1.2.1 Methodological Framework Overview

The methodology is organised around seven interconnected pillars (listed below) that move from diagnosis to assessment-ready evidence with a structure specifically designed to deliver the call’s requirements on scientific ambition, and pathways toward improved projections and assessments. Interdisciplinarity is a fundamental feature of the project, bringing together the modelling of surface hydrology, groundwater and routing, vegetation, carbon-cycle and climate, with Earth Observation, benchmarking, data–model fusion, and assessment-oriented climate science in a single workflow.

First pillar (WP4, WP10–12). The first pillar of the project establishes a genuinely interdisciplinary model-development framework that bridges the long-standing silos separating hydrology- and climate-oriented land-surface modelling approaches, bringing more detailed hydrological process representation into dialogue with the broader representation of ecosystem, agricultural and land–atmosphere processes in climate models. At these cross-disciplinary process interfaces, CHORALE will follow an explicit co-development methodology. It will first identify the common physical processes addressed from different disciplinary perspectives and distinguish the conceptual quantities and assumptions represented by each approach. It will then establish common or mutually informative process-level constraints, test whether alternative representations are complementary, mutually constraining or incompatible, and, where scientifically justified, co-develop consistent representations and interfaces. The resulting representations and interfaces will be evaluated across model architectures to determine which concepts and implementations are transferable, which require model-specific adaptation, and where alternative representations should be retained. This creates an iterative development loop across disciplinary communities rather than a sequence of parallel model developments. In this respect, the methodology will be operationalised through common process testbeds that combine model-specific reference configurations, fit-for-purpose observational constraints and shared evaluation criteria, progressing from offline process evaluation to reduced-cost coupled testing before configurations are selected for coordinated experimentation. CHORALE supports interdisciplinary co-development through an Open Science and FAIR shared environment combining version-controlled code, containerised software, Jupyter workspaces, federated HPC/storage and ESGF-compatible output. Common workflows will make model development, evaluation and transfer reproducible across the four ESMs. The shared workflow will benchmark hydrology, vegetation and carbon-cycle developments before coordinated experimentation. Vegetation access to subsurface water provides a key interdisciplinary interface (Section 1.4): hydrologically inferred root-zone storage constrains the accessible-water reservoir, while mechanistic rooting schemes determine how roots are distributed and respond to soil moisture. Joint evaluation will establish whether these representations are complementary, mutually constraining or model specific.

Second pillar (WP1). The second pillar establishes the observational basis for process understanding, model development and integration. WP1 will select and curate fit-for-purpose EO, in situ and complementary datasets according to process relevance, representativeness, provenance, independence and uncertainty. These resources will support repeatable benchmarking of targeted model developments and broader readiness validation coordinated in WP4. Regional testbeds—including Ebro/LIAISE and California as candidate regions—will be selected jointly for process relevance, observational richness and computational efficiency. The project will also draw on expertise, products and methodologies developed through ESA FutureEO and related activities, including evapotranspiration, irrigation and merged precipitation products from ESA 4DMED-Hydrology, ESA-WACMOS MED and ESA Irrigation+. A targeted diagnostic set will constrain Water–Energy–Carbon relationships across process-relevant timescales using complementary satellite and ground observations. Satellite products will characterise precipitation, soil moisture, evapotranspiration, land-surface temperature, vegetation condition and productivity, and terrestrial water-storage changes. Site measurements will provide soil-moisture profiles, meteorological forcing and, where available, groundwater levels and irrigation records. Eddy-covariance flux towers will quantify exchanges of carbon dioxide, latent and sensible heat between ecosystems and the atmosphere, supported by radiation and soil measurements. River-gauge records will characterise streamflow variability, seasonal runoff and catchment-integrated hydrological responses. Vegetation condition and its influence on water and energy exchanges will be assessed using diagnostics of root-zone soil moisture and energy partitioning between latent and sensible heat fluxes, complemented by novel indicators of plant drought-stress such as the Cooling Efficiency Factor Index (CEFI, Zampieri et al., 2025). The CEFI, currently computed using EUMETSAT MSG satellite, covers Europe, Africa, the Middle East and part of South America. Hydrological diagnostics will include runoff coefficient–soil-moisture coupling to quantify antecedent-wetness controls on event runoff (Crow et al., 2019); nonlinear runoff-generation thresholds derived from satellite soil moisture, rainfall and observed discharge (Ortenzi et al., 2025); water-cycle closure and partitioning among evapotranspiration, runoff and storage change using precipitation, evapotranspiration, discharge and GRACE/GRACE-FO observations (Pellet et al., 2019); and irrigation-induced soil-moisture variability (Massari et al., 2026). CHORALE will make use of Copernicus data and services as core observational and reanalysis references: the C3S ERA5 and ERA5-Land reanalyses for offline forcing and evaluation; C3S climate data records of Essential Climate Variables, including soil moisture, land surface temperature and leaf area index/fAPAR; Copernicus Land Monitoring Service vegetation and land-cover products; and CEMS GloFAS river-discharge reanalysis, complemented by Sentinel-based and ESA CCI products. Observational uncertainty will be characterised from data selection onwards and propagated into derived diagnostics where feasible, including through bootstrapping. The resulting documented reference resource will specify uncertainty, assumptions and domains of applicability and underpin the common WP4 benchmarking and validation protocols.

