Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Fixed

  • Initialization of LateralSSF variables ssf and ssfmax with vertical hydraulic conductivity profile exponential_constant. Removed parameter khfrac from the computation, as it is already part of parameter kh_0.
  • Mutating BMI model control functions (update, update_until and finalize) and extended mutating BMI functions (load_state and save_state) should return nothing.
  • Added downloading of testdata to Dockerfile, to ensure an image was able to build.
  • The reservoir (reservoir_index_f) and lake (lake_index_f) indices as part of network.river were not correct. These were mapped to their own index in the SimpleReservoir and Lake struct, and not to the corresponding river index. This resulted in incorrect surface water abstractions from reservoir and lake volumes, and surface water abstractions were set at zero at the wrong river locations.
  • Wflow ZMQ server: allow JSON reading and writing of NaN and Inf values to avoid a JSON spec error. For example, during initialization of a wflow model, some (diagnostic) model variables are initialized with NaN values.

Changed

  • Removed vertical concepts HBV and FLEXTopo.
  • Removed metadata macros exchange and grid_type. The macro grid_type is not required because this functionality is already part of BMI. The macro exchange is replaced by a function used by BMI. Remaining metadata macros get_units and grid_loc are only used by BMI.
  • Refactor the vertical SBM concept: divide the long struct SBM into different model components for interception, snow, glacier, (open water) runoff, soil, water demand and allocation stored in the struct LandHydrologySBM. Additionally, the atmospheric forcing and a shared vegetation parameterset are stored as separate fields in struct LandHydrologySBM (with soil model SbmSoilModel). The model component structs have model variables, parameters and boundary_conditions (if applicable), including associated functions for initializing and updating these model components. The original long update function of the SBM soil part has been split into separate functions.
  • Refactor the lateral (routing) components: as for the vertical SBM concept split the structs into variables, parameters and boundary_conditions (if applicable).
  • Refactor the sediment concept: similar to the refactor of SBM concept, the sediment model has been split into three main structures. SoilLoss was split into rainfall_erosion, overland_flow_erosion and (total) soil_erosion concepts. OverlandFlowSediment has been split into transport_capacity, sediment_flux and to_river concepts. RiverSediment has been split into transport_capacity, potential_erosion, sediment_flux and concentrations concepts.
  • Timestepping method parameters for solving the kinematic wave and local inertial approaches for river and overland flow are moved to a TimeStepping struct. The timestepping implementation for the kinematic wave is now similar to the local inertial method: a stable timestep is computed for each sub timestep (or a fixed sub timestep is used) as part of a while loop (for each model timestep).

Added

  • Support direct output of snow and glacier melt, and add computation of snow water equivalent (SWE).

v0.x releases

v0.8.1 - 2024-08-27

Fixed

  • Reduce allocations in update of vertical SBM concept.

v0.8.0 - 2024-08-19

Fixed

  • Added missing BMI function get_grid_size, it is used for unstructured grids, for example to get the length of arrays returned by BMI functions get_grid_x and get_grid_y.
  • Added a check for the solution of the quadratic equation as part of the Modified Puls approach for lake outflow. Lower limit should be zero (very small negative values can occur).
  • Limit lake evaporation (added variable actevap) and lake outflow to prevent negative lake storage. The variable actevap has also been added to the reservoir module.
  • The set_states function for model type sbm with local inertial routing for river and land component.
  • Inflow to reservoir and lake locations for local inertial routing: 1) with floodplain routing, the floodplain discharge was not added to the inflow of these locations, 2) the to_river variable from overland flow and lateral subsurface flow was not added to the inflow of these locations.
  • Close netCDF NCDataset with state variables in extended BMI function save_state.
  • For the computation of Gash interception model parameter e_r multiply the precipitation input with the canopy fraction (this was only done for the potential evapotranspiration input).
  • Function BMI.update_until could throw an InexactError: Int64 caused by a not whole number, representing how many times (steps) an update of a wflow model is required. This is fixed by using divrem for the computation of the number of steps in this function. An error is thrown when the absolute remainder of divrem is larger than eps(), or when the number of steps is negative.
  • Fixed internal and external broken links in docs.
  • The internal time step of the local inertial model (stable_timestep function) can get zero when LoopVectorization is applied (@tturbo) to the for loop of these functions. This issue occured on a virtual machine, Windows 10 Enterprise, with Intel(R) Xeon(R) Gold 6144 CPU (2 processors). This has been fixed by replacing @tturbo with reduction of Polyester.jl.
  • Fixed required states of the model type sbm_gwf: added h_av for the river and land domain.
  • For the constanthead boundary of GroundwaterFlow the head should not be changed (was set to top elevation of the aquifer if head > top), and exfiltwater should be 0 for these boundary cells.

