on Long Short-Term Memory (LSTM). This model represents vegetation changes as a function of hydrometeorological variables, such as precipitation and temperature. The LSTM-derived vegetation properties—fraction of vegetation cover, Leaf Area Index, and albedo—are then incorporated into the hydrological model. Using this novel approach, we simulated future hydrometeorological scenarios in India based on downscaled NEX-GDDP-CMIP6 dataset projections from the MIROC6 General Circulation Model (GCM). Our findings highlight the significant impact of changing vegetation properties on future ET fluxes, a factor often overlooked when employing constant vegetation parameters. We restricted the analysis to a single GCM to demonstrate the importance of our approach, which considers changing vegetation dynamics—a critical yet neglected aspect in current hydrological impact assessments of climate change. This study not only addresses the inadequacies of prevailing methodologies but also presents an advanced framework for integrating machine learning into physics-based hydrological models to account for evolving vegetation properties in future simulations.

Urban weather and climate modeling is challenged by the highly heterogeneous and dynamic nature of cities. It exhibits a persistent trilemma between spatial granularity, spatiotemporal coverage, and physical interpretability. We articulate this challenge and propose a hybrid framework integrating physics-based models, urban observations, and machine learning. Framing this challenge as an integration problem across methods and scales, we provide a structured guide for next-generation, decision-relevant urban weather and climate modeling.

Rainfed crops account for approximately 40% of India’s food production and support 60% of its livestock. Although linked to oceanic monsoon rainfall, their productivity also depends on terrestrial evaporation, particularly in the non-monsoon season. However, the degree to which rainfed crops also rely on moisture sourced from upwind irrigated areas, remains largely unknown. Using a combination of models and observations, we show that upwind irrigated crops contribute 7%±6% of the rainfall over rainfed areas, rising to 15%±10% during the pre-monsoon months. In the absence of this input, water stress experienced by rainfed crops can increase by 5–10% during the crucial mid to late crop growth phases, potentially affecting yields. Our results reveal an unrecognized atmospheric link between irrigated and rainfed agriculture, which is overlooked in current agricultural policies. Planning and managing these systems in an holistic manner can help strengthen regional food and water security under future climates.

Rainfed crops account for approximately 40% of India's food production and support 60% of its livestock. Although linked to oceanic monsoon rainfall, their productivity also depends on terrestrially-sourced rainfall, particularly in the non-monsoon season. The degree to which rainfed crops rely on moisture sourced from evaporation in upwind irrigated areas remains largely unknown. Using a combination of models and observations, we show that evaporation from upwind irrigated crops contributes 7% (mean) +/- 5% (spread) of the rainfall over rainfed areas annually, rising to 15 +/- 10% during the pre-monsoon months (averaged over the years 2000-2020). In the absence of this input, water stress experienced by rainfed crops can increase by 5%-10% during the crucial mid to late crop growth phases, potentially affecting yields. Our results reveal an unrecognized atmospheric link between irrigated and rainfed agriculture that is overlooked in current agricultural policies. Planning and managing these systems holistically can help strengthen regional food and water security under future climates.

Global warming has intensified extreme precipitation globally, stressing urban infrastructure. While Numerical Weather Prediction (NWP) models have improved, they remain inadequate for forecasting extreme events at local scales, especially in the tropical urban regions. Mumbai, a coastal Indian city with 21 million residents, experiences multiple extreme precipitation events with significant spatial variability during the monsoon season (June to September). The Global Forecast System (GFS) model struggles to forecast extreme rainfall in Mumbai, resulting in low hit rates and high false alarms in predictions. To address this, we performed statistical downscaling of GFS output to 36 stations across Mumbai using a Convolutional Neural Network (CNN). While the CNN improved overall GFS skill, improvements for extreme events were minimal due to their rarity in training data. We enhanced this using CNN-Transfer Learning (CNN-TL), pre-training the model on the full rainfall dataset and fine-tuning it on heavy rainfall events. This approach improved threat scores (TS) by 12.94% to 498.92% and 76.47% to 616.67% for Lead Days 1 and 2, respectively, at the 95th percentile across different stations. On Lead Day 3, the CNN-TL model achieved TS >= 0.2 for most stations at the 95th percentile, while GFS forecasts failed to capture extreme rainfall events at this lead time. Correlation coefficients also showed significant improvements. The CNN-TL approach outperformed Mumbai's operational forecast model. This method is replicable in other cities with available long-term station data.

