Abstract. Evapotranspiration (ET) is a key component of the terrestrial water and energy balance, and numerous global gridded ET products are routinely used to assess historical variability and trends. However, differences in forcing data, model structure and physics in these products complicate robust ET trend analyses. Here, we present a systematic intercomparison of 14 global terrestrial ET datasets for the period 2000–2019. We introduce a topology framework that categorizes ET datasets according to their trend signatures within multi-product ensembles, providing insight into the structural role of each dataset and revealing how certain products consistently amplify or oppose dominant trends, patterns that are not evident from standard ensemble statistics. We find that products which amplify negative trends consistently oppose the dominant ensemble trend direction, whereas products that amplify positive trends tend to produce statistically significant trends where most datasets indicate weak or non-significant change. We quantify the magnitude, direction, and statistical significance of ET trends across products and evaluate their spatial consistency. The analysis reveals substantial divergence among datasets. While many products indicate predominantly positive ET trends, agreement on the magnitude and direction of change is lacking across many regions. In many regions, trends differ by more than an order of magnitude, and the spatial patterns of significant trends are highly product-dependent. The resulting harmonized trend estimates and classification provide a reference resource for evaluating current and future ET products, assessing uncertainty in trend studies, and guiding the use and improvement of ET datasets. More broadly, the topology framework can be extended beyond ET to geoscientific data product ensembles in general, enabling fitness for purpose evaluation, uncertainty assessment, and more systematic intercomparison across datasets.

Accurate initialization is a critical step in fully distributed ecohydrological and soil biogeochemical modeling applications, yet is often hindered by the computational cost of achieving steady-state conditions across large spatial domains. This study presents a novel initialization framework that combines a flux-tracking 1D spin-up with a random forest (RF) algorithm to efficiently generate spatially heterogeneous and topography-informed initial conditions accounting for lateral fluxes of water, carbon, and nutrients. The framework first performs a limited number of 1D simulations to obtain steady-state conditions in a subset of representative cells, then uses RF to extrapolate these results across the catchment. Applied to T&C-BG-2D, a fully coupled distributed ecohydrological-soil biogeochemical model, the scheme reconstructs over 90 % of the spatial variability in soil carbon and nutrient patterns in terms of probability distribution similarity while reducing computational demands by up to 90 % compared to a fully distributed spin-up procedure. A sensitivity analysis across multiple simulation scenarios reveals that the number of tracked cells required, varying from 20 % to 40 % of total domain grid cells, depends on the catchment's spatial complexity and the environmental covariates embedded in the RF predictors. The framework developed here can be applied to other spatially distributed models that explicitly account for lateral transfer fluxes, enabling large-scale distributed ecohydrological-biogeochemical model initialization under constrained computational budgets.

Water quality monitoring networks face an inherent trade-off between measurement precision and spatial-temporal coverage. We present an open-source smart water quality buoy designed to explore the potential of maximising deployment density and sampling frequency through low-cost instrumentation combined with AI-enhanced analytics. The stable buoy enclosure was developed using computational fluid dynamics, water flume validation, and extensive field testing. Initially designed for 3D-printing, it houses three sensors (temperature, turbidity and conductivity) with an ATmega328P microcontroller, real-time clock, flash logging, and/or LoRaWAN connectivity. Laboratory calibration established measurement reliability suitable for network-scale deployment. Field deployments have demonstrated autonomous operation with a relatively light monthly maintenance protocol. This platform enables novel monitoring approaches that leverage density over individual sensor accuracy. Initial Machine Learning models trained on national databases (millions of observations) convert basic sensor measurements into estimates of complex parameters — nutrients, dissolved oxygen, and bacteria — with encouraging accuracy. The high-frequency data from dense sensor networks enables automated pollution detection by analyzing concentration dynamics and comparing them against patterns learned from a large database of water quality measurements.By combining accessible hardware with AI analytics, we investigate whether prioritising spatial-temporal resolution can advance water quality monitoring capabilities, particularly for early pollution detection and regulatory compliance in under-resourced catchments.

