A two-way coupling method and system of a multi-scale heterogeneous grid land surface hydrological model

Abstract

The application relates to a two-way coupling method and system of a multi-scale heterogeneous grid land surface hydrological model, which comprises the following steps: S1, multi-source data collection and preprocessing: collecting and preprocessing multi-source data required for coupling simulation; S2, model and coupler initialization: initializing a land surface process model, a hydrological model and an OASIS3-MCT coupler; embedding OASIS coupling interfaces in the two models respectively, and completing the definition of coupling grid information, task partition mode and a coupling variable list; S3, grid interpolation and time synchronization: realizing cross-model transmission of coupling variables between CLM5.0 and SHUD; S4, coupling simulation under a main time loop; S5, checking of simulation results, including energy balance checking and convergence checking; and S6, ending simulation and releasing memory resources. The application effectively solves the problems of model compatibility and data exchange under heterogeneous programming languages, different grid structures and parallel communication mechanisms, and improves the coupling calculation efficiency and simulation accuracy.

Description

Technical Field

[0001] This invention relates to the fields of Earth system science, hydrological simulation, and high-performance computing, and particularly to a two-way coupling method and system for land surface hydrological models using multi-scale heterogeneous grids. Background Technology

[0002] Given that ecosystems and cryosphere elements directly influence the energy, water, and carbon cycles of the land-atmosphere system, current numerical models urgently need to move from single-process simulations to multi-factor collaborative characterization, and further develop an integrated modeling system that tightly couples land surface, hydrology, and ecology. Land surface models (LSMs) focus on energy balance, while hydrological models focus on water budget; both are key tools for understanding the evolution of water resources and the water cycle response under the influence of climate change and human activities. Promoting their synergistic coupling is of great significance for improving the physical consistency and predictive power of watershed-scale hydrological-energy process simulations.

[0003] SHUD (Simulator for Hydrologic Unstructured Domains) is a tightly coupled surface-subsurface distributed numerical hydrological model characterized by multi-scale, multi-process, high temporal resolution, and flexible spatial resolution. The model employs unstructured triangular grids to adapt to complex terrain and watershed structures, enabling detailed characterization of key hydrological processes such as surface runoff, vertical soil water flow, river flow, and groundwater dynamics. Vegetation interception processes before precipitation reaches the surface and snow accumulation processes are described using a water balance bucket model and a snow cover diurnal model, respectively. Evapotranspiration of surface, soil, and groundwater is based on potential evapotranspiration calculated using the Penman-Montes formula and constrained by the soil moisture stress coefficient, thus emphasizing the physical representation of hydrological processes at fine spatial scales.

[0004] Community Land Models (CLMs) primarily focus on the exchange of energy, water, and carbon and nitrogen fluxes between land and the atmosphere, describing geophysical and biogeochemical processes at the Earth's surface through energy and mass conservation equations. Compared to hydrological models, land surface models typically provide more complete and detailed depictions of vegetation physiological processes, snow cover evolution, and soil freeze-thaw processes, and often employ regular rectangular grid structures to serve regional and global climate simulations. Although the hydrological module in land surface models has evolved from early bucket models to more complex land surface water and energy budget models, its description of surface and soil water movement remains primarily vertical, with relatively simplified depictions of lateral flow and groundwater dynamics. This poses certain limitations in applications involving complex terrain and watershed scales.

[0005] The OASIS3 (Ocean Atmosphere Sea Ice Soil) coupler provides a general framework for synchronous information exchange between different Earth system components. Its latest generation product, OASIS-MCT, significantly improves parallel computing performance by introducing the MCT (Model Coupling Toolkit), supports direct reading of offline-generated interpolation weights, and enhances compatibility with unstructured grids. OASIS-MCT no longer relies on a centralized coupling control process; instead, it achieves point-to-point data exchange between model components through the MCT communication library. It can also be used in conjunction with interpolation weight generation frameworks such as SCRIP, ESMF, and XIOS, providing efficient and flexible technical support for high-resolution hydro-land surface coupled simulations.

[0006] Traditional land surface-hydrological model coupling methods generally face the following key limitations:

[0007] (1) Grid mismatch and data exchange challenges: Land surface process models and distributed hydrological models typically employ drastically different numerical frameworks and grid systems, making efficient and conserved data exchange difficult. The two types of models differ significantly in time step and spatial resolution, lacking a unified spatiotemporal synchronization and scheduling mechanism, which makes it difficult to precisely control the timing of coupled variable transmission. Furthermore, cross-grid interpolation processes often violate water conservation, and the lack of systematic conservation checks and error constraint mechanisms leads to the continuous accumulation of errors in long-term simulations, affecting the reliability of the results.

[0008] (2) Primarily unidirectional drive with a lack of feedback mechanism: Existing coupling schemes mostly adopt loose coupling or unidirectional drive methods, usually only using the energy or meteorological flux of the land surface model as an external forcing of the hydrological model, while ignoring the reverse modulation effect of hydrological processes on land surface energy-water exchange. In particular, the spatial redistribution process of soil moisture (the advantage of SHUD) is difficult to feed back to the land surface model to dynamically constrain key processes such as evapotranspiration (such as the core module of CLM5.0), thereby weakening the physical consistency and simulation accuracy of the coupled system, and also limiting the effective use of fine hydrological information.