Changed

  • Stop exposing scalar variables through BMI. The BMI.get_value_ptr function was not working correctly for scalar model variables (a view was applied). Only a few scalar model parameters are defined, and it is not expected that exposing these variables is required (e.g. for model coupling) while code changes for these variables (including struct fields) are required.
  • The local inertial routing (constant) river boundary condition depth at a ghost node (downstream river outlet) was set at 0.0 in the code and can now be provided through the model parameter netCDF file (riverdepth_bc). In addition, the boundary condition river length dl at a ghost node should be set through the model parameter netCDF file (riverlength_bc), providing this boundary condition through the [model] settings in the TOML file (model.riverlength_bc) has been deprecated.
  • As part of the vertical SBM concept: 1) add variable net_runoff (land runoff minus actual open water evaporation ae_openw_l) and 2) change the definition of variable runoff by removing the subtraction of ae_openw_l (total land runoff). The variable net_runoff is now input to the kinematic wave for overland flow (replaces the original runoff variable as input). The lower limit of the original runoff variable was set at zero, which may result in water balance errors. This lower limit was implemented in the Python version of wflow_sbm, to avoid a very slow kinematic wave solution for surface flow computation (no distinction between a kinematic wave for overland and river flow). When in the Python version of wflow_sbm the kinematic wave for surface flow was split into a river and land component, the lower limit was removed for river runoff (net_runoff_river), but not for land runoff.
  • Always use fractions for the computation of potential evapotranspiration (interception and transpiration) and potential evaporation (open water and soil). Replaced variable et_reftopot by crop coefficient kc (use of et_reftopot has been deprecated).
  • For improved code readability it is now discouraged to use non-ASCII characters for the names of variables, structs, functions and macros. Using the non-ASCII character for built-in operators is still allowed. This change in naming convention is now in effect and all invalid uses of non-ASCII characters have been replaced by ASCII equivalents. This change also impacts the TOML keys kv₀, θᵣ and θₛ. These keys have been replaced with the ASCII versions kv_0, theta_r and theta_s in v0.4.0 and the old keys are not supported anymore.
  • Docs: 1) improved description of different model configurations in model-setup.md, also in relation to hydromt_wflow in docs, 2) citing info related to wflow_sbm publication in Geosci. Model Dev. (from in review to published).
  • Root water uptake reduction (Feddes): 1) extend critical pressure head parameters with h3_low and h3_high, 2) all critical pressure head parameters can be set (values for h1, h2 and h3 were fixed) and 3) the root water uptake reduction coefficient \alpha can be set at 0 (default is 1) at critical pressure head h1 (through the model parameter alpha_h1).
  • For the actual transpiration computation in SBM, the potential transpiration is partitioned over the soil layers with roots according to the model parameter rootfraction (fraction of the total root length in a soil layer). Previously, for each soil layer (from top to bottom) the actual transpiration was computed, and the remaining potential transpiration was used in the next soil layer.

Added

  • Total water storage as an export variable for SBM concept. This is the total water stored per grid cell in millimeters. Excluded from this variable are the floodplain, lakes and reservoirs.
  • Checks to see if all states are covered in the .toml file. If not all states are covered, an error is thrown. If there are more states specified than required, these states are ignored (with a warning in the logging), and the simulation will continue.
  • Support different vertical hydraulic conductivity profiles for the SBM concept. See also the following sections: The SBM soil water accounting scheme and Subsurface flow routing for a short description.
  • Local inertial routing to sbm_gwf model type.
  • Water demand and allocation computations for model types sbm and sbm_gwf.

