The emergence of AI-driven Earth observation systems promises to broaden access to petabyte-scale geospatial data beyond domain specialists. However, translating this vision into operational scientific infrastructure requires addressing fundamental challenges in data virtualization, code transparency, and domain-specific reasoning. We present Hydrology Copilot, a cloud-native AI framework for natural-language-driven analysis of Earth observation data. To demonstrate operational capabilities at scale, we implement the system using NASA's North American Land Data Assimilation System version 3 (NLDAS-3), which provides surface meteorological forcing and land-surface model output across North and Central America at 1-km resolution, from which drought diagnostics are derived.The system integrates five core contributions: (1) scalable data virtualization using Kerchunk-based cloud-optimized access, achieving a 1.5 to 4.6 times improvement in I/O latency across benchmark queries spanning regional single-day extractions (4.6 times speedup) to continental monthly aggregations (1.5 times speedup); (2) transparent code generation through Microsoft Azure AI Foundry agents that expose executable Python workflows for scientific verification; (3) persistent conversational memory enabling multi-turn analytical discourse across sessions; (4) intelligent query validation that enforces dataset boundaries and resolves ambiguous requests before execution; and (5) a multi-agent architecture coordinating query parsing, code generation, and visualization. We evaluate the system through drought-monitoring workflows, demonstrating reliable code generation, accurate results validated against reference computations and the operational U.S. Drought Monitor, and efficient operation across increasingly complex tasks.By bridging natural-language interfaces with rigorous hydrological analysis, Hydrology Copilot advances beyond proof-of-concept demonstrations to provide a deployable framework for operational Earth science applications.

Quantifying the influence of vegetation dynamics on carbon uptake and its climatic drivers is critical for regional carbon-cycle assessments, yet most land surface modeling studies still parameterize static vegetation. This limitation is particularly relevant in South Asia, where widespread vegetation greening has altered ecosystem productivity. This study focuses on the spatiotemporal variability of gross primary productivity (GPP), evapotranspiration (ET), and the resulting ecosystem water-use efficiency (WUE = GPP/ET) to understand the long-term trends and drivers across South Asia. Simulations are performed over South Asia for 1985-2023 using the Indian Land Data Assimilation System (ILDAS) with a dynamic vegetation scheme and hybrid meteorological forcing involving local IMD (India Meteorological Department) precipitation while maintaining transboundary consistency. Results reveal significant regional variations with high GPP ( >2500 g.C/m(2)/year) and WUE (> 2 g.C/kg H2O/year) over the lower Himalayan and northeastern regions, and low values over arid north-western region. Seasonal variability peaks during the monsoon for GPP (sigma = 95.44 g C/m & sup2;) and ET (sigma = 39.60 mm), whereas WUE varies the most in the pre-monsoon (sigma = 12.48 g C/m & sup2;/mm), reflecting vegetation adaptation under water-limited conditions. Non-parametric elasticity analysis indicates temperature (epsilon = -21.78) and pressure (epsilon = 76.30) as dominant climatic drivers of WUE, while soil moisture and leaf area index (LAI) are the primary internal drivers. Trend analysis reveals significant increases (p < 0.05) in GPP and WUE across large parts of South Asia, particularly in northern and central agro-ecological zones, consistent with vegetation greening within existing forest cover. Model evaluation shows stronger agreement for GPP with FluxSat (median R = 0.70; KGE = 0.43) than with MODIS, and robust ET performance against MODIS (median R = 0.81; KGE = 0.44) over vegetated regions. In the absence of direct observations, cross-validation of modeled WUE with MODIS-derived WUE indicates moderate agreement (median R = 0.36) but low KGE (0.21), reflecting uncertainties in ratio-based diagnostics. Overall, the spatial coherence of errors supports the utility of the current ILDAS framework with dynamic vegetation scheme for assessing long-term carbon-water interactions across South Asia.