Abstract. High-mountain wetlands (hereafter wetlands) have the potential to store and release large volumes of water. They provide substantial contributions to water supply for highland communities and receiving lowlands and their extended residence times can buffer water demand during drier periods. Despite their hydrological significance, major gaps remain in understanding the connection between wetlands and streams, which prevents us from accurately conceptualising them in models. We used a combination of conservative fluorescent tracing and monitoring of rainfall-runoff and wetland water levels to develop a perceptual model of connectivity between wetlands and streams. The experiments were conducted during the wet or dry season in five study catchments (0.38 km2 – 12.58 km2) from northern Ecuador to southern Peru. Fluorescein was introduced into wetlands and monitored downstream with activated carbon samplers for 5–12 months. Our results suggest travel times from less than 1 week to upwards of 3 ½ months, with one experiment seeing little to no response. Results indicate that wetlands have a stronger hydrological connection to streams in the wet season than in the dry season, where in some cases we observed no connection at all. Multiple peaks in fluorescein concentration from a single injection during the wet season may suggest that the wetlands contribute to streamflow via multiple pathways and processes. The results demonstrate a complex connection between wetlands and streams, controlled by topographic configuration and season, amongst other factors. However, persistent contributions from wetlands to streams observed several months after dye introduction support their significance to downstream, year-round water supply. The results provide improved understanding of wetland-stream connectivity that informs the development of models to address the need for site-specific evidence of wetlands’ influence on water supply and the ability to scale this information in space and time.

Evapotranspiration (ET) plays a central role in the terrestrial water cycle by coupling water, energy, and carbon exchanges between land and atmosphere. A Recent intercomparison of global ET products (Thomson and Markonis, 2024) revealed substantial uncertainties in estimated ET trends, including strong product dependence in magnitude, spatial patterns, statistical significance, and even trend direction.Here, we extend this work by introducing a topology framework that categorizes ET products based on their trend signatures. We processed and harmonized 14 global ET products—derived from reanalysis, remote sensing, synthesis approaches, and land surface models—onto a common 0.25° × 0.25° grid for the period 2000–2019. ET trends and associated significance were estimated using a block-bootstrapped Theil–Sen estimator at the grid scale and across meaningful spatial groupings, including IPCC reference regions, biomes, land-cover classes, Köppen–Geiger climate zones, elevation classes, and evaporation quantiles.Using this catalogue of recent ET trends and trend indices, such as the dataset concurrence index (DCI), we construct product-specific topologies by ranking the area fraction associated with characteristic behaviors including positive and negative signal boosters, and several forms of opposition.Globally, we find that “top negative signal boosters” are also “top outliers”. This means that top outliers are products that produce significant negative trends where all other significant trends are positive. This is caused by a majority of products producing positive trends. However, “top positive signal boosters” tend to be “top signal opposers”. These products have significant positive trends where the majority of products have nonsignificant trends. Both tendencies are true for a range of p-value thresholds. As a result, apparent large-scale ET trend signals are often driven by a limited number of products rather than by broad inter-product agreement.These topologies transform complex multi-product trend information into intuitive categories, enabling systematic identification of product-specific uncertainties and agreement patterns in large-scale ET trend assessments. This framework provides a new basis for categorizing ET products supporting interpretation of large-scale ET changes and data selection.Thomson, J. and Markonis, Y.: Multi-source analysis of recent changes in global terrestrial evapotranspiration, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-917, https://doi.org/10.5194/egusphere-egu24-917, 2024.