[0009] (3) Limited definition of coupling region: Most existing methods are difficult to flexibly set the coupling range. They can only perform overall coupling on fixed regular grids or large regional scales, making it difficult to achieve accurate coupling simulation at the watershed scale or local area. This results in non-target areas being forced to participate in the calculation, leading to low efficiency in the utilization of computing resources.

[0010] (4) Insufficient computational efficiency and scalability: The communication, interpolation and data recombination processes between model components have large computational overhead and limited scalability in a parallel environment, making it difficult to support the simulation needs of long-term series and high spatiotemporal resolution regional or watershed scales.

[0011] In summary, there is an urgent need to develop an advanced coupled simulation method that is compatible with heterogeneous grids, supports efficient bidirectional feedback mechanisms, and strictly ensures the conservation of physical quantities during data exchange, in order to meet the requirements of refined integrated simulation of hydrological-land surface processes. Summary of the Invention

[0012] The technical problem to be solved by the present invention is to provide a two-way coupling method for land surface hydrological models with multi-scale heterogeneous grids.

[0013] Another technical problem to be solved by the present invention is to provide a two-way coupling system for land surface hydrological models with multi-scale heterogeneous grids.

[0014] To address the aforementioned problems, the present invention provides a two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids, characterized by the following steps:

[0015] S1 Multi-Source Data Collection and Preprocessing:

[0016] Collect and preprocess the multi-source data required for the coupled simulation, including meteorological forcing data, surface parameters, spatial raster files, and grid files;

[0017] S2 mode and coupler initialization:

[0018] Initialize the land surface process model and the hydrological model; the land surface process model is the CLM5.0 model, implemented in the Fortran programming language, and uses a regular rectangular mesh; the hydrological model is the SHUD model, implemented in the C++ programming language, and uses an unstructured triangular mesh.

[0019] During the initialization phase of CLM5.0 and SHUD, input data reading, mesh generation, and initial state settings are completed synchronously, and the OASIS3-MCT coupler is initialized. Subsequently, the OASIS coupling interface is embedded in both modes to define the coupling mesh information, task partitioning method, and coupling variable list in sequence.

[0020] S3 grid interpolation and time synchronization:

[0021] Based on the OASIS-MCT configuration parameters in the namcouple file, determine the transmission frequency, interpolation method, and time synchronization interval of the coupled variables; and realize cross-mode transmission of coupled variables between CLM5.0 and SHUD through grids.nc, masks.nc, areas.nc, and rmp_*.nc files.

[0022] Coupled simulation under S4 master time loop:

[0023] In the main time loop, the simulation calculations of the land surface process model and the hydrological model are performed in parallel;

[0024] S5 performs checks on the simulation results, including energy balance checks and convergence checks:

[0025] Before each coupling time synchronization or the end of the main loop, an energy balance check is performed on the land surface process model (CLM5.0) side and a convergence check (SHUD) is performed on the hydrological model side.

[0026] S6 terminates the simulation and releases memory resources:

[0027] After the simulation is completed, the coupled simulation results are output and verified, the time series of coupled variables and the coupling diagnosis report are generated, and memory resources are released at the same time.

[0028] The meteorological forcing data in S1 comes from CMFD (China Meteorological Forcing Data), including precipitation rate, atmospheric temperature, atmospheric pressure, wind speed, specific humidity, downward shortwave radiation, and downward longwave radiation, with a spatial resolution of 0.1° and a temporal resolution of 3 hours; surface parameters include vegetation type and soil properties; spatial raster files include digital elevation model (DEM), watershed boundary, and river network data.

[0029] The definition of coupling mesh information, task partitioning method, and coupling variable list in step S2 shall be performed as follows:

[0030] In the coupled mesh definition, CLM5.0 provides the center point coordinates and four corner point coordinates of the regular rectangular mesh, all of which are one-dimensional variables; SHUD provides the center point coordinates and three corner point coordinates of the unstructured triangular mesh, all of which are one-dimensional variables. In SHUD, the local coordinate system is converted to the EPSG:4326 coordinate system with the help of the Proj library. When the coupled mode is run for the first time, the system automatically identifies and matches the mesh topology based on the mesh center point and corner point coordinates, and generates mesh description files grids.nc, masks.nc, areas.nc and interpolation weight files rmp_*.nc.

[0031] In the definition of task partitioning, CLM5.0 supports Orange partitioning or Point partitioning and allocates the regular grid to each core according to the number of computing cores; SHUD adopts Serial partitioning, with a single computing core completing the calculation of the unstructured triangular grid of the entire watershed.

[0032] When defining coupling variables, the coupling variables output by CLM5.0 to SHUD include: liquid precipitation rate, snowfall rate, snowmelt rate, canopy interception evaporation rate, vegetation transpiration rate, and soil evaporation rate, in mm / day; the coupling variable output by SHUD to CLM5.0 is the vadose zone thickness, in meters.

[0033] The cross-mode transfer of coupling variables between CLM5.0 and SHUD in step S3 is implemented as follows: according to the definition in masks.nc, interpolation is performed only on grids within the watershed. The transfer from regular rectangular grids to unstructured triangular grids uses the SCRIP bilinear interpolation method, and the transfer from unstructured triangular grids to regular rectangular grids uses the SCRIP nearest neighbor distance weighted interpolation method.