v0.7.3 - 2024-01-12

Fixed

  • Documentation: add leakage term to the wflow_sbm figure, document external input parameter ksathorfrac and fix description of adding external inflow to the kinematic wave.
  • Fixed BMI functions (e.g. BMI.get_value) that deviated from BMI specifications (BasicModelInterface.jl), including function arguments, return types and the BMI specification that arrays are always flattened (this was not the case for variables stored as 2-dimensional arrays or as vector of SVectors).
  • Bump compat for NCDatasets to 0.13, 0.14.
  • The solution for lake outflow as part of the Modified Puls Approach. The inflow and outflow variables are defined for period Δt, and not at t1 and t2 (instantaneous) as in the original mass balance equation of the Modified Puls Approach. Because of this, the terms of the quadratic equation (and solution) were fixed.
  • Use kvfrac for the computation of vertical saturated hydraulic conductivity at the bottom of the soil layer, since kvfrac is also used for the computation of vertical unsaturated flow.

Changed

  • For cyclic parameters different cyclic time inputs are supported (only one common cyclic time (for example daily or monthly) was allowed).
  • BMI: 1) added grid information (type and location) and whether a variable can be exchanged to metadata Structs, 2) extend model grid functions for wflow components that store variables on edges (local inertial model) with get_grid_edge_count and get_grid_edge_nodes.

Added

  • Functions for loading and saving states and getting the starttime in Unix time format. This is convenient for coupling with OpenDA (and other external) software. The set states functionality from the initialization function has been moved to a separate set_states function for each Model type, to support loading states independent of initialization.

v0.7.2 - 2023-09-27

Fixed

  • Water balance of modified Rutter interception model. The sum of the stemflow partitioning coefficient pt and free throughfall coefficient p could get larger than 1, resulting in an overestimation of stemflow and throughfall amounts and negative net interception amounts. And the first drainage amount dd, controlled by a change over time in canopy storage capacity cmax, should not be subtracted from precipitation to compute net interception.
  • The netinterception (net interception) computed by the modified Rutter interception model was stored as interception in SBM, while this should be the interception (interception loss by evaporation) output of the modified Rutter interception model. The interception of SBM is used to compute the total actual evapotranspiration actevap.

v0.7.1 - 2023-06-30

Fixed

  • State initialization of 1D floodplain volume. In the initialization function the wrong field name of type FloodPlainProfile was used (area instead of a).

v0.7.0 - 2023-06-12

Fixed

  • BMI.get_time_units now gets called on the model rather than the type, like all other BMI functions, except BMI.initialize. Also it returns “s” instead of “seconds since 1970-01-01T00:00:00”, in line with the BMI specification.
  • Added the interception component to total actual evapotranspiration actevap of SBM (was defined as the sum of soil evaporation, transpiration and open water evaporation).

Changed

  • The time values returned in the BMI interface are no longer in seconds since 1970, but in seconds since the model start time. This is more in line with standard BMI practices.
  • The starttime was defined one model timestep Δt ahead of the actual model time (the initial conditions timestamp (state time)). As a consequence this was also the case for the current model time. To allow for an easier interpretation of wflow time handling, either through BMI or directly, the starttime is now equal to the state time, resulting in current model times without an offset.
  • Using more than 8 threads can result in too much overhead with Threads.@threads. After performance testing, this has been changed for kinematic wave routing and the vertical SBM concept to spawning tasks with Threads@spawn for number of threads <= 8, where each task iterates over a chunk of size basesize. For more than 8 threads the low overhead threading Polyester.@batch (including the minbatch argument) is used. For local inertial routing the use of Threads.@threads has been changed to threaded loop vectorization (river and 1D floodplain local inertial momentum equation) and Polyester.@batch.

Added

  • For (regulated) lakes with rating curve of type 1 (H-Q table), lake storage above the maximumstorage (based on maximum water level from the H-Q table) is spilled instantaneously (overflow) from the lake.
  • Added support to use sum as a reducer function for csv and scalar output options.

v0.6.3 - 2023-03-01

Fixed

  • Removed error when _FillValue is present in the time dimension of the forcing netCDF file. The simulation is allowed to continue with the attribute present, given that there are no missing values in the time dimension. This is checked by the code, and an error is thrown if this is the case.
  • Column index of daily lake rating curves. This was incorrectly based on dayofyear with a maximum of 365. The column index should be based on julian day (leap days are not counted).