We present a robust, global method to reconcile absolute water surface elevations (WSE) from the Surface Water and Ocean Topography (SWOT) mission with a large-scale hydrodynamic model by iteratively correcting river bathymetry. After removing systematic sensor and representativeness biases, we update riverbed elevations under DEM and min-depth feasibility constraints and interpolate corrections along reaches. Applied worldwide at 0.05° resolution over 2023–2025, the approach reduces the global median absolute WSE bias (3.10m→0.18m), preserving the same global median correlation (0.38). A modest ~4% increase in the standard deviation ratio (1.07→1.11) indicates slight variance inflation. Evaluation over 2015–2025 at 16,746 Hydroweb sites and 2123 GRDC gauges, restricted to reaches with updated beds, confirms substantially reduced bias, negligible changes in correlation and variability, and no significant change in global median discharge skill. When combined with models, this bias-aware bathymetry will enable a refined numerical representation of floods globally.

Abstract. The fine scale distribution of snow is important for avalanche forecasting and biological refugia. Here, we test how historical context provided by 3 m snow cover observations derived from PlanetScope commercial satellite imagery could be used to downscale fractional snow covered area (fSCA) observed by the MODerate resolution Imaging Spectroradiometer (MODIS), Visible Infrared Imaging Radiometer Suite (VIIRS), and the Harmonized Landsat and Sentinel-2 (HLS) product using previously published approaches. We evaluate this over Colorado and California montane meadows using 1) probabilistic snow cover maps, and 2) random forest machine learning models. We then compare versus a downscaling benchmark that relies only on terrain characteristics, eliminating the need for PlanetScope observations. Provided these three approaches, downscaling snow cover using random forest models performed best on average, largely because these models corrected annually consistent snow cover biases between PlanetScope and coarser resolution fSCA observations. The approach used to downscale influenced 3 m snow cover most, followed closely by the accuracy of the fSCA estimate that snow cover was downscaled from. Snow cover downscaled from HLS using only terrain indices was often similar or better than snow cover downscaled from MODIS and VIIRS using context from PlanetScope. This demonstrates how a limited historical record of commercial satellite observations can be used to estimate the fine-scale pattern of snow cover in many regions, but also when publicly accessible remote sensing retrievals and information about the terrain may obviate the need for commercial observations.

Tropical peatlands, which cover 13–15% of the global peatland area, play a vital role in the global carbon storage, yet their key hydrological processes influencing carbon dynamics are not well accounted for in most land surface models. The Cuvette Centrale wetland of the Congo Basin is the world’s largest continuous tropical peatland area, which is governed by a complex hydrology. It is partly driven by river-peatland interactions and a spatially variable bimodal annual precipitation pattern across the Congo Basin, as well as by an unknown influence of deeper groundwater on the phreatic water levels (WL) in the peatlands. Accurate estimation of water and carbon dynamics for peatlands necessitates advancements in land surface models by incorporating peatland-specific modules to simulate key hydrological processes. In this study, we enhance the Noah-Multiparameterization (Noah-MP) land surface model to incorporate peatland hydrological processes, by including peat soil hydraulic parameters, microtopographic integration, and new runoff and evapotranspiration schemes. The peatland-specific scheme and the default TOPMODEL scheme of Noah-MP (further called “reference”) are applied across the Cuvette Centrale domain using two different meteorological forcing datasets, MERRA-2 and MERRA-2 with CHIRPS precipitation. The simulations were evaluated using in-situ WL observations and terrestrial water storage anomaly (TWSA) observations from the GRACE and GRACE-FO missions. Preliminary evaluation with in-situ WL observations shows overall improvement for peatland-specific simulations compared to the reference. Especially for the experiment with CHIRPS precipitation, which generally showed the better skill metrics, the new peatland scheme shows 5.79% increase in correlation and 86% reduction in RMSE compared to the reference driven by the same forcing. The terrestrial water storage anomalies simulated by the peatland-specific model in conjunction with altimetry-based river water storage estimates, also show an increased anomaly correlation with GRACE mascon-derived water storage anomalies. By more realistically representing the peatland hydrology, this work lays the groundwork for improved predictions of tropical peatland carbon–water interactions, with future scope for coupling with a river-routing model to better understand the peatland-river interactions for the Congo peatlands.