[0034] The simulation calculation of the land surface process model in step S4 refers to the simulation calculation of surface processes and subsurface processes by the land surface process model (CLM5.0), including soil evaporation, snowmelt, vegetation biogeochemical processes, soil freeze-thaw, and vertical discretization of soil moisture; the simulation calculation of the hydrological model refers to the simulation calculation of surface hydrological processes and groundwater processes by the hydrological model (SHUD), including river flow, surface runoff, unsaturated and saturated layer flow, and groundwater level changes. Among them, the calculation of snow accumulation process, canopy interception process, and evapotranspiration surface flux is undertaken by the land surface process model.

[0035] In step S5, the coupling time step for time synchronization is 1800 seconds, and the driving data time interval is 3 hours.

[0036] A system using the method described above, characterized in that: the system comprises:

[0037] The data collection module is used to collect data from multiple sources;

[0038] The initialization module is used to initialize the CLM5.0 land surface process model, the SHUD hydrological model, and the OASIS3-MCT coupler;

[0039] The main time loop module is used for parallel execution mode simulation calculations and performs bidirectional data exchange through the OASIS3-MCT coupler.

[0040] The mesh interpolation module is used to perform conserved interpolation between regular rectangular meshes and unstructured triangular meshes;

[0041] The time synchronization module is used to control the transmission frequency and time synchronization interval of the coupled variables;

[0042] The inspection module is used to perform energy balance checks and convergence checks;

[0043] The termination module is used to end the simulation and release memory resources.

[0044] The initialization module includes:

[0045] The land surface process model initialization submodule, implemented using a Fortran interface, is responsible for reading meteorological forcing, surface parameters, and initial conditions, and performing regular rectangular mesh generation.

[0046] The hydrological model initialization submodule, implemented using a C++ interface, is responsible for reading watershed data and hydrological calibration parameters, and performing unstructured triangular mesh generation.

[0047] The OASIS3-MCT initialization submodule provides a Fortran / C dual-language interface with the same functionality, responsible for defining task partitions, coupled meshes, and coupled variables.

[0048] The grid interpolation module includes:

[0049] Mesh definition and matching units support automatic topology recognition and matching of regular rectangular meshes and unstructured meshes;

[0050] Intelligent interpolation algorithm unit, realizing bidirectional heterogeneous mesh interpolation and mass conservation guarantee mechanism;

[0051] The coupled region control unit defines coupled regions and manages dynamic mask generation based on geographical boundaries.

[0052] The main time loop module includes:

[0053] The land surface process simulation submodule performs calculations of land surface physical processes, including soil evaporation, snow melting, vegetation biogeochemical processes, soil freeze-thaw cycles, and soil moisture movement.

[0054] The hydrological process simulation submodule performs calculations of watershed hydrological processes, including river flow, surface runoff, unsaturated layer flow, saturated layer flow, and groundwater level changes.

[0055] The coupling interface submodule is responsible for data synchronization and exchange between the two modes.

[0056] Compared with the prior art, the present invention has the following advantages:

[0057] 1. This invention achieves efficient bidirectional coupling variable exchange between heterogeneous (different programming languages, different grid systems) modes (CLM5.0 and SHUD) by adopting OASIS3-MCT coupling technology.

[0058] 2. This invention relies on the parallel communication mechanism of OASIS3-MCT to optimize the cross-grid data transmission and conservation interpolation process; through dynamic region definition and mask generation based on geographical boundaries, it realizes adaptive optimization of computing resources, which greatly improves the computational efficiency of large-scale simulations.

[0059] 3. This invention establishes a multi-timescale coordination mechanism, which supports automatic optimization of coupling parameters (such as LAG) and precise control of data transmission timing, thereby enhancing the numerical stability and physical consistency of coupled simulation, and further improving the control accuracy of the coupled process.

[0060] 4. By integrating multi-source geographic and meteorological data inputs, this invention enables the model to accurately characterize complex underlying surfaces and variable meteorological conditions, significantly broadening its application scenarios and reliability in watershed-level refined water-energy process simulation. Attached Figure Description

[0061] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0062] Figure 1 This is a schematic diagram of the overall process architecture of the present invention.

[0063] Figure 2 This is a spatial interpolation diagram illustrating the transfer of coupling variables from land surface to hydrology and from hydrology to land surface in an embodiment of the present invention.

[0064] Figure 3 This is a time-exchange sequence diagram of the land surface-hydrological coupling system in an embodiment of the present invention.

[0065] Figure 4 The diagram shows the time series of runoff simulations for the standalone land surface model, the standalone hydrological model, and the land surface-hydrological coupled model in the embodiments of the present invention, including a comparison of runoff at the daily (a) and monthly (b) scales. Detailed Implementation

[0066] This invention, based on the OASIS3-MCT coupler, introduces a unified coupling interface into heterogeneous numerical models, enabling bidirectional coupled operation between a land surface process model (CLM5.0) based on a regular rectangular grid and a distributed hydrological model (SHUD) based on an unstructured triangular grid. In the main time loop, time synchronization and cross-grid interpolation of coupled variables are completed according to the coupling configuration file, achieving the conservation and transfer of key variables such as evapotranspiration components, groundwater level, and meteorological forcing between different models.