Changed

  • NCDatasets version. Reading the time dimension of multifile netCDF file became very slow since NCDatasets v0.12.4, this issue has been solved in v0.12.11.
  • Store the time dimension of the forcing netCDF file as part of the struct NCreader instead of calling dataset["time"][:] each time step when loading forcing data.

Added

  • Show total duration of simulation in the log file (info), and show the current time at execution of each timestep (debug).
  • Support for exponential decline in horizontal conductivity in the sbm_gwf concept. This can be enabled using the conductivity_profile setting (either “uniform” or “exponential”). If set to “exponential”, it exponentially reduces the kh0 (or conductivity) based on the value of gwf_f to the actual horizontal conductivity (k).
  • An optional 1D floodplain schematization for the river flow inertial model, based on provided flood volumes as a function of flood depth per river cell. See also the following sections: SBM + Local inertial river and River and floodplain routing for a short description, and the following section for associated model parameters.

v0.6.2 - 2022-09-01

Fixed

  • Two issues related to reservoir and lake locations as part of local inertial model: 1) added as boundary points to the update of overland flow, 2) fixed check reservoir and lake location in update river flow.
  • Limit flow (qx, qy and q) in local inertial model when water is not available (set to zero).
  • In the grid output netCDFs, don’t set the _FillValue attribute, since the CF conventions don’t allow it.
  • Glacier parameter g_sifrac needs to be converted during initialization (time dependent).
  • The check that the sum of adaptive timesteps (Δt) of the local inertial model (1D and 2D) does not exceed the model timestep.

Changed

  • Changed depth h for reservoir and lake locations as part of the river local inertial model from bankfull_depth to zero.

Added

v0.6.1 - 2022-04-26

Fixed

  • Fixed an error with the log file, when writing to a folder that does not (yet) exists. Now, the folder is created prior to writing the log file.
  • Fixed a MethodError for read_dims, thrown when reading netCDF data with NCDatasets.jl 0.12.3 or higher.

v0.6.0 - 2022-04-14

Added

  • The FLEXTopo model.
  • Set a (different) uniform value for each index of input parameters with an extra dimension. A list of values (instead of one uniform value) that should be equal to the length of the extra dimension can be provided in the TOML file. See also Modify parameters.
  • Modify input parameters with an extra dimension at multiple indices with different scale and offset values, through the TOML file. See also Modify parameters.
  • Added support for netCDF compression for gridded model output, through the option compressionlevel in the [output] section in the TOML file. The setting defaults to 0 (no compression), but all levels between 0 and 9 can be used.

Changed

  • Re-organized the documentation: moved explanation of different model concepts to a model documentation section, added a user guide to explain setting up the model, added new figures to the description of wflow_sbm.
  • The unit of lateral subsurface flow ssf of LateralSSF is now m^3 d^{-1}. The unit was m^3 t^{-1}, where t is the model timestep. Other flow variables are already stored independently from t, this allows for easier interpretation and to use states independently of t.
  • Changed the reference level of water depth h and h_av of 2D overland flow (ShallowWaterLand) for cells containing a river from river bed elevation zb to cell elevation z.

Fixed

  • Fixed calculation of average water depth h_av of 2D overland flow (ShallowWaterLand) with the local inertial approach. The summation of h was not correct, resulting in too low values for h_av. For river cells of 2D overland flow h_av was only updated as part of the river domain (ShallowWaterRiver), this value is now also updated as part of the land domain (ShallowWaterLand).
  • Fixed the following two flow width issues for 2D overland flow of the local inertial model: 1) The flow widths xwidth and ywidth were mapped incorrectly (reversed) to the flow calculation in x and y direction, respectively. 2) For river cells the effective flow width for overland flow was not determined correctly: the riverwidth vector supplied to the function set_effective_flowwidth! covered the complete model domain and should have covered the river domain, resulting in incorrect river widths and thus incorrect effective flow widths for river cells.

v0.5.2 - 2022-02-03

Changed

  • Model types sbm_gwf and hbv use the same approach for the calculation of the drain width dw as model type sbm.
  • Renamed h_bankfull parameter to bankfull_depth for consistency between kinematic-wave and local-inertial river routing. When using the old name under the [input.lateral.river] section of the TOML file, it will work but it is suggested to update the name.