Sap flow (SAPFlow) is a critical process governing plant water use, influencing ecosystem hydrology, and regulating plant physiology through the exchange of water between the biosphere and atmosphere. Understanding the causal drivers of SAPFlow is essential for improving our ability to predict plant responses to environmental stressors, including drought and climate change. However, the causal relationships between environmental drivers and SAPFlow across diverse biomes remain poorly understood, particularly on a global scale. The present study uses 15 sites from the global SAPFLUXNET dataset to examine SAPFlow dynamics in three forest biomes: evergreen broadleaf forests (EBF), evergreen needleleaf forests (ENF), and mixed forests (MF). Using information theory-based process networks, we identify key causal drivers of SAPFlow. Our findings reveal that vapor pressure deficit (VPD) and soil water content (SWC) form a coupled system with SAPFlow, mediated by land-atmosphere feedback between soil and atmospheric aridity. VPD emerges as a dominant causal driver across all forest types, highlighting its pivotal role in the soil-plant-atmosphere continuum. Wavelet analysis further highlights distinct temporal interaction scales, with VPD exerting immediate influence on SAPFlow and SWC showing delayed effects. Our findings provide critical insights into plant-water interactions, with significant implications for refining land surface models and improving the accuracy of ecosystem response predictions under climate change scenarios.

Precipitation nowcasting over the immediate 10-90 min period is important for flood management and real-time decision-making in urban regions. Conventional short-range forecasting with high-resolution numerical weather prediction requires frequent data assimilation, model initialization, and spin-up, introducing computational latency. Machine learning provides an alternative by learning storm evolution directly from high-frequency observations and producing forecasts quickly after training. This is particularly relevant for Mumbai, India, where monsoon convection, land-sea interactions, and localized intense rainfall make short-term prediction difficult. Here, we develop a compact radar-only nowcasting framework that combines multi-elevation reflectivity, Doppler radial velocity, and radial-velocity-gradient proxy features within an encoder-decoder U-Net. Using the most recent radar volume scan, the model predicts 12 future composite reflectivity fields at 7.5-min intervals up to 90 min lead time. The derived velocity magnitude, divergence-like, directional-shear, and vorticity-like channels represent kinematic signatures associated with convergence and boundary interactions without requiring full wind-field retrieval. A high-reflectivity attention module improves sensitivity to convective cores, and physics-guided attribution examines whether the learned sensitivities are meteorologically meaningful. The model is trained using Mumbai Doppler radar observations from May to August 2023 and evaluated on temporally independent events. At 90 min lead time, Critical Success Index values are 0.437, 0.332, and 0.193 for ≥10, ≥20, and ≥30 dBZ thresholds, respectively. Compared with persistence, the model gives lower RMSE and higher spatial correlation at longer lead times. Once trained, it runs on a standard computer, generating nowcasts within seconds for real-time use.

The interannual variability of All India Monsoon Rainfall (AIMR) is of paramount importance for the country’s socioeconomics. The government classifies years with AIMR above 110

Micrometeorological variability significantly impacts the structures, functions, and dynamics of ecosystems. However, the assessment of feedback and causal relationships among microclimatic drivers and various ecosystems in the Himalayan region is rarely evaluated. Here, we studied the micrometeorological drivers controlling the variability in the net ecosystem exchange (NEE) of Himalayan Oak (Banj-Oak/Quercus leucotrichophora) and Pine (Chir-Pine/Pinus roxburghii) dominated ecosystems, as NEE is an indicator of ecosystem functioning. We used half-hourly eddy covariance flux data of CO2 fluxes from two sites established over Pine and Oak dominated ecosystems in Uttarakhand, India. We conducted the analysis with the information theory-based Temporal Information Partitioning Networks (TIPNets) approach to generate weekly process networks. TIPNets represent directed lag-structured causal graphs to identify the causal relationships and capture the temporal association among the variables. Our analysis aimed to capture fluctuations in variables with up to 6 h of memory. Based on the data availability, we generated the weekly networks at both the sites for the monsoon and post-monsoon seasons of 2016 and 2017. In both ecosystems, the sub-daily scale variations among the micrometeorological variables are responsible for the fluctuations in NEE. The Pine ecosystem is found to be more sensitive to changes in air temperature (TA) and uptakes more CO2 as compared to the Oak ecosystem throughout the study period. The transfer entropy links show that the NEE of the Oak ecosystem is moisturedriven (precipitation and relative humidity), while the Pine ecosystem is heat-driven (TA and net solar radiation) in both seasons. The influence of precipitation is not observed within a short memory of 6 h in the Pine ecosystem. This is because lesser fine roots take time to show the precipitation signature on NEE through infiltration, soil moisture, and root water uptake, compared to Oak. However, the impacts of moisture stress are evident in the network structure of both ecosystems, with more causal links observed in the network during dry periods compared to wet periods.