[0067] like Figure 1 As shown, a two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids includes the following steps:

[0068] S1 Multi-Source Data Collection and Preprocessing:

[0069] Collect and preprocess the multi-source data required for the coupled simulation, including meteorological forcing data, surface parameters, spatial raster files, and grid files.

[0070] The meteorological forcing data comes from CMFD (China Meteorological Forcing Data), including precipitation rate, atmospheric temperature, atmospheric pressure, wind speed, specific humidity, downward shortwave radiation, and downward longwave radiation, with a spatial resolution of 0.1° and a temporal resolution of 3 hours. The surface parameters include vegetation type and soil properties. The spatial raster files include digital elevation model (DEM), watershed boundaries, and river network data.

[0071] S2 mode and coupler initialization:

[0072] Initialize the land surface process model and the hydrological model; the land surface process model is the CLM5.0 model, implemented in the Fortran programming language and using a regular rectangular mesh; the hydrological model is the SHUD model, implemented in the C++ programming language and using an unstructured triangular mesh.

[0073] During the initialization phase of CLM5.0 and SHUD, input data reading, mesh generation, and initial state settings are completed synchronously, and the OASIS3-MCT coupler is initialized. Subsequently, the OASIS coupling interface is embedded in both modes to define the coupling mesh information, task partitioning method, and coupling variable list in sequence.

[0074] The definition of coupled mesh information, task partitioning method, and coupled variable list shall be performed as follows:

[0075] In the coupled mesh definition, CLM5.0 provides the center point coordinates and four corner point coordinates of the regular rectangular mesh, all of which are one-dimensional variables; SHUD provides the center point coordinates and three corner point coordinates of the unstructured triangular mesh, all of which are one-dimensional variables. In SHUD, the local coordinate system is converted to the EPSG:4326 coordinate system with the help of the Proj library. When the coupled mode is run for the first time, the system automatically identifies and matches the mesh topology based on the mesh center point and corner point coordinates, and generates mesh description files grids.nc, masks.nc, areas.nc and interpolation weight files rmp_*.nc.

[0076] In the definition of task partitioning, CLM5.0 supports Orange partitioning or Point partitioning and allocates the regular grid to each core according to the number of computing cores; SHUD adopts Serial partitioning, with a single computing core completing the calculation of the unstructured triangular grid of the entire watershed.

[0077] When defining coupling variables, the coupling variables output by CLM5.0 to SHUD include: liquid precipitation rate, snowfall rate, snowmelt rate, canopy interception evaporation rate, vegetation transpiration rate, and soil evaporation rate, in mm / day; the coupling variable output by SHUD to CLM5.0 is the vadose zone thickness (aquifer thickness - groundwater level), in meters.

[0078] S3 grid interpolation and time synchronization:

[0079] Based on the OASIS-MCT configuration parameters in the namcouple file, determine the transmission frequency, interpolation method, and time synchronization interval of the coupled variables; and realize cross-mode transmission of coupled variables between CLM5.0 and SHUD through grids.nc, masks.nc, areas.nc, and rmp_*.nc files.

[0080] Cross-mode transfer of coupling variables between CLM5.0 and SHUD is implemented as follows: According to the definition in masks.nc, interpolation operations are performed only on grids within the watershed, avoiding the participation of regions outside the watershed in the coupling calculation. The transfer from regular rectangular grids to unstructured triangular grids uses the SCRIP bilinear interpolation method, while the transfer from unstructured triangular grids to regular rectangular grids uses the SCRIP nearest neighbor distance weighted interpolation method.

[0081] like Figure 2 As shown, CLM5.0 uses a coarse, regular rectangular grid, which cannot finely characterize the spatial heterogeneity of watershed boundaries and interiors. SHUD, on the other hand, uses a fine, unstructured triangular grid. This grid closely fits the irregular boundaries of the watershed and is denser at river networks, reflecting the fine control of topography such as river channels and ridges over hydrological processes. The mapping from low resolution to high resolution uses bilinear interpolation, a method suitable for continuously changing variables (such as precipitation). Using the values ​​of the four nearest points in the source grid (CLM5.0), the centroid value of each triangle in the target grid (SHUD) is calculated based on distance weights. The mapping from high resolution (SHUD) to low resolution (CLM5.0) uses nearest-neighbor interpolation distance weights, averaging or weighting the states of multiple small triangular cells and assigning them to a large rectangular grid.

[0082] Coupled simulation under S4 master time loop:

[0083] Within the main time loop, simulation calculations for the land surface process model and the hydrological model are performed in parallel. Specifically:

[0084] The simulation calculation of the land surface process model refers to the simulation calculation of surface processes and subsurface processes by the land surface process model (CLM5.0), including soil evaporation, snowmelt, vegetation biogeochemical processes, soil freeze-thaw, and vertical discretization of soil moisture. The simulation calculation of the hydrological model refers to the simulation calculation of surface hydrological processes and groundwater processes by the hydrological model (SHUD), including river flow, surface runoff, unsaturated and saturated layer flow, and groundwater level changes. Among these, the calculation of snow accumulation processes, canopy interception processes, and evapotranspiration surface fluxes is undertaken by the land surface process model.