Climate and land use changes can significantly impact ecosystem water quantity as well as quality. To understand the corresponding physical processes within a catchment, an ecohydrological-biogeochemical model that can simulate water, nutrients and vegetation dynamics simultaneously is therefore required. The recently developed spatially distributed model T&C‐BG‐2D has enabled the simulation of coupled vegetation, hydrological, and soil biogeochemical dynamics within catchments. However, its potential in exploring the impacts of different scenarios and interventions on ecosystems can be limited by computational costs due to grid representations. In this work, we present a semi-distributed abstraction framework for T&C-BG-2D to simulate hourly river discharge and chemistry (C, N, P, K, Ca, Si, Mg) at the (sub)catchment outlet with minimal computational costs. Leveraging recent available remote sensing and reanalysis datasets, an algorithm was developed to enable a novel calibration procedure for all key model parameters (e.g., soil properties, land cover, and vegetation traits) within the catchment across representative hydrological response units. The newly developed semi-distributed version T&C-BG-SD was benchmarked against the fully distributed T&C-BG-2D model in the Hafren (Wales, UK) and the Erlenbach (Swiss pre‐Alps) catchments. To further evaluate the suitability of the framework in representing different catchment characteristics, the performance of the model was then examined across multiple catchments in the US and UK spanning diverse climatic conditions and land covers.

Spatial heterogeneity in water and energy fluxes drives patterns of vegetation productivity and soil carbon and nutrient cycling across landscapes. However, most ecohydrological models either neglect lateral transfers or treat biogeochemical processes in a spatially decoupled manner, limiting their ability to reproduce observed catchment-scale patterns. We address this gap by extending the mechanistic ecohydrological model Tethys–Chloris–Biogeochemistry (T&C-BG) to a fully distributed configuration (T&C-BG-2D) that explicitly represents lateral routing of soil carbon and nutrients. The model is evaluated against long-term hydrological and biogeochemical observations from the Hafren catchment (UK) and the Erlenbach catchment (Swiss pre-Alps), where it successfully reproduces observed dynamics of several river solutes, including dissolved organic carbon, ammonia, and nitrate. To overcome the computational bottleneck of distributed model initialization, we further introduce a hybrid spin-up framework combining flux-tracking one-dimensional simulations with a random forest–based spatial extrapolation. This approach efficiently generates spatially heterogeneous and topography-informed initial conditions while reducing computational costs by up to 90%. Together, these advances enable efficient, spatially explicit ecohydrological–biogeochemical modeling across complex landscapes.

Abstract Tropical cities are frequently exposed to short and intense rainfall that poses a significant flood risk to rapidly growing populations, making it essential to predict how these extremes will evolve under climate change. However, local projections remain limited by the lack of high‐resolution climate simulations and long‐term meteorological records. To address this, we present a globally scalable framework that downscales gridded daily rainfall data to hourly resolution using a stochastic disaggregation model. The model is calibrated using satellite‐derived hourly rainfall statistics, and rainfall intensity‐temperature scaling is applied using General Circulation Model‐projected dew point temperatures, eliminating the reliance on local hourly records. By linking dew point temperature‐rainfall relationships to projected changes from climate models, we generate future intensity‐duration‐frequency curves across 20 tropical and subtropical cities. Results show hourly rainfall intensities increase by 10%–17% by 2080–2100 relative to 1995–2014 under SSP2‐4.5 and under the assumption of Clausius‐Clapeyron scaling, with rare high‐impact events becoming up to six times more frequent under SSP5‐8.5. Temperature‐rainfall scaling emerges as the dominant source of uncertainty, surpassing climate model and emission uncertainty. We further show that using only mean temperature shifts rather than changes in the entire temperature distribution may significantly underestimate rainfall intensification. These findings underscore the need to accurately quantify local temperature‐rainfall scaling and offer a pathway for improved global projections of rainfall extremes, particularly in data‐scarce regions, supporting more resilient urban infrastructure planning for a future climate.

This work investigates the impact of microscale surface variability on the surface-energy balance (SEB) in vegetated urban areas, evaluates the suitability of several commonly used morphological variables (, , and ) in terms of their ability to parameterise the SEB, and investigates a set of alternative parameters. This is done by performing large-eddy simulations for an ensemble of randomly generated urban geometries (produced using a landscape generator) and comparing the results. The geometries would be considered identical by most numerical weather prediction (NWP) land-surface models (LSMs), but are uniquely different when resolved at the microscale. We find that variations in the SEB fluxes are relatively small, suggesting that assumptions made by existing LSMs are valid-namely that , , and alone do give a reasonable parameterisation of the urban surface from a mesoscale perspective. Using Gaussian-process-based machine learning to enable formal sensitivity analysis, we find that the maximum building height is important in determining all of the fluxes, with the exception of the latent heat flux, which depends predominantly on the fraction of shaded green space. The average building height and albedo of sunlit surfaces are also found to be of importance.