Added

Fixed

  • Model type hbv: the surface width for overland flow was not corrected with the river width.
  • Fixed use of absolute path for path_forcing in TOML file, which gave an error in wflow v0.5.1.
  • Fixed a crash when glacier processes are simulated as part of the hbv concept (Δt was not defined).
  • When the surface flow width for overland flow is zero, the water level h of the kinematic wave should not be calculated, otherwise this results in NaN values. When the model is initialized from state files, q and h are set to zero for indices with a zero surface flow width.
  • Fixed how number of iterations its for kinematic wave river flow are calculated during initialization when using a fixed sub-timestep (specified in the TOML file). For a model timestep smaller than the fixed sub-timestep an InexactError was thrown.
  • Fixed providing a cyclic parameter when the netCDF variable is read during model initialization with ncread, this gave an error about the size of the netCDF time dimension.
  • Fixed CSV and netCDF scalar output of variables with dimension layer (SVector).

v0.5.1 - 2021-11-24

Fixed

  • Fixed calculation of exfiltwater as part of the sbm_gwf model type. This was based directly on groundwater head above the surface level, without multiplying by the specific_yield, resulting in an overestimation of exfiltwater. This is required since the groundwater model estimates the head by dividing the volume by the specific yield or storativity of the aquifer. So, should the groundwater table rise above surface level, the head above surface level does not represent a water column one to one. (This also means the groundwater model (slightly) overestimates heads when the head rises above the surface level. However, this water is immediately removed, and the head will be set to surface level.)

Added

  • Optional dir_input and dir_output keys in the TOML, which can be used to quickly change the path for all input or output files that are given as a relative path.

v0.5.0 - 2021-11-12

Changed

  • Scaling of potential capillary rise is replaced by a common approach found in literature, based on the water table depth zi, a maximum water depth cap_hmax beyond which capillary rise ceases, and a coefficient cap_n. See also Capillary rise. Multiplying the scaling factor with the ratio of model time step and basetimestep in the original approach resulted in (much) smaller capillary fluxes at sub-daily model time steps compared to daily model time steps, and is not used in the new approach. Parameters cap_hmax and cap_n can be set through the TOML file, parameter capscale of the previous approach is not used anymore.

Fixed

  • Conversion of GroundwaterFlow boundaries \(\SIb{}{m^3 d^{-1}}\) as part of model concept sbm_gwf to \(\SI{m^3 s^{-1}}\) for sub-daily model time steps. For the conversion the basetimestep (86400 s) should be used (and not the model time step).

v0.4.1 - 2021-11-04

Changed

  • The \alpha parameter of the kinematic wave has a fixed value now and is not updated because of changes in water height (this could result in large water balance errors). See also Surface routing.
  • Cyclic input for other structs than vertical are also now supported (for example cyclic inflow to the river).
  • Moved update_forcing! and update_cyclic! functions to the run function. It is now easier to implement a custom run function with custom loading of input data (forcing and cyclic parameters).

Added

  • Check if reservoirs and lakes have downstream nodes. Without downstream nodes is not supported and in that case an error message is thrown that is easier to understand than the previous one: “ArgumentError: Collection is empty, must contain exactly 1 element.”
  • For the input.path_forcing TOML setting we use Glob.jl to expand strings like "data/forcing-year-*.nc" to a set of netCDF files that are split in time.
  • External water inflow (supply/abstractions) added to the kinematic wave inflow in m3/s. It is added/removed to sf.qlat[v] before computing the new q[v] with the kinematic wave equation.
  • Fixed values for forcing parameters are supported, see also Fixed forcing values.

Added

Fixed

  • River inflow for reservoirs and lakes in the kinematic wave. This inflow was based on sf.q[v] at the previous time step, and this has been fixed to the current time step.

v0.4.0 - 2021-09-02

Changed

  • Changed length units for lateral subsurface flow component from millimeter to meter. This means that state netCDF files from previous versions can only be reused if ssf is divided by 10^9.
  • Add snow and glacier processes to wflow_sbm figure of the documentation.