[0085] like Figure 3 The diagram illustrates the time integration and data interaction mechanism of the coupled system. In this scheme, atmospheric forcing data serves only as a unidirectional input to CLM5.0, with an update frequency of 3 hours. To achieve model synchronization, the coupling time step is set to 30 minutes, meaning the simulation step size of CLM5.0 is consistent with the maximum simulation time step size of SHUD. At each coupling moment, CLM5.0 runs first and transmits the calculated surface fluxes to SHUD; after SHUD completes the simulation of the current step using the surface fluxes from CLM5.0, it feeds back the updated hydrological state to CLM5.0 as the initial state for its next land surface process simulation, thus achieving close bidirectional coupling.

[0086] S5 performs checks on the simulation results, including energy balance checks and convergence checks:

[0087] Before each coupling time synchronization or the end of the main loop, an energy balance check is performed on the land surface process model (CLM5.0) side to ensure the conservation of system energy budget; a convergence check (SHUD) is performed on the hydrological model side to ensure the convergence and stability of the numerical calculation process, thereby ensuring the stability and accuracy of the entire simulation.

[0088] Among them, the coupling time step for time synchronization is 1800 seconds, and the time interval for driving data is 3 hours.

[0089] S6 terminates the simulation and releases memory resources:

[0090] After the simulation is completed, the coupled simulation results are output and verified, the time series of coupled variables and the coupling diagnosis report are generated, and memory resources are released at the same time.

[0091] A system using the above method, the system comprising:

[0092] The data collection module is used to collect data from multiple sources;

[0093] The initialization module is used to initialize the CLM5.0 land surface process model, the SHUD hydrological model, and the OASIS3-MCT coupler;

[0094] The main time loop module is used for parallel execution mode simulation calculations and bidirectional data exchange via the OASIS3-MCT coupler;

[0095] The mesh interpolation module is used to perform conserved interpolation between regular rectangular meshes and unstructured triangular meshes;

[0096] The time synchronization module is used to control the transmission frequency and time synchronization interval of the coupled variables;

[0097] The inspection module is used to perform energy balance checks and convergence checks;

[0098] The termination module is used to end the simulation and release memory resources.

[0099] The initialization module includes:

[0100] The land surface process model initialization submodule, implemented using a Fortran interface, is responsible for reading meteorological forcing, surface parameters, and initial conditions, and performing regular rectangular mesh generation.

[0101] The hydrological model initialization submodule, implemented using a C++ interface, is responsible for reading watershed data and hydrological calibration parameters, and performing unstructured triangular mesh generation.

[0102] The OASIS3-MCT initialization submodule provides a Fortran / C dual-language interface with the same functionality, responsible for defining task partitions, coupled meshes, and coupled variables.

[0103] The grid interpolation module includes:

[0104] Mesh definition and matching units support automatic topology recognition and matching of regular rectangular meshes and unstructured meshes;

[0105] Intelligent interpolation algorithm unit, realizing bidirectional heterogeneous mesh interpolation and mass conservation guarantee mechanism;

[0106] The coupled region control unit defines coupled regions and manages dynamic mask generation based on geographical boundaries.

[0107] The main time loop module includes:

[0108] The land surface process simulation submodule performs calculations of land surface physical processes, including soil evaporation, snow melting, vegetation biogeochemical processes, soil freeze-thaw cycles, and soil moisture movement.

[0109] The hydrological process simulation submodule performs calculations of watershed hydrological processes, including river flow, surface runoff, unsaturated layer flow, saturated layer flow, and groundwater level changes.

[0110] The coupling interface submodule is responsible for data synchronization and exchange between the two modes.

[0111] The following example, using a full-year simulation of the Heihe (Yingluoxia) watershed in 1990, illustrates the specific implementation process and parameter settings of the land surface-hydrological coupling simulation method described in this invention. This example aims to verify the feasibility of the coupling mechanism described in this invention using known historical hydrological data; the selection of the year for the input data does not affect the applicability of the method in other time periods. This example is only used to illustrate the technical solution of this invention and does not constitute a limitation on the scope of protection of this invention.

[0112] In this embodiment, the simulation range of the land surface process model CLM5.0 is 37-40°N and 98-102°E, with a spatial resolution of 0.1°, and contains 1200 regular rectangular grids; the simulation range of the hydrological model SHUD is limited to the Heihe (Yingluoxia) basin, with 181 corresponding 0.1° grids, and is discretized using unstructured triangular grids, containing 1831 triangular grid cells and 974 river segments.

[0113] 1. Data Collection Module

[0114] The data collection module is used to acquire and manage the multi-source input data required for coupled simulations, specifically including:

[0115] Meteorological forcing data: The China Meteorological Forcing Data Set (CMFD) was used, including precipitation rate (mm / s), atmospheric temperature (°C), specific humidity (kg / kg), wind speed (m / s), atmospheric pressure (Pa), downward shortwave radiation, and downward longwave radiation (W / m²). 2 The spatial resolution of the data is 0.1°, and the temporal resolution is 3 hours.

[0116] Spatial data: including digital elevation model (DEM), river network data and watershed boundary data, used to construct the model computation grid and hydrological topology.