Causal and attribution studies are essential for earth scientific discoveries and critical for informing climate, ecology, and water policies. However, the current generation of methods needs to keep pace with the complexity of scientific and stakeholder challenges and data availability combined with the adequacy of data-driven methods. Unless carefully informed by physics, they run the risk of conflating correlation with causation or getting overwhelmed by estimation inaccuracies. Given that natural experiments, controlled trials, interventions, and counterfactual examinations are often impractical, information-theoretic methods have been developed and are being continually refined in the earth sciences. Here we show that transfer entropy-based causal graphs, which have recently become popular in the earth sciences with high-profile discoveries, can be spurious even when augmented with statistical significance. We develop a subsample-based ensemble approach for robust causality analysis. Simulated data, and observations in climate and ecohydrology, suggest the robustness and consistency of this approach.

As climate and terrestrial ecosystems are closely coupled, climate variability can significantly impact the vegetation dynamics. Large-scale circulation patterns, such as El Niño-Southern Oscillation (ESNO), impact the spatial distribution of rainfall and temperature, and their extremes, which further affect vegetation productivity. ENSO is one of the primary drivers of Indian summer monsoon rainfall (ISMR), accounting for about 40% of its interannual variability. Some of India's most severe summer monsoon droughts are associated with the El Niño events. Pacific meridional mode (PMM), tropical Atlantic Niño and the surface temperature/pressure over the Middle East are also gaining attention as potential drivers of Indian summer monsoon rainfall and climate extremes over India. However, the control of ENSO and other teleconnections-induced climate variability on terrestrial ecosystem productivity is poorly understood, especially in terms of the spatial extent, strength, and underlying mechanisms. Here, we examine the relationship of Indian vegetation productivity with large-scale teleconnections such as ENSO and PMM. We use frequency decomposition and principal component analysis (PCA) to reveal the dominant timescales of variability in vegetation productivity and quantify its association with the large-scale features of climate variability. We find that while ENSO is the most significant driver of the vegetation productivity which causes ecological droughts over core monsoon region, PMM also has a significant control primarily on low frequency variability of Indian vegetation. Our findings quantify the primary climatic controls of variability in Indian vegetation and reveal PMM as a significant modulator of low frequency variability.

An assessment of death rates in India’s coastal megacity of Mumbai reveals the impact of rain, high tides and flooding. Climate change will probably exacerbate the effects. An assessment of death rates in India’s coastal megacity of Mumbai reveals the impact of rain, high tides and flooding. Climate change will probably exacerbate the effects.

Terrestrial vegetation, particularly forests and savannas, is vital for regulating water and carbon cycles; however, its resilience to climate change remains largely unexplored in India. Our study projects that global warming will increase the likelihood of a shift from forest to savanna-like conditions in major Indian forest regions by the late 21st century. Between 2001 and 2020, India’s tree cover grids declined from 30

The Indian Summer Monsoon Rainfall (ISMR) holds pivotal importance in the predominantly agrarian landscape of India, occurring from June to September. While the influence of large-scale atmospheric circulation on the monsoon has been extensively studied, the role of land surface processes remained largely unexplored until the early 2000s. In order to assess the impact of land surface evapotranspiration on monsoon rainfall in terms of recycled precipitation from June to September, we employed a dynamic recycling model. Our investigation revealed that land evapotranspiration contributes 20-25% of moisture to the late monsoon rainfall in September, specifically in Central and Northeast India. It is important to note that, as the recycling model relied on a reanalysis dataset, it does not account for contributions from land water management practices, such as irrigation. To comprehensively address the human water management component within the human-natural water cycle in South Asia, we implemented a coupled land-atmosphere regional framework using the Weather Research Forecasting – Community Land Model (WRF-CLM). The primary drawbacks of the state-of-the-art irrigation schemes in land surface models, applied to Indian case studies, included the absence of provisions for considering paddy fields, which typically maintain a submerged cropland. Additionally, they did not consider India-specific irrigation practices, driven not by soil moisture measurements or agricultural necessity but by electricity and water availability, often uncontrolled and characterized by randomness. Moreover, existing irrigation datasets developed and used in the literature for South Asian regions considered the wrong crop season, focusing on pre-monsoon summer rather than the monsoon crop season, leading to misleading findings. By incorporating India-specific scenarios, our study demonstrated that alterations in land irrigation practices induce changes in atmospheric circulation, consequently influencing monsoon rainfall patterns, primarily in September. To substantiate these observations, we constructed a causal network among land-atmosphere variables across different river basins in India. Causal discovery methods revealed connections from land to atmosphere within a basin, atmosphere to atmosphere across basins, and atmosphere to land within a basin. This highlights that two neighboring river basins, traditionally assumed to be hydrologically independent when designing water management practices, are, in fact, interconnected through intricate land-atmosphere-land connections. These findings underscore the necessity for a systematic evaluation of India’s proposed large-scale river interlinking project, emphasizing the importance of addressing land-atmospheric feedback to ensure the project's success and sustainability.