Climate-driven glacier retreat exposes newly ice-free terrain that is progressively colonized by plants, driving ecological succession and altering hydrological and biogeochemical processes in high-mountain ecosystems. Although hydrological impacts of glacier shrinkage have been widely explored, the effects of post-retreat vegetation succession remain poorly quantified. In this study, we apply a mechanistic ecohydrological model (T&C) that explicitly simulates plant migration and species range dynamics to assess hydrological responses to glacier retreat and vegetation succession from 1981 to 2099 under multiple climate change scenarios in a deglaciating ecosystem in the Swiss Alps. The results show that glaciers exert a first-order control on the hydrological cycle, particularly on runoff. Vegetation succession following glacier retreat plays a relatively minor role in hydrological processes initially, but its importance increases over time as glacier cover declines. The combined interactions among glaciers, vegetation and climate significantly modify hydrological regimes, with important implication for projecting future water resources, including changes in terms of magnitude and intra-annual (seasonal) variability, and water quality in high-mountain regions under continued global warming.

Enhanced rock weathering (ERW) is an emerging carbon dioxide removal (CDR) strategy that can support net-zero emission targets. However, current ERW modelling efforts rely on assumptions that introduce substantial variation in CDR estimates across varying ecosystems and hydroclimatic conditions. They typically ignore or oversimplify plant-soil interactions and high-frequency hydrological dynamics, obscuring short-term weathering responses and biotic feedbacks to soil moisture dynamics. Here, we introduce an integrated, process-based modelling framework, T&C-SMEW, which represents ecohydrological and ERW dynamics, along with microbially explicit biogeochemical processes. We compared framework simulations against a controlled mesocosm experiment and long-term field observations, demonstrating its ability to reproduce feedstock cation release, soil pH dynamics, gross primary production, and CO2 fluxes. T&C-SMEW reveals hydrological constraints and vegetation effects on ERW-mediated CDR by quantifying impacts on ecosystem respiration, net ecosystem exchange, and alkalinity export, emphasising the importance of ecohydrological modelling for ecosystem-level CDR estimation. These advances provide a modelling framework for identifying optimal deployment scenarios to establish ERW as a viable and operationally feasible CDR approach.

Agrivoltaic systems are characterized by the co-existence of photovoltaic panels on agricultural land, allowing simultaneous solar energy and food production without need for further land. Agrivoltaic installations alter the local microclimatic conditions of the land surface, impacting the performance of the agricultural systems embedded in them. In this study we develop an ecohydrological modeling framework combining a module that simulates changes in micrometeorology due to photovoltaic panel installations with a state-of-the-art model that resolves land surface water, energy, and vegetation dynamics (i.e., the terrestrial biosphere model T&C). We demonstrate that the modeling framework is capable of reproducing grassland dynamics across a broad range of climates and agrivoltaic architectures. With the use of the model we evaluated grassland performance across the Mediterranean for two most commonly used architectures, namely mixed mounted solar panels and rotating solar tracking panels. We found that C3 grassland yields can be significantly enhanced only in climates where annual potential evapotranspiration exceeds annual rainfall. Changes in grassland productivity were attributed primarily to changes in the light environment at the land surface, with changes in surface aerodynamic roughness and rainfall redistribution due to drainage on panels playing a smaller negative role of comparable magnitudes.