Table 1.3: Overview of the global ESM and regional configuration as CHORALE modelling frameworks.

Figure 1.2 makes the multi-model strategy of Box 1 explicit and verifiable. Every development has at least one reference implementation (green) and reaches at least two modelling frameworks, and every framework commits to at least two developments, except ICON-TSMP, whose committed contribution is the ParFlow-based groundwater representation (T2.3) that serves as the physically complete reference for the portable closures. No result therefore rests on a single model. Blue cells are transfer tests through the common WP4 testbed, and their outcome (directly transferable, transferable with adaptation or as concept or constraint, or model-specific; T4.4) is itself a project result. Split cells mark models that implement their own formulation in the task and also test the task’s main scheme through WP4, giving a like-for-like comparison of alternative representations within the same model. Dashed lines link WP2 and WP3 developments sharing a process interface (Section 1.4); their potential cross-cutting combination will be investigated in WP4 (T4.2) on implemented and, where possible, transferred developments, and may enter the Phase 2 configurations (MS4.4). Each implementation is first run as a standalone land-surface scheme offline simulation and analysed in WP6 before entering the coupled configurations (T5.2). The Matrix is a living, public record: initiated at MS4.1 (M9), updated at the configuration decisions (MS4.2, M21; MS4.4, M34) and released as D4.3 (M36) in the project repository, so that groups outside the consortium can see which development was tested in which model, with which outcome, and reuse it (Section 2.2).

Figure 1.2. CHORALE Development and Transferability Matrix. Rows: WP2 and WP3 developments; columns: global Earth system models and regional configurations (Table 1.3), with their land components. Green: implementation planned in the task (Section 3.1); blue: transfer attempted and evaluated through the WP4 testbed (T4.2–T4.4); split: implementation planned, with another scheme also tested through WP4; grey: not implemented. Dashed lines: envisaged WP4 cross-cutting developments (T4.2). The Matrix is maintained in T4.4, hosted in the project repository (T10.5) and delivered as D4.3.

Third pillar (WP2). The third pillar (and first model-development pillar, see Box 1) strengthens structural hydrology through explicit representation of shallow groundwater and unconfined-aquifer dynamics, capillary rise, water-table interactions with vegetation and rivers, and human-managed irrigation. Many land-surface schemes use a free-drainage lower boundary, equivalent to a water table at infinite depth (Zeng & Decker, 2009), removing subsurface storage and saturated-zone memory and suppressing capillary rise. The state of the art is integrated surface–subsurface modelling such as ParFlow (Jones & Woodward, 2001; Kollet & Maxwell, 2006; Maxwell, 2013), whose cost still confines it to regional domains (Fan et al., 2019). CHORALE will follow two complementary routes (T2.3): explicit ParFlow-based coupling of groundwater dynamics and lateral subsurface flow where computationally feasible, which also provides a physically complete reference, and portable simplified closures coupling the unsaturated soil column to an unconfined-aquifer storage, which replace the free-drainage assumption across the ESMs. Unlike earlier single-model water-table schemes (Niu et al., 2007; Miguez-Macho et al., 2007), these closures will be designed for transfer across ESMs and linked to deep-root groundwater uptake (T3.4) and irrigation abstraction (T2.4). This will improve representation of water-limited and transitional regions, where shallow groundwater and phreatophytic vegetation sustain evapotranspiration during dry periods (Fan et al., 2017; Miguez-Macho & Fan, 2021). River routing and inundation (Yamazaki et al., 2011) and irrigation with water-source constraints (Yao et al., 2022) complete this structural set. The representation of feedback, including those between humans and water for irrigation (Di Baldassarre et al., 2015), will improve model robustness over longer time scales. Transferability and computational suitability across ESM land-surface schemes will be evaluated in WP4.