[0117] Surface parameter data: CLM5.0 uses the mksurfdata_map tool to generate surface parameter data, including underlying surface type, vegetation attributes, and soil physical parameters; SHUD uses the Global Hydrologic DataCloud for modeling. After inputting the Heihe (Yingluoxia) watershed boundary, it automatically completes the extraction of meteorological driven data (not used in coupled mode), generation of unstructured triangular meshes, and construction of the river segment topology. The aquifer thickness is set to 8.6 m, corresponding to the bottom boundary of the 20th layer in the CLM5.0 soil layer.

[0118] 2. Initialization Module

[0119] The initialization module includes the CLM5.0 initialization submodule, the SHUD initialization submodule, and the OASIS3-MCT coupler initialization submodule:

[0120] The CLM5.0 initialization submodule, implemented using a Fortran interface, is responsible for reading meteorological forcing data, surface parameters, and initial state variables, and for generating regular rectangular grids. Simultaneously, it embeds the OASIS coupling interface within CLM5.0 to define coupled partitions (oasis_init_comp, oasis_get_localcomm, oasis_def_partition), register coupled variables (oasis_def_var, oasis_enddef), and write coupled grid information (oasis_start_grids_writing, oasis_write_grid, oasis_write_corner, oasis_write_mask, oasis_write_area, oasis_terminate_grids_writing).

[0121] SHUD Initialization Submodule: This submodule is implemented based on a C++ interface and is responsible for reading watershed spatial data, soil parameters, hydrological initial conditions, and hydrological calibration parameters. It initializes the variables on the unstructured triangular mesh and river network topology, and embeds the OASIS coupling interface in SHUD to define the coupling partition (oasis_c_init_comp, oasis_c_def_partition), coupling variables (oasis_c_def_var, oasis_c_enddef), and coupling meshes (oasis_c_start_grids_writing, oasis_c_write_grid, oasis_c_write_corner, oasis_c_write_mask, oasis_c_write_area, oasis_c_terminate_grids_writing).

[0122] The OASIS3-MCT initialization submodule provides a Fortran / C dual-language interface for uniformly defining coupled task partitions, coupled mesh description information, and coupled variables that need to be exchanged between different modes. Regarding coupled task partition definition, OASIS3-MCT supports multiple partitioning methods, including Serial partitioning, Apple partitioning, Box partitioning, Orange partitioning, and Points partitioning. Each partitioning method allows for flexible partitioning of the coupled mesh in a parallel computing environment by setting the global starting offset and local length of the partition. In this embodiment, since SHUD needs to combine watershed mask mesh constraints when transferring coupled variables to CLM5.0, Orange partitioning and Points partitioning methods offer higher flexibility and controllability in spatial subdomain control and mask processing. Therefore, Orange partitioning or Points partitioning is selected for the coupled task partitioning on the CLM5.0 side to facilitate precise control of the regular rectangular meshes participating in coupled computation. On the other hand, the SHUD model currently uses a serial computing architecture and does not yet support parallel task partitioning. Therefore, OASIS3-MCT adopts a serial partitioning approach, where a single computing core completes the hydrological simulation and coupled variable output for all unstructured triangular meshes within the watershed. Through this partitioning configuration, effective collaborative coupling between the parallel land surface model and the serial hydrological model is achieved.

[0123] When the coupled system is run for the first time, a coupled mesh description file and an interpolation weight file are automatically generated for variable transfer in the subsequent main time loop.

[0124] 3. Main Time Loop Module

[0125] The main time loop module is the core of the system's operation flow, and performs the following operations sequentially within each coupled time step:

[0126] The CLM5.0 simulation submodule performs calculations of land surface physical and biogeochemical processes at each coupled time step, including soil evaporation, snowmelt, vegetation physiological processes, soil freeze-thaw cycles, and vertical soil moisture movement. After completing the land surface process calculations for the current time step, it calls `oasis_put` through the OASIS coupling interface to output relevant liquid precipitation rates, snowfall rates, snowmelt rates, canopy interception evaporation rates, vegetation transpiration rates, and soil evaporation rates to SHUD. Simultaneously, after SHUD completes the hydrological calculations for the corresponding time step and returns the results, CLM5.0 calls `oasis_get` through the OASIS interface to receive key state variables such as the vadose zone thickness (aquifer thickness - groundwater level) returned by SHUD. These received state variables are used to update the initial soil moisture conditions in CLM5.0 and serve as important inputs for the land surface process calculations in the next time step, thus achieving dynamic constraints of hydrological processes on land surface energy and moisture exchange processes.

[0127] The SHUD simulation submodule: At the beginning of each coupled time step, SHUD receives coupling variable information from CLM5.0 via the OASIS coupling interface by calling `oasis_c_get`, and uses this information as the external driving condition for the watershed hydrological process simulation. Based on this, SHUD performs watershed-scale hydrological process calculations, including key processes such as river flow, surface runoff, unsaturated and saturated layer flow, and groundwater level changes. After completing the hydrological calculations for the current time step, SHUD calls `oasis_c_put` via the OASIS coupling interface to feed back the updated groundwater level and other key hydrological state variables to CLM5.0 for subsequent land surface process simulations. Through this interactive process, continuous feedback and dynamic updating of hydrological state variables to land surface processes are achieved.