In this study we propose a new methodology for pluvial flood risk estimation, combining stochastic rainfall modelling, climate projection based adaptations of the rainfall frequency-intensity relations and DEM data sets, along with hydrodynamic modelling. New global precipitation datasets, such as CMORPH, GSMaP or MERRA2 offer an affordable and accessible solution for water resource and water-hazard risk management in data-scarce regions and enable comprehensive global comparative studies. However, these datasets, often derived from satellite observations and coarse-scale climate modelling, consistently underestimate short-duration, high-intensity rainfall events, particularly those lasting one hour or less, that belong to the tails of the distributions (i.e., return levels higher than 30-year). This underestimation goes beyond spatial scale considerations, commonly addressed by areal reduction factors. Consequently, utilizing these global datasets for pluvial flood risk analysis results in conservative flood risk estimates. The availability of global terrain models and mapped man-made structures like buildings, channels, and roads enables the generation of wide-coverage digital surface models. These can be used for flood inundation modelling in combination with corrected extremes of the global precipitation data sets, allowing near-global rough flood risk estimates. In this study, we introduce a methodology for estimating pluvial flood risk using openly available global datasets. To achieve this, we derive hourly-scale Intensity-Duration-Frequency (IDF) curves suitable for pluvial flood inundation modeling in ungauged areas using global precipitation datasets. The first step uses high temporal resolution satellite remote sensing rainfall data (GSMaP) to train a stochastic rainfall generator model - the point process Bartlet-Lewis model. Subsequently, the weather generator is used to disaggregate daily global precipitation data (GPCC) through stochastic ensemble simulation. The resulting disaggregated ensemble data is then utilized to generate more accurate IDF curves including uncertainty, forming the basis for pluvial flooding risk assessments. Our approach integrates the openly available FabDEM terrain model with OpenStreetMap to generate digital surface models for flood risk modeling analysis. Discrepancies in flood inundation risk estimates in urban environments, attributable to underestimated rainfall intensity, are demonstrated using CADDIES, a 2-dimensional hydrodynamic model. The workflow allows the IDF curves for the current climate to be adapted based on climate model projections of temperatures using the Clausius–Clapeyron relation, and to study their impact on future flood risk. A comparative risk analysis is presented for several tropical coastal cities, including future pluvial risk projections. All analytical steps adhere to FAIR principles, utilizing publicly available datasets. The proposed workflow provides globally applicable first order estimates of pluvial flood risk, especially in data-poor areas, with better quality than existing global IDF studies or IDF curves derived directly from global precipitation datasets.

Soil-plant-atmosphere interaction (SPAI) plays a significant role on the safety and serviceably of geotechnical infrastructure. The mechanical and hydraulic soil behaviour varies with the soil water content and pore water pressures (PWP), which are in turn affected by vegetation and weather conditions. Focusing on the hydraulic reinforcement that extraction of water through the plant roots offers, this study couples advances in ecohydrological modelling with advances in geotechnical modelling, overcoming previous crude assumptions around the application of climatic effects on the geotechnical analysis. A methodology for incorporating realistic ecohydrological effects in the geotechnical analysis is developed and validated, and applied in the case study of a cut slope in Newbury, UK, for which field monitoring data is available, to demonstrate its successful applicability in boundary value problems. The results demonstrate the positive effect of vegetation on the infrastructure by increasing the Factor of Safety. Finally, the effect of climate change and changes in slope vegetation cover are investigated. The analysis results demonstrate that slope behaviour depends on complex interactions between the climate and the soil hydraulic properties and cannot be solely anticipated based on climate data, but suctions and changes in suction need necessarily to be considered.

The study of the water cycle at planetary scale is crucial for our understanding of large-scale climatic processes. There have been numerous studies that quantified the water cycle and its components, i.e., precipitation, evaporation, and runoff, over the land and the ocean. However, very little is known about how water fluxes are distributed across regions with different climatic or land properties. Here, we address this gap by providing robust estimates for terrestrial precipitation over a suite of land cover types, biomes, elevation zones, and precipitation intensity classes. We achieve this by estimating the mean annual precipitation of a 17-dataset ensemble between 2000 and 2019 at 0.25° spatial resolution. Our estimate of annual terrestrial precipitation is at approximately 114 000 ± 9 400 km3, with about 70% falling over one third of the grid cells, 80% over the 0 – 800 elevation zone, and two-thirds over forested regions. Our results also highlight that despite the current progress in the development of global scale data products there are still substantial uncertainties over the arid and/or high-elevation areas. Bigger discrepancies appear within the reanalysis data products, while remote sensing estimates show a better agreement with the in-situ ground truth. These results help to detect regions of high observational fidelity and pave the way to further explore and improve observational uncertainties. At the same time, we provide consistent estimates that can be used for benchmarking the precipitation partition in the climate models, and most importantly that can be used to assess future changes in global precipitation.