Irrigation is the major consumer of freshwater resources on Earth and represents the largest human intervention in the water cycle. Most irrigation modeling frameworks rely on simplified assumptions regarding the timing of irrigation that result in significant error in the irrigation water use estimation. In this study, we developed a generalized data-driven approach for estimating these irrigation timing attributes using a change point detection algorithm applied to thermal remote sensing data. Land surface temperature at cropland pixels was compared with hydrologically similar natural land cover pixels nearby to extract irrigation attributes. The approach was evaluated over two areas: Nebraska (NEB) and Mahabad, Iran (MAH), for which we had the in situ irrigation data. The method detected the start and end of the irrigation season with reasonable accuracy, exhibiting errors of 18 % and 15 % in estimating the duration of season, in NEB and MAH respectively. The cloud cover either at the start or the end of season was the primary source of error in both cases. Irrigation event detection accuracy across 10 NEB sites yielded F1-scores (Precision & recall combined score) of 0.59-0.74, varying with change point detection algorithm parameters. To optimize these parameters, extensive hyperparameter tuning was performed, leading to specific suggestions tailored for different irrigation practices. The results presented here demonstrate that the LST-based approach can be effective in characterizing interannual variations in irrigation timing attributes.

Evapotranspiration strongly couples land and atmosphere to regulate water, carbon and energy fluxes across tropical South America. Ongoing deforestation and fires reduce the capacity of deep-rooted trees to recycle moisture, while intensifying droughts further alter the timing and magnitude of evapotranspiration. Here we present a high-resolution, data-constrained hydrological modelling analysis to isolate the effects of anthropogenic disturbances and droughts on evapotranspiration and vegetation function across the Amazon and adjacent biomes from 2003 to 2020. We find that evapotranspiration declines from deforestation persist 21-22% longer than those caused by fire or drought alone. When these stresses co-occur, evapotranspiration losses intensify by 36% and persist 66% longer than the average impact of individual stressors. Across neighbouring biomes, we find that grasslands and savannas in the Cerrado are most vulnerable to droughts, with evapotranspiration recovery often exceeding seven years, while Pantanal wetlands recover rapidly due to sustained moisture availability. Furthermore, vegetation productivity declines under compounding stresses despite concurrent greening trends. Our findings reveal that recurrent human disturbances erode ecosystem resilience, threatening long-term ecological stability. Isolating the human footprint on evapotranspiration is pivotal to guide sustainable land-use transitions that preserve land-atmosphere coupling in South America's tropical ecosystems.

Soil moisture plays a crucial role in weather forecasting and climate prediction, and provides useful information for flood and drought monitoring, and agricultural irrigation scheduling. In recent decades, soil moisture measurements have been enabled by space-borne passive microwave sensors. Nonetheless, the application of satellite remote sensing soil moisture data in operational meteorological and hydrological contexts is constrained by their prolonged latency and relatively coarse spatial resolution. This study describes the development of an operational enhanced Soil Moisture Active Passive (SMAP) soil moisture retrieval processor, referred to as SMAP_E_OPL, which addresses these limitations. SMAP_E_OPL produces near-real-time (NRT) soil moisture data at a 9-km grid resolution utilizing the single channel retrieval procedure employed in the long-latency SMAP Level 2 Enhanced soil moisture products (i.e., SMAP_L2SMP_E). In contrast to SMAP_L2SMP_E, the SMAP_E_OPL processor uses the Inverse Distance Squared Weighted interpolation approach to resample the NRT SMAP L1B brightness temperature (TB) data from a 33-km resolution to the target grid for enhanced computational efficiency. Moreover, SMAP_E_OPL also allows the interoperable use of soil temperature and snow depth estimates from various operational land surface model (LSM) configurations. A bias-correction process is implemented to ensure consistency in the effective soil temperature (Teff) between SMAP_E_OPL and SMAP_L2SMP_E. Evaluation based on in situ soil moisture measurements across the continental United States demonstrates that SMAP_E_OPL delivers the NRT soil moisture at 9-km grid resolution with accuracy comparable to the official (but high latency) SMAP enhanced soil moisture products. The use of different LSMs in SMAP_E_OPL to estimate Teff yields similar quality in soil moisture retrievals, highlighting the flexibility of SMAP_E_OPL with respect to LSMs. These results confirm the viability of the SMAP_E_OPL processor in providing high-spatial-resolution soil moisture data with low latency for operational forecasting, data assimilation, and monitoring systems.