[0128] The coupling interface submodule, through OASIS3-MCT, enables interpolation transformation and time synchronization control of coupled variables between heterogeneous grids, ensuring the accuracy, stability, and physical consistency of data transmission between the regular rectangular grid land surface model and the unstructured triangular grid hydrological model. For spatial interpolation, OASIS3-MCT can utilize interpolation libraries such as SCRIP, ESMF, and XIOS, supporting interpolation transformation between regular rectangular grids, Gaussian reduced grids, and unstructured grids. Interpolation methods include nearest neighbor weighted interpolation, bilinear interpolation, bicubic interpolation, and conserved remapping. In this embodiment, differentiated interpolation strategies are adopted based on the variable characteristics and grid structure differences of different coupling directions. For the transfer of coupled variables from the regular rectangular grid to the unstructured triangular grid, a bilinear interpolation method based on SCRIP is used to ensure spatial continuity and variable smoothness. For the transfer of coupled variables from unstructured triangular meshes to regular rectangular meshes, a nearest neighbor distance weighted interpolation method based on SCRIP is adopted to avoid introducing non-physical smoothing errors during the mapping process from sparse or irregular meshes to regular meshes. Regarding time synchronization, the original time resolution of the meteorological forcing data is 3 hours. During the internal calculation of CLM5.0, it is automatically downscaled to a 30-minute time step for land surface process simulation. To maintain the consistency of the coupled system in the time dimension, the SHUD model synchronously uses a 30-minute time step to perform hydrological process calculations and completes the exchange and updating of coupled variables at this time scale, thereby avoiding numerical instability and physical deviations caused by inconsistent time steps.

[0129] 4. Inspection Module

[0130] This module is used to verify the physical plausibility and computational stability of the simulation process.

[0131] Energy balance check (CLM side): Ensures that the system's energy balance is maintained.

[0132] Convergence check (SHUD side): Ensures that the numerical calculation process converges and is stable.

[0133] 5. End module

[0134] When the simulation reaches the preset time or termination condition, the termination module is responsible for outputting the simulation results, generating coupled diagnostic information, releasing system memory resources, and safely terminating the operation of CLM5.0-OASIS-SHUD (oasis_terminate, oasis_c_terminate).

[0135] Through the coordinated operation of the above modules, this embodiment successfully achieved efficient bidirectional coupling between the regular rectangular grid land surface model and the unstructured triangular grid hydrological model, significantly improving the physical consistency, computational stability and simulation accuracy of the simulation of hydrological-land surface processes in complex watersheds.

[0136] Table 1 shows the performance of different simulation schemes in multi-scale runoff simulation in the Heihe (Yingluoxia) basin. On a daily scale ( Figure 4 a) The standalone SHUD model performed best (Nash efficiency coefficient reached 0.791), while the standalone CLM5.0 model, lacking a refined confluence mechanism, had a negative Nash efficiency coefficient (-0.337). The coupled model (CLM5.0-SHUD) significantly corrected this deficiency, increasing the Nash efficiency coefficient to 0.521, demonstrating the effectiveness of incorporating hydrological processes in improving high-frequency runoff simulations in land surface models. On a monthly scale ( Figure 4 (b) CLM5.0 alone performed best in vertical water balance calculation (Nash efficiency coefficient of 0.799), but the accuracy of the coupled model actually decreased (Nash efficiency coefficient of 0.407). This indicates that although the coupled system successfully captured diurnal hydrological dynamics, there are still systematic biases in long-term water distribution or parameter adaptation, resulting in a weakening of monthly-scale simulation capabilities.

[0137] Table 1. Runoff simulation performance of standalone and coupled models in the Heihe (Yingluoxia) watershed.

[0138]

[0139] Through the coordinated operation of the above modules, this embodiment successfully achieved efficient bidirectional coupling between the regular rectangular grid land surface model and the unstructured triangular grid hydrological model, significantly improving the physical consistency, computational stability and simulation accuracy of the simulation of hydrological-land surface processes in complex watersheds.