Added

  • Multi-threading of vertical SBM concept and lateral kinematic wave components (overland, river and subsurface flow) of wflow_sbm model SBM + Kinematic wave.
  • Improved error message for CSV Reducer.
  • The TOML keys kv₀, θᵣ and θₛ have been replaced with the ASCII versions kv_0, theta_r and theta_s, to avoid encoding issues with certain text editors. The old keys still work as well.

Fixed

  • Calculation of volumetric water content of vertical SBM (soil layers and root zone).
  • Update of satwaterdepth in SBM (evaporation was only subtracted from a local variable, and not from sbm.satwaterdepth).
  • Fixed the index lowerlake_ind of NaturalLake. Linked lakes were not simulated correctly (flows between upstream and downstream lake).
  • Interpolation function interpolate_linear(x, xp, fp) for CSV tables lakes. When x was larger or smaller than xp, maximum(xp) or minimum(xp) was returned instead of maximum(fp) or minimum(fp).
  • Fixed model timestep of reservoirs (SimpleReservoir) and lakes (NaturalLake) in relation to precipitation and evapotranspiration fluxes. This was set to the fixed wflow basetimestep of 86400 s, and should be set to the actual model time step from the TOML configuration file.
  • Add flux from Drainage (GroundwaterFlow) in the sbm_gwf_model to the overland flow component instead of the river component of the kinematic wave.
  • Fixed option constanthead = false (TOML file), constant_head field of GroundwaterFlow was not defined in this case. Fixed this by initializing empty fields (Vector) for struct ConstantHead.
  • Fixed return max(0, boundary.flux[index]) to return max(0, flux) the flux should be returned when cell is dry, no negative value allowed.
  • Fixed path to initialize lake to: dirname(static_path)
  • Fixed outflow = 0, when lake level is below lake threshold. Before a negative value could enter the log function and model would fail.
  • Fixed the lake storage initialization. For continuation runs (reinit = false), this caused the lake to be reset to the initial conditions instead of the saved state.

v0.3.1 - 2021-05-19

Fixed

  • Ignore extra dimensions in input netCDFs if they are size 1

v0.3.0 - 2021-05-10

Changed

  • New deposition process for coarse sediment in the reservoirs with a new parameter restrapefficiency in the sediment model.
  • New variables added to the LandSediment and RiverSediment structs in order to save more output from the sediment model.
  • Added variables volume and inwater to SurfaceFlow struct, this is convenient for the coupling with the water quality model Delwaq.
  • River water level (h) and discharge (q) forced directly into the RiverSediment struct (instead of using the OverlandFlowSediment struct first).
  • Require Julia 1.6 or later.

Added

  • Modify model parameters and forcing through the TOML file (see Modify parameters).
  • Run wflow_sbm (SBM + kinematic wave) in two parts (until recharge and after subsurface flow) from BMI, including the option to switch off the lateral subsurface component of wflow_sbm.
  • Support more netCDF dimension and axis order variants.

Fixed

  • Corrected a bug in sediment deposition in the river (case when incoming sediment load is more than the river transport capacity).
  • Fixed update of snow and glacierstore fields of HBV and SBM concepts by the glacier_hbv module.

v0.2.0 - 2021-03-26

Changed

  • Removed dependency of the f model parameter of wflow_sbm on the parameters \theta_{s}, \theta_{r} and M. This approach is used in Topog_SBM, but not applicable for wflow_sbm. The f parameter needs to be provided as part of the netCDF model parameter file.
  • Grid properties as cell length and elevation now stored as part of the model.land.network component and not as part of the vertical model components, as it is not used by these components. altitude (elevation) should now be provided as part of the [input] section of the TOML configuration file, and not as part of the [input.vertical] section.
  • Removed parameter \theta_{e} from SBM struct (not used in update). Parameters \theta_{s} and \theta_{r} included separately (instead of \theta_{e}) in LateralSSF struct, now directly linked to SBM parameters.
  • Improve error messages (netCDF and cyclic flow graph).

Added

  • Export of netCDF scalar timeseries (separate netCDF file from gridded timeseries). This also allows for importing these timeseries by Delft-FEWS (General Adapter).

Fixed

  • Model parameter Manning’s n now used during the update of the struct SurfaceFlow, to change the related \alpha parameter of the kinematic wave for channel flow.
Back to top