Abstract The absence of a definitive follow‐on strategy for the Soil Moisture Active Passive (SMAP) mission, which has already surpassed its planned mission lifetime, presents a significant challenge for the hydrological science community. GNSS‐Reflectometry (GNSS‐R), which exploits reflected satellite navigation signals, has emerged as a rapidly growing alternative for soil moisture retrieval, increasingly positioned as a low‐cost path to global observation. The two approaches, however, differ fundamentally in how they sense the surface, and the hydrological science community should weigh these differences carefully before decisions are made that could compromise the continuity of this essential climate variable's data record. In this study, we compared soil moisture products from these two systems across seven perspectives: (a) observation characteristics, (b) spatial resolution, (c) coverage and frequency of observations, (d) vulnerability to radio frequency interference, (e) complementarity potential, (f) performance, and (g) cost considerations. Our assessment suggests that GNSS‐R is better understood as a complement to SMAP than as a replacement. A GNSS‐R constellation can deliver sub‐daily observations, and lower upfront costs, capabilities that genuinely extend what single‐platform missions can offer. At the same time, GNSS‐R retrievals exhibit variable spatial sampling and canopy attenuation, and the highest‐performing products are typically trained, calibrated, or benchmarked against SMAP (an implicit cost that the low‐cost framing of GNSS‐R does not fully capture). Taken together, these findings underscore the need for sustained investment in both SMAP‐class L‐band missions and the expanding GNSS‐R constellation, each contributing complementarity strengths to a resilient global soil moisture observing system.

Land surface and crop models both simulate irrigation, but they differ in their approaches, primarily because they were originally developed for distinct purposes and scales. Through an example case study in a highly irrigated region, this research helps to better understand the gap between these models and the complexity of irrigation modeling. More specifically, irrigation was estimated over the Po Valley (Italy) at a 1 km(2) spatial resolution using (i) a crop model, AquaCrop, and (ii) a land surface model, Noah-MP. Both models were run with sprinkler irrigation using a similar setup within NASA's Land Information System, i.e. forced with the same meteorology and constrained by the same soil texture and generic crop parameterization. Irrigation estimates were evaluated at the pixel and basin scale, using in situ reference data. In addition, surface soil moisture (SSM), vegetation, and evapotranspiration (ET) estimates were compared with satellite retrievals.Noah-MP has on average higher annual irrigation rates (434 mmyr(-1)) than AquaCrop (268 mmyr(-1)), mainly because Noah-MP simulates more irrigation water losses (not consumed by transpiration) via runoff, interception, and soil evaporative losses, whereas AquaCrop only accounts for soil evaporative losses. When adding representative application water losses to irrigation estimates from AquaCrop, and conveyance water losses to the estimates from both models, the irrigation estimates from both models fall within reported ranges of 500-600 mmyr(-1). For the field-based evaluation, Noah-MP presents large irrigation events (>100 mm per event) and less interannual variability than AquaCrop. Two-week averaged SSM estimates from both models agree well with downscaled estimates from the Soil Moisture Active Passive (SMAP) mission, with spatially averaged unbiased root mean square differences of 0.05 and 0.04 m(3)m(-3) for AquaCrop and Noah-MP, respectively. Both models show limitations in terms of vegetation and ET modeling, mainly due to simplistic vegetation modules and suboptimal parameterization in both models. The results highlight the complexity of irrigation modeling due to its anthropogenic nature, and also show the need for better observations to validate and guide model estimates: reference irrigation data are sparse and satellite retrievals under irrigated conditions are quite uncertain.