Fourth pillar (WP2). The fourth pillar develops scale-aware and time-variable parameterisations for runoff generation, root-zone storage capacity, groundwater access and irrigation. Sub-grid variability in these processes, including runoff-onset thresholds and the water-table depth controlling capillary rise and plant access, will be related to observable landscape properties derived from high-resolution data on topography, soils, vegetation, groundwater and land management (Vreugdenhil et al., 2022; Blöschl & Sivapalan 1995; Skøien and Blöschl, 2006; Blöschl et al., 2013). Scaling relationships will preserve nonlinear thresholds and hydrological connectivity across model resolutions (Yang et al., 2024), including the distinction between irrigation activation and effective water application within heterogeneous grid cells (Modanesi et al., 2026), and embedding also human-water dynamics describing the interactions among water proximity, human settlements and irrigation (Palazzoli et al., 2024; Fang et al., 2018). Root-zone storage capacity, currently a static parameter, will become a prognostic variable updated from the vegetation's memory of past water deficits (van Oorschot et al., 2021; Gao et al., 2024). Parameterisations will be tested with a rainfall-runoff model in pilot catchments and multi-resolution simulations using common observation-based diagnostics and reproducible workflows to reduce biases in runoff, evapotranspiration and energy fluxes. At the interface with the fifth pillar, CHORALE will assess consistency between prognostic root-zone storage capacity, dynamic rooting and groundwater access to support compatible hydrology–vegetation representations across ESM architectures.

Fifth pillar (WP3). The fifth pillar targets the refinement of water-energy-carbon coupling and feedback in ESMs. Two separate developments in modelling plant and ecosystem processes have emerged in recent years. In combination, they will allow CHORALE to achieve large improvements in the realism and robustness of the participating land surface models. First, eco-evolutionary optimality (EEO) theory (Harrison et al. 2021) predicts the acclimation of ‘optimal’ plant strategies as environmental conditions change, with respect to fitness criteria that plants might in principle maximise to enhance their chances of survival. Second, demographic models that track plant age and/or size can capture the processes of ecosystem assembly that allow the distributions of plant strategies to emerge from competition (Fisher et al. 2018). EEO theory underlies the widely used ‘P model’ (Stocker et al. 2020), which replaces plant functional type parameterizations with a universal, extensively tested representation of gross primary production – essential for accurate modelling of evapotranspiration, due to the close coupling of water loss with CO₂ uptake through stomata (Tan et al. 2023; Gan et al. 2026). More recent EEO developments include universal prognostic models for annual maximum leaf area index (LAI: Cai et al. 2025) and its seasonal phenology (Zhou et al. 2025), the leaf economics spectrum (Wang et al. 2023), C₃-C₄ competition (Lavergne et al. 2026) and respiration by and carbon allocation to leaves, transport tissues and fine roots. EEO theory also underlies schemes to represent vegetation root-system adaptation to environmental changes, via the optimality-based representation of vegetation-accessible subsurface water volumes (van Oorschot et al 2021) – which can be extended to groundwater-accessible water where the water table and its capillary fringe lie within reach of deep roots. Similarly, EEO-based schemes represent plant responses to hydraulic stress (Mengoli et al 2025), which emerge from a combination of the potential for physiological acclimation and the limits of drought tolerance. Linked to EEO theory, the application of demography models will allow process-based upscaling of plant CO₂ and water exchanges and the implementation of hydraulic mortality thresholds (Robbins et al. 2025) to probe how hydrological extremes affect plant survival, competition and vegetation dynamics and feed back to the surface energy balance and the hydrological cycle.

CHORALE will implement major improvements in the participating land surface models, via EEO theory for plant physiological and carbon allocation processes, using two leading ecosystem demography models (LPJ-GUESS and CLM-FATES) that also include process-based representations of hydraulic mortality thresholds. The resulting suite of alternative theories for plant survival and adaptation will be assessed using curated data-rich testbeds designed to probe responses to ecohydrological stress and extremes. Specific developments will include investigation of the impact on the hydrological cycle of a diversity of plant hydraulic strategies, in both competitive and monodominant configurations, and validation/calibration of the simulation of modelled vegetation composition and function in water-limited systems.

Sixth pillar (WP5). The sixth pillar delivers coordinated land-only, coupled and targeted regional simulations across four ESMs. Alignment with relevant CMIP7 and IRRMIP protocols enables comparison with the wider assessment baseline, while paired global and convection-permitting simulations test the scale dependence of land–atmosphere coupling and provide data for AI-enabled downscaling. The coordinated design uses BASE (standard CMIP7 configuration), CHOR (including CHORALE developments), NOIR (CHOR without irrigation) and REGR (CHOR with policy-informed regreening). BASE-to-CHOR measures the aggregate effect of developments; CHOR-to-NOIR isolates the role of irrigation; and REGR quantifies regreening responses where irrigation is applied only when foreseen by the relevant initiative. Where possible, BASE simulations already performed within CMIP7 are reused. New global production comprises historical CHOR and NOIR runs and future CHOR, NOIR and REGR projections under two contrasting scenarios. Offline simulations are performed for each individual development and for the combined configuration including all developments, providing the basis for assessing their individual and integrated effects relative to the standard model configuration. Figure 1.3 summarises the CHORALE Coordinated Simulation Matrix. Two production phases use contrasting medium-to-high and low-emission scenarios (van Vuuren et al 2026). Scenario identities are fixed at project decision points using the best available AR7 and post-AR7 evidence and policy input; the costed number, structure and length of simulations remain unchanged. Regional time-slice experiments are driven by a global Phase 1 scenario run, selected based on the source model's performance over the historical period. Convection-permitting feedback experiments are nested within the selected regional simulations and can be initiated progressively as the required regional simulations become available. Two subdomain-cases will be simulated by at least two convection-permitting models, with the final cases selected where global and regional analyses identify sensitivities and feedback of greatest interest, subject to scientific relevance and computational feasibility.