Quantification of the impact of environmental stress on terrestrial vegetation photosynthesis is crucial for our understanding of the global carbon cycle, particularly under a changing climate. Vegetation responses to environmental stress manifest first as plant physiological changes, and at later stages through changes in canopy structure. Here we leverage CO2 and water flux data from 103 eddy covariance towers and satellite thermal images to assess whether current satellite reconstructions of solar-induced chlorophyll fluorescence capture these plant mechanisms. After removing seasonality using standardized anomalies (z-scores), we found that the relationship between tower-observed gross primary productivity and fluorescence reconstructions considerably weakened across a wide range of biomes. This loss of correlation results from a decoupling between stomatal responses and the physiological emission yield (ΦF) of fluorescence reconstructions during soil and atmospheric dry periods. The consequence is that productivity derived from fluorescence reconstructions will be progressively overestimated as dry conditions persist. Remotely sensed solar-induced chlorophyll fluorescence overestimates gross primary productivity under dry conditions due to the decoupling between stomatal responses and the fluorescence emission yield, as revealed by data from 103 eddy covariance towers and satellite thermal images.

Enhanced rock weathering (ERW) is a promising CO2 removal (CDR) strategy that aims to accelerate the natural process of silicate weathering to increase soil pore water alkalinity and sequester CO2. However, the measurement, reporting, and verification (MRV) of ERW remains challenging due to existing limitations of aqueous-phase sampling methodologies, such as passive and tension lysimeters, which may not fully capture weathering fluxes across varying soil moisture conditions. This study assesses the potential of a centrifugation-based pore water extraction method to improve the accuracy and reliability of ERW measurements. Using a forest ERW trial in Wales, UK, we compared the chemistry of soil pore waters obtained via lysimeters and centrifugation from feedstock-amended and control plots. The centrifugation method detected elevated total alkalinity and Ca concentrations in soil pore waters from feedstock-amended soils, whereas the effect of feedstock amendment was not detectable in pore waters extracted with lysimeters. The high tensions applied during centrifugation likely capture weathering products dissolved in meso- and micropore water, which lysimeters cannot extract. These findings suggest that centrifugation provides a scalable, low-cost approach for ERW monitoring, with implications for improving existing MRV protocols.

Ecohydrological models have progressively advanced in complexity by incorporating the latest knowledge from hydrology, ecology, and related disciplines. Recent developments include coupling hydrological processes with detailed dynamic vegetation responses to environmental cues, soil biogeochemical dynamics, and the integration of human activities. These activities range from the management of forests and croplands to the operation of built infrastructure such as reservoirs. Advancements in mechanistic approaches offer significant opportunities to translate fundamental knowledge into actionable strategies for a sustainable future, encompassing infrastructure planning and resilient water resource management. However barriers in their implementation include lack of a unified computational framework and lack of data to support the development of such a framework.In this study, we present a unified computational framework demonstrating how ecohydrological modeling can inform the design of a sustainable future. Our applications address key areas such as forest and cropland management, sustainable agriculture, climate-resilient infrastructure, and sustainable water resource management. Specifically, we introduce multiple new mechanistic hydrological, plant physiological, and infrastructure processes into the ecohydrological and ecosystem model T&C. We also address challenges related to model application in data-scarce contexts and propose a roadmap for leveraging mechanistic ecohydrological modeling to develop actionable design principles for achieving a sustainable, net-zero future.