In a warming climate, wildfires are becoming increasingly common, especially in semi-arid environments. Wildfires can disrupt forest ecosystems and induce changes to the land surface. Collectively, these impacts can alter the hydrologic response of a catchment following a fire, resulting in increased potential for surface runoff, reduced evapotranspiration, and, ultimately, a higher risk for flash flooding and mass wasting. The timescale of post-fire recovery of hydrological processes to return to pre-fire conditions is not well established due to the lack of ground measurements. Accurate characterization of the impacts of fire on hydrologic response is also challenging to simulate, given the complex interplay of various processes. Here, we present a generalized framework to quantify the impacts of wildfire on runoff generation. We consider the disturbances in the vegetation and soil as the two main factors contributing to post-fire floods. Using an ensemble modeling structure to account for parameter uncertainty, remotely sensed leaf area index (LAI) is assimilated into a land surface model (LSM) to simulate vegetation disturbance, and the maximum land surface saturation LSM parameter is decreased to parameterize the soil disturbance following observed fires. We consider the impacts of fire-induced changes to LAI and soil saturation on hydrologic states like runoff and evapotranspiration for two case studies. These case studies demonstrate the general applicability of hydrophobicity formulation to serve as a guideline for exploring the range of hydrologic responses post-fire.

The Hydrological Modeling and Analysis Platform, version 3, (HyMAP-3) is a state-of-the-art global hydrodynamic model integrated into NASA's Land Information System (LIS; Kumar et al., 2006). HyMAP was initially developed in preparation for the Surface Water and Ocean Topography (SWOT) mission, where the objective was to have a modeling system capable of assimilating SWOT data toward a near-real-time global estimation of water discharge. Since the initial HyMAP implementation in 2011 (Getirana et al., 2012), HyMAP-3 has undergone numerous improvements, such as its vectorization and parallelization, inclusion of human and nature-driven processes (e.g., river damming, water management, river bifurcation, compound coastal flooding), full integration into LIS, and coupling with a wide range of land surface models (LSMs), data assimilation (DA) and forecast schemes. This technical memorandum fully describes HyMAP-3's concepts and capabilities, as well as its overall architecture within LIS.

Extreme weather events cause significant societal impacts, in particular, when they behave unexpectedly. The "brown ocean (BO)" effect, describing the ability of the land surface via soil moisture to support tropical cyclone (TC) maintenance and intensification (TCMI) after landfall, remains poorly understood. Building upon our previous modeling framework utilizing the NASA Unified WRF (NU-WRF) system, this follow-on study explores the contributions of the dynamics of the soil moisture and advected water vapor to the TCMI of Tropical Storm (TS) Bill (2015) over the U.S. southern Great Plains (SGP). Impacts of various soil moisture conditions and surface enthalpy flux conditions on Bill's inland intensification were investigated by comparing their land-atmosphere interaction components of energy fluxes along with a backward trajectory analysis and three-dimensional visualization of low-level atmospheric moisture. Results demonstrate that the high antecedent soil moisture across the central United States (Great Plains and Mississippi Valley) from prior rainfall was crucial for the TCMI of TS Bill over the SGP. Without ample latent heat flux over land, even a moisture- laden low-level jet from the ocean rapidly dried over land, preventing intensification and causing storm dissipation in our simulations. Backward trajectory analysis suggests that high soil moisture content can enhance humidity within the storm's inflow, including the advection from the ocean, far inland, thus supporting the TCMIs. Ultimately, the inflow feeding the inland TC core is influenced by active land-air interactions within the boundary layer, where soil moisture and lower boundary conditions directly impact lower-tropospheric humidity, enabling or hindering the BO effect and subsequent TC intensification.