CHORALE builds transparent "what-if" regreening scenarios from existing restoration plans (Table 1.4) and land-eligibility constraints. The project quantifies the climatic consequences, water requirements, and hydroclimatic feasibility of the targets defined by the relevant initiatives and planning frameworks. Each scenario represents full attainment of the target as declared at the time the protocol is implemented. A common protocol will translate official restoration targets into harmonised land-cover and vegetation perturbations using transparent eligibility rules and CMIP7/LUH3 conventions. Irrigation is included only where foreseen by the initiative; model-specific implementation is fixed in T5.1 and D5.1. Land-management scenarios will be implemented in coordinated global and regional simulations. EURO-CORDEX and MENA-CORDEX experiments at approximately 12 km provide the regional component of the coordinated simulations, with two selected cases subsequently represented at convection-permitting resolution by at least two modelling systems. The CP cases will be selected based on the sensitivities and feedback identified from the coordinated global and regional analyses, subject to scientific relevance and computational feasibility. Final parameters are fixed in T5.1 and documented in D5.1.

Table 1.4. Regreening domains, their governing initiative and the planning document that defines the target. The harmonised restoration protocol and region-specific scenario parameters will be documented in D5.1 (M18), while the final selection of the convection-permitting subdomain cases will be fixed later through the WP5 production protocol and the configuration decision at MS5.2 (M39).

Seventh pillar (WP6). The seventh pillar translates model improvements into post-AR7 evidence on the feasibility, sustainability, trade-offs and unintended consequences of regreening and irrigation. It combines project diagnostics with community evaluation tools, compares historical and future responses, assesses extremes and agricultural resilience, and relates cross-sectoral outcomes to SDGs 2, 6, 13 and 15. To extend high-resolution information beyond the explicitly simulated regions and periods, CHORALE will train and benchmark AI downscaling models on paired global and regional CP-simulations. Harmonised training, validation and test datasets, process-oriented metrics and independent-period cross-validation will identify transferable approaches. Methods, models and workflows will be documented and released for reproducibility and reuse.

Best-performing AI approaches will be applied to longer low-resolution simulations to generate extended high-resolution datasets at substantially lower computational cost than explicit convection-permitting production, while dynamic simulations remain the reference for evaluation. Although ambitious, the proposal is feasible because it builds on well-established modelling frameworks and integrates recent and project-driven advances in hydrology, ecohydrology, and land–atmosphere science into ESMs for climate-change, projection and assessment. Its modular and cross-cutting approach - combined with access to state-of-the-art data, existing model infrastructures, and a multidisciplinary team - ensures that each component can be developed, tested, and implemented within a realistic timeframe. The proposed methodology is fully compatible with Open science practices that will ensure reproducibility and uptake: model developments, data, diagnostics, and workflows will be managed under FAIR principles, with a data management plan, documented evaluation protocols, and broad sharing of outputs. The gender dimension, which becomes relevant where results inform irrigation, regreening and adaptation decisions, is addressed in Section 1.5.

Figure 1.3. CHORALE Coordinated Simulation Matrix. Rows: coordinated experiments; columns: four global ESMs and the Regional Models. Filled cells are integrations produced in CHORALE. Cells marked BC are taken from the CMIP7 archive and not re-run. Offline land-only simulations on common forcing (T5.2) are run with each global model and with the ICON-TSMP land surface and hydrology components (the only regional system whose land component is not already used in a global model). Each CHORALE development is integrated offline separately. Coupled experiments (≥3 members) follow a shared-reference design: BASE is the standard CMIP7 configuration; CHOR includes the CHORALE developments, irrigation included; NOIR switches irrigation off; REGR adds policy regreening, irrigated only where the initiative foresees it. Brackets link each CHOR reference to its single-element perturbations. Future experiments are run for both scenarios, produced in sequence. Regional nests (including land-only offline) are single-member 20-year slices within the EURO-CORDEX and MENA-CORDEX domains, driven by the corresponding Phase 1 global experiments. Selected convection-permitting cases (at <4km) are nested within these regional simulations and can be initiated as the required regional output becomes available.