Claims

1. A two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids, characterized in that: Includes the following steps: S1 Multi-Source Data Collection and Preprocessing: Collect and preprocess the multi-source data required for the coupled simulation, including meteorological forcing data, surface parameters, spatial raster files, and grid files; S2 mode and coupler initialization: Initialize the land surface process model and the hydrological model; the land surface process model is the CLM5.0 model, implemented in the Fortran programming language, and uses a regular rectangular mesh; the hydrological model is the SHUD model, implemented in the C++ programming language, and uses an unstructured triangular mesh. During the initialization phase of CLM5.0 and SHUD, input data reading, mesh generation, and initial state settings are completed synchronously, and the OASIS3-MCT coupler is initialized. Subsequently, the OASIS coupling interface is embedded in both modes to sequentially define the coupled mesh information, task partitioning method, and coupled variable list; the definition of the coupled mesh information, task partitioning method, and coupled variable list is performed as follows: In the coupled mesh definition, CLM5.0 provides the center point coordinates and four corner point coordinates of the regular rectangular mesh, all of which are one-dimensional variables; SHUD provides the center point coordinates and three corner point coordinates of the unstructured triangular mesh, all of which are one-dimensional variables, and uses the Proj library to convert the local coordinate system into a coordinate system; when the coupled mode is run for the first time, the system automatically identifies and matches the mesh topology based on the mesh center point and corner point coordinates, and generates a mesh description file and an interpolation weight file; In the definition of task partitioning, CLM5.0 supports Orange partitioning or Point partitioning and allocates the regular grid to each core according to the number of computing cores; SHUD adopts Serial partitioning, with a single computing core completing the calculation of the unstructured triangular grid of the entire watershed. When defining coupling variables, the coupling variables output by CLM5.0 to SHUD include: liquid precipitation rate, snowfall rate, snowmelt rate, canopy interception evaporation rate, vegetation transpiration rate, and soil evaporation rate, in mm / day; the coupling variable output by SHUD to CLM5.0 is the vadose zone thickness, in meters. S3 grid interpolation and time synchronization: Based on the configuration parameters, determine the transmission frequency, interpolation method, and time synchronization interval of the coupled variables; Coupled simulation under S4 master time loop: In the main time loop, the simulation calculations of the land surface process model and the hydrological model are performed in parallel; S5 performs checks on the simulation results, including energy balance checks and convergence checks: Before each coupling time synchronization or the end of the main loop, an energy balance check is performed on the land surface process model side and a convergence check is performed on the hydrological model side. S6 terminates the simulation and releases memory resources: After the simulation is completed, the coupled simulation results are output and verified, the time series of coupled variables and the coupling diagnosis report are generated, and memory resources are released at the same time.

2. The two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids as described in claim 1, characterized in that: The meteorological forcing data in S1 comes from CMFD and includes precipitation rate, atmospheric temperature, atmospheric pressure, wind speed, specific humidity, downward shortwave radiation and downward longwave radiation, with a spatial resolution of 0.1° and a temporal resolution of 3 hours; surface parameters include vegetation type and soil properties; spatial raster files include digital elevation model (DEM), watershed boundary and river network data.

3. The two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids as described in claim 1, characterized in that: In step S3, the cross-mode transfer of coupling variables between CLM5.0 and SHUD is implemented as follows: interpolation operations are performed on the grid within the watershed area. The transfer from regular rectangular grids to unstructured triangular grids uses the SCRIP bilinear interpolation method, and the transfer from unstructured triangular grids to regular rectangular grids uses the SCRIP nearest neighbor distance weighted interpolation method.

4. The two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids as described in claim 1, characterized in that: In step S4, the simulation calculation of the land surface process model refers to the simulation calculation of surface processes and subsurface processes, including soil evaporation, snowmelt, vegetation biogeochemical processes, soil freeze-thaw, and vertical discretization of soil moisture. The simulation calculation of the hydrological model refers to the simulation calculation of surface hydrological processes and groundwater processes, including river flow, surface runoff, unsaturated and saturated layer flow, and groundwater level changes. The calculation of snow accumulation, canopy interception, and evapotranspiration surface flux is undertaken by the land surface process model.

5. The two-way coupling method for land surface hydrological models using multi-scale heterogeneous grids as described in claim 1, characterized in that: In step S5, the coupling time step for time synchronization is 1800 seconds, and the driving data time interval is 3 hours.

6. A system using the method as described in any one of claims 1 to 5, characterized in that: The system includes: The data collection module is used to collect data from multiple sources; The initialization module is used to initialize the CLM5.0 land surface process model, the SHUD hydrological model, and the OASIS3-MCT coupler; The main time loop module is used for parallel execution mode simulation calculations and performs bidirectional data exchange through the OASIS3-MCT coupler. The mesh interpolation module is used to perform conserved interpolation between regular rectangular meshes and unstructured triangular meshes; The time synchronization module is used to control the transmission frequency and time synchronization interval of the coupled variables; The inspection module is used to perform energy balance checks and convergence checks; The termination module is used to end the simulation and release memory resources.

7. The system as described in claim 6, characterized in that: The initialization module includes: The land surface process model initialization submodule, implemented using a Fortran interface, is responsible for reading meteorological forcing, surface parameters, and initial conditions, and performing regular rectangular mesh generation. The hydrological model initialization submodule, implemented using a C++ interface, is responsible for reading watershed data and hydrological calibration parameters, and performing unstructured triangular mesh generation. The OASIS3-MCT initialization submodule provides a Fortran / C dual-language interface with the same functionality, responsible for defining task partitions, coupled meshes, and coupled variables.

8. The system as described in claim 6, characterized in that: The grid interpolation module includes: Mesh definition and matching units support automatic topology recognition and matching of regular rectangular meshes and unstructured meshes; Intelligent interpolation algorithm unit, realizing bidirectional heterogeneous mesh interpolation and mass conservation guarantee mechanism; The coupled region control unit defines coupled regions and manages dynamic mask generation based on geographical boundaries.

9. The system as described in claim 6, characterized in that: The main time loop module includes: The land surface process simulation submodule performs calculations of land surface physical processes, including soil evaporation, snow melting, vegetation biogeochemical processes, soil freeze-thaw cycles, and soil moisture movement. The hydrological process simulation submodule performs calculations of watershed hydrological processes, including river flow, surface runoff, unsaturated layer flow, saturated layer flow, and groundwater level changes. The coupling interface submodule is responsible for data synchronization and exchange between the two modes.

Images