Climate change and anthropogenic disturbances have altered the global freshwater cycle, causing it to deviate from assumptions of stationarity. However, most current drought assessment methods do not consider such impact of nonstationarity. Here, we introduce a theoretical framework that quantifies errors in drought assessment under various nonstationary conditions, using bias-variance decomposition and Monte Carlo simulations. We then apply this framework to evaluate bias and uncertainty in drought estimation based on remote-sensing-informed global terrestrial water storage changes. Our analysis confirms that with over half of the world facing water storage declines, assuming stationarity can lead to underestimation of moderate droughts with high uncertainty depending on the chosen climatology reference periods, while overestimations of exceptional droughts remain consistent and highly certain, irrespective of climatology reference periods on average. Our framework offers insights into building appropriate climatology reference periods for drought estimation while emphasizing the need to recognize seasonal and regional differences.

Understanding the interactions between the atmosphere, the land, and the subsurface is fundamental to hydrology and is critical for a better assessment of the impacts of climate change and human management on hydrological systems. However, many land surface models simplify the subsurface hydrology and thereby these interactions. In this study, we couple the land surface model Noah-MP included in the NASA Land Information System (LIS) with the integrated hydrologic model ParFlow (ParFlow-LIS) using the Earth System Modeling Framework (ESMF) and the National United Operational Prediction Capability (NUOPC). This coupling improves the simulation of water and energy cycle processes by adding the three-dimensional variably saturated and heterogeneous flow in the subsurface using sophisticated and nonlinear physics-based equations as well as the advances in satellite remote sensing-based data assimilation of the land surface, thereby benefiting the integrated hydrologic modeling and data assimilation community. We use the High Plains aquifer, located in the central United States, as a testbed to evaluate the coupled ParFlow-LIS system. The new ParFlow-LIS system accounts for the effects of topographically driven flows on the land surface, producing more fine-scale patterns of land surface states and fluxes than standalone LIS. In addition, ParFlow-LIS enables the consideration of the effect of subsurface water storage on evapotranspiration. This is particularly important in areas and times with dry soils, such as during drought conditions or in the presence of a cone of depression due to pumping.

Soil moisture is a significant environmental factor that influences both the water and energy balance at the land-atmosphere interface. Therefore, proper assessment of the spatial and temporal distribution of soil moisture is crucial for many hydrological applications such as weather forecasting, agricultural water resource management and drought monitoring. This study involves the assimilation of Soil Moisture Active Passive (SMAP) soil moisture dataset within a land surface model and the evaluation of its performance in precise estimation of soil moisture by comparing the statistics with respect to standard European Space Agency’s Climate Change Initiative (ESA-CCI) soil moisture dataset. The Ensemble Kalman Filter technique has been used for assimilating SMAP soil moisture data using Noah-MP land surface model within NASA Land Information System (LIS) framework. The data assimilation (DA) framework includes Cumulative Distribution Function (CDF) matching for bias correction and twenty ensembles per tile. Meteorological forcings for the simulations have been taken from MERRA2 and IMD. Improvement or degradation due to DA has been analyzed in terms of the difference in anomaly correlation between open loop (OL) and DA soil moisture outputs with respect to the ESA-CCI soil moisture dataset over the entire Indian domain. The DA result shows improvement over larger areas in the case of MERRA2 forced simulations than IMD+MERRA2. The seasonal impact of DA in terms of the differences in DA and OL simulated soil moisture shows less variability in summer than winter. The results are validated with in-situ soil moisture datasets. Overall, the study shows that data assimilation is giving better results than open loop LSM simulation, which can be used for improved estimation of other water balance components.