Diffusive wave river routing, updated reservoir modelling, revised initialization of soil moisture

LISFLOOD-OS v5.0.0 includes the following changes:

1. A new optional module for river routing computations implementing the Muskingum-Cunge-Todini diffusive wave routing method (Todini (2007); Reggiani et al. (2014); Reggiani et al. (2016); Reggiani and Todini (2018)).
2. A new version of the reservoir module after Hanasaki et al. (2022) and Casado-Rodríguez et al. (2026).
3. A revised approach for the initialization of soil moisture state variables.
4. A patch in the module snow.py to promote snow redistribution and avoid non-realistically high snow cover values at very high altitudes.
5. A patch in the module waterabstraction.py for consistency with groundwater domain modelling approach.
6. A comprehensive workflow for the computation of the pre-run in temporal chunks
7. Full compatibility with python 3.10-3.11-3.12, new versions of netcdf and gdal libraries.

The conceptual changes are further detailed below:

1. Diffusive wave option for river routing computation: mct.py
   LISFLOOD v4.3.1: The modelling of river routing relies on the kinematic wave solution approximation of the shallow-water equations, with the optional Split Routing approach aiming to mimic floodplain behavior and flood wave attenuation. The kinematic wave approximation is computationally efficient and well suited for medium to high slope rivers (i.e. riverbed slope > 0.001). However, it does not include diffusion effects and this can lead to inaccuracies in the modelling of flood wave peak magnitude and arrival time in mild to low slope rivers (river slope < 0.001).
   LISFLOOD v5.0.0: The Muskingum-Cunge-Todini (MCT) wave routing method accounts for both convection and diffusion of the flood wave, while maintaining numerical efficiency (Todini (2007); Reggiani et al. (2014); Reggiani et al. (2016); Reggiani and Todini (2018)). The new optional module mct.py is designed to overcome the limitations of the kinematic wave numerical solution in the modelling of mild to low slope rivers.

2. New methodology for reservoir storage and outflow computations: reservoir.py
   LISFLOOD v4.3.1: Reservoir storage and volume followed the approach of Burek et al. (2013) in which reservoir outflow is a function of reservoir level of filling.
   LISFLOOD v5.0.0: Reservoir outflow depends on both revoir level of filling and inflow discharge. The revised module is an adaptation of the methodology of Hanasaki et al. (2022) according to Casado-Rodríguez et al. (2026).

3. Revised approach for the initialization of soil moisture state variables.
   LISFLOOD v4.3.1: Soil moisture state variables at the last time step of the prerun are used as initial values of the model run. In areas with very low precipitation amounts and thick soil layers, these initial soil moisture values might not be representative of the long-term conditions and lead to erroneous decreasing trends in soil moisture values.
   LISFLOOD v5.0.0: Initial soil moisture of the third soil layer values which are consistent with the local climatological and soil properties conditions are computed via an analytical pixel-wise solution. The implemented methodology is conceptually equivalent to the well-known strategy for the initialization of the lower groundwater zone.

4. Patch in the module snow.py to avoid non-realistic results in specific conditions.
   LISFLOOD v4.3.1: Snow melting can be underestimated in pixels at very high altitudes causing not realistically high values of snow cover.
   LISFLOOD v5.0.0: When snow cover exceeds a pre-defined threshold value of snow water equivalent, a fraction of the snow cover exceeding the threshold is ablated to the lower elevation fraction (from high to medium, or from medium to low elevation fraction) or added to snow melt (lowest elevation fraction). Both the threshold value (2000 mm water equivalent) and the fraction (1/100) were based on expert definition.

5. Patch in the module waterabstraction.py
   LISFLOOD v4.3.1: Industrial, domestic, livestock water demand are abstracted (also) from groundwater resources. While the consumptive water use leaves the system, the unused water volume is returned to the channels (return flow). This solution mimics reality (and it is consistent with LISFLOOD-EPIC approach). However, limitations of the groundwater modelling approach (i.e. lack of upper boundary for groundwater abstraction and lack of modelling of horizontal groundwater fluxes) can lead to overestimation of return flow and spurious river flow values.
   LISFLOOD v5.0.0: Only consumptive use for domestic, industrial, livestock purposes is abstracted from groundwater (return flow is zero). This patch accounts for the limitations of the groundwater modelling approach, and it avoids overestimation of return flow and spurious values of river flow.

6. Comprehensive strategy for the computation of the pre-run in temporal chunks
   LISFLOOD v4.3.1: LISFLOOD prerun simulation can be accomplished in temporal chunks to accommodate limitations (e.g. time wall) of the computational infrastructure. The prerun simulation can be split in a cold start and one or more warm starts. Average variables for the initialization of the LISFLOOD run must be computed by the user based on the results of each chunk of the prerun, and prior to the set-up of the LISFLOOD run.

LISFLOOD v5.0:0: LISFLOOD prerun simulation can be accomplished in temporal chunks, using a sequence of a cold and one or more warm starts. All the variables required for the start of the LISFLOOD run are available in the output folder of the last chunk of the prerun simulation.

LISFLOOD 5.0.0 modelling approach was used for GloFASv5 and EFASv6 calibration.