+ Summary
+ The water cycle is highly interconnected; water fluxes in one part
+ depend on physical and human processes throughout. For example, rivers
+ are a water supply, a receiver of wastewater, and an aggregate of many
+ hydrological, biological, and chemical processes. Thus, simulations of
+ the water cycle that have highly constrained boundaries may miss key
+ interactions that create unanticipated impacts or unexpected
+ opportunities
+ (Dobson
+ & Mijic, 2020;
+ Liu,
+ Dobson, & Mijic, 2022). Integrated environmental models aim
+ to resolve the issue of boundary conditions, however they have some
+ key limitations
+ (Rauch
+ et al., 2017), and in particular we find a significant need for
+ a parsimonious, self-contained suite that is accessible and easy to
+ setup.
+
+
+ Statement of need
+ Traditional approaches to water system modelling broadly fall into
+ highly numerical models that excel in representing individual
+ subsystems, or systems dynamics models that create broad
+ representations but that lack a physical basis. Early attempts at a
+ physical representation of the water cycle combined existing numerical
+ models through an integration framework
+ (Rauch
+ et al., 2017). While successful, this approach has an
+ incredibly high user burden because each subsystem model is so
+ detailed, and as a consequence is also difficult to customise. To
+ illustrate, SWAT is one of the most widespread models of the rural
+ water cycle
+ (Arnold
+ et al., 2012), while SWMM is the same but for the urban water
+ cycle
+ (Gironás,
+ Roesner, Rossman, & Davis, 2010). It has been demonstrated
+ that these two software can interface using the OpenMI integration
+ framework
+ (Shrestha,
+ Leta, De Fraine, Van Griensven, & Bauwens, 2013). Despite
+ this seemingly powerful combination of two near-ubiquitous models,
+ integrated applications have been limited, and we propose that this is
+ for the same reasons presented in
+ (Rauch
+ et al., 2017): user burden and customisation difficulty.
+ Because of this need, we provide a parsimonious and self-contained
+ suite for integrated water cycle modelling in the WSIMOD Python
+ package. It brings together a range of software developed over the
+ course of three years on the
+ CAMELLIA
+ project. Urban water processes are based on those presented
+ and validated in the CityWat model
+ (Dobson
+ et al., 2021;
+ Dobson
+ & Mijic, 2020;
+ Dobson,
+ Watson-Hill, Muhandes, Borup, & Mijic, 2022;
+ Muhandes,
+ Dobson, & Mijic, 2022), while hydrological and agricultural
+ processes are from the CatchWat model
+ (Liu
+ et al., 2022,
+ 2023).
+ WSIMOD also provides an interface for message passing between
+ different model components, enabling all parts of the water cycle to
+ interact with all other parts. The result is a simulation model that
+ is easy to set up, highly flexible and ideal for representing water
+ quality and quantity in ‘non-textbook’ water systems (which in our
+ experience is nearly all of them).
+ The package provides a variety of tutorials and examples to help
+ modellers create nodes (i.e., representations of subsystems within the
+ water cycle), connect them together with arcs (i.e., representing the
+ fluxes between subsystems), and orchestrate them into a model that
+ creates simulations.
+
+ Limitations
+ We highlight that WSIMOD is not intended to be a substitute for
+ sophisticated physical models, nor for a system dynamics approach.
+ In applications where detailed hydraulic/hydrological process
+ representations are needed (e.g., informing the design of specific
+ pipes, cases where processes are hard to quantify such as
+ representing social drivers of population growth, etc.) there are
+ likely better tools available. Our case studies highlight that
+ WSIMOD is most useful in situations where physically representing
+ cross-sytem processes and thus capturing the impacts of cross-system
+ interactions are essential towards the questions you ask of your
+ model. Secondary benefits are that the parsimonious representations
+ utilised are computationally fast and flexible in capturing a wide
+ range of system interventions.
+
+
+
+ Acknowledgements
+ WSIMOD was developed by
+ Barnaby
+ Dobson and
+ Leyang
+ Liu. Theoretical support was provided by Ana Mijic. Testing
+ the WSIMOD over a variety of applications has been performed by
+ Fangjun Peng, Vladimir Krivstov and Samer Muhandes. Software
+ development support was provided by Imperial College’s Research
+ Software Engineering service, in particular from Diego Alonso and Dan
+ Davies.
+ We are incredibly grateful for the detailed software reviews
+ provided by
+ Taher
+ Chegini and
+ Joshua
+ Larsen and editing by
+ Chris
+ Vernon. Their suggestions have significantly improved
+ WSIMOD.
+ The design of WSIMOD was significantly influenced by
+ CityDrain3
+ (Burger
+ et al., 2016),
+ OpenMI
+ (Gregersen,
+ Gijsbers, & Westen, 2007),
+ smif
+ (Usher
+ & Russell, 2019;
+ Usher
+ et al., 2018), and the following review
+ (Belete,
+ Voinov, & Laniak, 2017).
+ We acknowledge funding from the CAMELLIA project (Community Water
+ Management for a Liveable London), funded by the Natural Environment
+ Research Council (NERC) under grant NE/S003495/1.
+
+
+